Is EQ'ing headphones worth it?

[bobbooo]

Several have been posted here - sites that host a high quality recording, play it back in trials starting with high distortion, incrementally reducing with each trial. Search and you will find them, and try it yourself if you like. However, in testing trained listeners, audible distortion threshold is generally found to be around 3% in the bass, 1% in the mids and a bit lower in the treble. These distortion thresholds are reflected in the ITU standard for audio equipment used for blind testing (BS.1116, I believe).

The ideal confidence level of an ABX test depends on the goal of the test. An ABX test cannot have both high precision and high recall. High confidence percentiles are high precision but low recall; they reduce false positives but they increase false negatives. Low confidence percentiles (still above 50% of course) are low precision but high recall; they reduce false negatives but they increase false positives. This has been discussed at length on this forum, search and you'll find it.

Do you have a source for studies that have shown trained listeners can detect those levels of distortion? And at what confidence level? I would have thought the levels in the ITU standard for a reference system would be well below audible levels in order to err on the side of caution. You didn't say what your just audible thresholds and confidence level were when you did your ABX tests? As you are claiming to be able to hear distortion differences from positive gain when EQing different headphones (ignoring the fact that negative preamp gain can be applied to partially negate this), the onus is on you to show this audibility with a high confidence-level ABX result that has very low chance of false positives (which would diminish your claim). If you can't do this, then your claim has no scientific merit. And an important distinction - high confidence levels do not 'increase false negatives' (or decrease false positives). It's the probability of these occurring that are increased and decreased respectively. In this case you have made a positive claim of audibility which needs to be backed up by positive evidence with a high confidence level - the fact that the chance of false negatives is higher here is not relevant to your claim. I understand ABX tests are not perfect, but they're the best tests we have at determining actual audible thresholds. If you really want to eliminate false negatives, one potential test would be a null listening test - take your distorted track waveform, minus the original undistorted track's waveform, then ABX this difference track against one of pure silence. This would require no short-term memory, and reduces a complex relative audible difference test to one purely of absolute audible SPL threshold.

EDIT: On second thought, that last idea would probably be too easy of a test to pass, as it doesn't consider the effective perceptual auditory masking of the distortion when the actual distorted track is listened to normally. Best to stick with standard ABX tests, which have proven to be our most reliable tool for investigating psychoacoustics for decades.

 

[MRC01]

Do you have a source for studies that have shown trained listeners can detect those levels of distortion? And at what confidence level? I would have thought the levels in the ITU standard for a reference system would be well below audible levels in order to err on the side of caution. You didn't say what your just audible thresholds and confidence level were when you did your ABX tests?...

The audibility of distortion at those thresholds is well established. You can read evidence of this, and just look at some of the threads here where people took ABX tests and published their results. Here's another study it took me less than 1 minute to google: https://www.axiomaudio.com/blog/distortion/
Not surprisingly, they found a lower threshold of audibility of about 1% or -40 dB for distortion in the midrange and treble. This is typical. I intentionally didn't mention my own thresholds because I don't want to make claims that turn this discussion into a "golden ears" contest. This is about minimum thresholds of distortion audibility for trained listeners, not about any specific individual.

Incidentally, this ties back to an earlier claim in this thread that FR accounts for all of what we hear in a headphone, that other aspects of sound like distortion and spectral decay are inaudible, so if you EQ any 2 headphones the same, they will be indistinguishable to a listener. This is an unusual claim - in fact, an incorrect claim. As I said earlier, FR accounts for most of the difference we hear, but not all of the differences we hear.

... And an important distinction - high confidence levels do not 'increase false negatives' (or decrease false positives). It's the probability of these occurring that are increased and decreased respectively. In this case you have made a positive claim of audibility which needs to be backed up by positive evidence with a high confidence level - the fact that the chance of false negatives is higher here is not relevant to your claim ...

This is incorrect. Due to the statistics of trials, it's easier in ABX testing to get high precision, and very difficult to get high recall. As a sonic difference approaches any individual's perception thresholds, his ABX scores don't instantly go from 100% to 50% (random guessing). Even if the scores drop rapidly, they also drop smoothly. No matter where you set the confidence threshold, there is always a range where he detects the sound better than random guessing, but below that threshold. These are false negatives, and are inescapable. But the only way to eliminate these false negatives is to reduce the confidence threshold, which creates false positives - people who get it right by guessing.

