Saturday, 10 April 2021

REVIEW / MEASUREMENTS: AKG K371 (closed-back, over-ear, dynamic driver) - An affordable, modern, standard studio headphone. (And on using the miniDSP [H]EARS rig.)

Hey guys and gals, the post here today will be on the longish side since this is the first time I'm writing a headphone review with some measurements. Best to take the opportunity to discuss the testing itself as well as product impressions.

As you can see above, today we'll be considering the AKG K371, an inexpensive (~US$120) "studio" 50mm dynamic headphone that's easy to drive, with a sound signature based on Harman research into headphone tonality - see this nice presentation from 2017 containing a review of the various research over the years.

I bought these headphones from the usual retail channels.

First let me show you some pictures of these particular headphones, and later, we'll get into measurements with the inexpensive miniDSP (H)EARS test fixture.

In the image of the opened AKG K371 above, we see that it comes as a rather complete package. You can detach the headphone cable from the drivers easily with its small locking DIN-type connector that comes out of the left earcup (the cable connector has a spring push-button to unlock):

The headphones come with 3 separate cables - a short straight 1.2m length, a longer straight 3m, and a convenient 3m coiled wire. There's a canvas carry case. Small pamphlet/manual, and 3.5mm to 1/4" adaptor. Not a bad set of accessories for something that's just a bit over 100 bucks!

The memory foam earpads can be removed and third party pads can be used (like the Brainwavz Ovals) if you prefer a change.

I found the stock soft synthetic "protein leather" pads quite comfortable even after an hour of listening and they seal well for bass frequencies. As I write this, it's only early springtime so the temperature isn't warm yet. Being a closed headphone, I imagine that in hot weather, the ears might get a bit sweaty. These are not large headphones by any means. I probably have average-sized ears and they were comfortable; those with large ears might find the cups a bit tight. Likewise, the clamping force was OK for me and loosened nicely over a couple weeks of use but I think could be tight for those with a big cranium.

The sizing mechanism slides up and down with moderate resistance. I did not experience any problems although I have heard folks complain that it could get stuck. The hinge area also is able to fold backwards with a clicky ratcheting motion 180° to a compact size when transporting.

Construction is mostly plastic although the connector from earcup to headband is metal. The headband is made of a matte-finished plastic and has a rubberized texture. Headband is padded and feels reasonably comfortable and I didn't feel any bad pressure points on the top of the head in extended use.

The right/left labeling is simply a small "L" and "R" on the inner side of the headband like this which could be hard to spot:

I might just have to stick a piece of red/blue tape on the outside (color coded like with Sony studio headphones) to remind me of which side is which! Overall, I find the headphone quite sturdy and should not break easily assuming it's not abused.

I. Headphone Measurements using the miniDSP (H)EARS...

Let's talk about measuring headphones for a bit and introduce you to the miniDSP (H)EARS. Most of the time the "H" is dropped from this measurement jig although on the front of the unit, we see the label Headphone & Earphone Audio Response System which I think is a nice acronym:

Front DIP switches used to change microphone gain. I'll just keep it at the default +18dB.

For those with miniDSP (H)EARS, here are other useful links discussing the measurement jig and how to capture results: miniDSP EARS manualHeadphonesty guide.

As a headphone measurement jig, it looks like a metal headphone stand with two silicone "cheek" side slabs with "ears" sprouting out - a type of "ear-and-cheek simulator". This is not standard IEC 60268-7 coupling hardware but at least allows headphones of various sizes and shapes to be applied as would a human listener. The pinnae are a bit stiffer than my ears on account of the silicone material's lower compliance, and while I presume the ear shape including the molding for the concha (that deep part just before entry into the external auditory canal) must have been based on some "human" model, the helicis crus is a bit small and recessed while the tragus a bit protuberant compared to most ears I think.

If you have one of these, be mindful of the 4 screws holding down the synthetic ear/head in place; these areas of depression can make it more difficult to create a good seal. I know some folks have removed the screws and I've also had some success with using thin foam tape (like the soft skin-hugging tape used in Band-Aids) to smooth out the irregularities as well. Generally, with a bit of care, I have not found this to be a big problem. The higher density and lower compliance of the pinnae will also make it a bit more difficult to measure on-ear (supraaural) headphones with a proper seal however, leading to inaccurate poor bass response. Getting good, consistent on-ear headphone measurements has been by far the most difficult task I've tried thus far (hint: use some large, thick rubber bands around supraaural headphones to clamp down on the pinnae helps).

