Master Complaint Thread About Headphone Measurements

[Deleted member 16543]

Where it is supposed to be more accurate is where there is most variation. So the extra accuracy is not usable. I evaluated it for a few months: https://www.audiosciencereview.com/forum/index.php?threads/headphone-measurements-using-brüel-kjær-5128-hats.15352/

The biggest issue with it is that there is no research target curve for it. Without it, you are completely in the dark. So you get a response with ups and downs. Now what? It is like being lost in a jungle without a compass. With no reference you don't know which is north. This is why I did not purchase it even though the company was wonderful to work with and highly supportive.

I think 5128 is useful in research where you can build fixtures and such to have repeatability for the same headphone. For reviewing many headphones that is not practical and you are just stuck with all the negatives of the fixture and none of its pluses.

I totally understand about the target curve.
That's the difficult part, but also the most rewarding, I think.
This all ties up in a big circle to me and the work I'm doing.
I started with building a binaural microphone, but then I needed to equalize it.
Equalize to what? I needed to find out what a binaural microphone measures at the listening position of a well balanced pair of speakers.
To do that I needed very balanced speakers in quasi-anechoic listening conditions (still working on that although I have been using a working prototype for years now).
For music recordings I made do with off the shelf speakers and off the shelf EQ plugins, but that obviously won't do for precision measurements.

In all of this my main goal shifted from making the most accurate binaural microphone possible to making the most accurate pair of speakers possible (I am much more of a music listener than of a recordings maker, after all).
When the speakers will finally be ready it won't be much of a problem for me to measure them with my binaural microphone and share the target.
It will be a personalized (to the specific binaural microphone) target, but I think still educational to compare with other target curves of other rigs.

 

[zenki]

@amirm IEM when?
. lolz

 

[DarthShader]

Are there any plans to do more CSD / Waterfall measurements? In practice, these charts make it easy to tell how "fast" a headphone would sound, and if there are any sibilances or resonances.

 

[solderdude]

When the speakers will finally be ready it won't be much of a problem for me to measure them with my binaural microphone and share the target.
It will be a personalized (to the specific binaural microphone) target, but I think still educational to compare with other target curves of other rigs.

Fun idea.
The correction curve of Harman is 'averaged' and also includes an average of tonal preferences obtained through research. It will not resemble yours most likely. Certainly when it concerns the lows unless you use the same basic filters used for determining bass level preference.
It is also why the correction is totally off from 10kHz and higher and won't fit all headphones. All headphones will excite pinna resonances differently.

In your case you only want to create a speaker that measures acc. to the target curve you created based on 1 or more speakers in a specific room and distance (HATS/speaker) and position. Move the darn thing a meter and all will differ yet again.

Getting the reference right is the difficult part. certainly when using various unknown bandfilters in front of the micrphone that are also differing in directional sense.

 

[Deleted member 16543]

Fun idea.
The correction curve of Harman is 'averaged' and also includes an average of tonal preferences obtained through research. It will not resemble yours most likely. Certainly when it concerns the lows unless you use the same basic filters used for determining bass level preference.
It is also why the correction is totally off from 10kHz and higher and won't fit all headphones. All headphones will excite pinna resonances differently.

In your case you only want to create a speaker that measures acc. to the target curve you created based on 1 or more speakers in a specific room and distance (HATS/speaker) and position. Move the darn thing a meter and all will differ yet again.

Getting the reference right is the difficult part. certainly when using various unknown bandfilters in front of the micrphone that are also differing in directional sense.

It will be based on a speaker's target curve as described here.
Then, by placing the binaural microphone where the speaker measurement mic was, one can derive the correspondent response at the binaural mic's capules.

The difference in the bass region will be there, as a consequence of how much the Harman curve differs from the one used in that text, but it will be more or less the same for every headphone. By comparing the 2 different target curves and how a few pairs of headphones measure on both rigs, it should be possible to derive the fixed part of the target curve difference in the bass region, underneath the difference variable part due to headphones specific responses.
One could then account for that when normalizing the two different measurements from the two rigs.
It would be better if exactly the same pairs of headphones were measured, to take the production tolerance variable out of the equation.

