Jump to content
  • mitchco
    mitchco

    Dirac Live 2 Digital Room Correction Software Walkthrough

    Dirac 2 is the latest version of Dirac’s Digital Room Correction (DRC) technology. In this article, I walk through the steps of using Dirac to optimize the response of the Purifi SPK4 demo kit that I recently reviewed here on Audiophile Style.

     

    Can we make an already great sounding speaker sound better in my room? And by better I mean smoothing out the low frequency response below the room’s transition (Schroeder) frequency, making small, broad band tonal adjustments in the midrange and top end to be a bit smoother and finally a timing correction of the impulse response.

     

    The short answer is yes we can. We can see the improvement in the measured frequency and timing response. I can also hear an audible difference with a tighter, more clear sounding bass, smoother overall frequency response and a more coherent timing response (i.e. stereo imaging and depth of field) across a larger sweet spot.  

     

    With Dirac 2, designing and generating a partial correction from about 600 Hz on down below the room’s transition frequency is as simple as it gets. The sonic benefits are instant, much smoother bass response with no huge peaks and dips that plague virtually every room below Schroeder frequency. Full range correction comes with improved phase response (i.e. imaging and depth of field) covering a wider sweet spot, more on that in the subjective listening section.

     

    Before the walkthrough, let’s talk a bit about Dirac’s room correction technology.

     

     

    Technology Discussion

     

    First, what audio problem is room correction technology designed to solve? The unfortunate reality is that below a room’s transition (a.k.a. Schroder) frequency, the room controls the bass response arriving at your ears, not the loudspeakers. Due to modal resonances we get wildly different bass responses depending on where we place speakers in rooms relative to the listening position:

     

    image1.png

     

     

     

    The above chart is from Floyd Toole’s Audio Science Article.  As Floyd says, “In the investigation of many rooms over the years, I would estimate that something like 80% have serious bass coloration.” For those of us that take acoustical measurements of audio systems, this comes as no surprise. The peaks and dips in the low frequency response is typical and exists in virtually every room that was not specifically designed using proper room ratios, aside from construction and acoustic treatments.

     

    Room modes are calculated based on one’s room dimensions where room resonances take control of the low frequency response regardless of the speakers being used. I like this Room Mode calculator as you can move the cursor along the frequency scale and it will output a tone at that frequency. If your computer is hooked up to your speakers (careful with the volume!), you can hear the resonances in your room by hovering the cursor over the modes in the graph. It is an ear opening experience and great for training ones ears to know what to listen for.

     

    Room modes can sound like “one note bass” where the room mode is so bad, all you hear is the single droning bass note that drowns everything else out. Or the “where did the bass go?” when certain bass notes are played. Or simply just uneven bass, some notes are there, some are gone, and some are too loud.

     

    Dirac 2 is designed to smooth out the peaks of the low frequency room response and fill in some of the dips across the listening area. Our ears are more sensitive to peaks than dips. A good primer on why we hear what we hear in small room acoustics as it relates to room correction is James (JJ) Johnston’s, “Acoustic and Psychoacoustic Issues in Room Correction.” Download the PowerPoint if you can. The first 31 slides are worth the read.

     

    Dirac uses a “mixed phase” filter approach which is a combination of IIR filters and FIR filters for efficiency. The white paper linked goes into detail behind the design, and can be summarized as:

     

    A first main aim is to control not only the frequency domain properties of the system but also the time domain properties: The impulse responses as measured at different listening positions. In particular, we strive to reduce the “pre-ringings” (pre-echoes) that would otherwise result in an un-natural sound experience. 

     

    Secondly, we use dynamic models of the sound system that are based on measurements at multiple listening positions. This is important for obtaining a robust design that works over an extended region to provide a large spatial area with good sound quality. 

     

    Third, we may jointly optimize multiple loudspeakers to better control the sound pressures at different listening positions. This is done by precise phase control of the individual loudspeaker transfer functions at low frequencies. Joint optimization of a set of loudspeakers results in more distinct bass performance, better robustness of the compensation and better control of the impulse responses at different listening positions.

     

    Room correction is a complicated (and mostly misunderstood) subject area. If there is interest, I might write another article to discuss this in more detail as it is not just “eq” that is being applied here (i.e. the time domain properties are equally important).

    Dirac 2 consists of two pieces of software that work together. The Dirac Live Processer (DLP) is a VST plugin that “hosts” the correction filters in a convolution engine. The Dirac Live application is used to measure your loudspeakers in your room, design the filters, and upload the correction filters to the DLP.

     

    One interesting difference to other convolution engines is that the DLP accounts for any latency issues, so when switching the DLP on or off, the audio change is instant as opposed to a ¾ second silence gap with a 65K tap FIR filter. There is also no latency when switching between filters, which makes A/B comparisons easy so you can “bracket” in on your preferred frequency response in a shorter period of time.

     

    Since the Dirac Live Processor is a VST plugin, any software music player or Digital Audio Workstation (DAW) that supports the VST plugin architecture can simply load the DLP. In this article, I am using JRiver Media Center 64 Bit DSP audio architecture to host the DLP:

     

     

    image2.png

     

     

     

    I clicked on Manage Plug-ins in JRiver and navigated to where the DLP VST was installed and loaded it. Here we are seeing the DLP loaded into JRiver and active with a correction uploaded into the first of eight slots available.

     

    The 2nd piece of Dirac Live is an application that you run on your computer that is mostly “wizard” driven, and steps you through the process of:

    1. Connecting and configuring audio I/O 
    2. Adjusting playback and record levels 
    3. Select the type of measurements you want to take (i.e. chair, couch area, etc.) 
    4. Taking the measurements 
    5. Designing a correction filter whether partial or full range using a target of your preference  
    6. Examining the results of the correction in a simulation
    7. Uploading the correction filter to Dirac Processor
    8. Managing measurements, targets, corrections, etc.

     

    So let’s perform step 1 of the Dirac Live application wizard in the objective measurements section

     

     

    Objective Measurements

     

    image3.jpg

     

     

    For the measurements, I moved the couch and coffee table out of the way and the sub woofers are not active.