Another area of concern is that the ear canal is just a straight tube leading to the measurement microphone inside. The canal is also rather short. Human external auditory canals are more complicated than this which would affect the resonance characteristics leading to the tympanic membrane. Be careful with IEMs, especially those with longer tips that could jam against the microphone. Regardless, a compensation curve will need to be applied to correct for the resonances (miniDSP includes their IEM Diffuse-Field "IDF" compensation which we'll look at another time).

I can certainly understand Brent Butterworth's concerns and article about the miniDSP EARS back in 2018. Nonetheless, I do believe the inexpensive US$199 price has indeed brought "headphone measurement capability to the masses" and opened up the opportunity for the hobbyist audiophile to explore relative sound quality of headphones without aiming for absolute "industry standard" results.

It's interesting when we look around at headphone measurements online and notice just how variable the frequency response graphs presented are among sites like In-Ear Fidelity/Crinacle, InnerFidelity, DIY-Audio-Heaven, Rtings, Reference Audio Analyzer, and Audio Science Review. Unlike speakers where we can generally agree on measurements based on how a speaker would perform anechoically, headphone performance by definition is affected by the interface or "coupling" between it and whatever measurement system it's being applied to, whether just a flat plate, some kind of simulated "ear" system, or even live human subjects in some instances (I believe Rtings does this, check out their "Challenges & Solutions" video). The variability among humans and measurement techniques are also well laid out in this 2008 Stereophile article by Keith Howard.

By the way, do not forget the variability between the same model of headphones as well! Quality controls will affect the variability between each headphone made. Also, it seems that some manufacturers can change headphone characteristics even within the same product line - for example, check out the variability with the Campfire Andromeda IEM.

It's tempting to look at the prices for some of these Head And Torso Simulators (HATS) and ear-simulators thinking that one needs this level of equipment in order to test headphones. However, with all these variables, the "circle of confusion" when it comes to headphone measurements is relatively large. As such, I think it's important to not obsess on an exact graph or single result and recognize that there is only so much actual precision we need be concerned about and there is an element of "skill" that needs to be developed by whomever is running the measurements in achieving the consistency using whatever equipment one has (something I'm in the process of building through gradual experience with this stuff). There is a point where the measurement accuracy is also "good enough" as hobbyists and consumers. Look for trends and understanding of the broad strokes of what the measurements are telling us... 

Let's talk about the most important measurement - frequency response. Specifically, how should we display the frequency response graph?

The problem is that since the headphone has to couple with human ears, we all have our own head related transfer function (HRTF or Anatomical Transfer Function) which filters the sound that hits the tympanic membrane (microphone in the case of the measurement device) and this is in principle where we are measuring for an "accurate" sound. In fact, have a look at your own ears in the mirror. I bet you'll notice some asymmetry in the size and folds of the pinna - I can certainly see and feel these differences myself. It's remarkable how the brain through experience ultimately filters out the sound to make the stereo tonality perceived as "balanced"!

Since no two ears are exactly alike, what we can say is that the anatomical differences will lead to variation with usually a resonance peak between 2-5kHz. For a typical ear, around 3kHz one might see up to +17dB!  In the same way, the measurement jig with its synthetic ears will create a unique resonance pattern which might or might not conform to the resonance pattern of whatever "average" ear we ascribe to humans.

Logically, and simply being consistent with other device measurements, "ideal" should be aiming for a flat target graph IMO. Reading a "raw" graph is fine, but I don't see any benefit to readers "knowing" the raw shape since ultimately it's asking the reader to estimate variance from some "average" anyway. Therefore, when a frequency response is measured, how do we take this into account and convert to a "flat" looking standard curve that correlates with human preference for good sound? With what "compensation" curve then would we use?

Over the years, there have been free-field and diffuse-field targets of multiple forms applied to headphone measurements (see this Audio Precision note to see typical shapes of these). More recently, the Harman targets, particularly the Olive-Welti-Khonsaripour (2018) paper "A Statistical Model that Predicts Listeners' Preference Ratings of Around-Ear and On-Ear Headphones" has been commonly referenced. For us hobbyists, I think Tyll Hertsens' article a few years back before his retirement captures the essence of this complexity down to the level of human ear anatomy and those resonances.

So, the question for me to decide then was: "With the miniDSP EARS, what are you going to do when it comes to headphone measurements on this blog!?"

Well, until a better target comes around, let me stick with referencing the empirically-derived Harman 2018 AE/OE Target (as per the link above). For more on the Harman curve, check out Headphonesty's 3-part article, impressive work.

