Jump to content


  • Content Count

  • Joined

  • Last visited

Everything posted by mitchco

  1. @Kvalsvoll I think you are taking my article too literally 🙂 I said similar frequency response, not exact. As already caveated in the article, several times, the LS50's were placed in a room null whereas the JBL's were not. This accounts for the difference in the 50 to 60 Hz dip. Again, as noted in the article, had I moved the LS50's to the exact same spot as the JBL's, would have resolved this issue and perhaps a few others. Also noted in the article, the science shows our ears/brain are not too sensitive to narrow band dips in frequency response below Schroeder. The JBL has a near infinite baffle whereas the LS50 does not. This will cause a different Speaker Boundary Interference Response (SBIR) above 100 Hz to around 400 Hz (directivity related as the polar response will be different for both speakers in this frequency range, aside from the fact that the LS50's were not in the same physical position as the JBL's...). The JBL's "constant directivity" comes into play at around 400 Hz: Where as the LS50's does not. The rest of the differences on up in the audio band are due to the directivity differences between the two speakers. As Floyd Toole and Sean Olive have often said, one cannot eq directvity. So both the measurement and correction software are "blind" to directivity, as we are measuring and correcting the sound power in the room (i.e. steady state response). The point of the article is to show that two speakers eq'd "similarly not exact" sound remarkably close, yet the big difference being how much room sound is let into the recording by the wide directivity differences between these two specific speakers that represent the near opposite ends of the directivity index scale. The intent is to have folks listen to binaural recordings to hear the audible difference and determine what one's preference is with respect to how much room sound one likes mixed in with the direct sound. It is nothing more than that. Enjoy the music!
  2. Hello @Kvalsvoll I see you just joined CA and this is your first post. Welcome! Thanks for your feedback. The only fair comparison is the LS50 plus sub vs JBL with sub as they have a similar frequency response, but very different directivity index. The LS50 standalone was to let people listen to what it sounds like without a sub. Checking on the specs, The little speaker is rated at 106 dB SPL max output and 2nd & 3rd harmonics (90dB, 1m) <0.4% 175Hz-20kHz with 85dB (2.83V/1m) sensitivity. I don't think the speaker was overloaded. at 83 dB SPL at the LP. I follow these levelling best practices using pro gear. 83 dB SPL I do not find too loud and has the right balance of bass to treble. For overly compressed material I drop the level down to 77 dB SPL. Kind regards, Mitch
  3. Hi Ram @Ragera Thanks.. I have not reviewed the Genelec 8531, so can't comment. Ram, the main issue is that few speaker manufacturers publish their directivity specs. I don't see any listed for the XTZ for example. You can get some ideas from Soundstage measurements from the NRC: https://www.soundstage.com/index.php?option=com_content&amp;view=article&amp;id=16 While the LS50's are a steal for a grand, they are geared more towards near and mid-field listening and not best suited for your application given size of room and listening distance. Unless of course your preference is for more room sound, but you indicated you wanted to reduce early reflections... Tough to recommend due to lack of speaker manufacturer's directivity data. As a rule of thumb, which you already know, is speakers with waveguides tend to have a higher directivity index than domes... This is why I use JBL, not only for the waveguide, but they publish their directivity specs. Without the data, it is difficult to recommend... Regards, Mitch
  4. Here are two extreme opposite loudspeakers when it comes to sensitivity, size, directivity, topology and looks. Does the big speaker sound perceptually larger than the smaller speaker? Do the speakers sound the same if eq’d the same and level matched? Does the LS50 sound perceptually bigger when a subwoofer is added? These are some of the questions I attempt to answer in this article. While there are objective measurements, I have included binaural recordings comparing the level matched loudspeakers so you can hear over headphones what the differences are. Prepare for an ear opening experience. I love big, high efficiency loudspeakers. I feel they sound more “dynamic” than smaller low sensitivity “audiophile” loudspeakers. Is it my biased belief or does it prove out audibly? You be the judge by listening to the binaural recordings. Finally, consider this article a review of the lovely and talented KEF LS50, a precision engineered loudspeaker designed to sound good in your room. This pair is staying in my room. KEF LS50 There are many positive reviews about the KEF LS50 mini monitor from various publications like Stereophile and Soundstage. I am not going to repeat what has already been written, but I want to point out a key aspect on why the LS50’s sound so good, as evidenced by listeners and reviewers alike since its introduction in 2012 . KEF has written a white paper on their technology progression through the years called, The Reference. It is a wonderful 29 page read if you have an appreciation for fine engineering design. A key aspect of what makes these speakers sound so good is derived from this example set of frequency responses in this chart: “Figure 12 shows a set of frequency responses for an early prototype of Reference 5. The curves shown in this plot are the figures of merit for assessing loudspeakers as suggested by the work of Floyd Toole. The curves are measured at 96 data points per octave without any smoothing. Based on these figures of merit Toole was able to predict real listener preference with a remarkable accuracy.” The last sentence is the key idea I want to bring to your attention and expand on a bit. The chart above is often referred to as a “spin-o-rama” in which you can read about it here in detail (from 2002!) or if you simply want to know how to read the chart. The science behind this enables the key aspect of, “predict real listener preference.” I have written a summary about this before, “The Science of Preferred Frequency Responses for Headphones and Loudspeakers”, peer reviewed by Dr. Sean Olive. I bring this up as I took a science leap of faith based on Floyd’s (and Sean’s) research, which was leveraged and acknowledged by KEF in the Reference whitepaper. I purchased a pair of LS50’s without listening to them. But I already knew from product reviews and measurements, and in particular, Floyd’s predictive model, that the LS50’s will sound neutral in my listening room. As one will hear on the binaural recordings and shown in my in-room measurements, the LS50 is a remarkably neutral sounding loudspeaker. LS50 Objective Measurements: First, let’s look at a near field (30cm from driver on-axis) in-room frequency response measurement and compare to the NRC’s anechoic measurement: Courtesy Soundstage/NRC Remarkably similar. My room is certainly not an NRC anechoic chamber, as can be seen in the room picture above. But I can capture the sonic signature of the LS50 with a 30cm in-room nearfield measurement. Thinking of it another way, it also verifies my acoustic measurement setup and capture is reasonably accurate. Note the gentle rise starting around 2 kHz. If listening to the LS50’s nearfield on a computer desktop, you may need to dial in a high frequency shelf starting at 2 kHz to reduce the high frequency energy ever so slightly. If I wanted to use these as computer desktop speakers, I would still put them on stands behind the desk and raise them up, and if possible, some distance back from the front wall. Moving the mic back ~9ft to the listening position: Now we are seeing the room enter the picture, as I have done a fantastic job of placing the LS50’s directly in a room null at 70 and 80 Hz relative to the listening position. ☺ That’s a 20 dB dip and peak over a narrow frequency range. Yes, my room ratio is not good. As perceived by our ears, a 20 dB difference is 4 times as loud or quiet. However, given how narrow bandwidth the dips are, it is questionable on how much we really (don’t) hear it versus how much we just adapt to dips. According to this research, we are more likely to hear peaks over dips. It has been my experience as well. It seems my ears/brain fill in and/or adapt to the sound with dips in frequency response, but more sensitive to peaks. More research required. Take a look at the picture of my room and where the LS50’s are located. Ideally, I would have moved the big JBL’s out of the way, and placed the LS50’s in a known good position, relative to the listening position. I have done this in my previous loudspeaker reviews with no nulls. However, this setup will give us a good audible test of a) can we hear the null? b) can we hear digital room correction doing its job? Beyond 100 Hz we see a smooth frequency response, with the characteristic downward tilt of a perceptually neutral sounding loudspeaker, as per the Science link above. Let’s see how the LS50 compares to my “reference” JBL 4722: The LS50 frequency response is overlaid with my JBL 4722 system with dual subs, which uses the preferred in-room measured frequency response, as referenced from Toole’s and Olives research. Meaning, the 4722’s with dual subs have been DSP’d from 12 Hz to 21 kHz ±3 dB to my preferred target frequency response with a slope from 20 Hz to -10 dB at 20 kHz. The KEF LS50 hits that predicted in-room response based on its spinorama figures of merit quite closely. Validating Floyd’s predictive model… again. It is likely if I moved the LS50’s to the location where the JBL’s are, in an equilateral triangle, then the off axis response may drop those +1 dB bumps between 2 and 3 kHz, 4 to 6 kHz, and 20 kHz down a bit, as evidenced by the Soundstage NRC off axis measurements. I wrote a soapbox piece on why more loudspeaker designers aren’t using this publicly available Standard that is based on Toole’s research that clearly works for KEF (plus JBL, Revel, and a few others). Most speakers I listen to are too bright, as indicated in my other loudspeaker reviews. Only the Dutch and Dutch 8c came out of the box with nearly identical in-room frequency response matching the preferred target. The KEF LS50 is the 2nd speaker I have measured that matches the preferred in-room target frequency response for a neutral sounding loudspeaker, right out of the box. I would hazard a guess that most folks would enjoy its neutral response. Let’s listen. LS50 Subjective Listening A little preamble before we listen. I used the same binaural recording setup in this article for reviewing a set of neutral sounding headphones. The binaural mic specs and preamp frequency response measurements are there. The headphones measured ±3 dB from 20 Hz to 7 kHz and then rolling off smoothly after that. What goes on beyond 7 kHz is largely due to ones ear and how the mics are placed in the ear. My article linked above shows some of this variability. Point being, the binaural mics are quite flat from 20 Hz to 7 kHz with a smooth roll off. I wasn’t striving for ultimate recording fidelity, I wanted to have the binaural recording sound good enough, and with high enough resolution, to capture tonal differences between loudspeakers. This comparative objective and subjective analysis is what I want to focus in on. As one will hear in the binaural recordings, the top end “air” (i.e. top octave 10 to 20 kHz) is down a bit in response, as compared to when I listen to the speakers “live”. Just wanted to point out that difference between what I hear live, versus what’s on the binaural recording. Another psychoacoustic effect is that I hear more of the room sound mixed in the binaural recordings versus when listening live. Floyd Toole’s research has indicated that we have a tendency to “listen through the room” and quickly adapt. Anecdotally, it seems the same for me. When I listen to the speakers live, I don’t hear as much of the room, as I do on the binaural recordings. However, one interesting aspect to listen for when we get to comparing wide directivity (LS50) versus narrower directivity (4722) speakers, even though the tone quality is similar, you can hear more room sound in the LS50 binaural recording versus the 4722 which presents more direct sound due to the speakers (much) higher directivity index. The graphic above shows the measured directivity over frequency for a double 15” and 90 x 50 CD horn, which are the JBL 4722’s. The 15” woofer and 120 x 100 CD are the JBL M2’s and the Domestic cone/dome system is the Revel Salon 2. The KEF LS50 is a cone and dome system and likely similar to the Salon 2 from a directivity perspective. As we will hear, the directivity index is an important factor on how much direct versus reflected sound one prefers. In the case of this JBL M2 versus Revel Salon 2 shootout, most folks preferred the Salon 2, with the lower directivity, even though both speakers have very smooth and similar frequency responses. This later spawned a thread at diyAudio on The Preference for Direct Radiators. The audible difference is significant enough that with a little ear training on what to listen for, one can tune into that direct versus diffuse sound and start to hear the difference right away. The other aspect one can hear on the binaural recordings is that the left side of my room is a bit livelier than the right, as I moved the acoustic drum kit that was there, and heavy quilt I put over top of it when critical listening, downstairs. The kit used to sit to the left of the couch, but currently is an open reflective space: At 83 dB SPL reference level, which is the SPL I took the measurements and the binaural recordings at, while not overly loud, I can definitely feel the sound pressure level on my couch, not only bass frequencies, but lower mids as well. Of course, that feeling does not translate in the binaural recording. There is deep bass on the binaural recordings with the subs added, so one might want to check with Tyll’s headphone measurements to see if the headset being used can reproduce down to 20 Hz. I used a pink noise wave file and a calibrated sound level meter to set the reference SPL and subsequently level match the other binaural recordings. With that in mind, binaural recordings are intended to be listened to over headphones. For best effect, close your eyes and visualize being seated in front of the stereo and turn it up a bit, as the dynamic range of the recordings are as good as the source. KEF LS50 binaural recording (53 MB wav DR12) I chose Mark Knopfler’s Back on the Dance Floor for a couple of reasons. It is a rock recording, with DR10, downloaded from iTunes simply as a “typical” recording. The drums and instruments are well recorded with a few notes of Mark’s famous “Money for Nothing” guitar sound resonating throughout. But most importantly, it is a simple sparse dynamic recording with Mark’s vocal up in level in the mix. This gives us a good opportunity to focus on the vocal tone and make some direct judgements when we listen to comparison recordings. Disclaimer: The binaural audio clips are not full song recordings and conform to the spirt of “fair use” for copyright material for the purpose of education and research. The author nor the publisher derives any financial benefit from these audio clips. The binaural recordings were made at 44.1/16. How does it sound? Remember you are listening to the speakers plus room sound combined. The binaural effect is intended for one to perceive feeling like you are situated in the acoustic environment. To my ears, the LS50 has a neutral tonal balance with some decent (perceived) bass from a 5.25” woofer in a larger living room area (30 x 16 x 8 ft). Not overly bright (yay!). Not much low frequency response, but you don’t miss it due to the nice neutral balance of this mini monitor. Sounds full, not thin. I can perceive a little bit one note-ish or coke bottle bass character effect at 100 Hz, due to room acoustics, but otherwise nice and smooth. Mark’s vocal sounds clear and the drums and cymbals sound dynamic to my ears. For me, I always envision being at the club or concert hall listening to the band live, as how I want the auditory scene to sound in my living room. The LS50 mini monitors throw a huge image, as the off axis response is as good as the on axis response, especially in the vertical plane as well as horizontal. A key Toole figure of merit that results in a good sounding loudspeaker in a typical living room environment. The KEF whitepaper goes into detail on the variety of tech that makes the drivers and baffle subjectively disappear when listening. You can hear how big the image is on the binaural recording. KEF’s computer aided design has really paid off in making this mini monitor sound much larger than its physical size indicates. LS50 plus Rythmik L12 dual subs Objective Measurements I patched in the Rythmik dual L12 subs using Audiolense to supply the linear phase digital XO at 70 Hz and time alignment to the LS50’s. Additionally, Audiolense applies so called Digital Room Correction (DRC), both in the frequency and time domains. The combo of subs and DRC fill in the low end frequency response and gets rid most of the nulls at 70 and 80 Hz: Pretty good given that the LS50’s are located right in a deep null, relative to my listening position. It is just sheer laziness on my behalf to not have moved the JBL 4722’s out of the way and placed the LS50’s there, where I know there is no deep null, as I have measured it before, with a variety of speakers. Also, I spent all of 5 minutes with the Audiolense correction, with no fine tuning. But close enough for rock and roll, as it is a pretty narrow dip of 5 dB as opposed to 20 dB null relative to 100 Hz in the measurement of the LS50 by itself. I wonder if we can hear the null? While I could have set Audiolense to only correct below the room’s transition frequency, (i.e. Schroeder frequency) that is often recommended, I went for the full frequency correction, as we want to make a few comparisons. I used the exact same target frequency response design as my JBL 4722 plus subs, so it would have also corrected for those little 1 dB high frequency bumps as measured on the LS50’s with no eq. Here is the objective comparison: Still amazing to me how close the LS50’s are to my preferred