I had an opportunity to listen to the Audeze LCD-4 and LCD-5 headphones extensively over a two-week period while developing high resolution FIR equalization filters for them. While I have reviewed the LCD-4z, I have not had a chance to listen to both Audeze flagship headphones until now. A real treat!
Up front, I will say I am surprised at how different these headphones sound compared to each other. I would have thought that the next version of a company’s flagship headphone would be a refined version of the previous version. As it turns out, these headphones are pretty much opposites, especially in the upper midrange.
Listening Methodology Explained
As an ex 10-year pro recording/mixing engineer, I want to share with you what I listen for when evaluating headphones (or speakers for that matter). I think it is important to provide listening context before delving into the sonic comparisons so people can understand where I am coming from.
For me, it’s all about neutral sound. When I say neutral, I mean no one frequency or range of frequencies stand out compared to other frequencies. Technically, there should be a ruler flat frequency response, just like we expect from our digital music players, DACs, preamps, amps and loudspeakers (with no room). What’s on the recording is arriving at my ears with no frequency response unevenness and within the same tight tolerance as our music players, DACs and amplification. Almost all headphones I have listened to (and measured) are not neutral and some far from it.
Not only does the frequency response affect the tonal characteristic, it also affects the balance of the mix from a front to back “depth of field” perspective. It also affects the top and bottom of the soundstage from a height perspective.
For example, if the headphone has a midrange peak, not only are the lead vocals/instruments “colored” in tonal quality, they also appear too far forward in the mix, sometimes to the point of sounding disembodied from the rest of the mix, and adding height to the soundstage. This is not neutral/accurate sound reproduction.
And, it is not just “one” midrange peak, it may be followed with an upper midrange dip in frequency response for example. The peak in the midrange pushes the vocal (and instruments) too far forward in that range while the upper midrange is pushed too far back in the mix. This gives the vocals a “shouty” or “hollow” quality, but lack upper midrange definition. Therefore, the vocals are not well articulated and makes it harder to understand the lyrics, even though the vocals sound loud in the mix, which is a real dichotomy. As mentioned, not only is the depth of field affected, so is the perceived height. Too far forward in the mix increases the height in the soundstage and too far back in the mix decreases the height, to the point where the sound appears to be coming from below the horizontal soundstage. It is frequency dependent.
It could be the reverse as well. For example, a dip in the frequency response in the midrange will push the vocals too far back in the mix, lacking “body” to the voice. With a peak in the upper midrange, where our ears are the most sensitive, it can quickly go from sounding too forward to harsh, like a tinny transistor radio for those old enough to remember what that sounds like. So not only frequency coloration, but depth of field and vertical height of the soundstage are also affected by an uneven frequency response.
Another area I listen for is the bass response of the headphone. Not only subjective terms like “euphonic” or “tight” sounding, but the lack of lower bass frequencies. This is a fault of virtually every headphone I have listened to and measured. The Abyss 1266 TC’s have extended low frequency response to 20 Hz “out of the box,” albeit it also has uneven midrange issues. I listen for the “weight” of the deep bass of the headphone to hear if it is balanced with the rest of the mix. Too much deep bass is as annoying as too little.
As an aside, I don’t count headphones that have been designed with the Harman headphone target response or have been eq’d to the Harman headphone target. While it may be a popular preference for a specific demographic, it does not sound “neutral” to my ears. To my ears, there is too much low bass with the upper bass missing, accompanied by a big amplitude peak in the 2.5 kHz region, again, in our ears most sensitive frequency range.
“Boom and screech” is what I hear with this type of sonic profile. I am exaggerating a bit, but if this is what consumers prefer, it is not for me. I am a long time AES member and have read the research, including having Sean Olive review my NAD HP50 headphone article plus the science. I own a couple of headphones with the Harman target “built-in” and have tried over a dozen other headphones using Harman headphone target “PEQ” profiles, from a variety of sites, and none sound neutral to my ears. For one thing, the measurements for the headphones being used for eq purposes have not been properly prepared for that purpose. The measurement data requires a frequency dependent window (FDW) to be applied that better represents what our ears hear from a frequency resolution (i.e., spectral envelope) perspective. It is no wonder to me why none of these “PEQ” profiles sound neutral to my ears.