Keeping this next part short and sweet even though it took me about 3 weeks to get the job done to a reasonable level of satisfaction, broadly, here are the steps I took to create a compensation curve for my (H)EARS that I could be happy enough with:

1. Start with the miniDSP-supplied compensation curves called "HPN" and "HEQ", measuring a bunch of my headphones to see how these compare to other results posted online. It's quite clear looking at the "raw" results that the EARS has a strong concha/pinna/canal combined resonance up at 4.2kHz which is higher than the normal human ear resonance. While I like the HEQ's "flat" target in higher frequencies, it doesn't deal with the strong resonance as well as HPN. I ended up using a hybrid of the HPN below 5kHz blending with HEQ above 5kHz. I focused on the left compensation curve as a foundation and adjusted parameters to target some of the components of the Harman 2018 curve like the small sub-bass accentuation.

2. Using the left channel compensation from (1), now measure the same headphone earcup on the RIGHT side to check on right-left relative balance. Since this is the same driver but different silicone ears and microphone, tweak the mirror compensation curve to appear similar to the left. For this, I used my headphones with symmetrical ear cups like the Creative Aurvana SE and AudioTechnica ATH-AD700 - both relatively easy to measure with large enough earcups for proper seal.

3. With the general shape and left-right levels tweaked, do some final adjustments with a larger range of headphones. I drew inspiration from this Super Best Audio Friends thread for example. The AKG Q701 and Sennheiser HD800 are good headphones to use here. They have asymmetrical earcups/pads so not good for step 2, but easy to measure with consistent seal and large circumaural fit for final adjustments.

For your consideration, here is the final "Arch-Hybrid Compensation" (version 1.5) curves, both left and right plotted against the original miniDSP HEQ and HPN, along with Harman-2018 AE/OE:

Notice the strong resonance at 4.2kHz using the (H)EARS which is higher in frequency and magnitude than natural/standard ears.

The dotted green curve is the Harman 2018 target based on the GRAS 45CA coupler but with custom pinnae based on a modified IEC 60318-7 with softer Shore 00-35 silicone to reduce leakage (as described here). It's important I think to keep this in mind when we think about how the Harman curve was derived. Notice how different my compensation curves are compared to the Harman! This is a reflection of the significant differences between the (H)EARS and GRAS+custom pinnae and the price to pay for using a $200 hobbyist jig versus spending thousands on the industry-standard. ;-)

Due to variations between EARS units, the exact compensation curve will be different for each although I would imagine the general shape to be similar. As I said above, there's a significant amount of variability in headphone measurements and this is all part of the "art" we need to be mindful of regardless of what gear is used.

So then, in summary, the compensated frequency response measurements I publish will be using that solid "Arch-Hybrid" blue (L) and red (R) curves applied in order to achieve a "flat" response that would be Harman-like and sounds quite neutral to me when I listen with "flattish" headphone. I make no strong claims about "precision" other than a general attempt at a Harman 2018 AE/OE-like reference. With care, the measurements are quite easily reproducible and relative comparisons between headphones should not be a problem at all. Makes sense, I hope?

II. AKG K371 Measurements - Core

Without further ado, let's have a look at the AKG K371 "Core" results:

Using the excellent Room EQ Wizard software, we can examine a number of characteristics about the AKG K371 headphone. I've included in small font the serial number (anonymized last 2 numbers) for reference.

The top left quadrant of the summary chart is the all-important frequency response curve. You'll see in light blue and red the 3 "raw" readings I took averaging 95dB SPL volume for each side shifting the headphone on the (H)EARS slightly between captures (these are plotted -25dB to not obstruct the average, compensated result). The averaged and compensated frequency response curves are shown as solid red and blue lines. When making the measurements, I think it's better to just have 2 or 3 captures judged to be of good placement rather than too many because of all the possible variations which could easily skew the average. As noted in the graph, "flat" is defined as similar to the target Harman 2018 AE/OE as discussed above. I used the RME ADI-2 Pro FS headphone output (very low output impedance measured previously at <0.3Ω) as the standard DAC/amp for my measurements.

For the K371, we can see that indeed it does conform quite well to the Harman target. In the sub-bass, there's a bit of accentuation up to about +4dB if we take the average level at 1kHz as the baseline. The biggest deviation from target is a moderate dip maximally at 3.7kHz but affecting the 3-5kHz region. It's present in this DIY-Audio-Heaven report and the Audio Science Review post on these headphones show a similar upper-mid dip as well. 

