Neutralizer EQ - Opinions on Headphone EQ suitability and results..

[Cahudson42]

I find the concept intriguing..put 'everything inside the loop ' including your ears.. then EQ it.
Question is (for me at least) what am I ending up with? A 'personalized Harmon curve' based on my own hearing? An equivalent to 'flat' done otherwise? Or?

For those not familiar, after you install you create a 'profile' by listening to sine waves at frequencies from 32Hz to 16k, and adjusting a circular slider until the point you 'just' no longer hear the tone. 10 points as I remember..

A lot simpler than REW, a UMIK, and Equalizer APO..

Here is more detail (the app doc is useless. there isn't any):

https://www.techrepublic.com/article/how-to-perfect-your-androids-sound-with-neutralizer/

It's a bit frustrating that once you get the initial profile, you can't add the effects of, say, additional PEQs to tweak it..but that aside, at least testing on my 'everyday' SHP9500, it subjectively sounds pretty good with everything I tried so far on Amazon music.

Thoughts? Opinions? Criticisms? Pros? Cons?

All appreciated!

 

[BillG]

The developer doesn't give a enough information as what the goal of the app is for me to make a semi-educated guess as to what target curve they're aiming for. I'll speculate anyway, though: they're probably trying to create a perceptually flat EQ profile base upon an individual's hearing.

Perhaps an email to the developer would give us more precise information.

 

[Mad_Economist]

Seems like it's another of those "guess the HRTF from the threshold of hearing" type things. I'll give it a try tonight and post a comparison measurement with it active and not, although it may be a little noisy since I'll need to load some...well, noise on my phone to output, rather than doing a proper Farina sweep.

 

[pkane]

I find the concept intriguing..put 'everything inside the loop ' including your ears.. then EQ it.
Question is (for me at least) what am I ending up with? A 'personalized Harmon curve' based on my own hearing? An equivalent to 'flat' done otherwise? Or?

For those not familiar, after you install you create a 'profile' by listening to sine waves at frequencies from 32Hz to 16k, and adjusting a circular slider until the point you 'just' no longer hear the tone. 10 points as I remember..

A lot simpler than REW, a UMIK, and Equalizer APO..

Here is more detail (the app doc is useless. there isn't any):

https://www.techrepublic.com/article/how-to-perfect-your-androids-sound-with-neutralizer/

It's a bit frustrating that once you get the initial profile, you can't add the effects of, say, additional PEQs to tweak it..but that aside, at least testing on my 'everyday' SHP9500, it subjectively sounds pretty good with everything I tried so far on Amazon music.

Thoughts? Opinions? Criticisms? Pros? Cons?

All appreciated!

I tried to do something like this with my own software (Earful) and REW. The result wasn’t great. What worked better was to use in-ear microphones and to correct the response deficiencies measured that way.

 

[Mad_Economist]

Well, that's kind of boring. The EQ seems to be "what you see is what you get" in terms of the line shown on screen vs and the output EQ. After actually adjusting it by ear and finding its adjustments to my HD600 annoyingly small to capture in noisy plots (yes, yes, I should really get around to figuring out "Farina by wire", but mobile devices don't come up much for me), I simply started at 100% level for the bass and dropped it 10% each step. This produced this picture on the app:

And played back on the HD600 the result followed pretty closely (yellow trace/overlay = unEQ'd):

If you don't like squinting at two wiggly lines, here's the difference (which makes it a little harder to "look past the noise", but also makes the trend more obvious):

Repeating this starting from 0% at 32hz produced the same correlation of the plot and the difference

Excluding the intrusive impacts of low-frequency room noise; I knew I shouldn't have used -20dBFS white noise for this, my phone doesn't get the HD600 loud enough under those circumstances!

I can measure the different absolute levels by frequency required for it to spit out a flat line EQ, but my best guess is that it's using one of the ISO equal loudness contours; given the coarseness of the frequency bands and the limited adjustment range, it probably doesn't matter overly much.

Creative's ear photogrammetry thing is way cooler! It sounds way more objectionable, but it's cooler.

 

[Cahudson42]

actually adjusting it by ear and finding its adjustments to my HD600 annoyingly small

Fascinating! Given my lack of knowledge in this area, may I ask if what I conclude from your tests is correct?

1. Your HD600 is known to be one of the 'flatest' around in FR. Thus your result of small corrections makes sense. An additional test of a much poorer performing HP (with noticeable peaks/dips/resonances in FR) might confirm if this app has any utility in correcting for these - or not. (which may be unlikely because of the fixed nature of the sample frequencies and large steps between them, particularly in the upper range?)

2. Within it's limited +/- 5db EQ range, the app appears to do the corrections it shows on its display

3. But absent tests on further HP, it's premature to assume this app will "automatically" correct HP FR deficiencies, though it will provide - as any other EQ app, the corrections it shows.

Please correct me as necessary! Thanks!