To take a hypothetical example: suppose we set the confidence threshold at 99% as you suggest. The test subject scores 95% with distortion at 1%. Thus he fails and we say he can't hear the difference. Yet it's far more likely than not, that he actually did hear the difference. A test that is more likely to be wrong, than right, is not very useful! To account for this, suppose we lower confidence to the opposite extreme: 51%. Now, we correctly record this subject's performance, but other test subjects are likely to pass by random guessing. This illustrates the asymmetry of ABX testing: it's easy to get high precision and eliminate false positives: just set the confidence threshold as high as you want. A single test is now statistically significant. But the price you pay is false negatives. However, with ABX testing it's hard to get high recall and eliminate false negatives. Because of the high likelihood of passing by guessing, a single test is not conclusive and the results only become statistically significant if you conduct it over many trials and many different people.

The net is that ABX tests are never perfectly sensitive, so whatever thresholds we determine to be statistically significant with ABX testing, we know the actual thresholds are always somewhat lower (how much lower, we don't know). Yet ABX testing tells us that trained listeners can detect distortion at around 1% in the midrange and treble, with statistical significance.

 

[Robbo99999]

I have an update for you on my attempts at EQ'ing my AKG K702, this is Attempt #2 after my post a couple of days ago. Like @bobbooo & @Kouioui advised I downloaded Peace for Equaliser APO and applied Oratory's K702 Anniversary Edition profile using Peace and applying Oratory's equaliser settings on his pdf file. I did that and then fine tuned his profile to try to be more applicable to my K702's as they are slightly different than the K702 Anniversary Edition that Oratory actually measured for. After comparing the stock frequency response curves (thanks bobbooo for those) of K702 vs Anniversary Edition I came to the conclusion that the major difference between the two is the frequency response at 3500Hz which Oratory addresses by applying +4dB....now my K702 is already at the +10dB target of the Harman curve at that point so I decided to simply move the 3500Hz slider down to 0dB, but kept all his other settings the same...this removed the tinniness of the sound. Now my EQ'd sound vs stock is quite a bit better I think...it's removed some of the fake detail (brightness) of the K702 yet still retains what I think is 'real' detail in that female vocal area; the EQ has also increased detail in the bass line of songs so I can follow bass guitars more easily, and I think it's a more rounded/balanced sound overall and has removed some of the harshness of the K702 without reducing detail...a lot of genres are more enjoyable to listen to now, particularly stuff like Mettallica for instance - a lot more punch!

Here are some pics of my settings and graphs that I've annotated to hopefully help illustrate my approach to EQ'ing my headphones so far, and expands upon what I described in my previous paragraph:

Has anybody got any further advice for me re. refining my EQ, and what do you think to my approach with regards to how I've modified Oratory's Anniversary Edition profile to meet my standard K702 more accurately, does it make sense, have I missed anything?

 

[flipflop]

@Robbo99999, click the box with the green curve under the 'Effects' box to display how your settings affect the frequency response. Those bands will be overlapping, so it's difficult to imagine what it looks like.

 

[Robbo99999]

@Robbo99999, click the box with the green curve under the 'Effects' box to display how your settings affect the frequency response. Those bands will be overlapping, so it's difficult to imagine what it looks like.

Ah, didn't know about that button.

 

[Robbo99999]

Ah, and in addition to my questions in my long post above, does EQ'ing create any negative effects regarding what Amir measures on his DACS, eg SINAD / "distortion" for instance? One thing I'm aware of is that EQ'ing means running quite a hefty negative pre-amp, doesn't this mean that you're reducing the dynamic range? For instance my preamp is -5.2dB, and does it really need (should it be?) to be so large when in Peace when I play my music it seems to peak out at a maximum of about 50% on the Peak Value Meter of Peace? Couldn't I improve my SINAD/"distortion" by running not quite as large a negative pre-amp? How much difference does this really make to the variables that Amir often measures in DACs? (maybe @amirm too). And how much difference (if any) does any negative effect on those variables make to the noticeable sound/quality of the music?

 

[MRC01]

Ah, and in addition to my questions in my long post above, does EQ'ing create any negative effects regarding what Amir measures on his DACS, eg SINAD / "distortion" for instance? One thing I'm aware of is that EQ'ing means running quite a hefty negative pre-amp, doesn't this mean that you're reducing the dynamic range? ...

Two separate questions. Easy one first: digitally reducing gain should not cause any loss of information. The reason is simple: most music doesn't use anywhere near 16 bits of resolution, let alone 24. The digital bit depth is higher resolution than the music; the least significant bits aren't carrying musical information (though they're still useful for dither). However, these level reductions do reduce the analog S/N ratio by the amount of the reduction. With a clean analog stage this reduction should be immaterial or unnoticeable.