     

    This is the main screen you see once you have downloaded Dirac Live, installed it and connected to the DLP. Note that you are required to login to your Dirac Live account in order to connect to the DLP. Also, in order for the Dirac Live application to communicate with the DLP, the latter needs to be enabled and actively receiving an audio signal. Meaning that the DLP is enabled in JRiver and music is playing through Dirac Live Processor so it is “active.” Without the DLP being active, the Dirac Live application cannot connect to the DLP.

     

    image4.png

     

     

     

     

     

    Each step of the wizard comes with a help screen of what to do. It is concise and well written so you can quickly get the hang of what to do. Here we are selecting which recording device to use once we clear the instructions:

     

    image5.png

     

     

     

     

    Here I am using my Lynx Hilo ADC with a microphone preamp and calibrated measurement microphone. I have clicked into the area with the red square and loaded my measurement mics calibration file. In my case, the mic calibration file uses comma separated text whereas Dirac is expecting periods. Easy enough to open in a file editor and replace commas with periods. In most case, like using a UMIK-1 mic, the microphone calibration file will load without any calibration file modifications.

     

    Next step is to adjust speaker level and microphone gain settings:

     

    image6.png

     

     

     

     

    I wholeheartedly agree with “keep the master volume at the lowest setting and gradually increase it.”

     

    image7.png

     

     

     

     

     

    I have pressed play on the left channel and adjusting the levels to be in the “green.” If you have a sound level meter, 75 dB SPL C weighting at the listening position is loud enough and I would not go beyond 85 dB SPL, which to our ears, sounds twice as loud as 75 dB SPL.

     

    If you don’t have a sound pressure level meter (even a smart phone SPL app will do), then adjust to a comfortable playback volume. If you feel like covering your ears, then it is too loud.

     

    Clicking on next brings up the type of measurement you would like to take. These involve taking multiple measurements, the number of which depending on what type of listening arrangement is selected. The wider the listening area to cover the more measurements one needs to take. So it could be as few as 5 measurements and as many as 17:

     

    image8.png

     

     

     

     

     

    In this walkthrough, I tried two listening arrangements, a single chair with 5 measurements and the sofa wide imaging with 17 measurements. This is so I could also subjectively compare the two corrections. First lets walkthrough the single chair selection.

     

    Single Chair

     

    image9.png

     

     

     

     

    Click on Proceed to Measure:

     

    image10.png

     

     

     

     

    I decided to use the minimum amount of measurements for this setup to test out Dirac’s algorithms. Above, I just completed the 5th measurement. The idea is that you move the microphone relative to the placement as indicated in the diagram. In general, the mic placement is 40 to 60cm apart per placement and you can choose whatever order you want take the measurements by clicking into the location on the diagram. Note that the precise placement of the mic to the location is not crucial.

     

    As a side note, it is good to temporarily remove any objects between the speakers and the microphone positions, including the chair and or sofa at the listening position. We measure and correct for the loudspeaker and room with an unobstructed path and nothing around the microphone so we are getting a clean as a sound field as possible. We don’t want to correct for any reflections between the speakers and mic due to a coffee table or around the mic itself as that will lead to a colored sound. It is best to measure and correct for the unobstructed sound field first and then return the furniture to its original locations. Finally, I recommend getting a real boom mic stand for taking measurements. You can grab one at your local music store or order online. The little mic stand that comes with the UMIK-1 while cute, is not helpful as you need to place it on something to get ear height, which means reflections (i.e. colorations) are getting into the mic.

     

    Each time you take a measurement, Dirac will output a sweep signal starting with the left channel, then the right channel, then back to the left channel. So 3 sweeps per location times 5 locations equals 15 sweeps: 

     

    image11.png

     

     

     

     

     

    Here is the screen you see when taking a measurement, with the abort button in case you need to abort the measurement for whatever reason.

    Once you are happy with your measurements, click on proceed to filter design:

     

    image12.png

     

     

     

     

     

     

    Dirac suggests a target curve based on the measurements you took. Generally speaking, one is looking for a downward tilt of the frequency response at the listening position. This is because most forward firing drivers in loudspeakers start off as omnidirectional at low frequencies and become progressively more directional at higher frequencies. The rising bass energy yields a steady state room curve with a downward tilt. 

     

    How much tilt? The scientific research that Sean Olive (and earlier Floyd Toole) conducted over years shows that a slope of 20 Hz to around -8 to -10 dB at 20 kHz is what most listeners prefer as neutral sound. A high level overview of that can be found in, The History of the Harman Target. While the overview also discusses headphone target curves, which have the same preferred target (transfer function) as loudspeakers in rooms, it also discusses loudspeaker target preferences. More importantly, there are a listing of AES papers that one can get the details on. One example is the Subjective and Objective Evaluation of Room Correction Products:

     

     

    image13.jpg

     

     

     

     

     

     

    Of course, you are free to implement any target curve you want, but a good starting place is to use Dirac’s suggestion and if it is a bit too bright or dark sounding, one can always adjust the curve. Note that a flat in-room response is not the ideal target response as it will be exceedingly bright sounding. If you read through Sean Olive’s presentation referenced above, and in these slides, it is actually the tilting frequency response that our ears perceive as flat or neutral sounding at the listening position (see slide 25). Well worth the read.

     

    Looking at the analysis chart below, we are seeing the processed 5 measurements down to left and right, plus the suggested target curve:

     

     

    image14.png

     

     

     

     

     

    So these wonderful little 15L bookshelf Purifi speakers have a frequency response with the -3dB limits at 25 Hz and 22 kHz. Outstanding for such a small 2 way bookshelf speaker. Ctrl-click both checkboxes, top right to get both curves to display.