target frequency response right out the box. You can see the little high frequency bumps on the LS50’s, with no eq, at 2 to 3 kHz, 5 to 6 kHz and near 20 kHz. Had I taken the time to position the LS50’s correctly, I suspect those bumps would be gone. It will be interesting if we can detect an audible difference in a comparison. But before we get to that, let’s listen to the LS50’s with the dual subs, which we see measures almost identical to out of the box LS50’s beyond 100Hz. LS50 plus Rythmik L12 dual subs Subjective Listening Here is the same Mark Knopfler tune, at the same level, as with the LS50’s, except added dual subs and my preferred frequency response target, which is virtually identical to what comes out of the box with the LS50’s KEF LS50 plus subs binaural recording (56 MB wav DR11) How does it sound? Certainly on my NAD HP50 headphones I can hear the extended bass response. Let’s see how it sounds when we compare the LS 50’s directly with the LS50’s plus dual subs on the same track. The Rythmik L12 subs are Rythmik’s entry level sealed sub for about $540 each. Rythmik’s subs with direct servo tech sound tight and musical to my ears. LS50 versus LS50 plus dual subs subjective listening comparison A common task working with Digital Audio Workstations (DAW’s) is to import tracks, line them up and switch back and forth to compare. That’s what I did here. I took the two binaural recordings and imported them into my DAW, lined up the tracks to be close enough to sound synchronized: The top binaural stereo track is the LS50 and the bottom track is the LS50 plus dual Rythmik L12 subs. One is also seeing the DAW automation below each track where one track is unmuted for 10 seconds and then it mutes and switches to the other track for 10 seconds, and so on. In other words, every 10 seconds we alternate from one binaural stereo recording to the other. We start with the LS50, and then switch to the LS50 plus subs for 10 seconds, and then switch back on the LS50 for 10 secs and so on, ending on the LS50 with subs. Listen closely as the first 8 seconds is mostly intro and after two beats of the bass guitar, you are hearing the LS50’s plus subs. Meaning 10 seconds goes by quick as the first 8 seconds is the intro… LS50 compare to LS50 plus dual subs binaural recording (37 MB wav DR11) How does it sound? It may take a number of listens to tune into the differences. You may notice a “splice” or audible tick sound when the track switches. I left those in most places on purpose to give folks an audible cue when the track switches to the other binaural track. Some switches don’t make the tick sound, and requires a bit more concentration to hear the transition. It’s good ear training. In some switches, especially with the full on vocal in the chorus, it is tougher to hear the difference. But when it is just the drums and bass are playing, it is easy to hear the full range dynamic sound with the subs and fine tuning of the tone. Fuller sounding to my ears and not quite as boxy sounding. In some switches, I can hear the tone change. It almost sounds like a small pitch change. There is one switch near the beginning where the drums almost jump out when switching from the LS50 to the LS50 with subs. I checked the levels to make sure everything was properly level matched and could not find any errors. That’s also the same switch where I can hear the tone change with the drums are almost lower pitched when listening to the LS50 plus subs. Sounds more full range or “bigger” sounding to my ears. The vocals are a bit harder to tell apart, especially during the chorus with the layering and female voices, sounded more similar than different in a couple of switches to my ears. Now let’s compare this set up with the JBL 4722 with subs using the same target neutral frequency response. JBL 4722 plus subs Objective Measurements I have covered off detailed objective measurements of the JBL 4722 with dual Rythmik subs before and not going to repeat here. Here are the JBL’s overlaid with the LS50’s, both using the same subs (different XO point) and same target frequency response: That’s a close match. Getting into Bob Carver territory ☺ Had I moved the LS50’s right in place of the JBL’s, would have gotten rid of the last little bit of low frequency dip. Plus may have balanced out the 2 to 3 kHz bump a bit more. OTOH, at that frequency range, there is more reflected sound with the LS50’s and I would say it’s showing up in the REW measurement as added amplitude. Why is that? Audiolense’s frequency dependent correction windowing can be adjusted and in my case, at this frequency, (i.e. 2 to 3 kHz), Audiolense is correcting more of the direct sound, like a tone control. However, in REW, the acoustic measurement software used above, the time window is fixed at 500ms and letting more of the room into the measurement at higher frequencies, which is possibly reflected in the measurement. Which is correct? This is where ones ears comes into play as to what one prefers, a more direct sound or a more diffuse sound. Even if I lined up the KEF’s perfectly in place of the JBL’s, they would still have more room sound as their directivity is considerably wider than the JBL’s overall. And likely not quite the emphasis in the 2 to 3 kHz range either. When we listen to the compare, can we hear a difference in directivity, even though the loudspeakers are tonally (i.e. frequency response is) very similar? First, let’s listen to the JBL’s. JBL 4722 with dual subs Subjective Listening Here are the JBL 4722 with subs to get used to their sound: JBL 4722 with dual subs binaural recording (68 MB wav DR12) Depending on when folks listened to the other binaural recordings, and remember the room sound, this recording is different in the sense that there is (much) less room sound and more direct sound as the JBL’s have a much higher directivity index than the LS50’s. Really, these JBL’s are intended to be used in Cinema installations with hundreds of people in a much larger room than a living room. So in my living room, it almost sounds like headphones with little room sound. This binaural recording I find has a more focused stereo phantom image in the center than the other recordings with the LS50. I would say mostly due to the increased directivity. It is not that the LS50’s don’t have an outstanding stereo phantom image, it is simply that with their wider directivity pattern, and at my 9ft listening distance, has more room sound in the mix and sounds more diffuse, and as a result, not as focused. I spent quite a few years in pro sound and recording studio control rooms. The former used pattern controlled speakers and the latter was in rooms that were pretty absorbent. I got used to liking more direct sound than diffuse sound. If I was a classical music lover and frequented concert halls, likely my preference would be reversed. As an interesting aside, these JBL’s are horn/waveguide loudspeakers crossed at 630 Hz. I do not find them bright or fatiguing in anyway, no matter how many hours I listen to them. Again, I point to Toole’s and Olive’s research on what makes for a good sounding loudspeaker in a typical living room environment. Neutral sound is neutral sound regardless of the device used to present it. But as we will hear, the directivity index of a loudspeaker determines how much of the direct versus reflected sound one is listening to and can be used to determine ones listening preference. Let’s take a listen to David versus Goliath JBL 4722 versus LS50 both with dual subs subjective listening comparison Like before, I loaded up the two binaural tracks in my DAW and in ten minutes, edited it to swap tracks back and forth every 10 seconds: The JBL 4722 with subs starts first and then the LS50 with subs starts 10 seconds later while the 4722’s are muted and then switch back again 10 seconds later. It is tricky though, as it is only the first two beats of the bass in the first 10 seconds of the JBL 4722 with subs that one hears, before it switches to the LS50’s with subs. That’s because the first 8 seconds is the intro with no bass, and so it is hard to hear the initial transition. But over the course of recording, it is obvious in spots when hearing lots of room and then almost no room where one can readily identify the difference in speaker directivity. JBL 4722 compare to LS50 plus dual subs binaural recording (45 MB wav DR12) Tone wise, they are pretty close. What do you think? There is a small tonal change, as I can hear the room sound being added with the LS50’s. You can hear it almost like a slap echo (that’s an exaggeration) on the drum tone and vocal. It does open up the sound, as it is a bit brighter. Sounds more diffuse. The low frequency sounds similar on both, even though I cross the subs at 45 Hz on the JBL 4722’s and 70 Hz on the LS50’s. Focus in on the kick drum plus the bass line and try to notice any changes when the track switches. Should sound pretty consistent. For fun, go back to the LS50 and LS50 plus subs comparison and focus in on the kick drum and bass line again. One should start hearing the individual notes in the LS50’s with subs, as each one is equal loudness, but some low bass notes are missing on just the LS50, other than the 100 Hz overtone. Just like the other comparison, it may take a few listens or so to start repeatedly recognizing one speaker versus the other. While we may form an initial impression, repeated listening while focusing on different aspects of the sound starts to gather a collection of differences and more importantly personal preferences. For example, KEF has done an excellent job on making their mini-monitor sound bigger than it is with their computer aided design around speaker and baffle dispersion characteristics. Does one sound bigger or larger than the other on the binaural recordings? One area up for discussion is the lower mids. I do hear a difference in the lower mids between the 5.25” woofer and the double 15” JBL’s. Mark’s voice appears to have more weight or lower mid tone behind it. Of course, that is subjective on my behalf. It is very close sounding, but maybe the added room sound to the LS50’s is masking that. Will be interesting to hear what peoples thoughts are on the comparisons. Conclusion I hope folks find the binaural recordings fun and educational. For me it was an experiment on multiple levels and if done again, I would: have moved the JBL’s out of the way and moved the KEF LS50’s right in the same spot. This would have accounted for any low frequency null differences, as it is a known good location. However, I really did hear a difference with the LS50’s with subs with the -5 dB narrow dips at 70 and 80 Hz compared to the JBL’s with subs. Do you? moving the LS50’s to the proper location may also account for the small bumps in the high frequency response. However, the overall room amplitude would still be the same, but the frequency distribution smoother perhaps. I could have spent more time toeing the LS50’s in or out a bit more and re-measuring for proper set up with a few initial listening tests to check the tonal balance. I just placed the Skylan 24” stands and put the LS50’s on top – done. Try a smaller equilateral triangle with the LS50’s just to compare the direct versus reflected room sound to see if it was proportionally less, as one moves closer to the speakers. It should be… Replace my microphone preamplifier with a studio quality mic preamp. I ordered one based on these shootout results. The Lynx Hilo ADC is noted for its transparency and that has been my experience, but the mic preamp I have now is not as transparent. Try different music material. Open for suggestions. Having said that, the differences between David and Goliath (with subs) are not that big. No pun intended. The LS50’s are keepers. They sound excellent and a steal for $1000. I am happily listening to them now without subs but will be adding the Rythmik L12’s later when I upgrade the subs for my 4722’s. KEF, and the science, can predict a neutral sounding speaker in most listening environments. The LS50’s measure and sound neutral in my room. It is quite the engineering feat for a mini monitor to sound this good with this size image reproduction. Spend another grand to get a pair of subs and I think you would have a hard time beating that $2K combo as a full range system. One would have to spend a great deal more for something that may be incrementally better or simply worse if not based on sound engineering principles (i.e. spinorama). There is a point of diminishing returns to achieve neutral sound if that is ones goal. Of course, there are some SPL restrictions in the case of the LS50’s, but if you mostly listen at reference level (i.e. 83 dB SPL), and your room isn’t huge, these can easily do the job and sound neutral. I do think the LS50’s sound “bigger” (i.e. full range) with the dual subs. I point to near the front of the binaural comparison track between the LS50’s and LS50’s with subs where the latter jumps out when listening to one of the transitions that is mostly drums and bass. I hope I got the levels right, but could not find an error. It sure jumps out to my ears sounds more dynamic, full range, bigger… The big difference I hear when comparing the 4722’s to the LS50’s, both with the L12 dual subs, is the room sound due to the large directivity differences between the two speakers. This is after eq’ing (i.e. with DSP both in the frequency and time domains) both speakers and subs using the same target frequency response, which is pretty much exactly what comes out of the box with the LS50’s down to its low frequency cutoff. My JBL 4722’s are a fully optimized triamped system with linear phase digital XO’s, time alignment of drivers, tight frequency response tolerance across the bandwidth, yet the LS50’s with subs sound similar tonally. One can hear it on the binaural comparison recording. Does one sound bigger than the other? After repeated listening, there is quite a bit of room sound with the LS50’s, as compared to the 4722’s. Once ones ears are tuned to the differences, it becomes much easier to listen again and identify which speaker one is listening to. Both sound full range. Using the same subs for both, takes away the low frequency differences. However, I do notice on Mark’s vocal that it sounds fuller or deeper in tone on the JBL’s. I can’t be sure if that has to do with the cone/box size difference or the directivity difference where the room sound is changing the tone of the lower mids on the LS50’s. Next Steps There are many variables to making a binaural recording that is transparent and high resolution enough to convey the characteristics of a loudspeaker or any piece of audio equipment for that matter. Are the binaural mics in the ears correctly? Is my head in the same location from one recording to another? Level matched correctly? And so on. However, I think with care optimizing the recoding chain and looking at how to adjust for individual HRTF, binaural recordings offer another level of equipment review experience on an item that you may not get to hear in person and can shed some light on its audible performance. Even if only the relative comparison is used, it is still a useful tool. In the case of completely opposite loudspeakers eq’d the same, one can hear the audible difference between a speaker with wide directivity versus as speaker with narrow directivity and what your preference might be. Of course, there is no substitute to hearing the equipment “live” and perhaps in your own home. However, I am suggesting that it is time to kick up audio reviewing a notch by providing music lovers and gear listeners with more than just words. This approach can be used to identify fine details between gear not only over loudspeakers, but headphones as well. In the case where we might want to eliminate the room and listen to two different DAC’s for example. Sure, there are issues with HRTF’s and binaural mic placement, in addition to the myriad of different sounding headphones, but if there is a relative comparison change that is audible, then it would be captured in the binaural recording and available for all to hear. I am thinking of adding this ABX Switch Comparator to the mix to make it even easier to switch and record the comparisons, as opposed to having go through an editing stage. It would also make the process more reliable with less steps where things can go wrong. Based on this experiment of comparing the KEF LS50 with the JBL 4722, I feel I have only scratched the surface of what the possibilities are. This opens the door to compare digital signal processing as well, not only digital room correction products, but surround sound processing, upmixers, digital filtering of any kind, etc. As you may know, I am big on measurements, and will continue to do so, along with perhaps getting into electronics and digital signal processing measurements. More importantly, I wanted to find a way for folks to listen to the device under review, whether by itself or through a comparison, as added review information that is not currently available. I feel my first experiment comparing two very different loudspeakers and being able to audibly hear the difference is hopefully useful information. I have a few details to iron out to make the binaural recording process predictable and repeatable, but I think that it will be worth the effort. I always found it odd in “audio” equipment reviews that up until now, there was no way to listen to the device under test. Hopefully, this approach will change that. Happy New Year and Enjoy the Music! I wrote this book to provide the audio enthusiast with an easy-to-follow step-by-step guide for designing a custom digital filter that corrects the frequency and timing response of your loudspeakers in your listening environment, so that the music arriving at your ears matches as closely as possible to the content on the recording. Accurate Sound Reproduction using DSP. Click on Look Inside to review the table of contents and read the first few chapters for free. Mitch “Mitchco” Barnett. I love music and audio. I grew up with music around me, as my mom was a piano player (swing) and my dad was an audiophile (jazz). My hobby is building speakers, amps, preamps, etc., and I still DIY today. I mixed live sound for a variety of bands, which led to an opportunity to work full-time in a 24-track recording studio. Over 10 years, I recorded, mixed, and sometimes produced over 30 albums, plus numerous audio for video post productions, in several recording studios in Western Canada.
