[Resolve]

I agree with you that good driver symmetry should just be a "pass/fail" kind of thing. But I've found that's not quite the way things work in reality at the lower end of the price scale. And the only way to really change that circumstance is to educate people on the issue, through reviews (and discussions like this one), so that the manufacturers do a better job of addressing the issue in their products at all price points. This is my opinion anyway.

Yeah, I'm genuinely open to the idea because of this. I think in an ideal world with all the time and resources available, I would want to be able to show it in general. Like, maybe not by default, but kind of like how rtings does it. And in situations where the tolerances are exceeded, just have it not hit the database until we get another sample. There is also another positive element to that in that it puts pressure on brands to make sure their QC is good. So there's like a... 'greater good' aspect to showing it I suppose.

 

[ADUHF]

A graph that shows the average response of the left and right channels together as a single curve has its uses too, Resolve.

If you're using just a single EQ curve to correct the FR on both channels of headphone (which is what I currently do for my DT-770's), then all you really need for making those sort of general tonal balance adjustments is the average FR of the left and right channels.

So there can be some value in having both types of plots. Or in having an average of the left and right FR, and then maybe a separate graph (like the one in my previous post) that shows how the two channels deviate from that average response.

All of the above info can potentially be useful. Particularly to geeks like me.

If you asked me what are the top two pieces of info that I can use about a headphone's sound quality, which are actually measurable and quantifiable on a graph, then the first item on that list would be the headphone's overall FR and tonal balance. And the second item, not far behind that, would be its left-right driver symmetry. Because (unit variations & measurement issues aside) that will potentially tell me alot about the accuracy and precision of a headphone's stereo imaging.

There are other characteristics that I'll also look and listen for as well, such as clarity/distortion, extension, speed & impulse response, and so forth. But FR, and driver symmetry are both way up near the top of the list!

 

[Resolve]

And the second item, not far behind that, would be it's left-right driver symmetry.

Yeah this is near the bottom of my list of interests, because for me it's either good or unacceptable. There's not much middle ground in my opinion. Averaging for good results is possible and then you'd end up with something useful, but again when you have some imbalance going on it's more misleading than representative. And in that case... you end up needing another sample anyways - or at least I would ask for it. Once again though, I'm open to showing it, as long as it's not the default representation of the headphone's FR. I think that's really the biggest concern.

 

[ADUHF]

By not including the driver symmetry info though, what you're effectively doing is asking your readers to put all of their faith in your ears on that issue. Which is asking alot, imo, for some reviewers. And also unnecessary.

If you provide some sort of a graph that actually shows the differences in the two channels somewhere in your reviews though, then what that potentially does is provide an additional layer of objective and supportive information to back up any subjective impressions or observations that you may have to offer re a headphone's symmetry and stereo imaging capabilities. Which would only serve to further bolster your credibility on this subject.

 

[Resolve]

By not including the driver symmetry info though, what you're effectively doing is asking your readers to put all of their faith in your ears on that issue. Which is asking alot, imo, for some reviewers. And also unnecessary.

This is Crin's argument on it. It's a trade-off in my opinion. Again, this isn't a hill I will die on, I just think there's a lot of responsibility that comes with representing this stuff, and at the moment, with the current resources, I think it's safer to just ask for another unit when it's bad. I do think eventually I'll post that data too, but that will require the right way of representing it - again I like how rtings does it.

 

[jude]

By not including the driver symmetry info though, what you're effectively doing is asking your readers to put all of their faith in your ears on that issue....

Actually, @ADUHF, I have what is not so much a counter to what you're saying, as much as it is something else to consider.

While the readers/viewers would be expected to put their faith in you in this specific regard, they'd also need to put some faith in you and your equipment to be able to properly tease out and show symmetry. Even if a pair of headphones has perfect left-right symmetry, it does require both the instruments and methods to show that reliably, consistently. Even the slightest changes in position -- especially in the higher frequencies, but also at times in other parts of the audio band -- can show significant changes. Also, there's the importance of making sure the measurement system -- from end to end -- is as perfectly calibrated as possible.