     

    Note the room resonances showing up below 300 Hz in my room and if I were to choose a partial correction frequency, which I have done a bit later, I would choose 600 Hz and below to be corrected as we can see by 600 Hz the “up and down” peaks and dips are no longer an issue. In fact, having been using room correction since 2011, and having seen many other people’s measurements of loudspeakers in rooms, 600 Hz seems to be the most common point where the room starts to have an influence on the frequency response and gradually taking over completely by the time one reaches the rooms transition frequency, which in my room is at 200 Hz. As linked earlier, you can use this room mode calculator to calculate your room’s transition frequency.

     

    See the bottom right panel where it says on one side measured and corrected on the other side? Dirac has already calculated the correction in the background and by unchecking the measurement and checking on the correction, we get:

     

    image15.png

     

     

     

     

     

    As one can see, a considerably smoother response. So not only smoothing out below 600 Hz, but also bringing down the broadband rise from 2 kHz to 5 kHz when compared to the original measure. Still a little dip left at 100 Hz, but as folks may have read in JJ’s presentation, our ears are not as sensitive to narrow band dips as compared to peaks, so all good. Note the (very) slight channel level difference in the top end frequency response. As @flak has commented here, “Dirac Live 2.x algorithm might even slightly compromise on FR response of one of the speakers in favor of the best possible phase coherence between them that is important for imaging.” We will be exploring that in the subjective listening section.

     

    Next I clicked on “Proceed to filter export” and here I am exporting the correction filter to DLP. I can give it a name and description:

     

    image16.png

     

     

     

     

     

    Click on Export filter and Dirac takes a few seconds to upload the filter:

     

    image17.png

     

     

     

     

     

    Once uploaded, I can navigate to JRiver where I have the DLP VST loaded and enabled:

     

    image18.png

     

     

     

     

     

    It is recommended to leave the delay and gain compensation on. I also adjusted the gain slider to about -6 dB for headroom management. The meters will peak red if there is any clipping, which I experienced none. I could probably go to 3 dB of headroom management, but I have plenty of power on tap.

     

    OK, now listening to music! But before I comment on the sound, I also wanted to try a partial correction to 600 Hz as can be seen as a foreshadow in slot 2 ☺

     

     

    Partial Correction

     

    I can go back to my saved measurements, don’t forget to save!

     

    image19.png

     

     

     

     

     

    And back to the filter design section. Creating a partial correction is dead easy. All I need to do is grab the right marker that was previously at 22.1 kHz and drag the marker to the left to about 600 Hz:

     

     

    image20.png

     

     

     

     

     

    So the frequencies between 25.7 Hz and 587 Hz will be corrected and the rest left alone. Note, if I were using Dirac and Digital Room Correction for the first time, this is where I would suggest starting with a partial correction. This way one can focus your attention on just listening for an improved bass response and not a change in overall frequency response. 

     

    We proceed to filter export and load the filter into the 2nd slot:

     

    image21.png

     

     

     

     

     

    As one can see, I have several filter designs loaded, but here we are focused on the partial correction.

     

    Let’s move on to the sofa, wide imaging measurements.

     

    Sofa Wide Imaging

     

    image22.png

     

     

     

     

    After you have selected Sofa and Wide imaging, proceed to take the measurements:

     

    image23.png

     

     

     

     

     

    In this screen shot, I am on my last of 17 measurements. 3 sweeps per measurement times 17 measurements = 51 sweeps. Whew, that’s a lot and took me about 15 minutes to get through them all. I did not use a measuring tape. I just took a guess as to where the mic should go at each location, either raising or dropping the mic height and placing the mic at the front or back of my sofa, which I have temporarily moved out of the way. Note these 17 measurements covered a 6ft wide x 2ft deep x 2ft of height variability where my couch is.

     

    At any time you can re measure a position. Again we proceed to filter design. I just used the same default target as before and then switching to the corrected filter response:

     

    image24.png

     

     

     

     

     

    The response is so smooth I need to have a look without the target obscuring the correction:

     

    image25.png

     

     

     

     

     

    Wowser! That’s based on 17 measurements around my three seat couch area. To give an idea of all of the responses around the area, Dirac has a “spread” feature that allows you to see this if you check the box in the lower right hand:

     

    image26.png

     

     

     

     

    Those are 17 measurements made of left and right channels at the various locations around my 6’ x 2’ couch area. One can see the variability of the measurements based on position and again, by about 500 Hz, there is less variability as the loudspeaker is in control of the frequency response and not so much the room.

     

    Switching the view to impulse response:

     

    image27.png

     

     

     

     

    The corrected impulse is on the left and measured impulse on the right. As one can see on the left, the impulse or transient response is a cleaned up version of the measurement. Vertical scale is amplitude and horizontal scale in milliseconds. 

     

    The big vertical spike is the direct sound from the loudspeaker arriving at the microphone without any reflections. The smaller spikes, after the direct sound are early reflections. The corrected impulse response shows a clean spike and very little spikes (i.e. early reflections) after. 

     

    Sound travels roughly 1 foot per millisecond, so it becomes easier to understand that right after the direct sound, are the reflections (i.e. diffraction) off the speaker cabinet and then the speaker stand and anything else nearby like floor bounce or front/side wall bounce or ceiling bounce, all of those reflections are contributing to the sound, after the direct sound and our ears have a pattern to integrate those early reflections (i.e. the Haas effect) and understand amplitude and frequency response.

     

    It is a complicated topic as it involves so many aspects of why we hear what we hear in small room acoustics; standing waves, room resonances, early reflections, later reflections, room ratios, room construction and treatments, digital filtering, FIR, IIR, our ears non-linear response to both amplitude and frequency response, the Haas Effect, which reflections are good, which are bad, and on it goes.

     

    At this stage, I just wanted to show that Dirac works in the time domain. Given multiple measurement locations, one can understand that the sound and reflections arrive at different times and in different ways depending on the mic location relative to the loudspeakers. Using the concept of superposition and what we know about the physics of how standing waves respond in a rooms, we can work out the pattern and therefore correct for a wider sweet spot, which is what the 17 measurements are about. Dirac knows the phase relationships between the measurements plus taking the speakers as a pair into account, is used to calculate the correction filter.