  5. Integrating Subwoofers with Stereo Mains using Audiolense In this article, I walk through the steps of using Audiolense to create a digital crossover and time align dual subs with stereo mains. In addition, showing how to smooth the frequency response and reduce group delay at the listening position. This results in a smooth frequency response (12 Hz to 22 kHz ±3dB on my system) with all direct sound arriving at my ears at the same time. The phase response and group delay is mostly flat at the listening position. In addition, I walk through a time domain experiment designing two correction filters with the same frequency response, but one with time domain correction and one without. I discuss the audible differences through listening sessions. Why subs for music? In Audiolense Digital Loudspeaker and Room Correction Software Walkthrough, I was able to smooth the frequency response of my JBL cinema loudspeakers, time align the drivers, and achieve relatively flat phase and group delay at the listening position. I am happy with the results, but something is missing… My JBL speakers have solid output to 40 Hz and can extend in room response close to 30 Hz. Great high efficiency kick and punch, but missing a bit of weight on the bottom octave. I listen to rock, blues, and alternative music, most of which does not have deep bass (bass guitar low E is 41 Hz), so if I added subs, would I notice? Enter Rythmik Audio. A company that has been around for many years with a good engineering and “no-hype” reputation. Being a tech geek, I was intrigued by their direct servo technology and the most extensive FAQ I have seen from any sub manufacturer. Most importantly for me, one of the few sub manufactures that publish their own measurements and validated by 3rd party testers. I purchased two F12 entry level music subs direct from Rythmik. They arrived in a timely fashion and extremely well packed – thick cardboard box, within another thick cardboard box, with the sub floating in high density foam. Could drop off the end of the truck with no damage. Very nice. Sub Setup Much has been written about setting up subs in numerous configurations. One can even use room simulation software to determine best placement. However, since we are using DSP, try whatever sub setup and configuration you wish. If your ears (and measurements) disagree with the sound, then some fine tuning of placement may be required. However, using this advanced DSP, it is likely the measured results and listening experience will be more than acceptable. It is generally accepted for frequencies below 80 Hz, it becomes difficult to determine a sound's location. If you click on the link, you can try it in on your own system. As an ex-recording/mixing engineer, I can say for the recordings and mixes I worked on, low frequencies below 100 Hz, whether from bass guitar, drums, piano or synthesizer, were always in the center of the mix and never panned. This seems to be the case for most music, except from the 60’s and other recordings where it is intended as an effect. So, why “stereo” subs? I wanted more sub output, without having to buy a bigger sub. Whether it is one or multiple subs, it is important to be able to individually control the frequency and timing response of each sub with respect to the mains. This is the key takeaway from the article. In my case, I have control over the frequency and timing response of all drivers in the system. Setup: Is this the best location for my subs (i.e. between the JBL’s and electronics)? Probably not. It is more about convenience than anything else. Once measured and listened to, they sounded good no matter where I am in the room. However, I am mostly interested in how they sound across my three seat listening area. Sub configuration: I am letting Audiolense take control of adjusting delay/phase and XO duties for the best possible integration with the rest of the drivers. I level matched the subs to the mains with the volume control. Configuring Audiolense See my intro article to Audiolense for basic configuration and operation, as most of this article will focus on the differences, so as not to duplicate content. I have taken my existing stereo two way biamp (XO at 630 Hz) speaker setup, and turned it into a three way triamp setup, crossing the subs at 40 Hz: Why did I choose 40 Hz as the sub XO corner frequency? The JBL cabs with 2 x 15” drivers each, have solid bass output to 40 Hz, as will be seen in the measurements. I like the high efficiency punch and slam, but looking to supplement the bottom octave to give it the full weight. The source material I listen to does not have much output below 40 Hz (we will see about that), which allows me to get away with smaller subs, as the JBL cabs take the majority of the bass signal. As it turns out, even with 525 watts @ 4 ohms into dual 15”woofers per side, the music beat can trigger the limiters first before the subs run out of gas (at 300 watts per sub). This is at concert level of ~105 dB SPL continuous output at the listening positon with peaks above that. That’s just for short term fun, as hearing impairment begins at around 5 minutes at this continuous SPL, even though it sounds perfectly clean. Most of my critical listening is performed at reference level, i.e. the magic of 83 dB SPL. For lower levels, I calibrate using JRiver’s dynamic loudness control, which provides a more natural sounding volume control based on the frequency response characteristics of human hearing. The main Audiolense screen shows the newly designed digital XO: In my previous Audiolense article, there is more about XO choices, steep slopes, etc. Bernt has a really good section on XO choices in the help file. Also, Rod Elliott’s article on, “Phase, Time and Distortion in Loudspeakers” has a good read on crossover filters. Taking Measurements The detailed steps of setting up and taking measurements are covered in my previous walkthrough. Here, I am simply taking the measurement: A few things to note. See how the channel outputs are matched to which speaker. It is likely that you will need to enable Output Channel Override from the Advanced Settings menu. While I have a triamp config, a passive stereo mains with added sub(s) (i.e. biamp config), will require one or more additional DAC output channels, one per sub. Note the frequency sweep ranges for each of the channels. Keep this in mind when looking at the frequency response charts, relative to the crossover slopes. When a correction filter is made, the speaker’s raw response is convolved with the corresponding digital cross over slope. In the case of linear phase crossovers, which these are, the advantage is that all direct sound frequencies arrive at the same time to the listener’s ears, in phase. Check the time delay in the last column. This measurement has already been taken and these are the resultant delays between drivers, relative to the tweeter. If you look back at my previous article, you will see the same delay values for the midrange channels. While the subs appear to be in the horizontal plane, I did tape measure them, one can see the measured delays for each channel are different, which is why we want independent time domain control for each sub. Below is the Audiolense filtered frequency response that better represents what we hear versus the raw, unfiltered measurement. I used a custom filter procedure with True Time Domain (TTD) correction turned on, as well as TTD per driver and selective preringing turned on. See the Audiolense help manual or my previous article for details for designing your own custom filter procedure: I left the crossover slopes in so one could see how the each driver’s response would be convolved with their corresponding digital XO filter. Meaning, with the corner frequencies chosen, and steep XO slopes, each driver is working well within their normal operating range. Therefore, the acoustics slopes of the driver become the linear phase digital crossover slopes, and sum perfectly both in the frequency and time domain. Let’s focus in on the subs (Rythmik L12) response. Here I zoomed in on the horizontal frequency scale and left the 40 Hz low pass XO displayed so one can see how the subs measured response will be convolved with the XO. I.e. the acoustical slope will become the digital XO slope after 40 Hz: In my room, solid response down to 12 Hz (-3 dB). After 40 Hz, starting to see real room effects and then by 110 Hz, smoothing out and rolling off. Resembles the measured spec from Rythmik. Now let’s look at the bandpass (JBL 2 x 15” ported cabs): Good response to 40 Hz, which is the tuning frequency of the JBL vented cabs. Note in both the subs and bass measured responses, the left side of my system has nulls around 70Hz and 90 Hz and for the right speaker, non-minimum phase response at 110 Hz, 120 Hz and 140 Hz. In part, because my stereo is set up off center of the room so the left speaker is more in the corner and the right speaker positioned at the middle of the long wall. We will see Audiolense do its room correction job in these areas so both the timing and frequency response arriving at ones ears matches, even though the stereo is offset to one side of the room. 40 Hz looks to be a good XO point, again within the normal operating range of the woofers. Folks may choose a higher XO point if using bookshelf speakers. Same goes for 630 Hz XO point, well within the normal operating range of the compression driver. Let’s turn our attention towards the measured time domain (i.e. step response). In the Audiolense main form, select Impulse Response from the Chart View radio button group. Then from the Audiolense Analysis menu, select Measurement, then select Step Response: I have labelled the timing diagram to help identify which driver goes with which peak. We are looking at the direct sound plus the next 20 milliseconds of sound arrival at the microphone at the listening position. The first arrival at the listening position are the tweeters with positive polarity (i.e. the compression driver and waveguide). Second arrival is the bass and lower mids (i.e. the double 15” cabs) at .69 (left) and .71 (right) milliseconds later, again with positive polarity. Finally the subs arriving 2.75 (left) and 3.38 (right) milliseconds later after the tweeter, with negative polarity. The delay values are from the measurement window shown earlier. Note that Audiolense assigns 100 ms on the horizontal scale as the start of arrival of the sound. In our relative terms 100 ms = 0 ms. That’s approximately 3 milliseconds of delay for the subs, even though they are in approximately the same physical horizontal plane as the double 15” cabs. If we were to visualize that, sound travels ~1 foot per millisecond and would be as if the subs were physically placed 3 feet behind the mains from their current location. Certainly relative to the tweeter peak, the subs are delayed and have long wavelengths. Now let’s expand the horizontal scale to see 40 ms (i.e. ~ over 40 feet of sound travel in the room). Whoa! The left subwoofer has a huge reflection which shows up as an amplitude peak at 28 milliseconds (i.e. 128 ms on the chart). Its magnitude is bigger than the direct sound, which means it is maximum phase peak at 28 ms. Let’s look at 100 ms of sound travel to see if there are any other room issues: The tweeters spike is just a sliver compared to the subs peaks. No other major issues after 100 ms of sound travel. That’s quite the peak at 28 ms. Something to ponder. Let’s see what Audiolense can do about that. The other aspect of long sub wavelengths is where is the peak located? The peak can occupy several samples with the same amplitude values… and even peak higher later in time, as we see in the example above. Because we are using linear phase digital XO’s, the peak is half of the filter length in number of taps. In the case of our 65,536 tap filters, the peak would occur at sample position 37,268 – i.e. the peak of the waveform. For a minimum phase XO, the peak would occur at sample position zero, which would be the start of the rise of the waveform. Audiolense will automatically calculate and align the peaks of each of the drivers and ensuring all drivers are positive polarity. As a side note, there are several techniques and software tools available to measure the time alignment of drivers in a loudspeaker system. As mentioned above, it gets tricky in sub territory due the long wavelengths involved, especially with an XO of 40 Hz. I have tried most of the tools and techniques available. I must say that Audiolense has exhibited the best accuracy with a high degree of precision that is both predictable and repeatable for time alignment. I demonstrate that here with two sets of measurements taken months apart and can replicate the exact time alignment with the woofers and tweeters (see previous article). The fact that the process is automated is a real time saver. Manually time aligning drivers requires many steps and is prone to carbon unit failure. Finally, time alignment is not just for one mic location either. There are not enough pages here, but in my book I show time alignment of a three way triamped system maintains perfect time alignment after moving the measurement microphone to 14 different locations, covering a 6 foot by 2 foot grid area at the listening position. Basically the area of a 3 seat couch, whether sitting upright or back into the couch. Same goes for phase and group delay, virtually flat over the same listening area. Let’s see what Audiolense can do with this “typical” mess. Designing a Custom Filter Correction Procedure Designer (CPD): As mentioned above, this is the same procedure as described in the Audiolense intro article and not going to repeat here. Rather, let me share some tips. It is worth the time to read Bernt’s help file on what each of the CPD controls do, as it makes a big impact on the sound quality. Essentially one is defining how much correction both in the frequency (i.e. dB of correction applied) and time (over how long in milliseconds) domains, using a user defined frequency dependent window and psychoacoustic filtering that best represents what our ears/brain hear. This offers considerably more flexibility that any other type of eq, plus time domain correction is being applied. Not only time aligning drivers, but correcting for room reflections. There is an example of that in the Audiolense intro article where the group delay in the bass frequencies were greatly reduced. As shown in the above step response measurement, the left sub has a huge amplitude peak at 28 milliseconds that is greater in amplitude (i.e. maximum phase) than the direct sound. We will see Audiolense correct for that. It may take a few filter procedure iterations, similar to narrowing down the target response process, as described next, where one is happy with the sound quality. I encourage experimentation to try several graduated settings, generate/save filters and while listening to music, switch filters in real time and listen/compare. The workflow is fast, only taking a minute or two to cycle through it. Target Design: Let’s start with a quick tutorial on preferred target frequency responses. I have covered some of this in the Audiolense intro article and my series on measuring loudspeakers. In this article, I am going with Sean Olive’s and Floyd Toole’s research on The Subjective and Objective Evaluation of Room Correction Products and The Measurement and Calibration of Sound Reproducing Systems respectively. From Sean’s slide deck, is a preferred ranking of average magnitude responses, measured at the primary listening position: The top preference (red trace) is a flat, but tilted measured response. If 0 dB is 20 Hz, then it would be a straight line to -10 dB at 20 kHz. Note that this tilted measured response is perceived by our ear/brain as subjectively flat or a neutral response according to Sean’s research: A measured flat in-room frequency response is not the preferred target. See how an objectively measured response of 20 Hz and straight line to -10 dB at 20 kHz is subjectively perceived as a neutral or flat response to our ears/brain (red trace overlaid in the above chart). Reading the articles linked above and JJ’s research on Acoustic and Psychoacoustic Issues in Room Correction (See PowerPoint presentation) explains further. Armed with that knowledge, I designed a similar target response in Audiolense’s Target Designer: 0 dB = 20 Hz and a straight line to 24 kHz, so it is -10 dB down at 20 kHz. A good place to start and one can fine tune to taste by moving the 24 kHz red marker up or down in one or two dB increments at a time and listen. I use a “bracket” method by first making one target sounding too dull and the other too bright. Might take a few tries to establish this. Then between the two targets, I move the 24 kHz maker up or down, in 1 dB increments, generate and save the filter. Using JRiver’s Convolution engine, as an example, I open the file dialog and select a FIR filter while the music is playing. There is less than a second gap of silence as the filter is switched. It is fairly easy to hear the spectral differences between filters using 1 dB increments. One can cycle through the process fairly quickly and in no time, be narrowing it down to a couple of candidates. Once you are down to two candidates, this takes a bit more time, as you cycle through more music, switching back and forth a number of times listening to more and more tunes. But after a couple listening sessions over a few days, one will emerge as your top preference. Whether your preference is for a neutral tone, or whatever your preference is, you can narrow it down quickly using this method. As a side note, and not directly related to our subwoofer discussion, a loudspeakers directivity index and measured off axis frequency response is an important consideration when using DSP. The JBL 4722’s I use have a tight, but constant or controlled directivity polar pattern. It responds well to on-axis eq, as off-axis response is virtually identical, due to the constant or controlled directivity design of the waveguide used for this speaker. The Harman/JBL “spinorama” chart for my loudspeaker shows good constant directivity from about 400 Hz on up. With these speakers toed in an equilateral triangle of 10 feet, is just enough distance to illuminate the entire couch area with full range frequency response. Note these constant directivity waveguides require high frequency compensation by design. No audible issues were heard with the HF compensation engaged and the measured HF distortion is well within the capability of the JBL Pro 4” large format compression driver. One can also see that the target design follows the natural roll off of the subs in the room. This is best practice. It is worth spending the time on target design. For best accuracy, one has to zoom right in on the red marker (i.e. dot) to line it up exactly at what frequency and dB setting you want. I mean zoom way in. It will take multiple zooms by clicking down and dragging the mouse from top left diagonally to bottom right and releasing. And vice versa to zoom back out. Final guidance relative to custom filter procedure design and target design. Pick one and optimize first. Only adjust one variable at a time in order to train ones ears to know what to listen for. Personally, I optimize the spectral timbre (i.e. frequency response) first. In other words, the target design. For me it is about getting that neutral sound. Pick a target response, like the Olive and Toole one referenced above, use it and if you don’t like it, then fine tune to your preference. With the recommended bracket procedure, it won’t take long to zero in on what you prefer. If you look in my book or search online, there are several recording/mixing production guidelines, well known monitoring procedures and industry specs that one does try to attain as a professional in the industry. I spent 10 years in the recording/mixing chair and the “sound” is ingrained in my mind, as I heard each recording and mixes hundreds of times and on many systems outside the control room before committing to final mix for mastering. When your number one goal is to have the music “sound good” on a wide range of playback devices and environment’s, the pros try and mix and master on neutral speakers in a neutral environment (i.e. a control room acoustically designed to a specification). So whatever is artistically rendered, translates the intent as best as possible across a wide spectrum of sound reproduction systems. While there is quite a bit of variability in the sound quality of recordings, mixes and masters, I find that the vast majority of recordings sound good across my system, including the mixes I made in the recording studio. I should know how they are supposed to sound, as I was there and mixed it! All meaning to say that one target response does work well for the vast majority of recordings I have. For example, all of these tunes sound great on my system. Rolling Stones Top 500. It is not so much about the variability of sound quality that I object to. It is the excessive dynamic range compression that is crushing the ever living beat out of the music, is what I object to most. Filter Generation and Simulated Output Let’s look at the simulated frequency response, with the target: The -3 dB points are 12 Hz and 22 kHz, and within ±3 dB variance across the frequency range, plus a tighter tolerance than that through most of the range. It did take a number of iterations of CPD tuning and listening to achieve this result. Just like in the intro article, Audiolense’s simulation is virtually identical to the measured response, using a 3rd party acoustic measurement software like REW, for example. Not let’s look at the timing (step) response. Here is the simulated step response plus target over 100 milliseconds: Virtually text book timing response closely following the target. No preringing, perfectly time aligned, (that’s the vertical step showing no discontinuities) and the elimination of the maximum phase peak. Pretty much as good as it gets, for my speakers in my living room. Let’s zoom the horizontal scale to show the first 40 ms of sound arrival: A note on reading the chart, consider the 780 millisecond start of the step as reference = 0 milliseconds. Meaning the signal has reached the microphone at the listening position or our ears for that matter. This shows that all frequencies are arriving all at once (i.e. time aligned). Further, that nasty maximum phase peak at 28 milliseconds is gone and overall, the response follows close enough to the target for rock and roll ☺ A Time Domain Experiment An experiment I performed is AB’ing two different FIR correction filters, one with time domain correction and one without, but both having the same frequency response. The exact same target and correction procedures were used, except for TTD correction is not enabled, nor is TTD per driver, but all other settings remain exactly the same. This effectively turns the time domain correction off, but has the same frequency correction (i.e. tonal response), so when switching between FIR correction filters in real time while listening to music, one can start to tune into the difference it makes when a system is time domain corrected, versus one that is not, especially with subs. Why? Subs introduce milliseconds of delay and even at low frequencies, we can still hear the overhang or lag in the bottom end. Here we go, a new filter procedure, same as the previous filter procedure, but with time domain correction turned