It's one thing if we could reasonably expect people who've never done a headphone measurement to understand this, but most will just interpret any lack of perfect symmetry as being on the side of the headphone (which it may be), not the measuring process (which it also may be, but a variable often (and understandably) overlooked, especially by those who've not yet measured a headphone).

Regarding the equipment:

It definitely helps if you can measure both sides at the same time. Again, assuming a given headphone is perfectly symmetrical, that symmetry can be teased out measuring one ear at a time, but it would be much more challenging to do this way. It would also be much more challenging to do precisely one ear at a time over several seatings, absent the ability to monitor both channels in real time during placement.

Also, some of the more popular one-ear-at-at-a-time setups use GRAS KEMAR pinnae. While the latest KEMAR pinnae are available in symmetrical pairs, they're symmetrical from the pinnae flanges into the canals, but there's some asymmetry at the pinnae bases to accommodate for KEMAR's slightly asymmetrical head shape. While this shouldn't be an issue measuring in-ears, it could lead to asymmetry when measuring over-ears that do engage those differences in the pinnae bases. (I would expect these differences to be small, but, nevertheless, differences that result from this would certainly be attributed to the headphones by most readers.)

One advantage to the one-ear-at-a-time method is it's not as expensive, and you're guaranteed symmetry with the simulator itself (independent of the pinnae base differences mentioned above), as (obviously) you'd be using the same simulator/coupler for both sides. That said, at the level of precision (that you pay mightily for) of the type of simulators made by Brüel & Kjær and GRAS, you can expect symmetry in a pair. Keep in mind, though, that level calibration will still be necessary, as sensitivity varies even in shipped simulator pairs (even though their transfer functions are identical). In addition to the price of the additional simulator, then, there's the price of something like a piston phone to consider (which can also be quite expensive).

Regarding the method(s):

Even if you have the symmetry of the simulators accounted for, one must also keep in mind the methods (and instrumentation) necessary to take advantage of a hypothetically perfectly left-right symmetrical headphone. It helps to be able to see the frequency response (or some approximation of it) in real time for both channels as you're positioning the headphones.

We use an Audio Precision analyzer (with its APx software) which helps a great deal with that. With APx we can use a low frequency square wave during placement (which is not novel). What is novel is that APx lets us monitor at least three key things at once during placement:

• The RMS level of the low frequency square wave (to help confirm a seal);
• The view of the waveform in scope view (again, to help confirm a seal, and to give us one more thing to get symmetrical);
• And crucially it also gives us a real-time (downward-slanting) approximation of the frequency response of both channels at the same time in the FFT Spectrum Monitor window. NOTE: It helps to reduce the FFT length (2000 points is usually good) to force contours instead of spurs (of the odd-order harmonics), and also to engage a little bit of averaging to settle things down a bit. If you understand what square waves are, then you won't be surprised that the frequency response appears downward-slanted when doing this.

While it's a simple method, I had not seen it used before, and presented a slightly different version of it at the ALMA (now ALTI) annual conference back in 2018 to a roomful of engineers (from Harman, Brüel & Kjær, Audio Precision, GRAS, Knowles, and many other companies). I also showed it in a video here back then, too. Since then I simplified it a bit and have again shown it to many engineers in the industry with positive feedback.

Long story short, I pretty much agree with what @Resolve is saying about this.

Let's take a look at a headphone (the Dan Clark Audio Stealth) that has excellent symmetry, but whose excellent symmetry might be challenging to tease out measuring one channel at a time, or with a less than perfectly calibrated two-ear setup, or with a method/setup that doesn't allow for monitoring frequency response in real time.