     

    A note about target curves. While I have been using Dirac’s suggested target curve based on the measurements, you are free to create your own target curves. As mentioned above in the Subjective and Objective Evaluation of Room Correction products, the most preferred or accurate or neutral response is one that is 20 Hz to about -8 or -9 dB at 20 kHz. 

     

    Here I have dialed that in the target curve by grabbing the control points of the target and dragging them down a bit to get the desired target response. Of course, one can add and delete control points along the target (right click on the target). One can also save and load targets from the menu at the top left, beside the help icon.

     

    I was able to catch Dirac’s old correction computation as it was calculating the new correction along the new target:

     

    image28.png

     

     

     

     

     

    You can see how much I pulled down the target. It is not a lot, but is an audible difference. Just my personal preference. Of course, export to DLP and start listening:

     

    image29.png

     

     

     

     

    I have 7 of the 8 slots filled with a variety of filters to listen to. Before we get to the subjective listening section, here are a couple of “spot” verification measurements I made with REW.

     

     

    REW Verification Measurements

     

    This is just a “spot” verification of the correction. It is a loopback measurement where I am feeding the digital output of REW sweep test tone into the JRiver’s ASIO digital line input and it passing through Dirac Live Processor with the correction engaged and fed out my Hilo DAC to the Purifi amp and speakers, picked up by the measurement mic, through a mic preamp, Hilo ADC and routed to REW’s digital input. Then the measurement can be displayed in a variety of graphs to look at different viewpoints of the acoustic measurement of the room and loudspeakers. This is not a “gated” measurement as we do want the room’s low frequencies in the picture and using REW’s default impulse window of 500 milliseconds.

     

    One should note that this is not fully representative as it is one measurement made at the middle listening position. Ideally, I would take 17 measurements at the same or similar locations as I did in Dirac and vector average them in REW. And even then it still will not be quite the same as Dirac is applying more than just a vector average. However, it does verify that Dirac is doing what it is supposed to do:

     

    image30.jpeg

     

     

     

     

    On the top are the Purifi SPK4 speakers measured at the listening position from my previous article and the bottom with the Dirac correction enabled. The speakers are in a 9ft equilateral triangle. I just split the screens apart to make it an easier comparison. Let’s start with the low frequency response and work our way up to the high frequencies, calling out the differences.

     

    Dirac was able to extend the low end frequency response in my room to below 30 Hz. -3 dB down at 22 Hz for the left speaker (as it is more in the corner of the room) and 28 Hz for the right speaker, as it is almost center in the room and does not get the corner boundary reinforcement. Compare that to -3 dB down at 32 Hz on the left and 45 Hz for the right speaker without correction.

     

    Looking at the 30 Hz to 60 Hz region on the correction, both speakers are matched closer together in amplitude response for a more solid bass response. This is in addition to being smoother and less peaky at 36 Hz for the left uncorrected speaker and 55 Hz for the right uncorrected speaker. Remember, our ears follow the “envelope” of the frequency response with our ears more attuned to peaks rather than narrow dips in frequency response.

     

    The big correction comes at 80 and 90 Hz as one can see in the uncorrected response. The total swing from the lowest point at 90 Hz to the highest point at 110 Hz is 23 dB. To our ears, a bass note playing at 90 Hz will be ¼ as loud as compared to a bass note at 110 Hz and vice versa. I.e. a bass note at 110 Hz will sound 4 times as loud as compared to the bass note at 90 Hz. This is classic room modes and we all got em, albeit at different frequencies dependent on one’s room ratio. While the dip is still there, it has been reduced by over 50% to ±5 dB which is very close to studio tolerance of ±3 dB. I suspect it would be in this range if I vector averaged 17 measurements in REW. Good enough for a spot check.

     

    Note Dirac pulls downs the peaks at 234 Hz, 461 Hz and 1.2 kHz. We also see a smoother response from 2 kHz to 5 kHz which just pulls down the brightness of the AMT tweeter just a bit. Finally, we extend the response a bit past 10 kHz. From good to great goes the speaker’s in-room frequency response, not only smoother but extended on both ends of the frequency spectrum.

     

    Let’s look at the timing or step response. Dirac makes a claim that “misaligned drivers in multi-way loudspeakers can be corrected by automatically applying different delays to different frequency ranges.” So let’s verify this claim by first looking at the step response of the uncorrected Purifi SPK4:

     

    image31.jpeg

     

     

     

     

     

    So the spike (doublet) at time 0ms on the horizontal scale is the AMT tweeter and the negative going portion is perfectly blended into the start of Purifi’s PTT6.5 woofer. Classic or “textbook” passive crossover design. However, not truly time aligned. 

     

    Let’s look at the corrected response:

     

    image32.jpeg

     

     

     

     

     

    Here we see the drivers are perfectly time aligned with the straight vertical “step” as both drivers response are arriving at your ears at the same time. Claim verified. There is just a hint of preringing at the beginning of the step response. I know it is not audible, as I have experimented in detail with greatly exaggerated preringing examples in my book, “Accurate Sound Reproduction using DSP.” Completely innocuous with the main goal of time alignment of drivers achieved.

     

    While phase and time alignment are not supposed to be audible, in my listening tests I have performed over the years, they are to my ears. Excess phase correction in the low frequency response is audible to my ears as is time alignment of drivers. Perhaps a future article on more discussion around this, but suffice to say, Dirac is doing its job in the time domain.

     

     

    Subjective Listening

     

    image33.jpg

     

     

    Note the dual Rythmik F18 subs were not used in the evaluation. I did have to check a couple of times as the amount of sub 30 Hz output from the Purifi’s PTT6.5” (sub)woofer is astonishing. However, I do intend to use the subs in Part 2 of using Dirac Live Bass Management in an upcoming review.

     

    The first thing I wanted to try was use my reference bass recording, Madonna’s, “Power of Goodbye” to test out Dirac’s partial correction up to 600 Hz in my room.

     

    image34.jpgThe reason I like using this recording for testing bass frequencies is because it has long sustained bass notes. This makes it easier to hear the room’s resonances. As expected, there are some bass notes that practically disappear and other bass notes that dominate the sound in my room.