off, same target and all other settings identical. Here is the result: As one can see, the frequency response is virtually identical to the same frequency response with the time domain correction. Check. Now let’s switch to the time domain and look at the step response over 100 milliseconds: Does not track at all to the same target. Look back at the step response with the time domain correction turned on. Quite a difference. Let’s break it down a bit. I zoomed the horizontal scale to 60 milliseconds so we can see the time domain issues better: Just like the measured step response at the beginning of the article of the non-time aligned speakers, here in our modeled simulation, we can see the tweeter arriving first, bass cabs second, and the subs some 3 milliseconds later, with a negative going waveform. Of course, the reflection from the left sub is still there, higher in amplitude than the tweeter, but not quite the same height as the woofers from original measurement. Why? Well, we have applied frequency correction, so there is going to be some effect on the timing, but as can be seen, not much compared to the original measurement. This is a great example of how one can’t fix time domain issues with eq. Not only does the timing response resemble nothing like the overall target, but also can’t fix the high frequencies arriving at my ears before anything else and the subs arriving late. Given that the two frequency responses are identical, but the timing responses are not, means that Audiolense can adjust the timing response independently of frequency response. This is exactly what happens when Audiolense True Time Domain (TTD) is turned on. All drivers are time aligned and the timing response tracks closely to the target response while taming room reflection issues like that maximum phase peak in this article and reducing group delay at low frequencies as demonstrated in my previous Audiolense article. The question… is any of this audible? Personally, under blind conditions with my lovely assistant switching filters, I can distinguish between the two every time, even though it does take some concentration. It is not a night and day difference, but rather subtle. For me there are two audible tells. One is that more often than not, the tweeter or high frequencies are the first to arrive. This, to my ears, produces a more forward sound, being a bright brighter in tone, even though the frequency response is the same. That’s because the tweeter is almost always physically closer to ones ears than the other drivers. No amount of “eq” can fix the tweeter arriving first. Second, one can hear the lag on the other drivers, especially the subwoofer. It does take a while to tune into what is happening when switching between filters, but when you hear it, it is hard to forget about it. In addition, this may be what people perceive/confuse as “slow” bass, meaning bass overhang or simply the sub is still outputting, or not begun to output sound, even though the transient has passed and the mains have stopped outputting sound. It is easier to tell with transient material like drums that have a good high frequency cue and low frequency content, like a kick drum for example. Can you hear the click first and then the boom? Or does it sound integrated? Given the sophistication and power of today’s audio DSP software, I am hoping that the industry revisits the time domain of speakers in rooms, as it seems to me to be the missing half of what constitutes accurate sound reproduction. If the goal is to accurately reproduce the waveform of what is stored on the digital media to ones ears, then there can be no frequency or time domain distortions added to the waveform. Just like in my previous article where I show the flat frequency, phase response and group delay of my Lynx Hilo converter, I should be measuring the same as the sound arrives at my ears. This ensures, whatever is stored on the digital media is reproduced as close to as possible to my ears with little frequency or time domain distortion. However, there are a few major transfer functions along the way that really mess with frequency and timing responses – mainly non-time aligned speakers of all types, in wildly variable room acoustics of all shapes and sizes. This is where Audiolense comes into play as one can design a custom, high resolution digital FIR filter that contains the mathematical convolution of your specific speakers in your listening environment. This custom filter is designed to restore the music signal back to what is actually on the digital media, or as close as possible, again, specifically designed for your speakers in your room. It helps if your speakers are time aligned with constant directivity characteristics. It helps if the speakers have been designed and engineered using science like Harman’s spinorama system for example. Lots of controversy over whether to correct frequencies above Schroeder or 500 Hz as an upper limit. In my case, I am trading a more ragged frequency response for having high efficiency (or dynamic speakers). These speakers respond well to eq as verified by the measurements below. If my speakers were Salon2’s for example, they would have a smoother response beyond 500 Hz and may require no eq at all, but they are at least 10 dB less efficient than the JBL’s. Audiolense’s partial correction can be set for any frequency and independently controlled in both the frequency and time domains. Verification Measurements: As mentioned in my previous article, Audiolense simulations are virtually identical to real world measurements using a 3rd party acoustic measurement software like REW. While there is some variability based on smoothing algorithms used, my book shows in detail that these sophisticated DSP packages produce simulations that are virtually identical to their corresponding measurements. I can also overlay exactly what the subs contribute versus just the JBL cabs. Frequency response: -3dB points are at 12 Hz and 21 kHz and within a ±3 dB tolerance of the target design and better than that over most of the range. Step response: All direct sound arriving at the same time and well behaved over 100 milliseconds of sound travel in the room. That nasty peak at 28 ms is gone. Group Delay: Mostly flat with natural rising delay at the very bottom of the response. A little ripple at 350 Hz. Phase: Again mostly flat with natural rising phase in the low end. Here we can see a bit of ripple a 350 Hz. Again unlikely it is audible, but I will investigate. This Bruel & Kjaer application note on, Loudspeaker phase measurements transient response and audible quality, provides some insight for folks interested in this topic. The one limitation that is overcome with modern DSP software is the ability extract the minimum phase response, correct that, while independently correcting the excess phase response. What is interesting is that not only the frequency and timing responses match well between channels, but so does the phase response and group delay. All of which are responsible for a speaker’s ability to completely “disappear”. All one is left with is the stereo illusion presented in 3D with a rock solid stereo phantom center image. The most telling improvement is not only showing that the subs integrate seamlessly, both in the frequency and time domain, but how much the subs contribute to extending the bottom end of my JBL 4722’s: The red and green traces are with the subs integrated. The blue and purple traces are without subs as measured in the previous article. I used the “bracket” method to determine my target preference for each setup. Even over a span of several months’ between measurements, and target designs, I arrived at virtually the same tonal response, except for the bottom end extension. This was starting with blank target designs in each case. It is interesting to me that I consistently end up with the same tonal response. I know what I prefer ☺ Finally, a zoomed in scale from 10 Hz to 150 Hz showing the difference it made to my setup integrating subs: Listening Results Wow, I shouldn’t have waited so long to add subs to my setup. The added weight in the bottom octave really compliments the double 15” impact, giving it the full club or concert sound I have been looking for. I love live music and anything I can do in my home system to give me the feeling like being at the concert or club is all good to me. As a side note on the pic above. For critical listening, I move the coffee table out of the way, I took this pic while I was supposed to be working. Since this pic, I installed quiet curtains behind the speakers covering the windows. They do a really good job in quieting my overly live room to fit the upper RT60 spec limit for the size of my room. I must say I am surprised how much music material I listen to actually has output below 40 Hz – virtually everything I have has some content below 40 Hz. I can listen with just the subs turned on. I can see why the Rythmik L12’s are recommended by folks with planar or electrostatic speakers. To borrow an Austin Powers or Jake Peralta word, toit!! Here is a small sample of music I use to evaluate audio systems and simply enjoy the music. This subgroup contains tunes that have reasonably good dynamic range. But before I do, please allow me to make a short comment on the state of the recording industry: Michael Jackson – Thriller is #1 in worldwide sales, and still is today, along with AC/DC – Back in Black, #2 in worldwide sales. Both are DR 12. I want more of this dynamic sound and less of the hyper compressed music that makes up, unfortunately, the majority of my music collection. While I get artistic intent, I feel the DR scales above fairly represent what is good and bad sound from a dynamic range perspective. See the bottom of the DR chart, where DR = punch and impact. Yah, more of that please. There is no excuse for overly compressed music today. It’s just wimpy loud sound. If there is one thing we can collectively ask for as music consumers that would make the biggest impact on our sound reproduction systems is to allow the consumer to control the volume. Now back to our regular scheduled programming. I could go on about each one, but I think I have gone on enough. Most were a new listening experience for me, discovering for the first time how much low frequency content was on each recording. If there is good low frequency content on the media, it is reproduced unlike what I have heard before from my system. Great fun. Aside from a couple of later recordings, all of these recordings, mixes and masters are 16 years old or older. While my daughter bugs me that this is Dad music, it is sad to me that I have to go back 16 years or so to get a decently recorded, mixed and mastered rock album that has some dynamic range (check out the DR column in the playlist above). The more modern music I listen to on a regular basis, most unfortunately, is in the DR8 to DR6 range with too much dynamic range compression. To be sure, my sub application is for music. However, the subs are fun with movies too. For example, the Jumanji remake with the drums and rhinoceros stampede shook my house so much my daughter came running out from her room wondering if there was an earthquake going on. Mission accomplished. However, these are light duty home theater subs, Rythmik has several larger subs designed for LFE HT applications. For music, I find these more than loud enough for my particular scenario. Conclusion While several audio DSP or DRC products can smooth the frequency response, Audiolense’s True Time Domain (TTD) correction is something to experience. Not only accurately and precisely time aligning drivers but also taking care of room reflections. I know of no other DSP on the market that can do this with this level of workflow automation. One can set up, take a measure, and be listening to a first good corrected response in under 30 minutes. Fine tuning after that is to one’s preference. To me, accurate sound reproduction means the sound reproducing system (including room) is not altering the frequency or timing response arriving at my ears. Meaning a flat perceptual response within a ±3dB tolerance with no phase distortion or excess group delay. I want to hear the music arriving at my ears matching as closely as possible to the content on the recording. Most loudspeakers are not time aligned and the timing response (i.e. delay and phase) gets worse when adding subs due to the long wavelengths involved. In addition, room reflections are inevitable due to the physical dimensions of our listening environments. While there are several subwoofer configurations that can help smooth out the bass, Audiolense DSP can pretty much smooth out the response. I show two subs integrating perfectly with my stereo mains and arguably achieving a smoother frequency response than adding more subs alone would do. Having time aligned subs with mains really shows off the transient impact of having the entire music wavefront arriving at ones ears at the same time. I can’t emphasize this point enough. If you read JJ’s article linked earlier on, one can learn why we hear what we hear in small room acoustics. Audiolense takes advantage of this knowledge and programs the ability to control these parameters in a software DSP program. Includes user adjustable algorithms like frequency dependent windowing, which based on JJ’s research shows that the spectral balance (i.e. timbre) our ears care about is a blend of room interaction at low frequencies, and mostly direct sound in the mids and top end. Later arriving reflections have an influence on the perceived frequency response, and sometimes quite substantially. Therefore, a more psycho-acoustically correct frequency smoothing technique is used in combination. As a result, this is what Audiolense sees and corrects in the frequency domain. A frequency smoothing based on psycho-acoustic principles leads to a smoothed response that sits high in the comb filter region and avoids overcorrection of dips. These two psychoacoustic features are just the beginning of this very sophisticated and powerful audio DSP software program. At 64 bits of resolution, the calculations and FIR filter adds no distortion of any kind when convolved with the music signal. If you are going the whole nine yards with digital XO everything with True Time Domain correction, Audiolense has automated most of the workflow. Once the XO’s are designed and satisfactory, then the workflow is basically the same as if one is working with a passive loudspeaker. It is quite the time saver. I can recommend Rythmik subs and Audiolense to anyone looking to get the most out of their two channel or multichannel system. While I have been into DSP for quite a while, I should not have waited so long on adding subs. Using Audiolense, the subs integrate seamlessly with my mains as evidenced by the simulations and verification measurements – and my ears! Those subs are low frequency canons and really add weight below 40 Hz to give a deeper, but “toit” concussive sound quality. Those are Rythmik’s entry level subs. I am really impressed. If you can achieve objective measurements similar to what I have shown in this article, I don’t think you would be disappointed with the sound quality. You may find correcting the bass in room is all the partial correction one needs, if the loudspeaker exhibits really smooth on and off-axis mid and high frequency response. The time domain correction can also be independently set for whatever frequency. It can be set to the same frequency as the partial correction above. This will correct the low end in the time domain, like the two examples of reducing group delay in the previous article and controlling reflections in this article. Or you may choose to apply an overall time domain correction, if you can always hear the tweeter arriving first, but just a partial frequency correction to correct below 500 Hz. Experimentation is encouraged. Two other Audiolense features to be reviewed in a future article are user defined, mixed phase target design and multi-seat correction. For the latter, some folks feel is perceptually better than a single point measurement used for correction. Let’s see if that is true or not. Until then, enjoy the music! Note: Mitch Barnett's previous article titled "Audiolense Digital Loudspeaker and Room Correction Software Walkthrough" can be found via the link below. I wrote this book to provide the audio enthusiast with an easy-to-follow step-by-step guide for designing a custom digital filter that corrects the frequency and timing response of your loudspeakers in your listening environment, so that the music arriving at your ears matches as closely as possible to the content on the recording. Accurate Sound Reproduction using DSP. Click on Look Inside to review the table of contents and read the first few chapters for free. Mitch “Mitchco” Barnett. I love music and audio. I grew up with music around me, as my mom was a piano player (swing) and my dad was an audiophile (jazz). My hobby is building speakers, amps, preamps, etc., and I still DIY today. I mixed live sound for a variety of bands, which led to an opportunity to work full-time in a 24-track recording studio. Over 10 years, I recorded, mixed, and sometimes produced over 30 albums, plus numerous audio for video post productions in several recording studios in Western Canada.
  6. Thanks @isleofskye The 4722's ($2K US a pair) are bit narrower directivity than the 4429 or 4365. If you biamp the 4722's like I have, then you require digital or electronic XO. A multi-channel DAC, four amps, protection capacitors for the compression drivers. I upgraded the compression drivers on my 4722's based on a really long thread on AVSForum where one of the members tried several different compression drivers. Some DIY along with DSP is required to make these sound their best. Whereas, the 4429 or 4365's are ready to go as is. The JBL M2's are the closest to the 4722's, as they use similar tech, but with upgraded components in the M2. If I had the dough at the time, I would have gone with the M2's. But if you have the space, like huge dynamic speakers and willing to diy a bit, the 4722's are a good choice, if you can stand how fugly they are 🙂
  7. The Grace Design m101 is really nice kit! Gets great reviews. It was a toss up between the Grace and the ART, and I just emptied my pockets on a sub upgrade, and went with the ART. I am happy to hear it is working well for you. Good to know. Thanks. Mitch
  8. Allan @folzag, thanks for taking the time to listen to the binaural recordings and your response. It would seem our listening preferences are similar 🙂 I am still amazed at how much directivity difference there is between these two speakers, which is on the binaural recordings. I feel confident that the differences are predominately directivity related. My room has enough treatment to make the decay time (RT60) not only smooth across frequency range, but within industry guidelines for my size room. It does measure towards the livelier side of the spec... Thanks again, Mitch
  9. Happy New Year Dennis! Thanks for your kind words. Yes, I tried the up close and did not sound natural and was bested by the binaural recording. I got around to purchasing this mic pre and after a shootout, it still comes up on top against the more expensive ones. It's a discrete Class A design with 40 dB of gain before the tube stage. The binaural mics I use are high output and won't need the tube stage. Thanks for the approaches. Good idea on one speaker in one channel and the other in the other channel... Cheers! Mitch
  10. Hi @Tin When the KEF's were playing (even with subs) I put my ear up to the JBL cones and port and could not hear anything. If there was something, then it is too low in level and being masked by the SPL of the KEF's to my ears.
  11. Happy New Year Juergen @JR_Audio Thanks for the thumbs up. Yes, I write verbose 🙂 All the best! Mitch
  12. @maty Cool! Which recording did you prefer? You can find the amps I used in other articles...
  13. Hello @STC Each recording is a random length and have different start times... Perhaps I am misunderstanding your question?
  14. Thanks Arch! @Archimago Happy New Year man. I have some plans for this binaural approach moving forward. Keep up the excellent writings!
  15. Yes, works just fine. There are folks on the Audiolense support forum that have used this for both Maggies and electrostatic panels. Also, one of the reviewers of my book at Amazon uses it on an Open Baffle design...
  16. Hi Peter, Nice gear! And thanks for the information. That's not preringing you are seeing, but the result of using a linear phase target, which does looks correct. Try this, load your linear phase target, don't generate a correction, turn off all chart view details, but leave the target button enabled. Switch to the impulse view and then select from the Analysis menu, Simulation, Step Response Simulation plus target and you will see what I mean... I would use the same target, then click new target and switch to a minimum phase target, save that and use that for your correction. You can check what that looks like with the procedure above before you generate a correction. Then you should see a nice right triangle step response in the simulation. Give that a listen too. I ended up liking a mixed phase target the best. 80% minphase on bass and 100% on top. Look for mixphase menu item in the target designer. Up to you, but I would (initially) turn off any phase adjusts on the subs and let Audiolense do it's thing. The simulated step response looks good over time and you will see with a minphase target, the right triangle step will be there. Also, I would possibly drop the overall subwoofer levels down a bit to better match the level in the mains. You kinda want to shape the response with levels before you add a target. That way, there isn't so much filter insertion loss as the filter needs to attenuate the overall sub level. to be in line with the mains... Under the measurement menu, you should see Automatic Polarity Correction enabled, so no worries on polarity and won't cause any issues, regardless of polarity of subs. Under the Correction menu, is the Correction Procedure Designer. Click on that and select TTD measurement and click on new procedure and give it a name. I would uncheck Prevent treble and bass bass boost. I would up the Max correct boost to 12 dB. I would turn on TTD correction per driver in addition to TTD correction. From there I would enter some values in the TTD subwindow only. This where you want to play around a bit. Try some small values like 3/2 then 6/3 and note the frequency and step response differences and then give the filters a listen. The longer correction in the bass, will tighten up the bass considerably and may get rid of that dip you have, which is not likely audible anyway. Have a look at the manual for XO help with respect to width of the overlap. I went with the default values, but it is another area for a bit of experimentation. Lastly, you may want to reach out to Bernt on the Audiolense support forum with respect how to best handle your 3 subs from an Audiolense configuration point of view. What is the recommended approach and tradeoffs... Your results are looking good! Just a bit more fine tuning with the info above should get you most of the way there. Hope that helps. Mitch
  17. Hi @nefilim Yes, that's odd. It should be the same as this article. Maybe Chris @The Computer Audiophile has an idea as to what is wrong with the images? Chris, anything I can do to fix this?