PS: It's always safe to reduce gain by the full amount of the highest boost, but you don't necessarily have to reduce it that far. For example your biggest boost is 5.5 dB @ 110 Hz. Suppose you cut overall gain by only 2.5 dB instead of 5.5 dB. If you played a pure 110 Hz signal at full scale (or as low as -3 dB), it would clip. But music consists of a broad spectrum of frequencies, which means no single frequency can individually be at full scale else the overall signal would clip. Consider a worst-case full scale signal having at least 2 frequencies, say 110 Hz and 220 Hz. In this signal, each individual frequency is at -3 dB (which is why their superposition or sum is at 0 dB). So if you're playing music (e.g. signals having broad frequency spectrum), the 110 Hz portion of the spectrum can never be above -3 dB so it's safe to use at least 3 dB less attenuation than your biggest boost, which is 5.5 dB, so an overall cut of only 2.5 dB should be safe. Of course, such a setup WILL clip if you played a pure 110 Hz test tone above -3 dB. But it won't clip with any multi-frequency signal, like music.

Regarding distortion: experimenting with EQ and taking measurements with Room EQ Wizard, I find that boosting frequencies raises distortion at those frequencies. This may or may not be audible, depending on the level. However, boosting bass may increase distortion in the mids and treble where it is more easily audible. How can that happen? Most of the musical energy (amplitude) is in the low frequencies, with amplitude decreasing at higher frequencies. So you can think of the midrange & treble as a ripple riding on the bass wave. With most drivers, distortion is related to excursion (amplitude). And most headphones have a single driver reproducing the entire frequency range. So if you boost the bass, you increase the driver excursion, which increases distortion not only in the bass, but also in the midrange & treble since the same driver you're forcing to swing back & forth with larger excursions, is also reproducing those higher frequencies. This may or may not be audible, depending on how much you boost, at what frequencies, the headphone's distortion profile, and your overall listening level.

Overall with reasonable EQ none of this should be a problem. The EQ you have above has boosts below 6 dB, which is reasonable.

 

[Robbo99999]

Two separate questions. Easy one first: digitally reducing gain should not cause any loss of information. The reason is simple: most music doesn't use anywhere near 16 bits of resolution, let alone 24. The digital bit depth is higher resolution than the music; the least significant bits aren't carrying musical information (though they're still useful for dither). However, these level reductions do reduce the analog S/N ratio by the amount of the reduction. With a clean analog stage this reduction should be immaterial or unnoticeable.

Regarding distortion: experimenting with EQ and taking measurements with Room EQ Wizard, I find that boosting frequencies raises distortion at those frequencies. This may or may not be audible, depending on the level. However, boosting bass may increase distortion in the mids and treble where it is more easily audible. How can that happen? Most of the musical energy (amplitude) is in the low frequencies, with amplitude decreasing at higher frequencies. So you can think of the midrange & treble as a ripple riding on the bass wave. With most drivers, distortion is related to excursion (amplitude). And most headphones have a single driver reproducing the entire frequency range. So if you boost the bass, you increase the driver excursion, which increases distortion not only in the bass, but also in the midrange & treble since the same driver you're forcing to swing back & forth with larger excursions, is also reproducing those higher frequencies. This may or may not be audible, depending on how much you boost, at what frequencies, the headphone's distortion profile, and your overall listening level.

Overall with reasonable EQ none of this should be a problem. The EQ you have above has boosts below 6 dB, which is reasonable.

Thanks, sounds like it's probably not a problem for my setup then. I do reduce the gain digitally...the preamp is part of the Peace software and I've got a SoundblasterX G6 DAC...so I guess that's all digital - I also have 24bit selected for the DAC's bit depth regardless of whether I'm playing 16bit music...and I have iTunes set to 24bit playback too. Does that all mean that by choosing 24bit in those various areas that I'm not losing as much dynamic range with the digital gain reduction I'm using?

I can visualise what you're saying about single drivers and bass boosting and the midrange and treble riding the waves of the bass as the driver moves with the sound of the bass. Yes, my headphones are sounding good with this EQ, I don't think I have noticeable distortion...although I'm not sure how I would identify that...I don't feel like there's any detail missing from parts of the music, and I don't feel that there are any strange 'dips' or missing parts...just using my imagination as to how audible distortion would sound. EDIT: it sounds clean as in no incoherent booming...it's all got 'shape'.

EDIT#2: and thanks for you "P.S." edit to your initial post. I might consider using less gain reduction on the Pre-amp, but if there's no practical benefit to sound quality then I may as well leave it at a safe -5.2dB?