     

    Enabling Dirac brings the level of bass notes to the same amplitude and the bass response is solid and powerful on each sustained note. It is still amazing to me as when I was working in recording studio control rooms, the only way to control room modes was the geometric and acoustic design of the control room. Aside from the room controlling reflections and acoustic treatments, the bass traps required to deal with very low frequencies at long wavelengths meant for a special acoustic design: a room within a room. The bass would pass through the shell of the inner room and be “trapped” in the outer shell so as not to pass back into the inner room. You can imagine the construction costs. Now at the flick of a digital switch, for truly a fraction of the cost, one can have the same smooth solid bass response at every frequency. What a time to be alive!

     

    image36.jpgAnother fav track I have for listening to bass evenness is Rebecca Pidgeon’s “Spanish Harlem”. A very nice Bob Katz recording and master of an acoustic bass guitar, in addition to Rebecca’s wonderful vocal rendition. Spanish Harlem, in the key of G, uses the classic 1, 4, 5 progression. Here are the frequencies of the fundamental notes of the bass:

     

    image35.png

     

    So without Dirac enabled, the 62 Hz to 82 Hz range of bass notes are really down in sound pressure level and the 110 Hz bass note literally sounds like a “coke bottle” resonance it is so loud and dominating in my room. 

     

    Enabling Dirac, brings back the proper level balance in the low end where each bass note is heard succinctly and even in level. When your ears starting tuning in, it really becomes apparent the differences. Having nice solid, even sounding bass arriving at your ears is a real treat. The reason why I say our ears need to tune in is because we are so used to listening to uneven bass, we are simply used to it and don’t give it a second thought. Remember JJ’s presentation. But once you hear how smooth it can be with your existing loudspeakers in your room, it is impossible to go back. Just the ability to smooth out the bass response is well worth the price of the software, aside from the additional benefits we are about to discuss.

     

    Let’s try the full range correction, but for the single seat. I find it a great boon being able to switch on and off DLP while listening in real time. There is no delay or interruption of sound or glitches of any sort, just a smooth and transparent switch. This makes is really easy to compare uncorrected to corrected or comparing different correction filters. This makes it easy to hear the differences instantly and then appreciate them over a period of time.

     

    image37.jpgSRV and Double Trouble’s, “Tin Pan Alley” is one of my favorite blues recordings. It was the first song and first take of the recording for the album that captured the feeling so well. When I was recording, I was always trying to get bands to “practice” before we recorded, but I was secretly recording as invariably the first take is sometimes the best take. Case in point this tune, wow!

     

    Great dynamic range that pulses with the emotion of the song. Great bass line that sounds nice and smooth with Dirac enabled. 

     

    But what really stood out for me is the Stevie’s strat sound with the Fender Leslie Vibrotone giving it that rotating stereo effect. It sounds so clear and precise with Dirac enabled, almost spooky hearing SRV fingers sliding up and down the strings with the Leslie effect enabled. Goosebumps territory.

     

    I switched over to the sofa, wide imaging corrections and re-listened to the above tracks, switching from single seat to wide imaging, at the same time moving around on my couch. I wanted to see if the 5 measurements versus 17 measurements were audible in some way to my ears. I found the changes to be subtle with the 17 measurements offering a bit more clarity in the overall sound, especially moving to edges of my couch area. I must say the phantom center image was perfect between the speakers, no wandering of the center image versus frequency. Excellent spatial imaging with a 3D depth of field I only hear from speakers that are time aligned with both channels closely matching in frequency response.

     

    Even with the 5 measurements, Dirac’s room correction algorithm appears to be quite robust and does not require an overabundance of measurements to get good even bass and smooth overall frequency response. While I enjoyed the benefits of partial correction, my preference is for the full range correction in two areas. One advantage was smoothing any midrange and high frequency “peaks” which with the Purifi SPK4 demo kit were small, but makes a difference not only in the smoothness of the sound, but also the vocals and instruments in the midrange “sat” in the overall mix where I thought they should be rather than sitting in front of the mix being a bit forward in sound from 2 kHz to 5 kHz region.

     

    Finally, I like the full range correction as I can dial in a preferred target of simply what I am used to listening to. Personally, I find most speakers sound too “bright” to my ears and confirmed by my measurements. While I have performed some speaker reviews here on Audiophile Style, I have had many pass through my listening room and all too bright for my tastes. The only one that had the preferred response out of the box is the Dutch and Dutch 8c. The KEF LS50 a very close 2nd, but just voiced a little too bright in the 2 kHz range, but Dirac would do a nice job of toning that back just a bit. Maybe we will see that in a future article.

     

     

    Conclusion

     

    Well folks, it is 2020 and if you are not using digital room correction, at least partial correction to smooth the bass frequencies in your listening room, you are missing out on a simple but audible optimization. At a retail price of US $349, coupled with a US $100 calibrated measurement mic like the UMIK-1, this is a “no brainer” upgrade to your existing system that is both measureable and audible in a positive way. For existing Dirac 1.x customers, it is a free upgrade. In fact, while I am biased towards advanced DSP room correction, I know of no other way to achieve this level of sonic improvement for dollars spent.

     

    Dirac 2 has made the process of applying room correction as simple as possible. Setup the mic, take measurements and click next, next, next, through the wizard and be listening to the smooth bass tones in the shortest time possible. They have made it difficult for anyone to mess up ☺ 

     

    It is interesting that Dirac chose the VST plugin route, but I can see that their market is not only audiophiles but also the semi/Pro audio market as virtually every Digital Audio Workstation (DAW) uses Steinberg’s VST software plugin architecture. Dirac also provides AU and AAX plugins for Mac users and for DAW’s that use that plugin architecture like ProTools. You can find Dirac Live on the largest marketplace for plugins at KVRAudio.