I also find the Harman target profile to pretty much destroy the depth of field of the soundstage making it sound very 2D, i.e., lacking depth. There is too much deep bass with the middle and upper bass “suck out” compared to the rest of the mix. There is an unnatural “edge” to the sound that brings the upper midrange too far forward (and up in height) in the mix. Maybe this “boom-sis” boombox type of sound is a preference type for some, but I would think audiophiles are looking for a more neutral/accurate sound reproduction, approaching the same specs as their digital music players, DAC’s, and amps.
High frequency “s’s” on vocals is another area of sonic evaluation as it is difficult to reproduce. It is also difficult to record. There are pro audio “De esser” software tools designed to keep the s level under control. In most cases I find the recorded “s” sound on vocals to be overly sibilant (i.e., too much) sometimes on both the recording and headphone playback. Ouch! One can emulate it by just speaking or singing and listening to the “s” sound your voice makes and then compare that to listening to the s sound wearing headphones while listening to a tune. Check one, check one, sibilance, sibilance. Sound the same? Too much? Or is it non-existent? Or is the s sounding like shhhh?
People should know that “straightening out” (i.e., equalizing flat) the frequency response of a headphone not only makes the headphone sound neutral from a tonal perspective, but it straightens out the soundstage as well. I find the latter topic rarely discussed as folks may not be aware that pro audio mixers spend a great deal of time setting fader levels to get the “balance” to sound like the band or orchestra or personality of what is being recorded. There is definitely a depth of field, or front to back placement of vocals, instruments, delays, reverbs, etc., to mix or paint the audio picture that has a sense of depth. Therefore, uneven frequency response also distorts the depth of field (and height) of the mix and is another form of audible distortion.
In this context of sound evaluation, I am listening more to the sound of the headphone than I am to the music. I also know the tunes very well having heard them many times over many years. It provides me with a good frame of reference. Some of the tunes many of us have heard before and understand how the mix is supposed to sound. If one has a calibrated neutral system, whether loudspeakers or headphones, then it is fairly easy to hear the deviations from neutral.
With the listening context set, let’s get into subjective listening.
Tracy Chapman - Fast Car https://open.qobuz.com/track/2781545
The first thing I noticed was the lack of deep bass followed by the bright upper midrange edge on Tracy’s vocal. I have heard this track hundreds of times and the deep bass should anchor this song and give the feeling of weight and warmth. Tracy’s vocal is what I would already call punchy as it is not strangled to death with compression.
During the chorus her voice swells nicely, but the upper midrange pushes her voice too far forward in the mix. Same goes for the snare drum. The upper midrange “crack” of the snare is too loud relative to the lower midrange punch of the snare.
The mix is dominated by the upper midrange or her voice and then the crack of the snare drum, both masking the lower tones of Tracey’s voice and the snare drum “punch.” Not only pushed too far forward in the mix, but adding height as well. Turning up the volume is a bit too much upper midrange for my ears to handle. The s’s on Tracy’s voice are well recorded, but sound a bit unnatural as the upper midrange peak exacerbates the s’s to be lower in tone (almost shh sounding). Doesn’t sound quite right.
The deep bass, while down in level, sounds a bit more balanced with the rest of the mix because these headphones do not sound as bright as the LCD-5. A midrange peak pushes Tracy’s vocal a bit too forward. But the upper midrange is down a bit in level so Tracy’s voice doesn’t quite have the articulation that should be there. The acoustic guitars sound a bit muted. The s’s sound like they should. But the vocal does sound a bit strange as the midrange is pushed forward, so part of her voice has height whereas the upper midrange of her voice sounds below the band in the soundstage from a height perspective. That makes Tracey’s vocal sounds a bit disembodied from the mix. However, it is a smoother sound compared to the LCD-5.