Fig.1 (above): Dan Clark Audio Stealth, Seat 01, Measured 37 (unsmoothed).​

​

Fig.2 (above): Dan Clark Audio Stealth, Seat 02, Measured 44 (unsmoothed).​

​

Fig.3 (above): Dan Clark Audio Stealth, Seat 03, Measured 52 (unsmoothed).​

​

Fig.4 (above): Dan Clark Audio Stealth, Seat 04, Measured 56 (unsmoothed).​

​

Fig.5 (above): Dan Clark Audio Stealth, Seat 05, Measured 59 (unsmoothed).​

​

​

Fig.6 (above): Dan Clark Audio Stealth, five-seat average (smoothing = 1/12 octave) versus the five individual seatings (unsmoothed). The smoothed average is what appeared in our video about the Stealth.​

Fig.7 (above): Dan Clark Audio Stealth, five-seat average (unsmoothed) versus the five individual seatings (unsmoothed).​

As you can see in Seat 03 (Measured 52) in Figure 3 above, our left-right symmetry wasn't as tight as it was in the other seatings. Still, we found this a reasonable seating, and it was used in the average. If we published that seating alone, I can assure you most would conclude the right channel of the tested Stealth has more treble than the left. You can see from the other seatings that would be an unfair conclusion.

I know a couple of people found our measurement notes interesting (when I posted them in at least one other thread), so here they are for these seatings:

Measured 37: DCA Stealth FINAL - 5128 - RAW - 390.7 mVrms WN80 - seat 01 (brian) - test 01 - cups maybe slightly back of center - went back to 0 dBRG after last measure (Measured 36)​ ​ Measured 44: DCA Stealth FINAL - 5128 - RAW - 434.0 mVrms WN80 - seat 02 (jude) - test 02 - pulled forward from previous seating - pushed cups a touch vertically up​ ​ Measured 52: DCA Stealth FINAL - 5128 - RAW - 374.0 mVrms WN80 - seat 03 (brian/jude) - test 02 - neutral-back seating (brian's 5128 photo seating, jude squeezed it tighter)​ ​ Measured 56: DCA Stealth FINAL - 5128 - RAW - 402.2 mVrms WN80 - seat 04 (jude) - test 02 - neutral seating, maybe down a bit vertically from previous seating​ ​ Measured 59: DCA Stealth FINAL - 5128 - RAW - 404.4 mVrms WN80 - seat 05 (brian) - test 01 - neutral seating​

Anyway, again, I pretty much agree with what @Resolve is saying about this. There's definitely more to consider than might be obvious, even with something as seemingly straightforward as left-right channel symmetry.

2021-08-23 1724 EDT EDIT: I should have probably clarified (since we've not shown this before) that when I show "80 dBSPL white noise" in the measurements, we're using white noise to set the voltage level, after which we run a sweep to get the frequency response. I'll try to make this clearer going forward. Some of you probably figured out why I'm playing with this, and it's something I'll talk more about later.

---

The measurements in this post were made in our measurement lab at Head-Fi HQ using the following equipment:

• Brüel & Kjær High-frequency Head and Torso Simulator (HATS) Type 5128
• Audio Precision APx555 Audio Analyzer
• Audio Precision APx1701 Transducer Test Interface
• Benchmark HPA4 Headphone Amplifier
• Herzan custom-designed acoustic & vibration isolation enclosure

 

[ADUHF]

Thank you for diving in on this as well, and sharing the above info and graphs, Jude.

It's interesting to see some actual results on this. And I think it also helps to better illustrate a few of the reasons why this subject is even open to some debate. Because if these sort of driver symmetry measurements were easier to do, then presumably they'd be more or less of a no-brainer for most reviewers, since they can provide some potentially very useful info and insights into the quality or shortcomings of a headphone's stereo imaging.

When I made the following remarks, I wasn't really thinking specifically about the single-sided rigs and preview systems that you mention above. But those are other good examples of exactly why this type of measurement can be a bit more of a challenge to get right than some others.

