Resolve's B&K 5128 Headphone Target - you can try the EQ's.....

[amirm]

This really isn't true, though - the "RR1_G" target was produced in 2013, in one of the first few papers, and was strongly preferred!

It was an easy start because other target developed by other researchers were so poorly done -- the point I was making to Resolve in not chasing others just because they have published papers. The Harman study you cite only evaluated two headphones though. Much more work was done with a larger set across many other listening tests and for other types of headphones/IEMs. All in all it took about five years to write the last chapter. Here it is from Sean's recent presentation:

You seem to want to trivialize this with nary a single controlled study let alone what Harman put in.

 

[solderdude]

After testing some 300+ headphones and speakers, the story is told by the time you get to 5 or 6 kHz. I have applied EQ correction above that but the effect is incredibly small. There is not just not a ton of content above those frequencies and what is there is more to taste and reality. Get the response correct to 5 or 6 kHz and you are golden. Worst offences happen from 500 Hz to 3 kHz really.

Yes, for tonal balance I certainly agree.
From 5-7kHz is the sibilance area.
Between 8 and 12kHz is the 'sharpness' area.
To me sharpness and/or sibilance can make or a break listening pleasure and can even be a cause for listening fatigue.

A slight boost between 5 and 7kHz can make the sound a bit 'fresher/detailed' and between 8kHz and 12kHz can make sound more 'sharply detailed', but as with everything in life too much of something is never a good thing.
A substantial dip somewhere between 5 and 12kHz can make the sound dull.
Lots of peaks and dips in that frequency range can make treble sound 'coarse' and lacking in detail where an even response can make treble sound 'smooth'.
Those aspects, at least to me, are also important and equally important for enjoyment than getting the tonal balance at least decent.

I was kind of hoping the 5128 could shed a bit more light in that area. Maybe, to get the best picture, the idea of using a GRAS and 5128 might be giving some more clues on:
Seal and the 5-12kHz range.

Above 12kHz range I don't think there is much 'musical info' anymore but perhaps the 'airy' sound, some headphones do well, is in the treble extension. Well extended treble (to say 18kHz) does seem to sound 'better' than some headphones showing a steep drop above 12kHz.

 

[isostasy]

Yes, for tonal balance I certainly agree.
From 5-7kHz is the sibilance area.
Between 8 and 12kHz is the 'sharpness' area.
To me sharpness and/or sibilance can make or a break listening pleasure and can even be a cause for listening fatigue.

A slight boost between 5 and 7kHz can make the sound a bit 'fresher/detailed' and between 8kHz and 12kHz can make sound more 'sharply detailed', but as with everything in life too much of something is never a good thing.
A substantial dip somewhere between 5 and 12kHz can make the sound dull.
Lots of peaks and dips in that frequency range can make treble sound 'coarse' and lacking in detail where an even response can make treble sound 'smooth'.
Those aspects, at least to me, are also important and equally important for enjoyment than getting the tonal balance at least decent.

I was kind of hoping the 5128 could shed a bit more light in that area. Maybe, to get the best picture, the idea of using a GRAS and 5128 might be giving some more clues on:
Seal and the 5-12kHz range.

Above 12kHz range I don't think there is much 'musical info' anymore but perhaps the 'airy' sound, some headphones do well, is in the treble extension. Well extended treble (to say 18kHz) does seem to sound 'better' than some headphones showing a steep drop above 12kHz.

I currently have reservations about whether use of the 5128 will show us much about the 5-12kHz area though (although I would love to be proved otherwise).

I think this comes back to your question of what is 'reality'.

https://twitter.com/x/status/1630387955763388416

Here is the Twitter thread from Dr. Olive with examples of blocked canal measurements which most of us have seen before.

For the sake of another data point, here's my HD6XX on my FPC and in ear mic.

The frequency response in each person's ear above 5khz is all over the place and there are enormous variations between subjects.

Is there anything wrong with this though? Is it due to the headphone design or are we just seeing the response of each person's ear in a way that would be representative of any other input, e.g. sound coming from speakers? If this were so, are some headphones better at matching this than others? Does it tell us anything about perception or preference anyway?

Sorry if this doesn't make sense, finding it hard to verbalize my thoughts so posing questions was the best way I can do it right now.