To the top right, I've plotted the electrical impedance and phase for this headphone (usually I'll just use the left driver unless I see some significant discrepancies between channels). Minimum impedance is around 36.5Ω with a driver resonance peak at 24.5Hz at 42Ω. I've included a sensitivity reading as well with a 1kHz tone at 0.25Vrms. In the case of the K371, it's an extremely sensitive headphone! 106dB SPL/0.25V is approximately 118dB/V assuming good linearity (1Vrms is +12dB over 0.25Vrms). This is close to Hi-Fi News' measurement of 120.6dB/V done a few years ago. The official spec for the K371 is 114dB/V.

Bottom left we see the colorful waterfall plot. I've chosen a setting of 40dB Y-axis range from the peak, 100Hz to 20kHz along X-axis, and 10ms Z-axis timeframe. The waterfall has a nice and steep initial decline with no resonances to be found within 2ms from about 1.5kHz and above. Over time, there's some residual low-frequency resonance that bounces back like an echo that lingers below 1kHz out to 10ms. This is not uncommon with closed or semi-closed headphones I've tested. I wonder if more internal damping might help reduce cup resonance.

Bottom right is a look at the harmonic distortion profile with the headphone playing at my standardized level of 95dB SPL at 1kHz which is louder than my usual listening level. Distortion is plotted relative to fundamental at 100%. I've put the cursor at 1%. Given how forgiving the ears are when listening to music, I'm honestly not too worried about distortion causing much of a problem unless it significantly breeches that 1% threshold through much of the audible spectrum (remember the results of our harmonic distortion blind test a year ago). The K371 is nice and clean with mainly 2nd order harmonics (the amount of 2nd and 3rd harmonics particularly will vary for headphone with measurement jig interaction). We can see a bit of distortion poking above 1.5kHz but it's not much. Notice that I faded out the 3.5-5.5kHz portion. This area is where the EARS measurement jig has that strong 4.2kHz resonance and this artificially worsens the distortion result within that area; again, limitation of the test gear. However, notice that there's what looks like a second hump just above 5kHz which is likely a true area of increased distortion with these headphones.

While I might not do this regularly, we can look at the headphone linearity:

Notice that I've taken readings every 10dB SPL from an average of 65 to 105dB SPL. I've put the cursor at 1kHz and you can see how precisely the results are spaced 10dB apart in the legend values below (63.3, 73.3, 83.3...).

To take it one step further, we can then take that data and reference it to one of the levels like 75dB SPL and show the consistency with each reading!

Nice and flat, demonstrating good linearity with each output level (over a 40dB range) and across the audio spectrum.

Another result we can obtain without great difficulty is the impulse response:

It's a time-domain reflection of the frequency response. There's probably little point in showing this in general. However, what might be more interesting is to show what audio bandwidth-limited square waves look like out of the headphones:

Although it's a technical limitation, I think it might be a blessing that the miniDSP (H)EARS functions only up to 48kHz sampling rate (maximum 24kHz bandwidth). In order to avoid extraneous "ringing", the 100Hz square wave is bandlimited to ≤20kHz which I think is fine because in reality, we don't hear those ultrasonic frequencies anyway. Better to focus on actual audio frequencies when looking at these "microscopic" results.

Headphone drivers will never show a perfectly straight transition at the edge of the square waves as they have mass and take time to move. At times we'll also see roughness at the edge transitions due to overshoot. A "perfect" driver with this test would look like that blue tracing. The K371 is not bad and traced out the yellow waveform without any overshoot irregularity. Since these headphones are able to reproduce bass at least down to 100Hz quite well (in fact, with sub-bass accentuated), the waveform goes all the way to the next edge transition across 5ms as part of the "plateau" of the square wave.

For comparison, let me show you what these look like with the Sennheiser HD800. Note that my HD800 has been well-used over more than a decade and the original ear pads have long been replaced by some quite comfortable Accessory House Sheepskin/Protein Leather Hybrid ear pads so these might be a little different from stock measurements. 

Impulse response:

And here's the same 100Hz bandlimited "square":

Compared to the K371, notice that the Sennheiser has a little more irregularity at the peaks of the square waves perhaps suggesting some slight driver over-shoot. Also notice the "sagging" of what should be a horizontal squarewave "plateau" telling us that the HD800's frequency response has a bass dip. We could make a case though that the risetime is quicker with the Sennheiser HD800; suggesting "faster" transient response? If we look at how long it takes from the square wave transition point to peak of the squarewave, the Sennheiser HD800 takes around 0.5ms while it's closer to 1ms with the AKG K371.

We'll look at full Sennheiser HD800 measurements down the road.

While we can draw comparisons between different headphones, I'm not sure if these impulse response or square waves will add much to what we already know beyond the "Core" summary chart above. For these "microscopic" waveforms examining time-domain performance, the 100Hz square wave at least looks more interesting to me than the basic impulse response.