I understand that the subject of driver symmetry is a thorny one, which can present some special challenges to reviewers like yourself. For both the reasons you've stated above. And for other potential reasons as well. For starters, everything has to be working pretty much perfectly on your equipment just to ensure that both earpieces on the headphone are receiving the exact same signal amplitude at all frequencies. I would imagine that there are some ways of calibrating the rig to improve or refine its performance along these lines though.

There could also be issues related to the way a headphone interfaces or seals with a rig, that could potentially introduce measurement artifacts into one channel, but not the other. Averaging the response over multiple seatings (as you presently do) should help to mitigate the chances of that occurring though, I would think...

...And there are probably some other issues along a similar line that I haven't thought of, which could make it more difficult to accurately assess or measure a headphone's L/R balance.

Driver symmetry measurements are a bit like a stress-test on both the equipment used for doing headphone measurements, and also on a reviewer's or grapher's abilities to use that equipment effectively, and reliably. The upside to this (if you want to look at it that way) is that it can potentially also make you a better grapher, and maybe also help to work a few of the bugs and kinks out of the measurement gear.

If you don't feel that your gear or skills are up to the task of doing something like this though, there is no shame in admitting that. Because measurements are DIFFICULT TO DO!! And it can take quite a bit of time and practice to really get the hang of, and to figure out. That has certainly been one of my takeaways from reading many of Tyll's articles and watching his videos on the subject.

So, in spite of my apparent zeal on this subject, I really can appreciate where people are coming from when they say this not something they are comfortable doing. I'm rather thick-headed on alot of stuff,... but I get that. And I really have no desire to try to coax, convince or cajole anyone into doing something that they really don't want to do, or don't feel ready to do for one reason or another along these lines.

I'd rather have folks like you, and Resolve doing these kinds of measurements though, than some other potentially less thoughtful, or informed, or technically-minded individuals or reviewers... precisely because you seem to have a somewhat better grasp and appreciation for some of the potential pitfalls that are involved in doing something like this. And would probably take it more seriously, and do a much better and more accurate job of it than some other folks would.

If I'm interpreting the graphs that you've posted above correctly, it looks like all the data to calculate the separate curves for the left and right channels is probably already there. And you've simply combined it all together into one single average frequency response curve for the headphones. Which is fine, if that's all you want to show. And if you also make it clear that the data for that final curve is coming from two separate transducers in the headphones. As I already mentioned to Resolve above, this type of average info for both channels together can also be very useful for assessing and making adjustments to a headphone's overall tonal balance.

If you already have all the data and samples that are needed to do a reasonable job of showing the potential differences between the two channels though, then that sort of begs the question of why you wouldn't also want to put that out there for people to see and use as well?

All you can really do on something like this is the best you can do. And then you have to let the chips fall where they may. More than likely you'll learn something new from each attempt, and gradually begin to get better at it over time. (Just as Tyll seemed to do,) Until it simply becomes a matter of routine, like it currently is for some other graphers.

 

[ADUHF]

The other thing I was gonna mention on this is that you shouldn't underestimate a user's or reader's ability to interpret this type data. Most people who are into this kind of thing aren't dumb-dumbs. And have very likely looked at the (possibly somewhat less reliable) driver symmetry info that's already out there and available on some other review and measurement sites. And will probably already have at least a general idea about the limitations in this sort of info.

This is also the sort of subject that could make a good topic for some kind of an informational video which explains some of the myriad issues or hurdles that graphers often have to contend with when putting this sort of graphical info together. I'm sure there are some folks who've already gained a few new insights on this, just from the posts and info that have already been offered above.

Somebody else's turn to weigh in now, since that's about all I've come up with so far on this particular subject.

 

[jenkinsontherun]

It's one thing if we could reasonably expect people who've never done a headphone measurement to understand this, but most will just interpret any lack of perfect symmetry as being on the side of the headphone (which it may be), not the measuring process (which it also may be, but a variable often (and understandably) overlooked, especially by those who've not yet measured a headphone).