Social Distortion - Don’t Drag Me Down https://open.qobuz.com/track/40190
Maybe an unusual pick, but for a guitar player, this song rips!
There is a thumping kick drum that pounds pretty good, but the bottom octave is down in level and missing the weight behind the thump. Given the layered ripping guitars, with the LCD-5 being quite bright in that range, really pushes the guitar sound forward and up in height in the mix. Since the guitars are already mixed up to maximum, the guitars now dominate the sound with Mike Ness’s vocal accentuated over top. The guitars and vocals sound quite disembodied from the drums and bass and starts to sound a bit grating over time.
Gives a completely different presentation. The thump of the bass drum and running bass line is louder in the mix because these phones are not as bright sounding as the LCD-5. The midrange bump pushes the Mike’s vocals and guitars bit forward in the mix, but more of a midrange tone. With a dip in the upper midrange response, the guitars are sitting a bit too far back in the mix and missing Social Distortions signature ripping crunchy edgy guitar sound. With that midrange peak makes the depth of field sound a bit opaque. Hard to get a sense of how well layered this mix actually is. From a soundstage perspective, the guitars and vocals sound lower in vertical position than the rest of the mix. This is opposite of the LCD-5.
Massive Attack – Angel https://open.qobuz.com/track/1066043
Trip hop tune with pounding bass/drum line coupled with a wall of guitars. Great depth and beat to this tune.
The bass is super tight and concussive with the drum fills giving a cool stereo effect. Unbelievably transient sounding presentation. As the guitars come on, the brightness of the headphone really come into play pushing them forward past the drum/bass line and not equal in level as it should be.
The phaser sounding hi hat (and crashing cymbals) becomes too prominent in the mix with the upper midrange peak of these headphones. This domination loses the pounding/driving effect of the tune.
Again, a very different presentation. Because the upper midrange is pulled back versus the LCD-5, the hi-hat, cymbals and wall of guitars are a bit too laid back. The midrange peak gives a bit of a hollow sound, exacerbated by the upper midrange dip. But the overall response is more neutral than the LCD-5, which when switching headphones as quickly as I could, with rough level matching, really sounds like two completely different headphones, even though both are flagship headphones from the same company.
I listened to dozens of tunes daily, of all different genres, over a couple of weeks. In all cases, there was no mistaking which headphone I was wearing, aside from the weight and clamping differences. What I subjectively heard in the three tunes above was consistently heard across all music genres.
Let’s get into the measurements to see if what I was subjectively hearing matches the objective frequency response characteristics of each headphone.
This is my measurement setup. For listening, I use the headphone output of my Lynx Hilo. I also use a custom designed Class A headphone amplifier that I built many years ago. It has a hefty 160VA toroidal transformer used in a wide-band +- 15 volt regulated power supply with 3-amp regulators per channel. In addition to being low distortion, it has plenty of power to drive even the most inefficient headphones cleanly to well beyond levels of what my ears can take. While the Hilo can drive both headphones louder than I normally listen, the LCD-5 is a (much) more efficient phone than the LCD-4.
I spent quite a bit of time listening to each headphone before getting to the measurements. I wanted to not only get a sense of their sonic characteristics, but also fiddling with fit of the headphone and moving it around my head to find the sweet spot. For both phones, this did not take long.
The LCD-4 being heavier and bigger, still feels comfortable on my head and I soon forgot about them. The LCD-5 is much lighter, but with a pretty good “clamp” pressure on my head, but also soon forgot about them. I must admit, I am not that fussy. As long as I can get a good fit, that pretty much does it for me. 90% of my interest is in the sonic characteristic of the headphone.