Where this could be more educational and useful, in my opinion, is in the highs.
Aside the 10 kHz+ region, there is a potential to still get an extra octave of accurate target curve to compare headphones to.

I think that b&k missed an opportunity by releasing their 5128 microphone without some extra background work, which would make it a rig for increased accuracy in measurements, but at the same time make it possible to compare it to the previous closest rig out there, for which tons of measurements have already been made.
That and the ridiculous price is going to be a tough obstacle to make the 5128 become the standard measurement rig that I personally think it deserves to be.

 

[solderdude]

but it will be more or less the same for every headphone.

That part is not true in reality.
One could find 2 headphones that measure quite similar on rig A but much less similar on rig B while both rigs would measure very similar on a particular speaker in the exact same room/conditions (after being calibrated for this).

It is not headphone response unevenness but the way that headphone couples to a particular fake ear and fake ear canal which in itself differs from those of others. Of course there will be some folks who's ears are roughly the same as the fake ones. There will also be ones that differ.

 

[Deleted member 16543]

That part is not true in reality.
One could find 2 headphones that measure quite similar on rig A but much less similar on rig B while both rigs would measure very similar on a particular speaker in the exact same room/conditions (after being calibrated for this).

It is not headphone response unevenness but the way that headphone couples to a particular fake ear and fake ear canal which in itself differs from those of others. Of course there will be some folks who's ears are roughly the same as the fake ones. There will also be ones that differ.

I wasn't talking about that. The only measurement rig I refer to when measuring speakers is a simple measurement mic like ECM8000 or equivalent. Earthworks if one wants to be fancy.

Once the speakers have been equalized through DSP to a certain target curve (that is valid for a rig like ECM8000 and the likes and only for that types of mic), then a secondary rig, intended for headphone measurements, is placed where the ECM8000 was. I expect that response to vary depending on what type of secondary rig one uses to measure the speakers again to come up with a target curve intended for headphones on that specific rig.

As my discussion was limited, as far as this secondary rig goes, to b&k 5128 vs. GRAS 45CA with anthropomorphic pinnae+canal, I expect the measurements in the lows to differ by a quantity that is more or less constant among different pairs of headphones (plus the actual difference due to headphone bass frequency response on top of that, of course).
But yes, I wouldn't make the same assumption with rigs that are considerably different geometrically, though. Like the 5128 vs. a capsule that sits flat on a plane, like I seem to remember you use, if I'm not mistaken.
Or the 5128 vs. 45CA without the pinnae.

 

[solderdude]

I wasn't talking about that. The only measurement rig I refer to when measuring speakers is a simple measurement mic like ECM8000 or equivalent. Earthworks if one wants to be fancy.

Once the speakers have been equalized through DSP to a certain target curve (that is valid for a rig like ECM8000 and the likes and only for that types of mic), then a secondary rig, intended for headphone measurements, is placed where the ECM8000 was. I expect that response to vary depending on what type of secondary rig one uses to measure the speakers again to come up with a target curve intended for headphones on that specific rig.

That will give you an opportunity to measure that speaker's response. You can correct the output of that HATS so it measures the same as the speaker with normal mic. (I assume gated ?).
This basically is how HATs are calibrated but then in anechoic or specified conditions.
When you have calibrated your DIY HATS under your room conditions doesn't mean it will measure the same in nearfield, diffuse field and anechoic conditions. It will give a correction for only that HATS in that room.

A headphone couples completely different than speakers in any room, so that calibration is of little use but are kinds of calibration one can obtain.
This is where Harman research differs.

Yes, differences between BK5128 and 45CA are there. Amir has measurements of the same headphone done on 5128 and 45CA so you can compare. Also there are others that have compared the rigs. With headphones also seal comes into play for low frequencies. because of shape differences between HATS and 45CA fixture this too can lead to differences.

measuring transducers in all kinds of circumstances is a bitch. It pays to have standards but not all standards are equally good and relevant.