  18. In this article, I walk through the steps of using Audiolense 5.0 (AL5) to biamp my loudspeakers using a linear phase digital crossover. Followed by frequency and time domain correction, so the frequency response is tonally neutral and the direct sound is arriving at my ears all at the same time. Listening sessions and measurements provide feedback to optimize “room correction” and “partial correction.” Filter design verification measurements are provided. As a preview, Bernt Rønningsbakk, the author of Audiolense, has really hit it out of the park with his latest version. This is a significant improvement over AL4, which I have written a few blog posts here on CA. The core algorithms have been rewritten along with several enhancements, (see what’s new in AL5), in which I take full advantage of in this article. Audiolense is offered in three different versions: 2.0, which offers frequency correction; Surround, which offers frequency correction and time alignment of all speakers and includes a basic crossover for bass management; XO which includes True Time Domain (TTD) correction that fully synchronizes the first arrival for all frequencies and all speakers involved, plus a fully configurable digital crossover. The XO version (~$500 US) is what I used for this article. AL5 demo can be downloaded and provides the ability to listen to 90 seconds of correction over your system. The Gear I love high efficiency loudspeakers for their perceived dynamics. These 2-way JBL 4722 cinema loudspeakers are about as sensitive as one can get, rated at 104 dB SPL at 2.83 volts at one meter. I selected the biamp version and performed a compression driver upgrade, as suggested by user notnyt on the AVSforum. The 4722’s are fugly looking, but sound huge. Note one can use Audiolense with any passive speaker or virtually any loudspeaker for that matter, including surround sound systems. While there are differing opinions on Class D amps, I have been impressed by the Crown XLS DriveCore 2 series of pro amplifiers. I use a XLS 1502 that is rated 525 watts per channel into a 4 ohm load, which is what the JBL double 15” bass cabs are wired for. Coupled with a damping factor of 200 at low frequencies, the bass punches at any volume level. Above 630 Hz, the JBL 2453H-SL compression drivers are fed by Nelson Pass Class A single ended mosfet ACA kits that I assembled. Nelson’s “mu follower” sounds highly detailed, but never harsh. It’s a great match to a high sensitivity compression driver, as the amp has 14 dB of voltage gain so a resistive “pad” is not required to match the sensitivity level to the woofers, which are fed by a much higher gain amp, (e.g. 26 dB of voltage gain), like the Crown for example. What this means is there is no speaker attenuation nor any digital or analog attenuation. In my system, the amps are wide open, as are the analog outputs of the Lynx Hilo feeding these amps. Both amps are quiet, with a small amount of hiss when ones ear is pressed right up against the speakers. I use a Lynx Hilo AD/DA converter as I find great versatility in its uses and its transparent sound quality. JRiver MC 23 64bit Edition, which I find 15 to 30% faster in operation than the 32bit version, with the same sound quality. JRiver’s convolution engine to host the AL5 designed and generated FIR filters, plus JRiver’s 64bit internal volume control. JRiver’s variable loudness control with reference level set ~83 dB SPL at the listening position with volume protection enabled. Perspective on Why Digital Loudspeaker and Room Correction One of the reasons I posted what gear I use is to illustrate that everyone’s sound reproduction system components, including loudspeakers, and for sure, room acoustics, are different. However, the division of sound reproduction inaccuracy clearly errors on the side of loudspeakers in rooms. Especially compared to measurements of DAC’s and amplifiers which, if properly designed, have flat frequency responses, flat group delay and phase responses, with perfect step responses, and virtually no distortion. Loudspeakers in rooms’ measure significantly worse in every aspect, and typically by orders of magnitude worse, as compared to its electrical counterparts. Simply getting 20 Hz to 20 kHz out of a loudspeaker in a room to ones ears within 0.25 dB tolerance is virtually impossible, even at a single point in space, but that tight tolerance is a commodity in almost all DAC’s and amps. For example, here are the measured frequency/phase response, and group delay of my Lynx Hilo, using analog loopback. Meaning, I am measuring both the I/O line amplifiers along with the AD /DA conversion: Have a look at speakers timing responses (i.e. step response) and compare to the ideal step response of a loudspeaker (scroll down to the section on Time Coherence – Step Response). Looking at the loudspeaker step response measurements in the Stereophile article, and the ideal step response linked, one can see the timing distortion in most speakers isn’t remotely close to ideal. In fact, wildly distorted. Loudspeakers in rooms are the biggest offenders in distorting the frequency and timing response of the acoustic signal arriving at ones ears, again by orders of magnitude, as compared to upstream components. Audiolense is a state of the art, custom digital FIR filter designer that allows one to correct the frequency and timing response of your loudspeakers, in your room, so that the frequency and timing response arriving at ones ears is much closer to its electrical counter parts, and therefore more accurate or transparent sound reproduction. You will see measured smooth frequency response with flat group delay and phase response at ones ears in this article. Basically, the same frequency and timing response that is coming out of an amplifier, or as close as possible as one can get with an electroacoustic transducer in an acoustic environment. Folks may be surprised to learn how close one can get to exactly reproducing the audio signal that is encoded on digital media to ones ears at the listening position. Do no harm. The purpose of Audiolense is to undo the harm to the audio signal, introduced by loudspeakers in rooms, arriving at your ears at the listening position. This optimizes gear investment and maximizes musical listening experience on how good ones existing gear can sound. Think of it as a restorative process. Let’s get started. Configuring Audiolense At the end of the article is Bernt’s PDF help file for AL5. I am not going to cover the basics of measurement mic and preamp selection or how to set up for taking measurements. This is covered in the help file and in several articles and posts here on CA. The only reminder is make sure ones measurement microphone is calibrated. After launching Audiolense, select New Speaker Setup: Give it a setup name and in my case, it is a stereo setup, so I have selected 2.0 configuration from the dropdown list. Click OK: We want AllInOne measurement group and FullRange speaker type, if not already set. In my case, I am using AL5’s digital crossover, so I click on the Cross Over Configuration tab (note this step can be skipped if not using digital XO): Double clicking on each speaker row will enable a two way (i.e. biamp) configuration. Here I have set the crossover frequency to be 630 Hz, as recommended by JBL. I went with the default 2.0 Fixed XO Width – octaves setting. Click on Save Setup. The main screen shows the newly designed digital XO: Note the steep slopes. It is worth reading Bernt’s help file (attached) if interested in the audible impact crossover frequency and slope rates can have on one’s loudspeakers. All set. Before taking any measurements, make sure your microphone calibration file is loaded. Taking Measurements From the Measurement menu, select New Measurement: The first thing to realize is the Channel Output order. In my Lynx Hilo setup, Channels 1 and 2 are for bass left and right, Channels 3 and 4 are for tweeters left and right. From the Advanced Settings menu, select, “Output channel override” so that it is enabled: Back on the measurement screen, I can now reassign the output channels, which start at 0 instead of 1 on my Lynx Hilo, to match. Meaning, 0 and 1 are bass left and right, 2 and 3 are tweeters left and right respectively. Note I have already done so in the measurement window above. See the on the measurement screen? This controls the output level of the test tone (i.e. sweep). One may want to start at -40 dB or -30 dB, so that the volume is down in level to prevent both speaker and hearing damage. Once through the procedure below, and after the first measurement, it should be safe to turn up the output to the desired level. We can see the sweep range is set from 20 Hz to 24 kHz, but split between the two way digital XO with overlap, as determined by the width of the digital XO as per cross over configuration settings. I leave the “apply 10 dB attenuation for tweeters” checked to save on ears and tweeters. Also, go with the default sample rate of 48 kHz. The next step is to Check Speaker Connections, this allows one to verify that the channel routing is working correctly and that the test signal is going to the correct speaker and driver combination. E.g. the low frequencies are indeed going to the woofers and not the tweeters. Click on Check Speaker Connections: Click on Test Connection and once you hear the sound coming out of the correct speaker/driver combination, click on Connection is OK and continue with each driver until completed. So either it all passed, or one needs to go back and adjust the channel routing to ensure the signal is getting to the correct driver. Click on Run Measurement. Note with the “Notification and 10 seconds pause when recording” is checked, a dialog box will pop up to acknowledge, so one has time to get out of the way of the speakers and mic. For the first try, I am in a positon to quickly cancel the measurement if it is excessively loud. Optionally, if one has a sound level meter, adjusting the output to around 83 dB SPL C weighting at the mic position during the sweep is ideal, but not required. Once the test signal has completed, a dialog box opens with a Measurement Quality Report: What one is aiming for is a clean measurement with good or excellent dynamic range. This is very important as the filter design is based on the quality of the measurement (or multi seat measurements). It may take a few tries to optimize a good dynamic range by increasing the output level of the test signal through the speakers (but not excessively loud) and/or adjusting the microphone preamplifier gain. Click OK when complete. Click on View Measurement, if it is not already showing. Here is what mine looks like, zoomed in a bit by putting the mouse cursor to the top left, clicking and dragging to the bottom right and letting go. I have left the XO in the view as well: This is both left and right speakers with corresponding woofers and tweeters in a biamp setup. As can be seen, some low frequency trouble at around 70 Hz and high frequency anomalies around 12 kHz. The former is likely a low frequency room reflection or two. The latter is a common result of using large format compression drivers. However, both are important to note as we will revisit these in the fine tuning section. If happy with the measurement, the mic can be taken down and put away if one is not concerned about validating the final result. I will say that AL5’s simulations compared to actual measures are dead on. For the purposes of this article, I left the mic up, as I wanted to verify the results and show some comparisons between filters. I am able to setup for measurements, take measurements, create filters, and validate the filter designs for this article, all in under one hour. Took a few tries though… Measured Driver Delays As a side note, eagle eye folks may have noticed there is a delay value now filled in after taking the measurement: What this means is that relative to the tweeter’s acoustic center, the woofer’s acoustic center is delayed by .69 and .71 milliseconds respectively. But wait a minute, given that the compression driver, on the back of the JBL 2384 waveguide, is further back then the woofers, how can the woofers be this delayed behind the compression driver? What is the cause of the delay? It turns out the Crown XLS amp has on-board DSP that is engaged, even though it is in bypass mode, is the cause of the delay. Without this amp’s DSP delay, the woofers would need to be delayed, roughly by 30 samples at 48 kHz, to be lined up with the acoustic center of the compression driver. Now that we have a quality measurement, let’s start designing our custom FIR filter. Design a Filter – Correction Procedure Designer From the Correction menu, select Correction Procedure Designer: I would advise at this point to have a look in Bernt’s help file and read the section on Correction Procedure Designer, how it works, and the three default procedures. It may take a few reads to understand what is happening, like True Time Domain correction: “The True Time Domain correction cleans up the impulse response in the time domain and corrects the speakers towards the time domain behavior in the target. A good frequency correction goes a long way, but usually time domain correction improves the sound quality even further. In most setups, a true time domain correction will provide better perceived dynamics, 3D and air than the alternative. When True Time Domain Correction is checked, the whole speaker will be time domain corrected as one single entity. This is the only TTD option for passive speakers. The option “TTD Correction per driver” will also be visible when digital crossovers are used. Checking this option will also time domain correct each driver individually before the speaker is assembled and corrected as a whole.” Keep the first sentence on TTD in mind when we get to designing the target response. In my case above, I have clicked on New Procedure, gave it a name, and now ready to design my own custom digital filter. Note which checkboxes are checked. As described in help file, I leave the measurement and correction window default values. I also have selective preringing prevention checked. Refer to the help file for what it does and when to use it. Personally, except at gross levels of preringing far above what is in a typical time domain correction filter, I never hear any preringing artifacts. First, let’s try a shorter time correction window in the bass, so now I have 2 cycle window both in the low and high frequencies as entered in the True Time Domain Subwindow: Click Exit. Design a Filter – Filtering the Measurement On the main screen, click on Filter measurement button. This will bring up which filter procedure to apply. Choose the procedure that we just custom designed in the Correction Procedure Designer and click OK. Here I have zoomed in the result and left the raw measurement to show what the filtering does: A frequency smoothing based on psycho-acoustic principles leads to a smoothed response that sits high in the comb filter region and avoids overcorrection of dips. The measurement & correction window helps smooth the filtered measurement in the frequency domain and determines how much time the correction filter is given to complete the job. The settings also affect how the frequency response is smoothed before it is corrected. The Time Domain Window (TDW) is used together with other methods to smooth the frequency response. The ultimate goal of the procedure is to obtain a smoothed response that correlates well with what we hear and what we would want to correct. As can be seen above, even without the XO join of the low and high frequency sections, not the smoothest frequency response. Especially when compared to the frequency response of my Lynx Hilo above. Remember the time delay between drivers? Let’s look at the measured step response by switching to the chart impulse view and selecting from the Analysis menu, Measurement and Step Response Measurement: Here I have zoomed in on the first 10 milliseconds. While not an integrated view of the step response, one can see that the sharp spike is the tweeter (i.e. compression driver) followed by the woofer. Whether one is looking at the two peaks, or the start of each step, clearly, these drivers are not time aligned The tweeter will arrive at ones ears first, then followed by the woofer, .69ms left and .71ms right later. It should go without saying that no amount of eq can fix a time alignment problem. I have been playing with this type of FIR DSP since 2011 and to my ears, in multi-driver systems, time aligned drivers sound best. Scroll to section “Time Coherence – Step Response” for a detailed explanation. In AL5 it is a snap as it is fully automated in the software, and as we will see, seamless time alignment of the woofers to the tweeters, even though there is ~ 8 inches of z offset between the drivers. With both the frequency and time domain measurements completed, we can see that both are not ideal. Let’s move on to designing a target response. Design a Filter – Target Design If one is interested in various frequency response targets that listeners prefer, have a read of Dr. Floyd Toole’s excellent AES open access paper on The Measurement and Calibration of Sound Reproducing Systems. See Figure 14 on page 17, the subjectively preferred steady-state room curve targets in a typical domestic listening room: I prefer the “trained listener’s” room curve. However, I urge folks to try out different target curves to find ones preference. In AL5, it is simple to make multiple target curves, generate the filters and be AB’ing in a matter of minutes. Target or room curves have a substantial audible impact to the timbre (i.e. tone quality) of ones sound reproduction system. Experimenting with different target or room curves is well worth the effort. In the main screen, click on New Target: Use the mouse to click additional target points on the graph so as to draw the target to the spec. As mentioned above, target design is quite critical to the overall timbre (i.e. tone quality) and worth the time to play around with different slopes. Some folks like a bit of a bass bump, there is no right or wrong, simply what one prefers. However, trying the spec’d target to start with generally results in a perceived neutral tonal response to ones ears at the listening position. A good reference point before departure. Here is the finished result: This is a minimum phase target. I also created a linear phase target using the exact same target design. Linear phase correction will correct towards linear phase behavior where all frequencies arrive at the listener at the same time with the same phase. AL5 can design mixed phase targets as well. I recommend spending time reading Bernt’s help file as there are many variables to consider in target design, which have an audible impact on tone quality. As an initial starting point, tracing the low and high frequency roll off of the drivers is time well spent. Here is another popular target, rather than a tilt, it is flat out to 1 kHz and using 1 kHz as a hinge point, straight line to -6 or -7 dB or so at 20 kHz, depending on how much high frequency energy ones ears can take: Small variations in targets will impart a certain perceived timbre or tone quality. The tilting target, before this one, will have a bit of a recessed midrange, whereas flat to 1 kHz, the midrange will be a bit more out front. Again, worth creating a few targets and switching between to determine ones preference. For the purpose of this article, I went with the tilted target, but later listened to the flat to 1 kHz linear phase target. Click on Save when happy with your target design. One may revisit tweaking a target response several times during this workflow. A good naming convention goes a long ways to staying organized. Design a Filter – Filter Generation and Verification In the main window, click on Generate Correction Filter. In the main window, one should now see: We are looking at everything sans the original measurement. It is good to just look at the correction filter first to see if there are any anomalies. Here I unchecked everything in the main window, chart view details, except the correction: Looks good. Keep an eye on low and high frequency boost at the frequency extremes as this can lower (i.e. attenuate) the overall level of your correction filter by -10 dB. Additionally, we will revisit the high frequency correction in the fine tuning section on partial correction. OK let’s look at the simulated frequency response by unchecking everything but the simulated response and target: Looks pretty good and follows the target precisely until the frequency extremes. The -3 dB points are 25 Hz left speaker and 35 Hz right speaker. The reason is my stereo is offset to one side of the room where the left speaker is more in the corner and benefitting from that boundary position, whereas the right speaker is almost center line of the room and in just about the worst position possible. 