III. AKG K371 Measurements - Isolation and Leakage

Beyond the "Core" measurements, we can look at how the headphones interact with the outside world.

Fist, let's consider noise isolation which is how well, given a good seal, wearing the headphone is able to attenuate outside noise. We can do this comparison of relative noise level when a loudspeaker is playing at a preset SPL with the headphones on or off.

I used my Energy C100 speakers playing white noise at 1' away to create this graph:

As you can see, the headphones provide some isolation from outside noise above 400Hz; achieving about -9dB at 1kHz and around -20 to -25dB by 5kHz using my test methodology. The EARS will pick up noise from behind the headphone so in actual use, the noise isolation will be a bit better. I don't think it makes much sense to show too much detail here so I used 1/3-octave smoothing.

Conversely, we can measure how much headphone leakage there is (ie. how much of the sound can people nearby hear?). Ideally, we should be doing this in a quiet little anechoic chamber/box, but since I didn't feel like building one, here's what I did. Set the headphone output to an average of 100dB SPL as measured by the EARS. Then run a sweep measuring the sound at 1" away from the front-center "seal" between the earpad and measurement rig using an external miniDSP UMIK-1 measurement microphone. For a closed headphone like this, this area of the seal between the face and earpad seems to be where most of the sound is escaping from. It's not hard to put one's ear near the headphones to listen and select an appropriate place to capture a reading. Because of signal-to-noise limits of the UMIK-1, I chose to get closer than a more typical 6" or 1' position. With open headphones, the area of maximal leakage could be just behind the driver in which case I'll grab a reading from there instead. Not difficult to get a couple readings and find an average as well. I do this in a quiet room and also put a thick blanket over the headphone/microphone to reduce outside noise from contaminating the result.

Here's what the leakage spectrum looks like with these AKG K371 headphones:

Notice that I'm using REW's averaging tool between 100Hz to 10kHz. It's capturing ~55dB SPL average amplitude of the leakage at this microphone placement with a maximal intensity of 63dB at 4.3kHz. If we calculate the approximate dB at 1-foot (we can use this attenuation calculator), we come out with the leakage as 33.4dB SPL at 1'. This result is close to what Rtings found (33.85 at 1') with their K371 test measured at 6" and scaled.

Notice the angle of the average line tilts upwards, toward the higher frequencies. So leakage is primarily higher frequencies heard by bystanders.

IV. Subjective

In terms of look and feel in use, I found the K371 quite comfortable in the hand and on the ears. It's pretty well all black which is good for its intended studio purpose - don't buy these for impressive looking "bling" :-). The rubberized texture is comfortable. They're quite light weight (260g measured without cables) which is great for longer term listening. The plastic is not as creaky as other studio-type headphones I have (Fostex T50RP/Dekoni Blue, AudioTechnica ATH-M50, Sony MDR-V6 come to mind) and I like the soft synthetic ear pads as well which seal better than most headphones I've used but they do get a bit warm compared to the relative comfort of thick velour pads for example.

As for sound quality, I would say that these are quite "pleasant". I listened to them for a couple of weeks before these measurements, and indeed, I thought there was more sub-bass with these than most headphones. Given the closed nature and relatively easy-to-seal soft earpads (so long as you don't have huge ears!), one can appreciate the pressurized low rumble in bass-heavy music like Massive Attack's "Angel" (Mezzanine), Hans Zimmer's "Time" (Inception soundtrack), or the sub-bass-heavy "2049" and almost frightening intensity of "Flight to LAPD" from Blade Runner 2049. For those of us who love the bass experience in a home theater, headphone bass alas does not provide the powerful overall somatic experience of a subwoofer in the room and the thumping against the chest. But it can shake the noggin'. :-)

After running the measurements and noticing the upper-mid dip, I had a good listen to the headphones with a sine sweep and yes, I can hear that ~4kHz dip. I don't think it sounds as pronounced with my ears compared to what is found on testing. I wonder if human ear sensitivity as per the Fletcher-Munson curve has a significant part to play here where we are most sensitive between 3-4kHz which happens to also be the locale of this dip.

With a dip like this, there could be a positive effect of reducing harshness in some music. For example, the K371 does a good job with compressed modern pop. As a catchy tune, I think Dua Lipa's "Love Again" on Future Nostalgia sounds great and is my favourite track on the album. These headphones don't sound "sizzly" which is good for longer term listening on stuff like this. Likewise, that 3-4kHz dip does reduce sibilance and some of the brightness on tracks like the old audiophile acoustic nugget "Keith Don't Go" (Acoustic Live). Apart from this, the treble has adequate "brilliance" when called for.