 

[Deleted member 16543]

That will give you an opportunity to measure that speaker's response. You can correct the output of that HATS so it measures the same as the speaker with normal mic. (I assume gated ?).
This basically is how HATs are calibrated but then in anechoic or specified conditions.
When you have calibrated your DIY HATS under your room conditions doesn't mean it will measure the same in nearfield, diffuse field and anechoic conditions. It will give a correction for only that HATS in that room.

A headphone couples completely different than speakers in any room, so that calibration is of little use but are kinds of calibration one can obtain.
This is where Harman research differs.

Yes, differences between BK5128 and 45CA are there. Amir has measurements of the same headphone done on 5128 and 45CA so you can compare. Also there are others that have compared the rigs. With headphones also seal comes into play for low frequencies. because of shape differences between HATS and 45CA fixture this too can lead to differences.

measuring transducers in all kinds of circumstances is a bitch. It pays to have standards but not all standards are equally good and relevant.

We have a starting point for well balanced sounding speakers that I assume to be accurate and good enough: a target curve that is flat up to 1 kHz and then slopes down linearly to -6 dB @20 kHz, as measured according to psychoacoustics principles (so yes, gated, of course) by a standard measurement microphone like the ECM8000.

What type of pressure do speakers that measure in such way impart on the listener's eardrums?
We need a binaural microphone to find out. One that is geometrically as close as possible to the boundaries that the listener's eardrums see in reality when listening to speakers (canal, pinna, head), since those boundaries affect the pressure at the eardrum itself.
This is the reason why I think the 5128 is better than the 45CA with pinnae: it has an anthropomorphic head that would allow for a more precise measurement of that pressure, and therefore finding a target curve that is more meaningful.
But for 40 grand I would expect this work to be done by B&K already, not having to do it myself.

Since both speakers and headphones during listening are pretty much at a standard configuration in respect to the listener's head (give or take some variance in positioning), all we need to do is put the binaural microphone at the sweet spot of the speakers, measure (once again according to those psychoacoustics principles) and get our target response like that.
We don't need to be concerned about free vs diffuse field or anything like that.
However, measuring speakers in quasi-anechoic conditions would be strongly advisable for more precise measurements independent of the room.
Only if one wants to make binaural recordings would they need to equalize the binaural microphone. For headphones measurements there's no need for that.

Leaks aside, in the bass region, for rigs that are similar to each other from the point of view of the pressure produced at their capsules by headphones, such as the 2 rigs in question (as opposed to pressure produced by speakers, for which the shape of the entire head makes a difference), there will be a difference in the lows for the same headphone pair that will be the same for other headphone pairs.
Actually, capsule response difference aside (and assuming proper seal), for these two specific rigs in question I don't see any reason to expect much of a difference in measurements at all, as far as the bass response goes.
The main difference will be in the highs, and it's there that I expect the 5128 to have the edge... if only they provided a target curve measured as explained above or similarly.

And of course above 10 kHz there are all the other problems due to headphone placement that make it very tricky to get meaningful measurements without doing a lot of measurements and trying to find the variance in response due to the headphones and extract it out of the variance due to the different placement.
But below 8-10 kHz the measurements would be as good as they can possibly be, keeping in mind that an averaged canal geometry is definitely not 100% accurate for everybody but only the next best thing, short of an individual measurement of the pressure at the listener's eardrums (which is not possible because inserting a mic probe in the canal alters the geometric boundaries too much anyway).

 

[solderdude]

What type of pressure do speakers that measure in such way impart on the listener's eardrums?
We need a binaural microphone to find out. One that is geometrically as close as possible to the boundaries that the listener's eardrums see in reality when listening to speakers (canal, pinna, head), since those boundaries affect the pressure at the eardrum itself.
This is the reason why I think the 5128 is better than the 45CA with pinnae: it has an anthropomorphic head that would allow for a more precise measurement of that pressure, and therefore finding a target curve that is more meaningful.
But for 40 grand I would expect this work to be done by B&K already, not having to do it myself.

For measuring speakers the 5128 is suited, for measuring headphones the 45CA seems a more logical choice. The fact that results differ says enough.

Yes, HATS and everything that measures calibrated and professionally is screamingly expensive and for good reason.
The research and low sales volumes need to be paid.