     

    I have all but given up going to audio shows due to the fact that below the (hotel) room’s transition frequency, the room dominates the bass response regardless of the type of or how expensive the loudspeakers used are. As Floyd Toole’s research shows that bass subjectively accounts for 30% of how we judge speakers sound quality. And “ANY loudspeaker can sound better after room EQ, so long as it competently addresses the bass frequencies - this is not a guarantee, but really is not difficult for at least the prime listener” In the case of Dirac, the sofa, wide imaging selection extends the bass sweet spot across at least three seats. 

     

    To answer the inevitable question, which sounds better, Acourate, Audiolense or Dirac, since I have evaluated all three. I will say they all improve the low frequency response of virtually any system in any room. One can model the “ideal” loudspeaker in an “ideal” room and then it is a matter of which room reflections and how many are positive for the sound quality and how many are not, are taken care of. There is significant scientific research on this topic and with quite a bit of agreement. Having said that, all three DSP room correction software employ different strategies to arrive at the “ideal” result. In the case of Dirac, if you are looking for the shortest path to the result, this is it. That’s also a limitation. Other than different target responses, everything else is pretty much closed off and taken care for you with no user adjustable controls. 

     

    I did not try Dirac 1.x, so have no way of comparing other than reading through the Dirac 1.x 42 page manual. I must say Dirac has greatly simplified the use of their room correction software for Dirac 2. I did not have to open a manual, I just started using it. I found the improvements made to my loudspeakers in my room by Dirac were audible to my ears and measurable in a positive way. Recommended.

     

     

     

     

     

    image38.jpegMitch “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 working full-time in multiple 24-track recording studios. Over 10 years, I recorded, mixed, and sometimes produced over 30 albums. I wrote a book on, “Accurate Sound Reproduction using DSP” and run an Accurate Sound Calibration service.




    User Feedback

    Recommended Comments



    Hi Mitch,

    Another great article. Thanks.

     

    I gave the new Dirac VST plug-in a try. I found some bugs and it does not yet support multi-channel home theater. I'm now on their beta-tester list and will try it again in the future.

     

    I like the Dirac VST plug-in because I can insert it in JRiver DSP with other audio processing.

     

    I'm now a big fan of the Multi-Sub Optimizer (MSO) software. It's the best ever for low frequency correction assuming you have multiple subwoofer channels. I posted my experience here: Optimization of a Subwoofer Line Array with MSO

     

    I hope to try Dirac with my MSO subwoofers in the future.

    Share this comment


    Link to comment
    Share on other sites

    Hi @hulkss good to hear from you! Thanks for your comments. I have used MSO as well. Excellent software!  Would be curious to see the corresponding phase response that goes with the frequency response plots in your MSO post...

    Hope you are well.

    Kind regards,

    Mitch

    Share this comment


    Link to comment
    Share on other sites

    2 hours ago, mitchco said:

    Hi @hulkss good to hear from you! Thanks for your comments. I have used MSO as well. Excellent software!  Would be curious to see the corresponding phase response that goes with the frequency response plots in your MSO post...

    Hope you are well.

    Kind regards,

    Mitch

     

    Sure, I'll add a couple more charts in a post to the thread I started.

    Share this comment


    Link to comment
    Share on other sites

    Great article Mitch!  I’d love to jump down this rabbit hole but I can’t wrap my head around how to implement it in my system:


    Sources:

    Mac running iTunes/Roon

    Cambridge Audio CXC

    Technics SL-1200GR

     

    All run into a pair of KEF LS50W.  I’m not sure if an additional piece of hardware like something from miniDSP is the way to go or if I just need to look at other options.

     

    Bill

     

    Share this comment


    Link to comment
    Share on other sites

    Bill, great to hear from you! I hope all is well.

    Dirac suggests: http://diracdocs.com/Roon_&_Dirac_Live_2.0.pdf to work with Roon. 

    I don't know how your other sources are hooked up, but the idea is that the input signal would be routed through Audio Hijack to Dirac Live Processor and then out to your LS50W.

    Don't have a Mac or LS50W, so could not say for say for sure...

    Hope that helps.

    Mitch

     

    Share this comment


    Link to comment
    Share on other sites

    19 hours ago, mitchco said:

    Bill, great to hear from you! I hope all is well.

    Dirac suggests: http://diracdocs.com/Roon_&_Dirac_Live_2.0.pdf to work with Roon. 

    I don't know how your other sources are hooked up, but the idea is that the input signal would be routed through Audio Hijack to Dirac Live Processor and then out to your LS50W.

    Don't have a Mac or LS50W, so could not say for say for sure...

    Hope that helps.

    Mitch

     

    Mitch,

     

    Thanks for the response!  In order to get all of my sources to benefit from Dirac would a hardware solution like the miniDSP DDRC-24: https://www.minidsp.com/products/dirac-series/ddrc-24  be better?  Signal path would be:  Sources->miniDSP & Dirac->KEF LS50W

     

    As an aside I made a first attempt at measuring using a UMIK-1/REW to come up with filters for Roon to use and it seemed to take away bass.  Is it safe to assume those are just room modes I've become accustomed to?  My room is small 12L x 10W x 9H so it's an acoustical challenge to say the least.

     

    Bill

    Share this comment


    Link to comment
    Share on other sites

    Bill,

     

    The miniDSP DDRC-24 should work no doubt. I have never used one, but I know it has a limited amount of filter taps as compared to the processing on a PC. Less taps generally means less low frequency resolution. However, I can't speak about how Dirac 2 is implemented in the miniDSP unit and may or may not be an issue. Perhaps @flak can comment or a post on the miniDSP forum or maybe another member has experience with this combo and can comment...

     

    Wrt 2nd question, hard to say without seeing any before and after graphs. It could be room modes, but could also be the target is a bit too bright that makes the bass sound thin. If you could post some charts that would assist. Also, feel free to add a bit of bass boost like in this target from the miniDSP Dirac 2 user manual: https://www.minidsp.com/images/documents/miniDSP Dirac Live 2.0 User Manual.pdf  See page 36.

     

    Kind regards,

    Mitch 

    Share this comment


    Link to comment
    Share on other sites

    4 hours ago, mitchco said:

    Wrt 2nd question, hard to say without seeing any before and after graphs. It could be room modes, but could also be the target is a bit too bright that makes the bass sound thin. If you could post some charts that would assist.