19 kHz is the -3 dB HF point with these large format compression drivers. Not that I can hear that anyway, but they sure sound smooth. Now let’s look at the predicted or simulated timing (i.e. step) response as it would be at the listening position. To do this, uncheck all items in the main window, chart view details. Then in the bottom left hand corner, chart view, click on Impulse response. Then from the Analysis menu, select: On the main chart, you will see the new plot. It will take a couple of “mouse click top left, hold and release right bottom” zooms to get to this level – note the horizontal scale in milliseconds. We want to see the first 20ms, with the first 5ms being the most important to see that “right triangle” shape of a time aligned step response, and following the minimum phase roll off of the loudspeaker in the low end. Looks good. No preringing at the beginning of the vertical step with no discontinuities in the vertical step, meaning that all frequencies are arriving at ones ears at the same time. This is much closer to an ideal speaker’s step response and a major improvement over the measured step response in the previous “Filter the measurement” section. For a refresher on what an ideal speaker’s step response should look like, scroll to section “Time Coherence – Step Response” Without changing anything else, here is the step response, using the exact same target response, but now with a linear phase target: Looking a bit different and audibly, sounding subtlety different. Once we get into the listening sessions, I will indicate my preference. Either way, good looking step responses. I should point out that the corresponding frequency responses to the two different step responses are identical. The next step on the main screen is to click on Save Filter. This opens a dialog to save a .cfg file that contains the path names to correction wav files. Once you choose the save location and give the file a descriptive name, it will be followed with another dialog on which sample rates one wants to generate. Don’t see the sample rates you need in the list? Cancel out and go under the Setup menu, select Optional Sampling Frequencies, enter in the sampling frequency you wish and add. Then repeat the save filter and now checkbox the sampling frequency that one just created. In my case, I am using JRiver and in the convolution window, select browse and load the .cfg file and now ready to listen. No editing of the .cfg is required. Please refer to Bernt’s PDF help file for details or check out other articles and blog posting on CA for details on how to load .cfg files into JRiver’s Convolver. Let’s give it a listen and make a few unnecessary validation measurements. Listening Sessions and Measurement Verification Listening Sessions I wish folks could hear it. With those big fugly speakers, the experience is immersive 3D imaging. The depth of field is a real treat. It is almost if one can fall into the mix. And certainly belies having a 42” LCD display parked between the speakers. Closing my eyes, and as far as the audio presentation is concerned, it simply is transparent, as if the LCD screen does not exist. In fact, measuring the ETC with REW shows any reflections off the LCD screen to be below audible levels using Dr. Toole’s psychoacoustic research as reference. I critically listen at ~83 dB SPL C weighing, slow metering, give or take a bunch of dB’s depending on how over compressed the music material is. SRV gets cranked up. Oasis, turn that down! ☺ What do I listen for? Point source sound. Meaning if the stereo were playing mono material, ones perception of the sound should appear to be emanating from a point directly between the speakers, centralized at ear height. With the correction in place, the left and right responses are almost identical, both in the frequency and time domains. So not only is the sonic reproduction neutral and time aligned, any loudspeaker and room discontinuities between left and right channels are accounted for as well. Literally a mirror image. From a stereo perspective, the phantom center image should be dead center and not waver with frequency. Given the mirror image, stereo reproduction should be properly decoded to again closer represent the waveforms as encoded on the digital media. Many years ago when I recorded and mixed on time aligned monitors, every instrument had their placement, not only from a stereo perspective, but also depth of field. Using the Haas effect, digital delays and reverbs, one could make a band sound like they were recorded in a small club all the way up to a stadium gig. For timbre, can I hear preringing before the attack of the main sound? Does the tone quality sound neutral? Meaning, no one frequency or range stands out, or not too bright or dark sounding on a wide range of recorded material. Do the s’sss sound natural on the voice? Overly sibilant? Or sounds like shhh? Bass notes sound distinct? Some bass notes are louder? Softer? Etc. One of my favorite tunes to crank up is Stevie Ray Vaughn’s Tin Pan Alley (DR 18) and hearing the snares rattle with the bass guitar cab along with the ambience feels like being in the concert hall with the band playing live. Same with Cream Live in 2005, you’re in the best seat in the house, at the mixing console! Another test I perform is recording my acoustic guitar with the measurement mic and then playback the recording, and play along. It becomes clear quickly if the timbre is off in any way. Oh the other hand, it is amazing how close to the real thing it sounds, right down to the finger noise on the strings. I also have electric and bass guitars, and if you look in the picture at the beginning of the article, an acoustic drum set that I play along with the stereo. It’s nice to have speakers that can keep up with a live kit that sounds good, with low distortion. I am liking the sound at this stage and do have a couple of comments related to the trouble areas mentioned in the measurement section where there is some room disturbances around 70 Hz and the top end needs some consideration. But I will hold off until the fine tuning sections as I establish a baseline of the current filter design. With that in mind, let’s take some measurements with the correction FIR filter in the circuit. Measurement Verification Here I measure at the listening position using John Mulcahy’s REW shareware acoustics program. At the outset, I want to say that Bernt’s simulations are virtually identical to the measurements. The measurement path is, test tone output of REW is digitally routed through the Lynx Hilo to JRiver’s ASIO digital input, with no digital attenuation, passed through the Convolution engine, hosting the Audiolense digital XO and FIR correction filters, and then routed through the Hilo with no digital or analog attenuation to the 4 amplifiers, to stereo biamp speakers, to the mic at the listening position, through the microphone preamp, into the Hilo ADC and then routed to the digital input of REW, calculations performed and finally displayed on a graph. Here is the “unsmoothed” or full resolution measured frequency response at the listening position: Our ears don’t “perceive” that level of comb filtering, as displayed in this high resolution plot. Our ears perceive more of the envelope of the waveform, rather than every narrow rise and dip in the response. Our ears are more sensitive to peaks than dips. Traditionally, 1/6th octave smoothing is used to visually smooth the envelope to approximate what our ears would perceive. Today we have frequency dependent windowing (FDW) and psychoacoustic filtering. However, this level of technical detail is out of scope for this article. 1/6th octave smoothing is close enough for our illustrative purpose and is a reasonable graphic representation of what our ears perceive. Here is the same measured response, but with 1/6 octave smoothing: This matches very well with the AL5 simulated frequency response, including the same -3 dB points. The measured tilted response is perceived as neutral to our ears at the listening position. For fun, I suggest folks try a flat 20 to 20 kHz target and compare to this one by AB testing in near real-time like in JRiver. While not as flat as an amplifier’s frequency response, factoring in the tilt, it is much smoother than the filtered measurement and within a much tighter tolerance. Sounds really smooth to my ears. No one frequency stands out, sounds neutral. Now the timing (step) response: If you scroll up and look at the AL5 simulated step response, using the minimum phase target, and compare, they are pretty much identical. I did not bother to show the linear phase target, as it is also identical to the AL5 simulated step response. So frequency and timing response verified as it matches perfectly with AL5’s simulated responses. Furthermore, looking back near the beginning of the article and comparing to the uncorrected frequency and step response, one can see a major improvement. Sounds that way too. While I don’t show it in this article, my eBook shows that smooth frequency response and time alignment is valid at more than one mic position. I show 14 measurements of both frequency and timing responses are virtually identical around a 6’ x 2’ grid area that represents my 6 foot, three seat couch listening area. No matter where I sit on the couch, the sound is time aligned, and there is very little tone quality difference moving from seat to seat. I don’t hear any “phasiness” or comb filtering, when I move around, sounds virtually the same. However, controlled or constant directivity waveguides bring another challenge as the waveguide requires constant directivity horn equalization, but the compression driver already has phase plug correction built in. Applying a correction on top of a correction sounds, you guessed it, overcorrected. Overcorrected high frequency response tends to sound too compressed or processed. Another giveaway is moving ones head from side to side at the listening position and the center image tracks to one side of the speaker, instead of pivoting around the center phantom image. This is another sign of HF over correction. This is one area of interest for the fine tuning section below. The other area to revisit is that frequency response dip around 70 Hz. That is “room correction” territory and will be first up in the fine tuning section. Fine Tuning Tips and Tricks Room Correction My stereo is offset to one side of the room and firing across the short width of a 30’ x 16’ x 8’ room. This is going to cause low frequency reflections, which has an impact on the sound quality I am listening to. It is not obvious at first, unless one has heard it before, and verified with measurements, then what reference does one have to compare to? What to listen for? One giveaway is that certain bass note frequencies sound different, usually not as loud and/or focused. When listening to a bass response that has no room reflections interfering with the low frequency response, one notices that every bass note emanates from one point in space between the speakers, sounds solid as every note is distinct and does not vary in perceived loudness. You only have to hear it once to learn what it sounds like, but where to hear it? Another way to look at the time domain of a sound reproduction system, is group delay. A linear phase system’s group delay is constant, meaning a flat response. Flat group delay also means a flat phase response. Btw, so should a well-designed DAC and amp measure both flat group delay and phase response. See the Lynx Hilo measurement of group delay and phase near the beginning of the article. There is some consideration for a slowly rising group delay at the lowest frequencies, as this is the group delay introduced by the ideal loudspeakers minimum phase response, as speakers do not reproduce down to DC (i.e. 0 Hz). However, in my example, there are a couple of non-minimum phase room refection’s that are audibly changing the group delay (and phase) at the listening position. John Mulcahy, REW author, has a good write up on minimum phase. Let’s look at the group delay in the same measurement as above: We can clearly see group delay at 50 Hz on the left speaker and 65 Hz on the right speaker. Again, because the speakers are offset left from center of room. If I could set up the speakers to be perfectly symmetrical on either side of the center line in the room, the group delay would still be present, but at the same frequencies. How to get rid of it or reduce it, or restore to flat response? If you read John’s article on minimum phase linked above, one will read that it cannot be corrected with eq. I agree. Here what we need to do is fix the time domain response in the room. This is easily done in AL5 by increasing the TTD window size in the bass frequencies, via the Correction Procedure Designer, and rerun the correction, leaving everything else the same. We are only changing one parameter. So instead of a 200 ms correction TTD window starting at 10 Hz, let’s try a 500 ms correction window by entering 5 cycles at 10 Hz: In the True Time Domain Subwindow we went from 2 to 5 cycles of time correction at 10 Hz. The idea is that we increase the correction window size to envelope the non-minimum phase reflections in the bass. The impulse inside the TTD window is corrected towards the time domain behavior of the target response. Exiting, regenerating the filters using the same target curve, and all other settings the same, and re-measure, we get: We can see the two major group delay issues at 50 Hz and 65 Hz are now gone. We see the slope of the group delay rising as it reaches lower frequencies as it follows the minimum phase response of the target. The only thing I changed was increasing the amount of low frequency time domain correction. Once ones ear’s tune into how solid and stable the bass sounds at any low frequency and sits “right in the pocket”, you won’t be going back. There are multiple ways of dealing with low frequency room issues below Schroeder frequency, like multiple subs (that need to be time aligned IMHO) or a roomful of bass traps, however, both solutions also alter the frequency response. That’s what’s cool about the power and sophistication of today’s DSP software, it can surgically remove low frequency room reflections, yet the measured frequency response stays the same. Here is the frequency response from the two measurement runs of TTD settings 2 and 2, and 5 and 2 cycles, overlaid from 20 to 100 Hz: One can see the left and right pair’s frequency response is virtually identical. Overlaying the same measures, but change the view to group delay: This shows the effectiveness of “room correction” in action by reducing group delay independently of frequency response. Here we are seeing the group delay being reduced by over 30 milliseconds at 50 Hz and 65 Hz respectively. If one wants to lower the group delay even further, using a linear phase target will assist. Here is the group delay measurement using the same target response as above, but now a linear phase target: And for fun, corresponding phase response: While not perfectly flat like the Lynx Hilo, for a loudspeaker in a room, at the listening position, considerably better than without correction. As shown in my eBook, I was able to measure mostly flat group delay and phase throughout the 6’ x 2’ listening area, so again, it is just not at a single mic position. We can see in addition to a perceptually smooth and neutral, and time aligned response, we can also achieve a relatively flat group delay and phase response arriving at ones ears at the listening position. Note that it varies from system to system how close you can get to a perfect time domain behavior. In another living room with less benign acoustics, an attempt to remove the group delay peaks may be ignored (when the selective pre-ringing prevention is engaged) or lead to more than desirable pre-ringing. Sometimes these time domain anomalies are practically uncorrectable although it can be done mathematically. AL5 has been designed to enable the advanced user to dial in the best possible compromise in his/her own listening environment There is controversy over the audibility of group delay at low frequencies, as the current research does not take into consideration frequencies below 500 Hz. Here we are discussing frequencies below 100 Hz. For music, that has a range of bass frequency notes, I can AB the difference every time, as I know what to listen for and prefer the bass without low frequency room reflections (i.e. non-minimum phase behavior). I agree that non-minimum phase behavior cannot be equalized in the frequency domain, however, as my measurements bear out, it certainly can be removed in the time domain. Sounds amazing to my ears as I only experienced this type of low frequency response previously in acoustically designed control rooms at recording studios. As a side note, I hope audio manufacturers and/or industry society’s, sponsor more blind or double blind listening tests on the audibility of speaker time alignment and group delay below 100 Hz. Given the power of today’s computers and sophisticated audio DSP software, which was not available a decade ago, it would be good to update the research. At this stage, I am happy with the timing of the bass response of my system. Now to move on to high frequency tweaking. Partial Correction Partial correction is great if one just wants to correct frequency and timing response independently up to a certain frequency. Like Schroeder or whatever, you choose. As mentioned in the section, Listening Sessions and Measurement Verification, I pointed out the waveguide requires frequency compensation due to its design, but the compression drivers high frequency response already has been optimized (i.e. corrected) using a phase plug to extend its high frequency response. This is a point of departure depending on one’s gear and goals. More often than not, the natural extension of the driver, and in this case, a large format compression driver, typically sounds more relaxed and not compressed or processed. From the Correction menu, select the Correction Procedure Designer again. Don’t forget to select the right Procedure Name in the drop down before proceeding to making any changes. Here I have already clicked on Partial Correction checkbox in the Designer and then clicked on the Partial Correction tab: I have entered a full frequency correction for waveguide compensation, but up to 5 kHz time domain correction. Above 5 kHz, it is the natural time domain response of the compression driver. Note that I have left “use target as EQ in no correction zone” unchecked. Then one exits, regenerates the correction filters, save and load into JRiver and give a listen. Takes a couple of minutes to do so and one can be listening and AB comparing to a previous filter, in near-real time. As before, here is the new correction filter response based on the partial correction: Now here is the same correction, nothing else has changed, but with partial correction turned off: Compare the two correction filters past 10 kHz. To my ears, a little more “relaxed” high frequency response due to the “quieter” filter. I am letting the phase plug in the compression driver use its own time domain correction as opposed to adding on top of it. Now when I move around on the couch, the stereo phantom center stays anchored in the center and my ears are pivoting around it as opposed to phantom center moving to one side of the speaker as my head moves. This really is fine tuning one’s high frequency preference, not only target slope or tilt, but also how much high frequency and time domain correction is or is not required. It is worth spending a bit of time playing with partial correction to get a sense of what it does and what you prefer. In some cases with dome tweeters with excellent minimum phase response, one may prefer the full TTD correction. And with that final tweak, I am happy with the end result. As an update, I have been listening for four days and feel no need to make any further adjustments. I was going to use the global tilt control in the Target Designer to tilt the overall target or room curve a little more downward, as I initially thought the sound reproduction was a little too bright. After extended listening to a wide range of music and movie material, I am still happy with the original result. However, I must admit, I did go back to the flat to 1 kHz linear phase target, which I have been listening to for a number of years. Currently listening to it now. Conclusion The AL5 XO version that I used is approximately $500 US. I don’t know of any other $500 investment one can make that has this much positive sonic impact on what our ears perceive at the listening positon. I view “accurate sound reproduction” as the music arriving at my ears matches as close as possible to the content on the recording. Technically, we are matching waveforms with a high degree of accuracy and precision, while taking into consideration real world transfer function limitations. Just like a good quality DAC or amp, the frequency response is flat (tilted for psychoacoustic perception for loudspeakers in rooms), with flat group delay and phase response, save the minimum phase response of the loudspeaker that does not reproduce down to DC. Perfect step response where first arrival frequencies arrive at ones ears, all at the same time. Again, slightly different in the acoustic domain to take into account the loudspeakers low frequency roll off. We can accurately and precisely match the digital test signal all the way through the system to ones ears as demonstrated by the measurements of my fugly cinema loudspeakers. Note one can route the test signal through JRiver’s digital input and 64 bit convolution engine, DAC, analog output, preamp, amp(s), loudspeakers, room, and to ones ears. That means that Audiolense is correcting the entire playback chain, digital, analog, electroacoustic, and acoustic. So any anomalies or channel imbalances or other potentials for frequency and/or time domain distortions, are taken care of, all the way to ones ears. Something to ponder as Audiolense’s filters are very powerful with 65,536 filter taps. This is 2000 times more powerful than MQA’s 32 filter taps and most DAC’s few hundred filter taps, albeit for different filtering purposes. This is all a bit technical and does take some effort to get it right, but the payoff is unlike most people have heard. That’s the issue. Most folks don’t have a baseline to compare to. Hence measurements, to verify smooth frequency response, flat phase and group delay, removal of low frequency room reflections, and time aligned sound arriving at your ears, in your environment. If one can replicate the measurements I have here, which is certainly doable, even on an industrial loudspeaker designed to be in a cinema like mine, you will be rewarded with accurate sound reproduction. Once you hear it, there is no turning back. Look for an updated version of my eBook with additional fine tuning techniques in addition to more detail about the basics, such as multi-seat correction and using more of the measurement and analysis features. In the meantime, don’t forget to check out Bernt’s PDF help file attached at the end of this article. There is an Audiolense User Group to search and ask questions. In the end, it is the linear phase target that I enjoy the most and my preference. I hope folks give the demo a try and hear for yourself the improvements one can make to restore the response to what’s on the recording arriving at your ears with very little frequency or time distortion. Update It’s been two weeks since I wrote this article and thought to provide an update. I have not made any further changes since where I left off. Still the same flat to 1 kHz and straight line to -6 or -7 dB at 20 kHz linear phase target. Still the same TTD (5&2 cycles) and partial correction parameters (20 kHz freq, 5 kHz TTD). I have listened to a wide variety of source material for hours each day. Very happy with the result. I don’t think I can make it sound any better. Or can I…? I like the sound of 15” high efficiency woofers in big cabinets because they sound… big! I like the feeling of being at a rock concert in my home with this type of big sound. The 4722’s pound the rock and blues that sooths my soul. The 4722’s kicks and punches, but rolls off a bit early in the bottom octave, even for rock and blues music. Most rock, blues, and certainly distorted alt, typically do not have much content below the low E on a regular 4 string bass guitar (i.e. 41 Hz). And is dependent on how much the drums have been rolled off in the bottom octave during the recording/mix/master process. So a pair of music subwoofers are on the way to add a bit of weight from 20 to 40 Hz. Stay tuned for another article dialing in a pair of subs. Enjoy the music! Audiolense Help Document (PDF 3.2 MB) Note: Mitch Barnett's article titled " Integrating Subwoofers with Stereo Mains using Audiolense" can be found via the link below. Mitch “Mitchco” Barnett. I love music and audio. I grew up with music around me, as my mom was a piano player (swing) and my dad was an audiophile (jazz). My hobby is building speakers, amps, preamps, etc., and I still DIY today. I mixed live sound for a variety of bands, which led to an opportunity to work full-time in a 24-track recording studio. Over 10 years, I recorded, mixed, and sometimes produced over 30 albums, plus numerous audio for video post productions in several recording studios in Western Canada. I wrote this book to provide the audio enthusiast with an easy-to-follow step-by-step guide for designing a custom digital filter that corrects the frequency and timing response of your loudspeakers in your listening environment, so that the music arriving at your ears matches as closely as possible to the content on the recording. Accurate Sound Reproduction using DSP. Click on Look Inside to review the table of contents and read the first few chapters for free.