I hope this make sense though? I guess I see a paradox arising where, in order to demonstrate meaningful results in the treble, you'd have to make a target curve with very little smoothing so you can see all the interesting peaks and dips. But then as soon as you've done that you've just created a target which doesn't match most people. So you have to smooth it more and more until it isn't miles away from the average of what most people hear anyway.

I re-watched Amir's old video on understanding headphone measurements and that helps to understand where he's coming from in this discussion. "You have to allow for a certain amount of uncertainty. Nothing about headphone analysis or even equalisation should be about precise assessments".

 

[Robbo99999]

I currently have reservations about whether use of the 5128 will show us much about the 5-12kHz area though (although I would love to be proved otherwise).

I think this comes back to your question of what is 'reality'.

https://twitter.com/x/status/1630387955763388416

Here is the Twitter thread from Dr. Olive with examples of blocked canal measurements which most of us have seen before.

View attachment 284160

For the sake of another data point, here's my HD6XX on my FPC and in ear mic.

The frequency response in each person's ear above 5khz is all over the place and there are enormous variations between subjects.

Is there anything wrong with this though? Is it due to the headphone design or are we just seeing the response of each person's ear in a way that would be representative of any other input, e.g. sound coming from speakers? If this were so, are some headphones better at matching this than others? Does it tell us anything about perception or preference anyway?

Sorry if this doesn't make sense, finding it hard to verbalize my thoughts so posing questions was the best way I can do it right now.

I hope this make sense though? I guess I see a paradox arising where, in order to demonstrate meaningful results in the treble, you'd have to make a target curve with very little smoothing so you can see all the interesting peaks and dips. But then as soon as you've done that you've just created a target which doesn't match most people. So you have to smooth it more and more until it isn't miles away from the average of what most people hear anyway.

I re-watched Amir's old video on understanding headphone measurements and that helps to understand where he's coming from in this discussion. "You have to allow for a certain amount of uncertainty. Nothing about headphone analysis or even equalisation should be about precise assessments".

You bring up some interesting points that I'll let some other people answer if they want, but on the point of equalising a headphone accurately to say the Harman Curve from measurements, then I don't see a problem with that - it's the best target that we know of, and EQ'ing your headphone accurately to that will give you your best chance of your initial best sound (before you personalise it, mostly in the bass probably). As long as you don't use silly sharp (high Q) boost filters in the treble to EQ it accurately, then you're ok - Oratory uses a sensible approach to EQ'ing accurately to a target curve, so I'll use that as an example. Nothing wrong with EQ'ing accurately to a target curve as long as it's done right.

 

[MayaTlab]

From my POV this is pretty harsh - all predictive models are flawed, and there are always outliers, but the PPR model is pretty good! It gets the "shape" of things generally right, with good headphones trading places somewhat, and a couple of headphones rated as being much worse than they were heard to be - odds are you could avoid those "false negatives" with some refinements to the algorithm (and, indeed, the slightly more complex in-ear model has fewer of them).

You're very unlikely to build a perceptual model that's perfect in predicting human responses to complex stimuli, IMO - but one which can look at FR and spit out a generally not-too-far-off ranking of headphones is quite an innovation! An imperfect but still significantly better than guessing (and, if you want my two bits, probably better than "dead reckoning" based on plots) model is a useful starting point for future development!

I can understand that you see it as harsh, particularly given that I don't think that we know of any other predictive model that has been quite as well detailed in public, but the more I look at the underlying data, the more I believe that I'll stand my ground on this issue. I'm a little too occupied right now to write a longer post, but if you can wait a while, I think that I've found a way to explain in a visual way some of my misgivings about that model and to dig a little bit deeper into the rather complex interplay between predicted error (which distribution is not a question of a few outliers vs the rest BTW), delta between two predicted scores, the influence of neighbouring samples, etc.

One simple question that I would raise right now, that I hinted at when I deliberately used the rather cumbersome "doing a lot of predicting" syntax, is : since predictive models have a performative purpose, should the quality of a predictive model be evaluated simply by looking at its goodness of fit statistics, or by its capacity to actually "do" work, ie predict outcome ? In which case, what are we expecting such model to be able to predict, and what happens when we actually put the model to work ?

 

[IAtaman]

Crinicle's take on the topic.

 

[Jeromeof]

Playing around with @crinacle new Database and I must admit I like the Bass and Tilt options -

I could imagine a time when we could easily autoEQ to the diffuse field and have these 2 simple adjustments for preference - then I suppose if those adjustment preferences where "shared" to a central data gathering "authority" you could start to how a new best sounding "average" curve.