Edit: Perhaps a bit of an assumption: 4. If one has compromised hearing, hearing loss at certain frequencies, particularly in the upper registers typical with age, the app may be able to provide some EQ compensation, based on that detection threshold will be higher than typical, and so the EQ reflecting it.. (unfortunate that it does not detect 'by ear' and correct accordingly)

 

[Mad_Economist]

1. Your HD600 is known to be one of the 'flatest' around in FR. Thus your result of small corrections makes sense. An additional test of a much poorer performing HP (with noticeable peaks/dips/resonances in FR) might confirm if this app has any utility in correcting for these - or not. (which may be unlikely because of the fixed nature of the sample frequencies and large steps between them, particularly in the upper range?)

The response of the HD600 isn't *incredibly* non-flat, but relative to most targets, its FR deviations exceed the bounds of the app. I can test with something with a more eccentric frequency response, but see below.

2. Within it's limited +/- 5db EQ range, the app appears to do the corrections it shows on its display

3. But absent tests on further HP, it's premature to assume this app will "automatically" correct HP FR deficiencies, though it will provide - as any other EQ app, the corrections it shows.

Indeed so - and I'll note, it simply doesn't have enough adjustment bands to fix "most" headphone problems. It's probably most significantly useful in attenuating any subjectively massive peaks/filling in massive dips a bit, and the hearing threshold trick is I suppose cute, but I recall a Samsung stock application on an older Galaxy that did literally the same thing with hearing thresholds by frequency to pretty much the same end.

I may try using RMAA's asynchronous measurement mode to get some less noisy plots tomorrow (although they will, unfortunately, be in RMAA, but we all have crosses to bear) - or, of course, someone else could; all you need is an ADC connected to your smartphone output here.

 

[BillG]

I've always been curious as to the target curve the following is aiming for:

https://play.google.com/store/apps/details?id=com.sonarworks.soundid.mobile

@Mad_Economist would you care to access it, if you've the time and inclination to that is?

 

[Soniclife]

Indeed so - and I'll note, it simply doesn't have enough adjustment bands to fix "most" headphone problems. It's probably most significantly useful in attenuating any subjectively massive peaks/filling in massive dips a bit

To be fair I think that's all it's really trying to do, it's just a quick way to get to a 10 band EQ setting. I've found it to work well.

 

[Soniclife]

I tried to do something like this with my own software (Earful) and REW. The result wasn’t great. What worked better was to use in-ear microphones and to correct the response deficiencies measured that way.

Was your result worse than this app, or was your disappointment with the approach itself?

 

[JSHamlet234]

This is interesting and possible to implement in a useful way, but there are some problems with this approach that must be overcome in the software. Our hearing sensitivity varies with frequency and our sensitivity to lows and highs also partially catches up to our sensitivity to mids as the listening level increases. For this app to produce a flat curve, it would need to compensate for both of these phenomena. Otherwise, the final FR will be very scooped.

 

[pkane]

Was your result worse than this app, or was your disappointment with the approach itself?

I've not tried this particular app. My correction was based on the lower threshold of audibility using Earful and the headphones I was trying to correct. I can only judge by what I heard, but correcting headphone response using the in-ear microphones was much more satisfying than correcting using the lower threshold curve produced with Earful. In fact, the in-ear-mic-corrected curve is what I've been using ever since, with all the headphones I own (four as of this moment )

 

[Soniclife]

I've not tried this particular app. My correction was based on the lower threshold of audibility using Earful and the headphones I was trying to correct. I can only judge by what I heard, but correcting headphone response using the in-ear microphones was much more satisfying than correcting using the lower threshold curve produced with Earful. In fact, the in-ear-mic-corrected curve is what I've been using ever since, with all the headphones I own (four as of this moment )

I'm not sure if you are aware, but this app uses noise at a reasonable volume to mask the tone, you turn the vol up on the tone till you hear it emerge from the noise. I don't know if the noise is white, or has been filtered somehow. Obviously measuring using mics in ears is rather more scientific, but do those mics work with IEMs?

 

[pkane]

I'm not sure if you are aware, but this app uses noise at a reasonable volume to mask the tone, you turn the vol up on the tone till you hear it emerge from the noise. I don't know if the noise is white, or has been filtered somehow. Obviously measuring using mics in ears is rather more scientific, but do those mics work with IEMs?

No, the in-ear mics don't work at all with IEMs, but then I don't use IEMs for music, and rarely for anything else.

Earful has a choice of signals, single tone, warble, and band-limited white noise and works similarly to an audiology test.

 

[Mad_Economist]

I've always been curious as to the target curve the following is aiming for:

https://play.google.com/store/apps/details?id=com.sonarworks.soundid.mobile

@Mad_Economist would you care to access it, if you've the time and inclination to that is?

Spent some time dinking around with Neutralizer, SoundID, and asynchronous recording today. Here's the FR of the adjustments that I got from both using their recommended protocol with the HD600:

These plots were created by measuring the output of my smartphone (LG V10) into a consumer sound card (Creative Soundblaster) with and without the equalization active, then subtracting the unequalized results to leave only the adjustments. If people particularly want, I can measure this on my HATS, but it's not exactly a detailed adjustment.