  19. mitchco

    Dutch & Dutch 8c Loudspeaker Review

    The D&D 8c’s are the first loudspeakers to come into my room where I could achieve studio control room frequency response accuracy of 21 Hz to 20 kHz ±3 dB at the listening position without any external DSP. Also, the first loudspeaker where my preferred high frequency response (i.e. neutral) was just about perfect right out of the box. Other speakers I have tested required a -5 dB shelf from about 3 kHz to 20 kHz to sound neutral. Plus, the first full range bookshelf speaker that do not require subs, if set up correctly. The D&D 8c’s can reproduce studio control room frequency response accuracy in one’s room using the onboard DSP software. The 8c’s are perfectly time coherent, meaning the direct sound from all drivers are arriving at ones ears at the same time. Combined with a midrange cardioid dispersion pattern, and sub drivers big enough to not require external subs, to me, represent the state of the art in loudspeaker engineering design today. Let’s Talk Tech While the 8c is similar in concept to the Kii THREE I reviewed, the 8c’s overall design and engineering are different and unlike any other loudspeaker on the market today. The 8c is a largish, heavy bookshelf loudspeaker that weigh in at 53 lbs. The shape is more rectangular in depth than square. Worthwhile to have a 2nd person on hand when lifting these onto stands. Make sure your stands are rated for the weight. My sand filled, Monolith stands are rated for 75 lbs. The front baffle is a beautifully engineered CNC precision design. The dispersion pattern is just about perfect on and off axis. I reached out to Martijn Mensink, the designer of the 8c, to understand his thinking that went into the waveguide design: “The waveguide is designed specifically for this tweeter. We started out with an OS (Author note: OS = Oblate Spheroid) profile. It performed reasonably well, but there were some clear diffraction issues which resulted in comb-filtering, especially on-axis. An OSWG causes very little diffraction, but it assumes a perfectly flat wavefront (which of course has to be diffracted/curved from flat to spherical), which in practice is near impossible to achieve. The OSWG is a good starting point, but it needed quite some work to adapt to our tweeter. Our waveguide assumes an already somewhat curved wavefront due to the dome shape of the source, so that a minimum of diffraction is required. The profile is the result of theory and simulation, but ultimately also a lot of iteration - we built many many prototypes! Thank you CNC-router and 3D-printer. The eventual design has a small radius at the throat, then a very short Oblate Spheroid part (pretty much conical), then a very large mouth radius and finally a large baffle edge radius. The waveguide is optimized for constant directivity and response smoothness. For both it is a strict requirement to deal with diffraction effectively, because diffraction causes comb-filtering effects that change with angle.” For years, I have been researching constant directivity waveguides and listening to several designs in my sound reproduction system. The 8c’s are next level loudspeaker engineering design and execution. The Polar Map is virtually perfect from 100 Hz to 20 kHz: From 100 Hz on up, the on and off axis frequency response is incredibly accurate, best I have seen to date. The midrange driver is an 8” aluminum cone crossed over at 1250 Hz using a 4th order Linkwitz-Riley linear phase crossover to a 1” aluminum/magnesium alloy dome tweeter. Specs here. Each side of the cabinet has a cutout to the inside chamber of the midrange cabinet, which is used to passively control the midrange polar response to be a cardioid pattern. This also controls Speaker Boundary Interference Response (SBIR). The idea is that the passive cardioid midrange (i.e. 100 Hz and up to the tweeter XO) outputs very little side and rear sound pressure so there is little to no speaker boundary interference. Once you measure the physical distance from the rear of the subs to the front wall in centimeters, in the 8c’s Active Room Matching user interface, you select the same values from a drop down list of values in centimeters for both front and side walls. While using REW, I selected a range of values around the measured distance to see which yielded the flattest response. Each value selected made quite a measurable difference, so it is worthwhile to check a small range around the measured physical distance. For example, if the subs are 40 cm away from the front wall, also try 30 cm and 50 cm in the Active Room Matching panel and re-measure each time with REW. It does not take long to find the right value and usually it is bang on to the closest value that was physically measured. Impressive! On the back of the loudspeaker are two 8” aluminum cone subwoofers operating from 100 Hz on down in a sealed cab. Yes, that’s right, the dual subwoofers are built in. With their large linear max excursion design, these bad boys have an amazing amount of low frequency impact. That impact is further realized by being close to the front wall boundary, leveraging what is called the Allison effect, named after the late Roy Allison. By placing the loudspeaker close to the front wall (e.g. 40 cm to back of cabinet) one can gain roughly +6 dB in response below 100 Hz: Even more low frequency room gain can be had using perpendicular boundaries, but at the expense of increased speaker boundary interference response (SBIR) above 100 Hz. In the linked article, scroll down to bookshelf loudspeaker to see a great example of SBIR. A 6 dB increase below 100Hz is significant and relatively speaking is a 4 times increase in power (watts) and on the way to being almost twice as loud, as perceived by our ears (a 10 dB increase in SPL is perceived as being twice as loud to our ears). As Martijn Mensink, designer of the 8c’s explains, “Basically, the first comb-filtering dip caused by the front-wall reflection is shifted to a frequency above the working range of the subwoofers. For instance, when the distance is 40 cm, the first boundary dip will be at 215 Hz. The woofers only go up to 100 Hz. The wall increases the output by a maximum of 6 decibels and you get a very nice coherent first wave-front, with no smearing in the time-domain. Above 100 Hz the cardioid midrange takes over. Whereas the subwoofers are using the boundary sort of like a springboard, in the midrange it's as if the wall isn't actually there, because very little sound is radiated towards it by the cardioid. That's why you can place the 8c's so close to the front wall! “ Again, next level loudspeaker engineering to seamlessly integrate the room into the loudspeaker design that addresses common room acoustic issues that we all have in our listening rooms. Not only mitigating SBIR, but leveraging the Allison effect to maximize low frequency output/impact at the same time. Nicely done! All parameters for controlling the 8c’s in one’s acoustic environment are available through a web browser user interface that communicates with the loudspeakers onboard computer in real time: Each speaker (i.e. on board computer) is connected to the internet via a hardwired Ethernet connection. Each speaker’s serial number is identified in the user interface. This is where we will be making all of our configuration and calibration adjustments in the next section. The 8c’s use Analog Devices ADAU1452 DSP chip in which D&D developed their own DSP board and audio boards, along with all of the software developed for the 8c. Having the loudspeakers connected to the internet makes it easy to apply software “loudspeaker” updates, as new features are added to the on-board computer’s memory. This future proofs the loudspeaker, as the driver technology and physical design are as it good as it gets. This approach also allows D&D tech support to interface with the loudspeakers if ever required. The 8c has balanced analog and digital (AES3) inputs along with a range of input sensitivities to handle both consumer and pro level output devices. I didn’t spend too much time on the hardware side, as to my ears, sounds transparent. Mostly, I am interested in how the loudspeaker sounds to my ears and how accurate it is from an objective measurement perspective. Let’s get started. Setup, Configuration, and Calibration I pushed my large JBL speakers to the sides and set up the 8c’s on my 24” sand filled, Monolith stands and placed the 8c’s ~30cm from the front wall, as measured to the back center of the subwoofers. Vibrapods are used to isolate the speakers from the stands and the stands from the floor. I did not move the subs out of the way, but they were not hooked up for this evaluation. JRiver MC 24 is the software music player, connected to my Lynx Hilo via USB and then using the AES3 digital output of the Hilo to the AES digital input on the 8c on the left and then using an AES cable to link the left speaker to the right speaker. There is an included AES terminator with the 8c. For critical listening, I case the guitar, put a comforter over the drum kit and move the coffee table out of the way. During the evaluation, I also used the balanced analog outputs of the Hilo to the balanced analog inputs of the 8c’s and did not perceive any difference in sound quality as compared to the digital inputs. As a former recording/mixing engineer, I use industry guidelines from the ITU and EBU to set up my speakers in an equilateral triangle, speakers toed in, on axis, pointing directly at my ears. I also calibrate my listening level, so when I am performing critical listening, I monitor at ~83 dB SPL, C weighting, slow integration, using a calibrated sound level meter. Bob Katz’s article that I linked, provides an excellent overview of the process and why. Most recording/mixing/mastering engineers use the same equilateral triangle setup and monitor level calibration for producing the art. I use the same approach for reproducing the art. For levels below 83 dB SPL, JRiver’s dynamic loudness control is engaged. Remember folks, our ears frequency response changes with signal level. The calibration process also includes using REW, as I measure the frequency response at the same reference SPL and adjust the speaker’s frequency response to my preferred target frequency response at the listening position, using whatever comes with the loudspeaker. While I can easily and expertly apply Acourate or Audiolense DSP, the point of the review is to use only what comes with the loudspeaker. There is good scientific research on subjective listening tests correlating to objective measurements from Sean Olive and Floyd Toole on The Subjective and Objective Evaluation of Room Correction Products and The Measurement and Calibration of Sound Reproducing Systems respectively. From Sean’s slide deck is a preferred subjective ranking of average magnitude responses, objectively measured at the primary listening position: The top preference (red trace) is a flat, but tilted measured response. If 0 dB is 20 Hz, then it would be a straight line to -10 dB at 20 kHz. The “trained listeners” curve in Figure 14 in the Toole reference is similar. Note that this tilted measured response is perceived by our ear/brain, as subjectively flat or a neutral response according to Sean’s research: See how an objectively measured response of 20 Hz and straight line to -10 dB at 20 kHz is subjectively perceived as a neutral or flat response to our ears/brain (red trace overlaid in the above chart). Most participants in the study preferred a frequency response from 20 Hz with a straight line to -10 dB at 20 kHz. A measured “flat” in-room frequency response is not the preferred target, as it sounds too thin or lacking bass. “The Science of Preferred Frequency Response for Headphones and Loudspeakers” goes into more detail and provides links to further studies, which show the same preferences, for both loudspeakers and headphones, after repeated listening trials. As Dr. Floyd Toole says, preferred is synonymous with accurate. I have been using computer based software DSP since 2011 to custom design digital FIR correction filters in both the frequency and time domain for loudspeakers in rooms. In my own listening tests, I prefer the tilted response from 20 Hz to -10 dB @ 20 kHz. To my ears, sounds subjectively balanced or neutral from top to bottom. Whatever your preference is, this is a good place to start, as the subjective listening tests that Sean and team have performed, multiple times, with multiple participants, does correlate to a preferred in-room measured response, assuming good loudspeaker design with smooth directivity. After taking a few measurements with REW, I started playing with the Active Room Matching (ARM) settings. First, I set the ARM for each speaker independently. The left speaker measured 28cm from the front wall and the right speaker measured 38cm from the front wall due to the picture window. The following ARM settings are what I ended up with for the left and right speakers respectively: Notice the treble adjustments, no adjustment on the left and -0.5 dB on the right. As mentioned at the beginning of the article, the 8c’s out of the box hit my preferred high frequency target frequency response with virtually no adjustment. <Soapbox> I find most speakers unlistenable in the top end. They do not exhibit a downward tilt in high frequency response which results in sounding too bright, no matter what room they are in. Looking at my last three speaker reviews, each one needed approximately -5 dB high frequency shelf starting between 3 kHz to 5 kHz depending on speaker, on up to 20 kHz. That’s a considerable reduction in high frequency output to sound neutral to my ears and based on Toole’s and Olive’s subjective listening research. For loudspeakers that don’t have any way to control the high frequency output, one needs to resort to outboard tone controls or DSP to tame the top end. Why so much high frequency output? I find if confounding as to the reason why this is. Is it marketing worrying that if the speaker does not have enough high frequency sizzle that it won’t stand out from the rest – sound boring, or even sound… neutral ☺ Or is the standard “if it measures flat on-axis in the anechoic chamber” approach flawed? As alluded by Kevin Voecks in the link above, the modern approach, again from Toole and Olive, is the spin-o-rama which uses 72 anechoic measurements and computer processing algorithms to determine a neutral sounding speaker based on this anechoic data. Certainly described in Toole’s book and even this presentation from 2002. How to make neutral sounding loudspeakers has been known for some time… Today, loudspeaker manufactures can choose to make more neutral sounding loudspeakers by using what Floyd Toole and others have contributed to an open industry standard called, Standard Method of Measurement for In-Home Loudspeakers ANSI/CTA-2034-A.” This is just a preview, as the standard costs a hundred dollars or so to purchase. If you spend time doing the research, one can conclude that this measurement approach can predict with a high degree of accuracy what a good sounding loudspeaker will “sound like” in an in-home environment correlated to objective anechoic measurements. I sincerely hope more loudspeaker manufacturers leverage this “standard” measurement approach to produce more neutral sounding speakers that don’t require significant high frequency output reduction to sound neutral… </Soapbox> Kudos to D&D for engineering the first loudspeaker I have measured that requires no or minimal high frequency adjustment to sound neutral in my room. Next I dialed in some Parametric EQ below 500 Hz. Given the 8c’s mid and high frequency response is lining up with my preference, no eq is required. What we are doing here is bringing down the room modes or standing wave peaks a bit to provide a smoother response. It is room ratio dependent and will be different for each room. Using REW, I measured each speaker independently and here we are seeing the left and right onboard PEQ’s I ended up with, as I fine-tuned the response using multiple REW sweeps: Clicking on a filter will provide a details view of the parameters that can be varied in each PEQ: This is filter-3 in the list for the right 8c above. One can adjust the frequency, gain and Q factor. The Q factor refers to the width of the parametric filter. The higher the Q the narrower the width of the filter. With the ability to toggle the filter off and on in real-time, one can assess the impact by ear and/or measurement microphone whether playing music or test tones. With the measurement microphone at the listening position, I made small adjustments by bringing down the peaks and leaving the dips alone. The psychoacoustic science says our ears are more sensitive to peaks than dips in frequency response. And that our ears hear the envelope of the frequency response, like 1/6 octave visual display smoothing is reasonably close to what our ears/brain perceive. I am not showing the detail for every PEQ, as it is room dependent. All of them were small adjustments to smooth out the response below 500 Hz. With setup, configuration, and calibration complete, let’s look at the measurement results. Objective Measurements For this article, I thought I would take a couple of near field measurements first to show how good the frequency and phase response is for the 8c’s. This is with my calibrated measurement microphone pointing in between the tweeter and mid-range driver, 30cm away: The default impulse response measurement window is 500 milliseconds long, so a half second of room reflections are definitely in the result. Even still, look how smooth the frequency response is: The room’s standing waves are encroaching at low frequencies, but there is virtually no speaker