 

[Mad_Economist]

This is a pretty hefty accusation, and, paralleling my responses to folks like Jude when they've made such accusations against you, I've got to ask, is that really congruent with all that we know about Jude's interest in and involvement in measurements? It's certainly an easy narrative, and it makes things very tidy, but so does the (quite comical) narrative that you sabotaged Schiit measurements because you didn't like the firm - except that wasn't true.

Had someone DM me asking if I was saying Jude had accused Amir of sabotaging Schiit, so let me clarify on this one, that's two separate thoughts that got spot welded in composition:

1: Jude has (in my view rather comically) alleged that Amir is NwAvGuy - this is clearly not the case, because we know what Amir was up to during the era where NwAvGuy was posting (WhatsBestForum and Hydrogen), and he isn't an engineer. You can make a much more reasonable case that I'm NwAvGuy than that Amir is.

2: Posters on Head-Fi and Reddit have spuriously claimed in the past that Amir sabotaged Schiit measurements. This was never evidenced, and generally Amir's data matched Schiit's spec, so it was pure slander from a position of ignorance. Jude never did this, and I'd be extremely surprised if he ever stooped that low.

 

[IAtaman]

The politics and drama of measuring headphones - who would have thought!

 

[Robbo99999]

@Mad_Economist , seeing as you've posted up today, anymore news on how you folks are progressing?

 

[Mad_Economist]

@Mad_Economist , seeing as you've posted up today, anymore news on how you folks are progressing?

Unfortunately, the news is that there is no news yet - we're working on a bunch of projects simultaneously, and this one isn't top priority at least until next month.

 

[Robbo99999]

Unfortunately, the news is that there is no news yet - we're working on a bunch of projects simultaneously, and this one isn't top priority at least until next month.

Yeah, fair do's, let's check back later.

 

[Jimbob54]

I'm NwAvGuy

No, I'm NwAvGuy

 

[MrBrainwash]

Without beating around the bush: I created a preset for DT990 pro 250ohm. I'm happy with the result. It was a barbaric approach, but it worked. I equalized the DT990 to match the HD600 using the crinacle tool and added a preset for the HD600, tuned to TheHeadphoneShow's target. Something didn't quite click for me initially, but then I used the leftover 3 kHz filter and subtracted 2.60 dB, which solved the issue and gave me the desired sound. In the end, I just gently boosted the sub-bass by 2.20 dB, which was lacking a bit. The effect was a nice, a finishing touch, like the cherry on top.

To achieve complete satisfaction, I also enable Sonic Headphones for Windows. This makes the sound less harsh and enhances the spatial quality. It might not be necessary for the DT1990, but the acoustics here aren't particularly great, and the treble still tends to be a bit too sharp. If a specific album already has good acoustics and a softer sound, I turn this option off.

Without the correction at 3 kHz, the sound was too piercing, and that tiny detail was disrupting my perception. I had the impression that there wasn't enough bass, even though theoretically, it shouldn't have been the case. This situation once again illustrates how significant small things are in critical frequency ranges. A similar situation occurs with the sub-bass; we might not consciously hear it, but adding those ~2 dB somehow alters the perception of everything above it. I don't know why it happens, but it feels cleaner and more realistic to me.

 

[amirm]

1: Jude has (in my view rather comically) alleged that Amir is NwAvGuy - this is clearly not the case, because we know what Amir was up to during the era where NwAvGuy was posting (WhatsBestForum and Hydrogen), and he isn't an engineer. You can make a much more reasonable case that I'm NwAvGuy than that Amir is.

What do you mean I am not an engineer? I grew up with electronics being my hobby and went on to get my degree in electrical engineering. Professionally I haven't had to do hands on design but have managed teams of hardware designers building computers and VLSI ICs going into them. I also spent four years fixing audio and RF products so engineering is in my blood.

I also managed hardware and software video products where I think NWAVGUY worked. So there is commonality between us. But headfi folks didn't know any of this and just equated me with another technical guy they hated.

 

[iamsms]

Unfortunately, the news is that there is no news yet - we're working on a bunch of projects simultaneously, and this one isn't top priority at least until next month.