Dude you gotta learn to be more concise. Flowery language like this can deter people from getting your point.

 

[jude]

Dude you gotta learn to be more concise. Flowery language like this can deter people from getting your point.

I have to admit the excerpt you quoted is pretty bad, so you do have a point. I can only promise you (since you probably didn't get any further than that) that the rest of the post is probably no better.

...And I really have no desire to try to coax, convince or cajole anyone into doing something that they really don't want to do, or don't feel ready to do for one reason or another along these lines...

I don't believe you.

The other thing I was gonna mention on this is that you shouldn't underestimate a user's or reader's ability to interpret this type data. Most people who are into this kind of thing aren't dumb-dumbs. And have very likely looked at the (possibly somewhat less reliable) driver symmetry info that's already out there and available on some other review and measurement sites. And will probably already have at least a general idea about the limitations in this sort of info...

Of course it's not about assuming that those looking at measurements are dumb-dumbs. It does seem to me, though, that the moment there's an x and y axis and some squiggly lines, they're taken quite literally.

One headphone manufacturer used to include individual measurements with every headphone they shipped, and people would compare their measurements on the forums. As I recall, these were production test measurements, and not comprehensive, averaged sets. So when one headphone's measurement showed more energy or less energy at, say, 5 kHz than another, there was often concern. People would return and exchange headphones to get one that didn't have a measured dip or peak that they didn't see on another's measurements of the same model on the forums.

It was just as likely (probably more likely) that the differences that were posted were just the kinds of differences you get from one measurement pass to the next, even on the same headphone. As you can see in my measurements above, even two passes on the exact same headphone can result in notable differences.

So it's not about assuming people are doofuses, but knowing that measurements are interpreted very literally by many -- every dip and peak.

...This is also the sort of subject that could make a good topic for some kind of an informational video which explains some of the myriad issues or hurdles that graphers often have to contend with when putting this sort of graphical info together. I'm sure there are some folks who've already gained a few new insights on this, just from the posts and info that have already been offered above...

We actually do have some planned, and just have to find the time to put them together. The measurement process can be interesting and challenging.

I will occasionally post things like the above long-winded post -- yes, other long-winded posts -- to explain things about measuring that hopefully some people find interesting or surprising.

If you haven't already seen our Apple AirPods Max measurement steps, you should check out these links:

https://www.head-fi.org/threads/airpods-max.949152/page-98#post-16058062​

​

https://www.head-fi.org/threads/airpods-max.949152/page-127#post-16071445​

​

https://www.head-fi.org/threads/airpods-max.949152/page-153#post-16091082​

Here's a post about measuring headphone sensitivity (and headphone sensitivity specs):

https://www.head-fi.org/threads/sennheiser-hd650-massdrop-hd6xx-impressions-thread.197776/post-13803816​

I hope we can get to making videos about headphone measurements in the near future. There's a lot of ground to cover with the fascinating Brüel & Kjær 5128, we did (and are still doing) ANC testing and MOS (mean opinion score) testing with Head Acoustics, I'd like to show the fun that can be had with the Audio Precision audio analyzer, and more. It's just about finding the time.

 

[ADUHF]

As a reviewer, you also have to weigh the value and benefits that these types of measurements would have to your readers/users versus the time, energy and difficulty that it would take to do them well. I think this was also one of Resolve's main points.

For some reviewers and users it may not be worth the time and trouble. For someone who's trying to decide between this or that $150-200 studio headphone though, it may be a very useful and critical piece of info that could potentially effect their buying decisions.

This frankly wasn't something that I really even paid that much attention to, until I got my first halfway decent pair of headphones, the AKG K553 Pros. And then began taking them apart to see what was actually going on inside, in an attempt to figure out why the different channels seemed to have different timbres, and why the stereo images didn't always seem properly centered. Before that, I was mostly just using some $30 Koss portable headphones with titanium drivers.