boundary interference. The back of the 8c’s are (L) 28 cm and (R) 38 cm away from the front wall. Optimizing this distance in the Active Room Matching user interface, for both left and right speakers, virtually eliminates any SBIR. These measures are with no PEQ’s applied. How good is it? Check out the phase response over frequency: Virtually textbook. If you look back at the picture of the 8c in the room, you can see all sorts of boundaries, floor, side of subwoofer, angled partition to the left of the speaker, aside from the subs being back 28 cm from the front wall. The 8c’s loudspeaker design with built in DSP technology is so good we virtually get a flat phase response coming off the speaker and front wall. Put another way, the sound from the rear subwoofers bouncing off the front wall are in phase with the rest of the speaker. Something to ponder… If we looked at the step response, we would see the direct sound arriving at the microphone, all at the same time. Also known as a time aligned or time coherent system. An amazing technical feat given the number of odd boundaries to my already disadvantaged room ratio room. What is even more impressive is this is with a few parameters punched into the ARM section, with no PEQ’s, and already the performance is virtually textbook. Martijn and team at Dutch & Dutch have done an outstanding technical design job to effectively mitigate bad room effects (i.e. SBIR), yet at the same time leveraging good room effects (i.e. the Allison effect). This allows one to further enjoy the incredibly accurate sound quality these loudspeakers reproduce. Moving the measurement microphone to the listening position: Note I move the coffee table and couch out of the way. The issue with leaving a couch or chair near the measurement microphone’s proximity is that you are correcting for both the direct sound and comb filter response. Similar comb filtering occurs with a coffee table or any object between the speakers and measurement mic. You will get a much better sounding result with nothing around the proximity of the microphone or between the microphone and speakers during calibration and then adding the furniture back after. As mentioned in the calibration section, I kept sweeping REW’s test tone and looking at the charted response. First I adjusted the Active Room Matching (ARM) parameters for smoothest response that matched my preferred target response. Then, I worked on the low frequency room modes or standing waves below 500 Hz using the on-board PEQ’s. I arrived at the following frequency response measured at the listening position: That’s an excellent “in-room” measured frequency response using no external eq or DSP. It is just about ±3 dB from 21 Hz to 20 kHz. Studio control room frequency response accuracy. Given my unfavorable room ratios, it really says a lot about the 8c’s engineering design. I would expect most rooms to be better than mine and could achieve as good or even better result. How closely can I match the 8c’s to Olive’s research on preferred (neutral) frequency response? Wow, that is quite close compared to my measured reference frequency response using custom designed DSP FIR filters and subs. The 8c’s effectively hit the target 100%. That’s a first for any loudspeaker I have measured in my room without resorting to external eq or custom DSP. Fantastic job D&D for your loudspeaker to not only adapt to my room, but hit my preferred neutral in-room target frequency response. Let’s have a look at the dispersion characteristics of the 8c’s where I move the measurement microphone, 3 feet left of center from the LP, measure left and right speakers, then 3 feet right of center, and measure left and right again: Impressive. Simply the HF level is a bit lower with very little high frequency response deviation over a six foot listening area. Virtually perfect off axis high frequency response from 1 kHz on up. Inevitably the room comes into play down low, but the good news is that they are all dips and our ears don’t really hear those, as we are more attuned to the envelope. Putting a ruler on the graph along the slope of 20 Hz to -10 dB at 20 kHz, the envelope pattern holds, even across a 6 foot seating area. Confirmed in my listening tests, the 8c’s sounded incredibly smooth both at low and high frequencies, no matter where I shifted around on my six foot couch while music is playing. I believe we are seeing the future of loudspeaker design with fully integrated features, like being able to appropriately adapt to any room. It is simply a matter of time before FIR filters with enough taps applied to correct more of the room becomes a reality. I suspect Acourate or Audiolense or the like will soon be incorporated, using OpenDRC into these “all in one” loudspeakers. We have already reached the pinnacle on PC’s using powerful DSP software programs that can reduce low frequency group delay or eliminate room modes and standing waves. Well, the 8c’s are the most accurate loudspeakers that have come through my living room. There is no point in displaying the timing (i.e. step) response, as we already know the speaker is time coherent from the near field measurement shown earlier. The phase is virtually flat across the frequency band indicating that the 8c “system” is time aligned or linear phase. Put another way, we need a linear phase “system” to reproduce an ideal loudspeakers minimum phase response, as speakers, like microphones, are minimum phase devices. The trouble occurs when adding speaker crossovers that distort the phase, along with room acoustic issues below Schroeder when the response is no longer minimum phase. So… just about virtually every loudspeaker in any room, except when using custom DSP solutions or a pair of D&D 8c’s! Subjective Listening Results I spent over 100 hours listening to these speakers with a wide range of digital media content. The 8c’s are the most accurate speakers I have reviewed to date. The 8c’s sound great at any listening level from night time “don’t wake the kids” to “stupid loud.” While they can’t match my dual 15” cabs per side with dual 12” subs, they do put out a prodigious amount of musical sounding bass that totally belies their bookshelf size. Both my wife and daughter thought my external subs were hooked up (they weren’t). One of my favorite alt bands, the Brian Jonestown Massacre has several tunes I enjoy, including this dreamy/spacy sounding tune, “Dropping Bombs on the Sun.” Sounds staggeringly huge with solid low frequency response. Because the bass line has quite a few different, but long sustaining low notes, I find it a good tune to listen to how the bass sounds in the room. Nice and full sounding, without sounding boomy or “one note” resonance sounding. Just solid. Madonna’s Ray of Light, arguably one of her best, and while I am not a huge Madonna fan, I am blown away by this William Orbit produced adventurous album which has some incredible music and mixes. Solid low end response with Madonna’s voice perfectly recorded and not overly sibilant, as compared to some of her other recordings (e.g. Fever). Drowned World, the Substitute of Love is quite the mix. Not once did the 8c’s sound strained, as I kept turning up the volume to old school concert level of 95 dB SPL at the listening position. These speakers sound way bigger than they look. What you don’t see is the double 8” subs leveraging the front wall for some huge and impactful bass. At the beginning of Juno Reactor’s, Immaculate Crucifixion is a low frequency drum sound that comes across as if the entire front wall is flexing in the living room when turned way up. It is one of those mixes that makes you feel you are at a club with the strobe lights flashing and the dance floor pulsing to the beat. Moving up the dynamic range scale, ever since I first heard Peter Gabriel’s Security CD played over large format control room monitors at a recording studio, I was floored at how good it sounded, especially the drums. Shock the Monkey (DR16) has really good drum slam. Not only dynamic sounding, but the stereo mix sounds crystal clear and well defined across the 8c’s. It is as if the entire front wall of my room is the size of the stereo image. The polar response of these speakers and the mix of direct and indirect sound was just about perfect in my room. I normally listen to higher directivity speakers, as I like more direct sound and less of my (crappy) room sound. But I am really digging the 8c’s, which are a fraction of the size of my JBL Cinema loudspeakers. As a former recording/mixing engineer, creating the stereo illusion with mono mic’d in the studio multi-track approach is not an easy task. If you would like to see a visual representation of mixing, I would highly recommend David Gibson’s YouTube of The Art of Mixing. Prepare to have your mind blown. Why do I keep bringing up polar response and directivity index? While speakers can have the same on axis frequency response they will sound different based on their polar response or directivity index. Some folks really like the diffuse sound of omnidirectional loudspeakers or open baffle loudspeakers. Others like a more focused direct sound and less room acoustics. The latter is my preference. However, the 8c’s have a nice balance of direct versus reflected sound. I really enjoy the size of the image that come off these loudspeakers, certainly reminds me of the size of the sound of my large JBL’s, but at a fraction of the size. Really an amazing feat. Flim and the BB’s Tricycle DR19. Turning it way up is a great rush, as the band really punches it. I get a kick of how loud the band comes in at the top, right after the piano intro. The 8c’s sound incredibly dynamic and tight. As I mentioned before, the top never became shrill or harsh as I pumped up the volume. The horns timbre never altered, even up past 95 dB SPL, which is twice as loud as my critical listening level. The clarity of these loudspeakers are outstanding. This is because of the phase coherent wave front being sent to my ears with no speaker boundary issues. If you have the measuring tools, try using REW to measure your loudspeaker 30cm from the center of the speaker and look at the frequency and phase response. Compare to the 8c measured results above… One of my most favorite tracks is SRV’s Tin Pan Alley with DR18. When turned way up the dynamics feel like being at the concert hall, as I found the 8c’s to have a wonderfully enveloping sound when my front wall becomes a window into the performance. The tonal balance on these speakers is spot on. My teenage daughter likes Rihanna’s Shut Up and Drive. Quite the dance mix with bright vocals and a ton of low bass. Even cranked up twice as loud as my critical listening level with peaks beyond 95 dB SPL, the vocals still sounded smooth and the subs are really pressurizing the room: The 8c’s are so close to the front wall, I could only get one of the subs on video, but I think you get the idea. LOL. After listening to that in my 16’ x 31’ x 8’ living room, it is pretty convincing that these speakers don’t require subs unless you are complete bass head. Moving on to listening to some high frequency content, Marilyn Mazur’s, Bell Painting (DR25 on CD!). The attack, clarity, and decay of the percussion instruments sounded realistic to my ears. I have a number of percussion instruments at home, including a few bells and it is really surprising to hear the reproduction sounding virtually identical to the real thing. The “stick on bell or cymbal” transient response is realistic sounding, as I play sticks on the ride cymbal and bell on my drum kit in the living room and compare as the loudspeaker is playing. The purity of the tone quality (i.e. timbre) is unlike anything I have heard. I don’t know how quite to describe it, other than it is accurately realistic to my ears. The bells are accurately reproduced anywhere I move on the couch while the bells are playing. There is no polar response “drop outs” where the high frequency abruptly disappears and then one moves their head a few inches and it reappears. The CNC waveguide design matching to the magnesium alloy 1” dome tweeter has really paid off. There is no beaming or any anomalies across the high frequency range, as I listen to the bell transients and decays. Smooth as silk on and off axis. Best I have heard. While I am not much of a classical music fan, one piece I do love, aside from all the music on 2001: A Space Odyssey, is the Blue Danube played at the end credits. I just play the movie end credits to listen. Sounds completely enveloping to my ears, as if I am transported to the hall while I keep turning the volume up. The concert hall seems to have infinite depth. The music just floats perfectly in the halls reverb. So well recorded to get the right mix of direct and reverberant sound. Outstanding imaging, as the front wall of my living room becomes the concert hall. What more could I ask for. Conclusion What ingenious loudspeaker design and engineering: Studio control room frequency response accuracy from 21 Hz to 20 kHz ±3dB at the listening position. Perfectly smooth response on and off axis. I.e. perfectly controlled directivity from 100 Hz to 20 kHz. High frequency output is perfect right out of the box. No large HF attenuation required to sound neutral. Midrange cardioid polar pattern to avoid speaker boundary issues, yet at the same time leveraging the Allison effect for more bass output/impact off the front wall from the dual subs on the back of the cabinet. Perfectly phase coherent loudspeaker design, as the direct sound and the reflected bass off the front wall is arriving at one's ears at the same time. No time smearing. Sophisticated room calibration controls with Active Room Matching and access to 24 PEQ’s per speaker. Low latency mode for TV or movie watching to avoid any lip-synching issues. All in one “bookshelf” package. It is serious at 53 lbs. If you have made it this far in the review, obviously, you can tell these are my new favorite accurate loudspeakers. I am really not sure what could be better than these, especially how well they integrated into my challenged room ratio room. Folks will ask how the 8c’s compare to the Kii THREE’s. I did get to listen to both back to back. I even put a Three on one stand and an 8c on another stand and ran them in stereo. Other than low latency mode, both have a different delay and almost impossible to reconcile between them. But I did listen to them back to back. Because of the cardioid response for SBIR control, both of these speaker designs are ahead of the curve. The 8c’s have larger midrange and sub drivers, a more refined high frequency directivity response when comparing the two polar maps, and the advantage of more onboard room integration controls (e.g. PEQ’s). The latter resulted in a smoother or more tightly controlled frequency response down low for the 8c’s. The 8c’s larger drivers sound more dynamic. The high frequency directivity response, aside needing HF attenuation on the THREE’s, sounded a bit smoother on the 8c’s. The 8c’s dual 8” subwoofers in my room are enough to not require external subs… If I was to try and subjectively characterize the sound difference, both are very neutral performers, best I have heard. The THREE’s sound a bit drier than the 8c’s. I don’t know any other way to describe it. Objectively speaking, it may be a damping factor difference between the amplifiers. Both measure just about textbook response in the frequency and time domains. But with the 8c’s larger drivers, more onboard room calibration controls, and the best directivity polar map I have seen, edge out the Kii THREE’s to my ears. But that is just my preference. To me, a loudspeaker like the 8c is the future of loudspeaker design and engineering, arriving today. Take everything we know about the science of small room acoustics and psychoacoustics and equip the loudspeaker with tools so we as consumers can mitigate the fact that loudspeakers are listened to in rooms. It is just a matter of time that the loudspeaker will ship with a calibrated measurement microphone so the onboard computer can optimize loudspeaker room placement. For now, Martijn’s suggestion is placing the subs 40 cm from the front wall. He recommends that as a good starting point, if you can get them that close to the front wall. This will fully enable the Allison effect while mitigating SBIR issues. Worked out really well for me and why I say the 8c’s are pretty damn good without external subs. I can’t get over the polar response of these loudspeakers and the size of image they throw. It extends from floor to ceiling and beyond the width of the speakers. The stereo image or auditory scene or whatever you want to call it, totally belies the size of these speakers. It gives a new meaning to picture window ☺ Sorry to see these go… Now what? Enjoy the music! I wrote this book to provide the audio enthusiast with an easy-to-follow step-by-step guide for designing a custom digital filter that corrects the frequency and timing response of your loudspeakers in your listening environment, so that the music arriving at your ears matches as closely as possible to the content on the recording. Accurate Sound Reproduction using DSP. Click on Look Inside to review the table of contents and read the first few chapters for free. Mitch “Mitchco” Barnett. I love music and audio. I grew up with music around me, as my mom was a piano player (swing) and my dad was an audiophile (jazz). My hobby is building speakers, amps, preamps, etc., and I still DIY today. I mixed live sound for a variety of bands, which led to an opportunity to work full-time in a 24-track recording studio. Over 10 years, I recorded, mixed, and sometimes produced over 30 albums, plus numerous audio for video post productions, in several recording studios in Western Canada.