My routine is to take the headphone off and put it back on for each measurement. That way I can see how consistent the measurements are and identify any inconsistencies. I took 20 to 30 measurements per headphone.
As explained in the Abyss AB1266 TC article, typical headphone measurements regardless of measurement technique/apparatus used, starts losing accuracy above 5 or 6 kHz due the HRTF and internal reflections where the short wavelengths cause huge peaks and dips in frequency response. Therefore, the measurement data above 6 kHz does not accurately represent the measurement of the headphone. This is reflected in the measurement charts.
Here are both the LCD-4 and LCD-5 left and right frequency response overlaid in an REW frequency response chart:
LCD-4, green and red trace. LCD-5, blue and purple. The black horizontal marker should be considered a neutral frequency response target, i.e., flat. Both headphones measure pretty much identical from 20 Hz to about 2 kHz and then there is a (big) point of departure. The LCD-5 is about +6 dB in the 3 to 5.5 kHz range while the LCD-4 is -3 ish dB at 3 to 4 kHz.
To put that into context, there is almost a 10 dB SPL difference between the two headphones in the 3.5 to 4 kHz region (i.e., the upper midrange). Our ears/brain perceives a 10 dB SPL difference as being twice as loud or quiet depending which viewpoint is chosen. I.e., the LCD-5 sounds almost twice as loud in this frequency range then the LCD-4. That is huge audible difference, as noted in my subjective listening section.
Let’s look at each headphone’s frequency response individually, again using the black horizontal line as neutral sound reproduction, to zero in on the frequency deviations from neutral.
The LCD-4 starts to roll-off in the bass just past 100 Hz and is -4 dB down at 20 Hz. Nice and smooth lower midrange response that starts to rise past 300 Hz with the main peak between 1 and 2 kHz. This is responsible for the midrange peak sound I heard while listening. Then there is a bit of a dip in response from 2.5 kHz to 4.5 kHz. This is right in the ears most sensitive range, so it is no surprise that folks find these headphones a bit laid back as they are in this frequency range. There is about a 2 dB peak at 5 kHz which adds a little presence to the sound, but not annoyingly so.
The LCD-5 starts to roll off around 70 Hz and similarly to the LCD-4, about -4 dB at 20 Hz. The LCD-5 stays flatter than the LCD-4 until about 500 Hz where the response starts to rise with a broader peak from about 800 Hz to 1.5 kHz. But as one can see a sharp rise with +6 dB upper midrange energy from 3 to 6 kHz. Not only quite the departure from the LCD-4, but also brighter even than the Raal Requiste SR1a. I must admit the LCD-5 are the brightest headphones I have listened to and measured. Seems a bit unusual for a flagship headphone, especially given how relatively neutral the LCD-4’s sound and measure. Fortunately, headphone equalization is a thing. Let’s see what can be accomplished.
High resolution FIR filtering for headphone equalization
Eq’ing headphones is popular and there are many eq programs and “PEQ” profiles freely available from many sites, including Audeze built-in eq profiles that are available via Roon. All sorts of discussion on “dummy head” measurements, compensation curves, eq targets, etc. However, as mentioned previously, having listened to several of these eq profiles on various headphones, including the LCD-4, none of them sound neutral to my ears.
I have also measured several of these PEQ profiles using loopback. One wonders what measurement data were used in the development of these PEQ’s. As alluded to earlier, the raw measurement data needs to be processed with a Frequency Dependent Window (FDW) and I see no evidence of any of the measurement data processed in this way.
Not only is there considerable variability between eq profiles, the actual eq does not have enough frequency resolution to begin to bring a headphones frequency response curve to flat with any degree of accuracy or precision.
A “5 band” parametric eq (PEQ) can only apply a gross approximation compared to a fully linearized frequency response. A “31 band” graphic equalizer increases the frequency resolution a bit more, but neither of these equalizers can match high resolution Finite Impulse Response (FIR) eq filters.