 

[MRC01]

... I do reduce the gain digitally...the preamp is part of the Peace software and I've got a SoundblasterX G6 DAC...so I guess that's all digital - I also have 24bit selected for the DAC's bit depth regardless of whether I'm playing 16bit music...and I have iTunes set to 24bit playback too. Does that all mean that by choosing 24bit in those various areas that I'm not losing as much dynamic range with the digital gain reduction I'm using? ...

It's only an amplitude shift and the music doesn't use all 16 bits so you will not lose dynamic range. As long as the 16-24 conversion is properly dithered, it won't add distortion, it will only raise the noise level slightly. Again, with music (or any broad frequency spectrum signal) you can use 3 dB less reduction than your biggest frequency boost; in your case you only need -2.5 dB of gain reduction, not -5.5.

... I can visualise what you're saying about single drivers and bass boosting and the midrange and treble riding the waves of the bass as the driver moves with the sound of the bass. ... just using my imagination as to how audible distortion would sound.

If you're curious about playing with that, here's one recent ASR thread on this: https://www.audiosciencereview.com/forum/index.php?threads/distortion-listening-test.8152/

 

[Robbo99999]

It's only an amplitude shift and the music doesn't use all 16 bits so you will not lose dynamic range. As long as the 16-24 conversion is properly dithered, it won't add distortion, it will only raise the noise level slightly. Again, with music (or any broad frequency spectrum signal) you can use 3 dB less reduction than your biggest frequency boost; in your case you only need -2.5 dB of gain reduction, not -5.5.


If you're curious about playing with that, here's one recent ASR thread on this: https://www.audiosciencereview.com/forum/index.php?threads/distortion-listening-test.8152/

I guess I'm average at detecting distortion (-21dB was my threshold) (tested at normal listening volume)....yep and I don't think my EQ is causing distortion in my headphones:

 

[Robbo99999]

@MRC01 , actually my volume control is not all digital on my setup, I'm My Computer -> SoundBlasterX G6 DAC -> JDS Atom Amp -> AKG K702.
So the JDS Atom amp is not digital, I'm increasing the volume on the JDS Atom to compensate for the preamp gain reduction in PEACE for Equaliser APO. Do you see any issues regarding that setup and -5.2dB gain on the pre-amp? I'm also at 79% Windows Volume because Amir found that the G6 DAC has some clipping/distortion on the Line Out if run at 100% Windows Volume (-2dBFS cures the problem apparently). Do you see any issues there in light of all that?

EDIT: got some clipping according to the Peak Value Meter in PEACE when playing Muse's Supermassive Black Hole when at -2.5dB on Preamp, so I think I definitely need to run a bigger negative gain on the preamp, maybe not the full -5.2dB I was using before though.

 

[MRC01]

Just make sure to apply the preamp gain in digital domain before the EQ and the DAC. If you're using -5.2 dB then you shouldn't need to worry about 79% / -2 dB; you should be able to leave it at 100% because the incoming signal will already be attenuated.

Regarding clipping, most modern rock/pop is heavily compressed, some of it is already clipped, though they often use soft limiting which can make it look like it's not clipped when it really is. I took a 2nd look at your EQ settings. Because your +5.5 dB boost @ 110 Hz is a low shelf, it's affecting not just a narrow frequency range but everything below 110 Hz, so you may actually need the full -5.5 dB preamp gain.

On low gain, the JDS atom doesn't clip even at max volume. So set that volume as high as you want.

 

[Robbo99999]

Just make sure to apply the preamp gain in digital domain before the EQ and the DAC. If you're using -5.2 dB then you shouldn't need to worry about 79% / -2 dB; you should be able to leave it at 100% because the incoming signal will already be attenuated.

Regarding clipping, most modern rock/pop is heavily compressed, some of it is already clipped, though they often use soft limiting which can make it look like it's not clipped when it really is. I took a 2nd look at your EQ settings. Because your +5.5 dB boost @ 110 Hz is a low shelf, it's affecting not just a narrow frequency range but everything below 110 Hz, so you may actually need the full -5.5 dB preamp gain.

On low gain, the JDS atom doesn't clip even at max volume. So set that volume as high as you want.