     

     

    Mitch,

    I’ll have to try and attach some charts but I was thinking back to some the Hartman white papers I’ve read and if Dirac is trying to fill some of the dips in the bass and those dips are due to room modes isn’t any boost it applies going to make that dip worse?  Shouldn’t it instead focus on bringing down the peaks?


    What charts would help and what smoothing/windows would you want to see them in?  Would you want FR using REW’s sweep or using pink noise?  What time window would you want to see on a waterfall plot?

    Share this comment


    Link to comment
    Share on other sites

    Hi Bill, Dirac I believe is cut only filters, so if you lower the peaks to be even with the majority and not worry about the narrow dips. Narrows dips we don't hear.

     

    Sure. just use REW's default settings and sweep 10 Hz to 24 kHz using a 48 kHz sample rate of the left and right speakers at the listening position. Just PM me the .mdat or send to [email protected] and once we sort it out, you can post the results here if you wish.

     

    Mitch

     

     

    Share this comment


    Link to comment
    Share on other sites

    This sounds like something most “audiophiles” would want, however, am I correct in understanding that you need a computer based audio player to implement it? How would you use it if you just use a discrete streaming device like an Aries mini?

    Also, same question if you have a turntable?

    Share this comment


    Link to comment
    Share on other sites

    6 hours ago, Iansr said:

    This sounds like something most “audiophiles” would want, however, am I correct in understanding that you need a computer based audio player to implement it?

    Yup although some hardware players are now implementing it.  See NAD M10 https://www.stereophile.com/content/nad-masters-series-m10-streaming-integrated-amplifier

    6 hours ago, Iansr said:

    How would you use it if you just use a discrete streaming device like an Aries mini?

    Pipe it through such a player or, as I do, using Jriver on my PC-based audio player, pipe it through JRiver.

    6 hours ago, Iansr said:

    Also, same question if you have a turntable?

    The M10 has line-level inputs which would support the output of your phono preamp and, AFAIK, digitize the output from the TT.

    Share this comment


    Link to comment
    Share on other sites

    2 hours ago, Kal Rubinson said:

    Yup although some hardware players are now implementing it.  See NAD M10 https://www.stereophile.com/content/nad-masters-series-m10-streaming-integrated-amplifier

    Pipe it through such a player or, as I do, using Jriver on my PC-based audio player, pipe it through JRiver.

    The M10 has line-level inputs which would support the output of your phono preamp and, AFAIK, digitize the output from the TT.

     

    Similarly Kal, miniDSP offers hardware capable of implementing Dirac Live's filters to analog and digital sources.  I'm considering the DDRC-24 which is very reasonable considering you get a full Dirac Live license with it.

     

    Just sent my REW measurements off to @mitchco to get his thoughts as my small 12x10x9 room suffers from room modes which I've attempted to alleviate through a combination of a near field setup and DIY acoustic treatments.

     

    This article has me very interested in getting as much of the room out of my sound as possible to hear that level of detail I experience with my headphone setup.

     

    Bill 

     

    P.S. Very sorry to hear about Art.  His voice will be missed in this pursuit of ours.

    Share this comment


    Link to comment
    Share on other sites

    5 hours ago, wgb113 said:

    Similarly Kal, miniDSP offers hardware capable of implementing Dirac Live's filters to analog and digital sources.  I'm considering the DDRC-24 which is very reasonable considering you get a full Dirac Live license with it.

    I have great respect for miniDSP products but they often have i/o or processing constraints that deter me from doing more than experimenting with them.  

    Share this comment


    Link to comment
    Share on other sites

    I agree with @Kal Rubinson. My concern for any hardware DSP is the limited processing power available. Typically they are limited to 4096 FIR filter taps, which is not enough for low frequency control. For example, the miniDSP 2x4 HD in the datasheet talks about 4096 taps. But this is the total number. For 4 channels you have 1024 taps each channel available. The frequency resolution of a 1024 taps filter @ 48 kHz samplerate is 48000/1024 = 46.875 Hz. So below 100 Hz there are just 2 frequency bins at 46.875 and 93.75 Hz. This clearly means that you have no control over the lower frequency range.

     

    I don't know how Dirac is implemented in these h/w devices, but on a PC you typically use 65,536 taps per channel which gives full low frequency control. Remember the primary purpose of DRC is to correct the low frequencies below Schroeder. So my review of Dirac 2 is only in the context of using Dirac on a PC. I make no claims on how it works/effectiveness with miniDSP gear or any other hardware.

    Share this comment


    Link to comment
    Share on other sites

    Hi,

     

    great article with really useful information, Thank you for this!

     

    i've read all comments as well, but i have a question:

     

    should one remove, let say carpet of (7ft by 8ft) between the speakers and listening position as well as sofa before making measurements, or not.  what would be the best excercise?

     

    Pictures on the wall, bookshelfs (with books) and maybe some room treatment stuff at the corners should stay where they are when making measurements, right?

     

    thanks in advance,

    Tp

    Share this comment


    Link to comment
    Share on other sites

    I enjoyed your article and want to try it myself.  What is the name of the DSP VST file?  I cant seem to find it anywhere?

    Share this comment


    Link to comment
    Share on other sites

    Hi @Tp no need to remove the carpet or other objects on walls, etc. What we are trying to avoid is anything in the direct path between the speakers and microphone like plants coffee table, etc. Also, anything around the microphone can cause odd sounding corrections, hence moving the chair or sofa from around the mic.

     

    Hi @mlknez Do you mean DiracLiveProcessor.dll? On my Win10 computer it is located at: C:\Program Files\Common Files\VST2

    Share this comment


    Link to comment
    Share on other sites

     

    Firstly I’d like to say thank you Mitch for this excellent article (which I’ve only got around to looking at now with “benefit” of UK Covid lockdown!) 

     
    I do have a couple of questions further to both the article and following comments.
     