FIR filters typically have “32,768 bands” of graphical eq or greater. That is 1000 times greater frequency resolution than a 31-band graphic equalizer. This ensures that the frequency resolution is high enough to ultimately smooth the frequency response with a high degree of accuracy and precision.
Here we can achieve the same type of frequency response resolution/tolerance as one’s digital software audio player, DAC, and headphone amplifier: ruler flat +- 0.25 dB frequency response tolerance. If you’re going to eq, why not use the best eq possible?
For example, here are both the LCD-4 and LCD-5 frequency responses before and after FIR filtering.
I have offset the curves so we can be seen how flat the frequency response is.
For the LCD-4, the bass is brought up to 20 Hz with a -3 dB point at 17 Hz. Deep bass extension. The +2 dB from 1 to 1.5 kHz peak is brought down to flat and the -3dB dip in the 2.5 to 3.5 kHz range has been brought up to flat. Now the lead vocals/instruments midrange do not sound colored or forward and the upper midrange does not sound dark and sounds more articulate. The deep bass now gives the headphone weight (no pun intended). The soundstage now has a real 3D depth of field. The balance of some songs I have mixed in the past sound just like when I first mixed them, a spooky kind of déjà vu feeling.
For the LCD-5, truly can be characterized as a “night and day” difference. The increased low bass gives added weight to the sound and the taming of the +2 dB 1 to 1.5 kHz peak helps. But the huge difference is in the upper midrange. Going from barely listenable to, wow, these now sound really clear and full!
So with both headphones eq’d flat, do they sound the same? I am thinking of a way for folks to hear this. In the meantime, if I was to make some sort of analogy, I would say the LCD-4 sounds more like a tube amp is hooked up to it versus an ultra-fast solid-state amp driving the LCD-5’s. To my ears, the LCD-4 has a larger sound stage overall than the LCD-5, but that is just my impression, with no objective data to support it.
What I can say is that both headphones sound virtually perfect with the FIR filters. It is a matter of personal preference whether the bigger, rounder sound of the LCD-4 or the superfast/tight sound of the LCD-5. Both are highly detailed, but presented in a much different way.
What I am not covering here is the DSP software I use that is designed specifically for correcting the frequency and phase response of headphones. These are new DSP algorithms designed to extract the frequency/phase response of the headphone and apply a custom frequency dependent window that closely represents what our ears hear from a frequency resolution perspective.
I must say I was surprised at how bright the LCD-5’s sound overall and when compared to the LCD-4. Really a yin and yang experience. Both headphones lack deep bass response. The midrange bump in both headphones tend to push the lead vocals/instruments a bit too forward, but after that comes the big departure.
The LCD-4 is laid back with the upper midrange vocals and lead instruments sound a bit on the dark side or not as articulate. From a soundstage perspective, the vocals/lead instruments drop in height and are pushed back into the mix.
From a tonal perspective, the LCD-5 sounds similar in the bass and lower midrange, including the +2dB rise in the 1 to 1.5 kHz region. Then the LCD-5 heads North of neutral with a big rise in the upper midrange.
Without applying eq, the LCD-4 is more tolerable than the LCD-5. I really like the speed and clarity of the LCD-5, but they are simply too bright out of the box for my ears.
With both headphones eq’d flat, the tonal response is very similar, but the way it is delivered is quite different. The LCD-4 has a big soundstage with a rounded bass quality and a very detailed midrange top end, but never harsh, no matter what the volume. The LCD-5 have an incredibly fast transient response with rock solid bass, maybe even a bit on the dry side, but also very detailed midrange and top end. If you are a tube aficionado, you may like the LCD-4 more. If you are a solid-state amp fan, you may like the LCD-5 more.
Either way, I hope you are enjoying the music!
For those interested in acquiring these high-resolution FIR filters, contact [email protected]
The LCD-5 neutral FilterSet is US $200 and the LCD-4 is $175. Requires a convolver.
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 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.