Notably, while the SoundID results are more "smile curve" like, in spite of using the preset for the HD600, the resulting equalization definitely will not put its bass in a Harmanish range.

Edit: If I were to guess the target response, it'd be Sonarworks' general target. This would cohere with the lack of adjustments with the HD600 selected, as Sonarworks' test protocol considers it pretty close to the ideal other than the bass:

I'll do some sweeps with SoundID and no adjustments other than different headphones to see if it's actually making changes there which are substantial...assuming I can get it to go 5 minutes without crashing...

 

[Mad_Economist]

Some further messing around later, I think SoundID is...a bit broken. The response it showed for the DT770 when I selected it matched Sonarworks' plot of the HD600...and then when I selected the LCD2, it had the DT770's adjustments. I'll check this on the outputs, of course, but that might explain the lack of headphone-specific compensation...

Edit: Iterative testing has shown that at least 6 different audio playback apps for android crash SoundID when I play my log sine sweep. I'm going to try the desktop version instead...

 

[BillG]

I'm going to try the desktop version instead...

Thanks for your time and effort.

 

[Mad_Economist]

Thanks for your time and effort.

Happy to do it, although sadly the result isn't terribly interesting. Behold, roughly the corrections that Sonarworks would be expected to apply to the example headphones:

The desktop version sucks a lot less than the mobile app though - it didn't even crash once, and only had about three meaningful compatibility issues with ARTA and my sound card!

 

[Mad_Economist]

Unsure if this merits its own thread - I'd do it if I had more interest in continuing this line of inquiry, but as a "fire and forget" I'm not so sure - but since I had Sonarworks on my measurement PC already, I decided to measure the response of a number of headphones I possess for which they have presets, mirroring what Brent did here.

The result shows general conformity (note that all measurements were compensated to the DF-HRTF of my HATS) but with some notable areas of spread. The acoustically low impedance designs (the planars and the open headphones) are fairly well grouped below about 500hz (the scale has been reduced to 30dB on the vertical axis, "zooming in"), but the closed designs vary more. I took particular care in the placement of the closed headphones here, but as their responses can easily vary by 10dB with normal leak variation, the source of difference could be on my end, Sonarworks' end, or both.

The higher treble - past 10k or so - is its usual state of anarchy, but since for several of these headphones the same-unit variations in that band can be comparable for different positions, I'm not inclined to give Sonarworks any guff about it. Bearing in mind that the midrange falls around the 100dB line, the treble as a whole is somewhat recessed versus the diffuse field target, which is congruent with the Harmanesque approach of Sonarworks.

The upper midrange is somewhat interesting, however - there is meaningfully more variation than I would have expected in the 1-6khz area, exceeding what I can attribute to placement variation (most of the headphones are quite consistent in this band) or unit variation (again, not a high variance area in most cases). I'll note that Sonarworks uses an earless flat plate fixture - albeit with an atypical multi-microphone approach - for some of their measurements to derive EQ; perhaps they do not account as well as might be hoped for the lack of pinna's varying effects per headphone. A difficult task for sure.

Particularly, I'm surprised to see headphones which are close together at 500hz poorly aligned at 2000 or even 1000hz. The gulf is by no means vast, but this is among the most consistent response bands for the headphones in question. Higher up, some features (e.g. the high Q dip on the CAL and the small peak of the HD800, both around 6khz) likely simply reflect a lack of attempt to combat very thin and likely variable response errors, but things closer to the center of the midrange leave me a bit confused.

Overall, seems to me like a case for using something fairly leakage invariant for your Sonarworks headphones of choice - you'll already have plenty of bass boost, and this will let you actually keep it if you have the audacity to need glasses or have hair.

@amirm pinging because if and when you do get a HATS or similar in again, I'd be quite keen to see your results from Sonarworks with the stuff you have on hand.

 

[MrBrainwash]

It's a bit frustrating that once you get the initial profile, you can't add the effects of, say, additional PEQs to tweak it..but that aside, at least testing on my 'everyday' SHP9500, it subjectively sounds pretty good with everything I tried so far on Amazon music.

Thoughts? Opinions? Criticisms? Pros? Cons?

All appreciated!

I am playing with this app now. I treat it as smart 10 band EQ equivalent.

I actually add another effects. I am using it with PEQ/GEQ (qudelix/ES100) or other apps (wavelet/BuddyEQ). There is a lot possible combinations and options this way.

This don't have to be standarized for me becouse I don't treat it seriously. The only thing I care about is that what I hear, and process of tunning is quite fun. I use app procedure but not every time. I created presets for HD668B, Tygr, HD560s, AurvanaSE, X2 for now, and I like results.

I got good results with oratory1990 presets but I don't think that it has to be end word in this topic of qualization for headphones.