Any chance of you guys publishing (either on forum or youtube) some amp/dac blind testing results? Resolve and a few other headphones.com folks (not you) have made some (imo) outrageous claims about amps and dacs in the past, and I think many would like to see them try to validate their past claims. And they have shown the blind testing device in one of the videos in the past.

 

[amirm]

Crinicle's take on the topic.

What a mess....

 

[Mad_Economist]

What do you mean I am not an engineer? I grew up with electronics being my hobby and went on to get my degree in electrical engineering. Professionally I haven't had to do hands on design but have managed teams of hardware designers building computers and VLSI ICs going into them. I also spent four years fixing audio and RF products so engineering is in my blood.

I also managed hardware and software video products where I think NWAVGUY worked. So there is commonality between us. But headfi folks didn't know any of this and just equated me with another technical guy they hated.

Perhaps I should have said "practicing electrical engineer"- the note I was aiming to hit there was fairly identical to what I understand to be your official (and quite reasonable) statement on the matter: "Furthermore, anyone with any technical background reading his contributions and mine see that we are very different people. He has much more hands on electronic design experience than I do for example. On the other hand, I have much more experience in other fields such as signal processing, acoustics, computer technology, etc. And of course on the business and industry side."

Any chance of you guys publishing (either on forum or youtube) some amp/dac blind testing results? Resolve and a few other headphones.com folks (not you) have made some (imo) outrageous claims about amps and dacs in the past, and I think many would like to see them try to validate their past claims. And they have shown the blind testing device in one of the videos in the past.

Good question! @Resolve has only had negative results with the ABX tester, which I believe he's publicly referenced but which I'd love to do a video or short on! Cameron has had some interesting results (e.g. a >.001% blind matching of some filters in digital audio), which is part of what's delayed us - I really don't want to publish a false positive and it takes some doing to debug blind testing across an ocean. Expect stuff in this vein this year for sure though!

 

[Robbo99999]

Without beating around the bush: I created a preset for DT990 pro 250ohm. I'm happy with the result. It was a barbaric approach, but it worked. I equalized the DT990 to match the HD600 using the crinacle tool and added a preset for the HD600, tuned to TheHeadphoneShow's target. Something didn't quite click for me initially, but then I used the leftover 3 kHz filter and subtracted 2.60 dB, which solved the issue and gave me the desired sound. In the end, I just gently boosted the sub-bass by 2.20 dB, which was lacking a bit. The effect was a nice, a finishing touch, like the cherry on top.

To achieve complete satisfaction, I also enable Sonic Headphones for Windows. This makes the sound less harsh and enhances the spatial quality. It might not be necessary for the DT1990, but the acoustics here aren't particularly great, and the treble still tends to be a bit too sharp. If a specific album already has good acoustics and a softer sound, I turn this option off.

Without the correction at 3 kHz, the sound was too piercing, and that tiny detail was disrupting my perception. I had the impression that there wasn't enough bass, even though theoretically, it shouldn't have been the case. This situation once again illustrates how significant small things are in critical frequency ranges. A similar situation occurs with the sub-bass; we might not consciously hear it, but adding those ~2 dB somehow alters the perception of everything above it. I don't know why it happens, but it feels cleaner and more realistic to me.

You're a good guy @MrBrainwash but I don't think this is particularly relevant to the thread. What's the relevance to B&K5128?

 

[amirm]

2: Posters on Head-Fi and Reddit have spuriously claimed in the past that Amir sabotaged Schiit measurements. This was never evidenced, and generally Amir's data matched Schiit's spec, so it was pure slander from a position of ignorance. Jude never did this, and I'd be extremely surprised if he ever stooped that low.

Jude's position was that I must not know what I am doing. And that he does because he called Audio Precision for help to replicate my measurements. That aside, the issue was simple: post my measurements and finding quantization error (truncation from 24 to 20 bit resolution of the DAC), schiit fixed the firmware and handed a unit with that fix to Jude. Jude got different results and turned it into my not being competent. Whether Jude knew this had happened or not, I don't know. What I do know is that when I found the problem in a second Yggy, the owner sent the unit to Schiit. And first thing Schiit did was update the firmware before testing it. And then declared measurements matched Jude's and not mine. Anyway, none of it was above board.

Fortunate Schiit did a reboot and adopted objective measurements as a tool to build better products, proceeded to send me new products for review and one of their people participates here. Pretty sure they would not do this if they thought I had cooked measurements to make them look bad.