The K553 was a genuine learning experience that opened my eyes alot wider to alot of this stuff, including the problems and imbalances in my own hearing. And since then, this issue has become one the most important ones that I look at in any new headphone that I purchase.

If you've primarily owned headphones that are more in the $350 and up price range, then you may be somewhat less sensitized to this type of problem because the tolerances are likely much tighter. And the drivers are probably better matched as a general rule. That doesn't mean that an imbalance cannot exist though on more expensive headphones. And if it does, then it's something that I'd think that most users here would want to know about.

What I found with the AKG K553 was that many of the fit, seal and other issues with the filters in the headphones were user-correctable. And fixing alot of that stuff definitely helped to achieve some better results.

There were other imbalances though, which I think were likely due at least in part to the single-sided design of the headphone, that were not as easily fixable with physical modifications. So that's when I began experimenting with using separate EQs for the two channels. Most of that was actually done with a 2-channel, 31-band, analog DBX 231s graphic equalizer, which is a fairly cumbersome piece of hardware.

With a few changes to the pads, and tweaks to its overall tonal balance, and L/R balance, the AKG K553 was a pretty decent sounding headphone imo. And it was also alot more comfortable than most of the other options in my price range. Because it had almost no clamp. So that's why I decided to stay with it, and try to fix most of its issues, rather than simply getting a different pair headphones.

At the time, I think that was probably the right choice for me. Because I got a whole lot of use out of them, and enjoyed them quite a bit, after making alot of the above tweaks. And also learned a little bit about the design and construction headphones, and how to affect some useful changes to that. I'd rather not have to go through all that with my other headphone purchases going forward though.

 

[castleofargh]

@ADUHF , let me fix that for you:
"The other thing I was gonna mention on this is that you shouldn't underestimate a user's or reader's ability to misinterpret this type of data."



About publishing something or not, I’m 110% hypocrite. As a principle, of course I’m pro data. Having access to data is step one for most things, and should always be prioritized.
At the same, I make up all the excuses in the world to justify not posting my own graphs. It certainly helps that my measurement rig is a nonsensical, uncalibrated, frankensteined pile of cheap stuff.

in this conversation, I'm unofficially with @Resolve. I very much share his concerns and views. Officially, I'm with you. why doesn't he share the data with everybody so we can make our own decision? right?

 

[omegaorgun]

I see many, many overlapping measurements, then get dizzy...so is she (DCA Stealth) any good? LOL!

 

[ADUHF]

@ADUHF , let me fix that for you:
"The other thing I was gonna mention on this is that you shouldn't underestimate a user's or reader's ability to misinterpret this type of data."



About publishing something or not, I’m 110% hypocrite. As a principle, of course I’m pro data. Having access to data is step one for most things, and should always be prioritized.
At the same, I make up all the excuses in the world to justify not posting my own graphs. It certainly helps that my measurement rig is a nonsensical, uncalibrated, frankensteined pile of cheap stuff.

in this conversation, I'm unofficially with @Resolve. I very much share his concerns and views. Officially, I'm with you. why doesn't he share the data with everybody so we can make our own decision? right?

Thank you for the reply, castle.

I think my unofficial position on driver symmetry is probably about the same as my official position... only moreso. Your points above are taken though.

I like being able to consult multiple sources for different opinions and results on this type of thing though. So I'd generally come down in the more-info-is-better column when it comes to stuff like this.

Feel free to share some of your frankensteined results here as well, if you wish.

 

[ADUHF]

I see many, many overlapping measurements, then get dizzy...so is she (DCA Stealth) any good? LOL!

Some recent opinions and info on this from both Jude and Resolve.





More info, opinions and questions here...

https://www.head-fi.org/threads/dan-clark-audio-stealth-review-interview-measurements.959214/