More typical audiophile fare like female vocals such as Julia Fordham's Collection sounded natural with excellent separation of vocals from the background instruments.

Soundstage isn't expansive like open headphones, the sense of depth and any "outside the head" experience isn't there. While not really a fair comparison given headphone type and price differential, the Sennheiser HD800(S) clearly sounds much more expansive compared to these.

As a pair of studio monitoring headphones, these are great for resolution; listening to those little nuances. It's easy to pick apart poor recordings. Distortions from low-bit lossy codecs sound very obvious and ugly with these. A couple months back, I mentioned the Korean drama It's Okay To Not Be Okay soundtrack and the track "Her World (Moon Young's Theme)". I didn't notice just how much swirling intermodulation-type distortion was in the first 10 seconds (quite possibly intentional and fits with the emotional tone of the track) until I had a good listen with these.

V. Summary...

AKG K371 on miniDSP (H)EARS during measurements. Testing best performed with a shot of XO cognac... :-)

I hope this post gives you an idea of the kinds of headphone measurements and presentation format I'll likely be using in the days ahead. Also, for those considering using the miniDSP (H)EARS, I hope this touches on some important considerations (here's an excellent review with more info and comparisons). While one could just use the miniDSP-suppled calibration/compensation curves ("HEQ" and "HPN"), I think it's worth taking some time to do some "triangulation" using popular high quality headphones you may have to tweak the compensation curve. Regardless, whatever one does, there will be idiosyncrasies due to the non-standard gear. Comparing measurements using my Sennheiser HD800, AKG Q701, Creative Aurvana SE/Live!, AKG K371, and looking at what others have found is useful when doing some detailed tweaking.

As for the AKG K371 headphones, while I think it's too early to call these "classic" headphones (released in November 2019 so time will tell), I think they make an interesting modern comparison standard when talking about relative performance. They're easily available in the retail channels, relatively inexpensive, and the low ~35Ω impedance along with very high sensitivity measured around 118dB SPL/V at 1kHz makes this headphone very easy to drive even with low-powered mobile gear.

AKG, being a Harman subsidiary, has implemented the "Harman target curve" into these so you'll get a chance to listen for yourself if you like this kind of tonality. As discussed above, they do accentuate the sub-bass a bit. I don't know what the threshold is for those who consider themselves "bass-heads", but these headphones are certainly not bass-light and for me is tastefully done. If you're doing production work, be careful about the 3.7kHz dip and double check with other headphones and speakers to make sure that region doesn't sound too harsh. It think these would make for excellent "workhorse" closed headphone if you're into home recording with reasonable noise isolation and leakage. To put this into some context, when I have these headphones on and listening to music at normal levels, I can't hear my home phone ringing nearby (I have no problem hearing the phone ring while using a semi-open Dekoni Blue/Fostex T50RP, and definitely no issue with hearing what's happening around me with the open AKG Q701 and Sennheiser HD800). As for leakage, when my wife is sitting across from me about 1.5m away, she barely notices me listening to music nor can she hear the silly audiophile YouTube videos I'm listening to. ;-)

If I were to EQ, just a simple parametric 3.75kHz, +6dB, Q5 peak would be adequate.

Alright audiophiles. That's all for this week... Stay safe and hope you're enjoying some sweet tunes!


  1. Hi Archimago, thanks for a very detailed post, as usual. One thing that concerns me about EARS is the fact that it's a separate DAC with no world clock input. That means, there is a clock skew for sure between EARS and your RME used for output. You can check that by taking measurements in REW with different sweep lengths. And having a clock skew means you need to discard phase information when averaging multiple measurements.

    1. You mean 'EARS is a separate ADC'. That's right, and you can't measure phase at all - not only in averaging. But phase information on a one driver design (most headphones) seems not much useful.

      I find the un-human inner ear design much more concerning. There is no way to insert typical IEMs. Why did they do that?

    2. Thanks for correcting me, Techland. Yes, I've meant ADC.

      For IEMs, typical approaches for hobbyist measurements are either using a Chinese replica of an IEC 60268-7 coupler or the QuantAsylum's QA490 frontend. Doing IEM measurements is easier because you only need to create a pressure chamber between the IEM driver and the measurement mic and maintain the same distance between them between measurements to preserve frequencies of the resonances emerging inside the measuring chamber.

    3. Thanks Mikhail and Techland,
      Yeah, we can forget the phase information here. And I agree Tech, the ear design is much more of a concern, hence the look of those raw measurements and imperative to apply compensation.