Cheers for the response, I was thinking about it before you replied, regarding the 79% windows volume setting and the fact that yes I'm -5.2dB attenuated in the preamp...and I did consider that this could make the 79% windows volume unnecessary, but I remember reading in Amir's review of the G6 DAC that the G6 had the problem with pumping out the low frequencies (that's where the clipping/distortion was) when it was at 100% windows volume...and if you think about it then it's actually the low frequencies on my EQ that I'm boosting, to +5.5dB, so I think I need to leave it at 79% volume to prevent those low frequencies from rising above the -2dBFS that Amir identified as the clipping point. But I see your thinking behind it, because it's something that crossed my mind too.

You say to make sure that preamp gain is applied in digital domain before the EQ and the DAC....I'm assuming that's what PEACE does when I use the preamp built into PEACE, so I'm complying with your recommendations already right?

I'll keep the -5.2dB preamp gain like you suggest and also as written into Oratory's settings on his pdf.

I've been listening to some music tonight on these EQ'd K702 headphones, I'm really impressed...it feels even less like I'm listening to headphones than it did with the stock AKG K702 which already felt like a large soundstage when at stock....there feels like there's loads of space between the music and the instruments and kinda like you're inside of it...and the bass line is so much easier to follow which was less present before....and the harshness of the K702's has been rounded off and I think it sounds a lot more natural now...I'm really impressed.....I just voted "Yes" to the poll in this thread - EQ'ing headphones is worth it! EDIT: Obviously most of this groundwork is down to Oratory & Harman as well as everyone's input here in this thread that I've listened to...I couldn't have done it without it all!

 

[Robbo99999]

Hi, I have one question re Peace and Equaliser APO - are the changes reboot persistent, ie do I need to load up Peace or Equaliser APO at each reboot in order for the settings to take effect? I've tested this and it appears that the changes are reboot persistent, so I think there is no need to load Peace or Equaliser APO each time I turn on my computer, but I just wanted to ask here to be sure? Also, how come the changes can be reboot persistent if the programs are not running....does Peace & Equaliser APO rewrite something 'permanently' to disk until you go back into the program to turn it back to stock? I've tried googling this and amazingly there's nothing I could find!

 

[thewas]

Yes, the settings of EQ Apo are reboot persistent as those are just interfaces changing the direct sound settings of your PC.

 

[Fluffy]

EQ Apo always runs in the background. It's convenient to have a shortcut to the editor (I put it on the desktop and in the start menu) so you can easily see what changes it's doing. It actually does basically write something to disk – there is a file called "config" (usually located at C:\Program Files\EqualizerAPO\config) and the program is continually reading it to know what changes to make. On the EqualizerAPO Editor it's the tab called "config.txt".

 

[Robbo99999]

EQ Apo always runs in the background. It's convenient to have a shortcut to the editor (I put it on the desktop and in the start menu) so you can easily see what changes it's doing. It actually does basically write something to disk – there is a file called "config" (usually located at C:\Program Files\EqualizerAPO\config) and the program is continually reading it to know what changes to make. On the EqualizerAPO Editor it's the tab called "config.txt".

Thanks for the reply. Do you know the name of the process that EQ APO creates that runs in the background? I've tried using Task Manager to look for a EQ APO and there's nothing running. Maybe EQ APO just semi-permanently changes some "standard existing Windows audio config files" in the Windows audio chain...maybe that's how it implements it's changes even when the program is not running in the background. (Yep, I understand what you mean by the config.txt, but that doesn't explain how the changes are persistent between reboots when the program is not running).

EDIT: also there are no new startup items in the startup folder and no new Windows Task Scheduler tasks created since I installed Peace & Equaliser APO...so I'm pretty sure that Equaliser APO is not starting up any new software at boot time....I think the persistent changes must be from editing some "standard existing Windows audio config files" that I mentioned in the previous paragraph?

 

[Fluffy]

I am also curious how that works. I never managed to find any process by that name, so its puzzling. Someone who knows how windows direct sound works behind the scenes will probably know the answer. Note that in the configurator you need to install it per sound device, and it requires a reboot, so it might work like a driver or something (you can uninstall it for a specific device and it will stop working for it). Also, it stops working if you uncheck the box "enable audio enhancements" in the device properties (advanced tab) in the windows sound settings. And it also doesn't work if you are sending data straight to a device with ASIO for example. So it must be riding on something the windows direct sound is doing already.

 

[flipflop]

I've tried using Task Manager to look for a EQ APO and there's nothing running.

Check the 'Processes' tab. Peace.exe is running on my PC.
FWIW, you can also activate the tray icon by going to 'Settings' and tick the (you guessed it) 'Activate tray icon' box.

 

[thewas]

Check the 'Processes' tab. Peace.exe is running on my PC.

You don't need to even install Peace to run EQ Apo, Peace is just an interface for Equalizer APO.