Since both speakers and headphones during listening are pretty much at a standard configuration in respect to the listener's head (give or take some variance in positioning), all we need to do is put the binaural microphone at the sweet spot of the speakers, measure (once again according to those psychoacoustics principles) and get our target response like that.
We don't need to be concerned about free vs diffuse field or anything like that.

Harman and others have gone before sing professional equipment. They will have started out on the same premise. No need to reinvent the wheel unless you want to do it cheap and in your own time and get fairly accurate results.
I am all for that... in fact made my rig on a similar-ish way.

Leaks aside, in the bass region, for rigs that are similar to each other from the point of view of the pressure produced at their capsules by headphones, such as the 2 rigs in question (as opposed to pressure produced by speakers, for which the shape of the entire head makes a difference), there will be a difference in the lows for the same headphone pair that will be the same for other headphone pairs.

I have done tests that baffled me. 2 microphones (different brand, both electret) exact same rig, exact same headphone. NO pinnae, ear canals.
Measured headphone A with mic X and with mic Y. Calibrated mic Y so they match FR completely.
Now measured headphone B with both mics... small difference here and there.
measured headphone C .. substantial difference in the bass. Quite a few dB. ... hmmm.. measured A again on both mics.. overlapping. re-measured B... tiny difference. Measured C again... substantial difference again.
So... no it isn't as easy as one might think. Listening to the headphones with music using deep sub lows told me mic X gave results closer to each other when EQ'ed acc. to mic Y.

But below 8-10 kHz the measurements would be as good as they can possibly be, keeping in mind that an averaged canal geometry is definitely not 100% accurate for everybody but only the next best thing, short of an individual measurement of the pressure at the listener's eardrums (which is not possible because inserting a mic probe in the canal alters the geometric boundaries too much anyway).

5128 and 45CA differ substantially between 2kHz and 10kHz depending on the headphone. Surely one of them must be closer to a reality.
To evaluate that one has to have both and a good reference. Both are excellent for their usage but will not give the same results.
Someone can tell me this or that one is based on this or that research. I will never know. Not going to buy them nor play with them and then publish a nice report/paper about them.

I would rather use the time to enjoy music.

 

[Deleted member 16543]

Neither the 5128 nor the 45CA are used to measure speakers. Speakers are measured with different types of microphones altogether.
That seems to work and I take it at face value. I could try to justify it by what effect a room has on soundwaves, but as I said, I take it a face value and I don't want to reinvent that wheel.

The value of the 5128 is in finding out a response that is closer to the actual pressure sensed by the eardrums of a listener sitting in front of well balanced speakers. Finding that response means having a more reliable target response to equalize headphones to (provided one uses the same 5128 to also measure headphones). You may call it reinventing the wheel, I call it ongoing research.

The difference of headphones C between the two rigs you tested is indeed unexpected. What kind of rig and specific capsules did you use? If the difference between the two measurement rigs is only in the capsules, the only thing I can think of that would explain it is if one of the capsules is directional and different headphones excite it in different ways. Maybe a less than perfect seal in the capsule housing?

5128 and 45CA difference in measurements is to be expected. They have a different shape, and the length of the canal is also different. My money would be on the 5128 being closer to an actual average canal than the GRAS pinnae.

 

[solderdude]

The 5128 is designed for acoustic measurements, not specifically for headphones. More suited for measuring speakers.
The 45CA is specifically a headphone test fixture and not designed for acoustic measurements, not designed for measuring speakers.

The 2 capsules were used in my FP rig. A WM61 (original, not fake Chinese one) and an IMM6.
There was no leakage or less perfect seal and were in the same physical position. The WM61 is much smaller so after the imm6 was used I have to make a small adapter to fit it back in. This was not an issue as the WM61 A measured the same in the adapter as before (a small plastic bushing made to size)

 

[amirm]

The value of the 5128 is in finding out a response that is closer to the actual pressure sensed by the eardrums of a listener sitting in front of well balanced speakers.

What room are you going to use for that? The only one that makes sense is an IEC compliant one which is what Harman has done for 45C calibration. No such thing exists for 5128 and you are not going to be in a position to create your own.