      That QuantAsylum QA490 looks like a great rig for IEM testing with good consistency. I've tried a few IEMs with the EARS and some seem to work well with good consistency like the 1MORE Quad. But others, like the Etymotic ER-4B definitely needs more care to make sure to maintain consistency in that depth of penetration.

  2. I love your blog and this is my first time posting.
    Not a lot of love for the build quality on Amazon.
    These cans are $250 Australian dollars...... too much if the build quality is crap.

    1. Hey Unknown,
      Yeah, I can appreciate what they're saying online. These are mostly plastic so I think they're alright for around US$100, but don't throw them around ;-). Hmmm, $250 Aussie bucks is a bit steep. The Canadian dollar is about on par and I was able to get mine for around CAD$160 a couple months back.

      These days with the pandemic and sketchy electronics supply chains, I see a lot of variability in prices online... Today, the current price of CAD$258 on for free shipping is pure robbery and I would wait.

  3. Nice review. Now I have ANOTHER pair of cans to buy. I have the AKG K271's (2 pair) which are very strong in the HF region and bass shy to me, but what my old ears need for recording work to hear what others with normal hearing experience. I only track with these and would never master with them.

    I look forward to comparing these to my Audio Technica ATH-50X which do have very good bass. My AKG K-701 and the AT ATH70xxx (both open back) do have a greater sense of space and air.

    Right now my FAVs are my Beyer DT770 pros my son gifted me at Christmas, but at 250 ohms need more gain.

    It is good to have choices in the headphone game. The square waves look excellent for cans, but I am concerned about at dip at 3.7khz, a critical spot for piano and vocal work.

    1. Hey Jim,
      You guys in the US are very fortunate to have such good deals much of the time! :-)

      Sometimes, it's hard to find even relatively common items on Amazon here.

      Yeah, I'll be curious how you find these compared to K271. A friend lent me some old K260 (600Ω); haven't measured them yet but yikes, talk about mid-accentuation.

      Definitely you'll notice more bass than the K701. As for the AudioTechnicas, my ATH-M50 actually measures similarly - not as much sub bass and a tad less brilliance but it also has that dip around 3.5kHz.

      I don't have much experience with the Beyers but my friend also lent me a pair of 1991 "Made in W. Germany" DT 990s! Installed some "new old stock" Beyer velour pads and I'm impressed by the sound of these. Again, not measured yet but will later!

      Have fun with the headphones, Jim. ;-)

    2. Thanks for that insight into the AT 50's. I would not have known about the dip in those either. Amazon has a good deal right now on the 371's. Like I really need another pair of cans. lol

      As for the 271's, they are definitely hyped in the mids and highs, which makes them seem bass shy. . They do allow me to hear mistakes that I might miss if I was using the Beyer 770s.

  4. Glad to see you going into headphone measuring. I really like your approach. Especially the recognition of the natural variability between individuals' anatomy, test apparatuses, and frequency response preferences. I'm curious on your selection of 95db as the standard listening volume for your tests. Is that to pick an upper bound that a person is likely to listen to music? Perhaps I'm an outlier but I don't think that I'd listen at higher than 75db or so for any extended period of time. Do you think it makes sense to measure and a couple of different levels?

    1. Hi Doug,
      Thanks, and it has been fun! Other than trying to aim for a weekly article schedule (which isn't absolutely fixed in stone of course), hopefully these posts can be useful for hobbyists to follow along and learn together.

      I'm of the opinion that there is definitely such a thing as "good enough" objectivism :-). I appreciate the folks out there using industry-standard and expensive test gear since this helps me do stuff like double check my findings! I think it's fun for me to keep it "on the cheap" to see what the average audiophile can still accomplish while appreciating the variability in transducers especially for something like headphones.

      The 95dB SPL setting is basically a "round number". You'll see 94dB quite commonly as a microphone reference by the way because 1 Pascal of pressure works out to 94dB. Practically, on balance with the EARS microphones set at +18dB gain, 95dB gives me a good amount of overhead while also good signal-to-noise.

      Yeah, I don't listen at an average of 95dB SPL either but it does provide a reasonable level I think where we're not letting the headphones off too easy and not so loud that I can't have a listen for a short while without going deaf :-). If the headphone is able to stay well controlled with little distortion at 95dB, it'll more than likely be good at lower levels. I'll keep an eye on things like linearity as you can see in the graphs above at different output levels.