    Firstly the measuring bit looks fairly straightforward - hopefully a case of just getting used to the interface and software.
     
    Here, and further to your answer to Ted_b question; I get that moving tables, plants etc. whilst measuring makes sense, but I humbly suggest that (especially in smaller rooms - see final question below) the sofa is effectively part of the room and might best be left in situ; moving it could have quite a large impact?
     
    [As an aside I believe the acoustics of auditoria are modelled even to account for the absorption of people in seats - or standing!]
     
    What then is unclear to me (also having looked at Dirac’s website) is, having chosen target curve and generated filter, how to implement; in my case using Mac and Audirvana. I don’t see any mention of AU plugin but believe this would be the default for Dirac and A+?
     
    But then, screen shots seem to show the filters as for a specific sample rate and bit depth (is this correct?) A+ will happily playback whatever rate file is chosen (without SRC unless chosen) so would I need separate filters per SR/depth or would one one filter “adapt” to the rate chosen? Or something else...
     
    Back to your own adventure, once you have generated and implemented filters, I am slightly confused why you choose to use REW to verify? Is it not possible simply to measure again (Dirac) after you’ve applied your target eq so you’re looking at the changes in the same “format”? Or if this is possible, why then would you choose to use REW?
     
    Finally, a further question on a practical level. A friend of mine has recently moved house and has a small room (similar in size to that of Bill - wgb113) which he intends to use solely for audio, and into which he has shoehorned his existing system. At the moment his B&W floorstanders are almost wedged into corners, necessitating port bungs in place. Even so, really not sounding all that great (don’t worry, the chances of him reading this are minuscule!) and seems like the perfect candidate for testing. Would it make more sense to remove bungs, allowing the speakers to breathe and letting Dirac take care of everything? Or leave them in so less work for eq?
     
    Best wishes all. And Mitch, I for one would greatly enjoy a follow up article if you do have time.
     
    Andy

    Share this comment


    Link to comment
    Share on other sites

    On 4/17/2020 at 12:35 AM, mitchco said:

    I don't know how Dirac is implemented in these h/w devices, but on a PC you typically use 65,536 taps per channel which gives full low frequency control. Remember the primary purpose of DRC is to correct the low frequencies below Schroeder. So my review of Dirac 2 is only in the context of using Dirac on a PC. I make no claims on how it works/effectiveness with miniDSP gear or any other hardware.

     

    Hi @mitchco (maybe also @flak)

     

    Does Dirac Live 2.0 (or later) + Dirac Live Plugin on PC use 65k taps ?

     

    Has this been confirmed specifically for Dirac?

     

    I know Audiolense and Acourate do (and even higher number of taps)

     

    Share this comment


    Link to comment
    Share on other sites

    A great post as always Mitch.

     

    I have been using Dirac with a MiniDSP DDRC-22D for years now with my B&W 803D speakers and with Dirac V3.0 now out I finally decided to update from V2.3 and run my measurements again using your recommendation of removing furniture. In the past I have removed objects between the LP and speakers (if any) but have left the L shaped leather couch in place. I would describe my room as tending to be more on the dead side with carpeted floors and absorption on the wall approx. 1.5m behind the listening position. I guess the couch would have been a major point of reflections?

     

    One slight bugbear in the past was that whilst Dirac did a great job as far as frequency correction was concerned, the sound always seemed to lack the "openness" of the uncorrected playback and that is now back without losing any imaging focus. Historically I never liked recommended target curves of more than 6dB maximum tilt, as they invariably sounded dull. I am listening now with a straight 8dB tilt from 20Hz to 20kHz and it sounds great.

     

    I don't know if the improvement I'm hearing is due to the new software or the couch removal during measurement or a combination of both....and I really don't care 😀  Interestingly my first impression with the new measurements apart from the increased openness is that everything sounds just a touch slower and coherent. 

    Share this comment


    Link to comment
    Share on other sites

    On 4/16/2020 at 6:35 PM, mitchco said:

    I don't know how Dirac is implemented in these h/w devices, but on a PC you typically use 65,536 taps per channel which gives full low frequency control. Remember the primary purpose of DRC is to correct the low frequencies below Schroeder. So my review of Dirac 2 is only in the context of using Dirac on a PC


    Could one (Dirac) limit the frequency band below a certain frequency ? Like you concentrate the frequency taps in restricted area, and then achieve a more narrow frequency bin. 
     

    I think no, cause this is about frequency resolution, so you can’t shrink somehow. 
     

    Also some HW manufacturers is using 96 kHz, and then you actually need double the number  of taps in in order to achieve same resolution. Correct ?

     

    What would be an acceptable frequency bin or resolution ?


    I probably don’t ask the correct question, but I hope you somehow understand what I’m thinking about 😀

    Like adjusting more taps to lover frequencies. And less to the higher ones. Don’t think this make sense. 


    96000/65536 = 1.46 Hz. Would you like to see a lower number, or what is acceptable In your opinion ?


    I notice Dirac now have a basic version below 500 Hz for 3.x, and then you pay extra in order to get access to adjust frequency above that. 
    I still would think Dirac requires some sort of available processing power from the HW manufacturers. Unless Dirac supplies the board to the manufacturers. Do you happen to know anything about how this works ?

     

     

    Share this comment


    Link to comment
    Share on other sites

    Recently hired Mitch to optimize my system.  Intelligently applied DSP is the most important change to my system since I started with this hobby in the late 1970s while studying electrical engineering in college.   Mitch ended up using Audiolense.   We’re on the second set of filters, trying to dial in the “perfect” timbre for my taste.  Over the years like most audiophiles I made many changes, most resulted in incremental Improvement, but this is the first true game changer in how my system sounds.   Real science, measurable results, very talented, easy to work with remotely.   What else can I say.  

    Share this comment


    Link to comment
    Share on other sites




    Create an account or sign in to comment

    You need to be a member in order to leave a comment

    Create an account

    Sign up for a new account in our community. It's easy!

    Register a new account

    Sign in

    Already have an account? Sign in here.

    Sign In Now




×
×
  • Create New...