 

[amirm]

@amirm IEM when?
. lolz

I have to get into them but just procrastinating.

 

[amirm]

My money would be on the 5128 being closer to an actual average canal than the GRAS pinnae.

Being closer to average doesn't do you any good if you can't correlate that with listener preference. The goal here is not to just spit out some graphs. It is a means to an end of finding out if a headphone is good sounding or not. 45C is doing a superb job for me after testing a number of headphones. I see no reason for anything else and randomness that brings.

 

[amirm]

Are there any plans to do more CSD / Waterfall measurements?

No. Audio Precision charges extra to spit out the stupid graph! And stupid it is because the peaks of the frequency response tell you the same thing that waterfall shows. This is why I don't show them for speakers either even though my software there generates them.

 

[MZKM]

I have to get into them but just procrastinating.

That is probably going to be more challenging in terms of getting a good fit and having accurate bass response. We’ll have to wait and see.

If only headphones/IEMs were as “simple” to measure as speakers in terms of repeatability.

 

[Deleted member 16543]

The 5128 is designed for acoustic measurements, not specifically for headphones. More suited for measuring speakers.
The 45CA is specifically a headphone test fixture and not designed for acoustic measurements, not designed for measuring speakers.

The 2 capsules were used in my FP rig. A WM61 (original, not fake Chinese one) and an IMM6.
There was no leakage or less perfect seal and were in the same physical position. The WM61 is much smaller so after the imm6 was used I have to make a small adapter to fit it back in. This was not an issue as the WM61 A measured the same in the adapter as before (a small plastic bushing made to size)

Not sure where you get the idea that the 5128 is less apt for headphones measurements than the 45CA. While I wouldn't use the 45CA to find out the target response according to the procedure I described earlier (it distorts the HRTF too much due to the lack of the head), I don't see any problem with using the 5128 for headphone measurement. The capsules see the same geometry, and what's outside the headphones (head or no head) doesn't matter in that context.
So the 5128 is useful for both (but you don't really need it for speakers anyway because we already know what to aim for by using a simple $100 measurement mic), while the 45CA only for headphone measurements.
The problem is that I really wouldn't trust the target curve relative the 45CA, because of the HRTF distortion.
But on the other hand the 5128 doesn't even have a target response and it would need to be found through a quite tedious procedure, to make its headphones measurements meaningful.

As for your test, I'm really intrigued. I'm not an expert on omni capsule response in tight geometric quarters, but if there is one on the forum I'd like to hear their take. I see no reason whatsoever why the bass response would be different by only changing the capsule, if the chamber geometry remains unaltered.

 

[Deleted member 16543]

What room are you going to use for that? The only one that makes sense is an IEC compliant one which is what Harman has done for 45C calibration. No such thing exists for 5128 and you are not going to be in a position to create your own.

Either that or use speakers in the very near field (reflections below 15-20 dB for most of the spectrum) that have a response that matches the target response I described earlier. Room treatment and big room would help, but this is something that is not too far fetched to achieve in any regular sized room of anybody's house, in principle. The real problem is that speakers that allow for this kind of measurement are not something one can buy off the shelf.

 

[Deleted member 16543]

Being closer to average doesn't do you any good if you can't correlate that with listener preference. The goal here is not to just spit out some graphs. It is a means to an end of finding out if a headphone is good sounding or not. 45C is doing a superb job for me after testing a number of headphones. I see no reason for anything else and randomness that brings.

Exactly. What B&K doesn't provide you (after spending a whopping $41k!) is a target to correlate the response to. They could have done their homework better. It's like they did the difficult part of the job (making a product that is arguably the closest to a head replica that one can find, even bothering to match the eardrum impedance!) and then half assed the final part of providing a target that would let people know how to use it for headphones measurements.. weird.
Again, for that kind of money I would expect a plug and play product. Maybe they'll fix it sooner rather than later. And at that point I bet it would be a tool that even better correlates measurements to listeners preference. Whether the added correlation is worth paying 3 times as much probably has a different answer depending on who you ask.