  5. Nice to see you venturing into headphone measurements as well.

    1. Hey SolderDude!
      Great work on DIY-Audio-Heaven all these years. Your exploration of all those headphones have inspired me to have a look myself. I saw your graph of multiple output levels to gauge linearity and thought "I gotta give this a try also" when checking out these K371s.

      Keep up the great work and systematic comparisons!

    2. With some headphones you can see the bass levels dropping at higher SPL (3rd hrm dist also increases as a warning sign to measure.
      In the distortion plot I would not show more than 4th or 5th harmonics to keep the plots somewhat clean.
      Looks like your compensation curve for the EARS looks fine. How did you calibrate it ? The compensation curve it comes with is (or at least was) 'incorrect'.

  6. Hi Archimago.

    Great to see, that you are now also digging into headphones. Great. After „closing“ of Innerfidelity, it was a bit difficult for me (try and error) to find good headphones. Nowadays, the results of RTINGs good reliable for me, to decide which headphone / in-ear it is worth to try or not. Continue your good work.

    Stay healthy!

    1. Thanks for the note Juergen!

      Hope all is well...

      Interesting you mentioned InnerFidelity. The other day I was watching the movie Nomadland and could not help but think of Tyll.

      Yeah, the guys at RTINGS have a great site. More Canadians :-).

  7. Some people might be interested to compare frequency responses of headphones.
    all and headphone measurements are available here:

    You can feed this data to the online visualizer tool:

    You can also use this data to get a flat frequency response by using the SBAF (super audio best friends) curve. he SBAF curve should not be confused with the Harmon curve from Olive and Welti. The Harmon curve is a consumer preference target. Use the spreadsheet available here:

    For people who like to use headphones for creating or mixing music with hardware or DAW software these compensation curves are very useful. For casual listening YMMV. Most people i know including myself prefer the Harman Curve for that.

    1. I do flatten the frequency response somewhat (not 100%, 80% is enough for me) using the VST plugin called TB Morphit plugin.

    2. Thanks FastFreddy,
      Nice links. Will make sure to have a look at some of the results in the days ahead!

      Yeah, overall I do like that Harman curve although for casual listening, I actually prefer an even "brighter" sound.

  8. Hi Archimago,

    Do you have any interest/plan to try various VST plugins that optimize the sound specifically for headphones? I mean crossfeed.

    My favorite one is 112 dB Redline monitor. I've been using it as my default listening method for headphones at home (and for earphones outside) for more than 10 years.

    But there are also Meier, TB Isone crossfeed plugins.

    You can also quite easily build your own VST-chain for crossfeed purposes (I've done it).

    Do you have any experience with crossfeed technologies? Have you tried it before?

    1. Hi fgk,
      Admittedly I have not played with many of the crossfeed DSPs over the years. I remember trying a few a number of years back just in foobar (Meier is one).

      Cool that you've stuck with 112dB Redline Monitor. Have not heard it... Maybe will try installing a VST plug and give it a go!

  9. Hi all
    I didn't see any reference to Oratory1990's pretty long list of EQ presets on Reddit:
    Quite nicely done. I tried this and Amir's curves for the Koss ESP95x on my ESP950s and found both turned them into more conventional sounding phones (less detailed to my ears) with the Oratory curve definitely boosting bass into audible distortion. Need to experiment further.
    Hope you're all well, as Ontario enters a very steep 3rd wave. The mood is not good here, with fatigue leading to fatalism or recklessness. I've had my first shot, and only hope that the vaccination program steps up rapidly, especially among front-line workers.
    Last lap, I hope

    1. Thanks Phil for that link!

      Great stuff from Oratory1990 - really professional looking work there.

      Very cool electrostats you have there Phil. Yeah, I think it's important to note that EQ'ing for a specific tonality isn't always the best thing to do. For example, I know my Sennheiser HD800s accentuate certain parts of the treble and sure, I can EQ some of that down. But do I necessarily "want" to or "need" to? Maybe a little but definitely not to the point where it follows the Harman curve.

      When I want fun, let's just say the Harman curve is not what I follow. :-)

      I consider it as just a useful "average" / "standard" to compare headphones to which is a reason why I wanted to published the results of these K371s first.

      Yeah, the 3rd wave of COVID intense here in Canada. ICU capacity getting very limited this past week in BC as well. The Brazilian P1 is close to 1/2 of the variants we're seeing here these days and affecting the younger folks not immunized yet.

      Stay safe Phil. Hope things calm down a bit and the "social mood" improves soon...

  10. BTW guys,
    Look at the comments here:

    There's a filter file there provided by Mitch Barnett for use with a convolution DSP for the K371!

    Sounds great on my system.