Finding value in headphone measurements

[Jimbob54]

Like this one... ?
View attachment 79022

@Thomas savage - make this happen.

 

[pwjazz]

what do you want from headphone measurements? What do you go in looking fo

As someone who is starting to build headphones, I'm interested in measurements for two reasons:

1. As I modify tuning to achieve a particular sound, measurements can be a quick way to give me some idea of whether my tweaks are headed in the right direction

2. If I graduate to actually manufacturing and selling, measurements will help with QA, especially channel matching and making sure that I keep product variation under control

 

[bobbooo]

I suspected that there might be contention on putting electrical Z in the "not audibly significant" section - my reasoning, at the risk of being controversial, largely boils down to a position that for amplifier-headphone impedance interactions to significantly impact frequency response, a minimum of one of the amplifier designer and headphone designer must have made a terrible mistake...

All the more reason to include it then One of the great things about ASR is its measurements can reveal deficiencies and faults in devices that are not immediately obvious yet can impact performance, which in turn could influence manufacturers to improve on these deficiencies. A headphone may measure great under ideal conditions with a ~0 ohm output impedance source, but if the headphone itself has wildly varying or ridiculously low impedance (like the Campfire Andromeda), real-world usage will be audibly compromised. The fact is many multi-driver balanced-armature IEMs do have wildly varying impedance, so it would be useful to quantify the impact of this on measured headphones' frequency response for users. Similarly, many (probably the majority) of stereos and AVRs (plus some PC sound cards) have ridiculously high headphone out impedances which will affect most headphones. If these are common problems that impact end users, then I don't think them being bad design is a good reason to ignore them, rather I would say the opposite.

 

[Robbo99999]

Ah, just found this thread, and it's a useful one, I'll put across what I'm interested in from these headphone measurements. To me the proof of the pudding is in the eating, or in this case listening to the resultant EQ's that myself or someone else has created from the measurements......also to experiment with different Target Frequency Response Curves. For example theoretically calculated Target Curves like the ones from the Chris Struck Method (which is Mad_Economist's spreadsheet for instance) which allows you to recreate any speaker that has been measured by Amir in "any room of your choice". I'd also like to compare different Harman Target Curves that have been developed over the years (the different versions eg 2018, 2015, etc), and I'd try Diffuse Field & other targets. I'd like to experiment with that, so I find the practical application to the end user more interesting in terms of using the measurements to optimise our own headphones.

I am also, to a lesser extent, interested in the comparison of different headphones.....trying to evaluate/link quality of said headphones from the measurements - tricky to do for various reasons.

And thirdly and perhaps more interesting to me than the previous little paragraph...."the future of headphones", eg Smyth Realizer type angles or anything that is ground breaking that is pushing headphones forward in being able to emulate speakers in a room or an "environment" within which you find yourself, so 3D imaging I guess. Or anything that you can do DIY-style to come closer to some of this.

 

[Jimbob54]

Ah, just found this thread, and it's a useful one, I'll put across what I'm interested in from these headphone measurements. To me the proof of the pudding is in the eating, or in this case listening to the resultant EQ's that myself or someone else has created from the measurements......also to experiment with different Target Frequency Response Curves. For example theoretically calculated Target Curves like the ones from the Chris Struck Method (which is Mad_Economist's spreadsheet for instance) which allows you to recreate any speaker that has been measured by Amir in "any room of your choice". I'd also like to compare different Harman Target Curves that have been developed over the years (the different versions eg 2018, 2015, etc), and I'd try Diffuse Field & other targets. I'd like to experiment with that, so I find the practical application to the end user more interesting in terms of using the measurements to optimise our own headphones.

I am also, to a lesser extent, interested in the comparison of different headphones.....trying to evaluate/link quality of said headphones from the measurements - tricky to do for various reasons.

And thirdly and perhaps more interesting to me than the previous little paragraph...."the future of headphones", eg Smyth Realizer type angles or anything that is ground breaking that is pushing headphones forward in being able to emulate speakers in a room or an "environment" within which you find yourself, so 3D imaging I guess. Or anything that you can do DIY-style to come closer to some of this.

Do you reckon if you stuck to one or 2 hp models and tried loads of different target curves (or more accurately, eq to those curves) you'd find a target you were happy to adopt?

It's easy to say you prefer eq a to stock or vice versa, I suspect harder to say prefer a to b to c

 

[Robbo99999]

Do you reckon if you stuck to one or 2 hp models and tried loads of different target curves (or more accurately, eq to those curves) you'd find a target you were happy to adopt?

It's easy to say you prefer eq a to stock or vice versa, I suspect harder to say prefer a to b to c

I'm not sure how easy I'd be able to tell apart Target Curves, but I have tried Diffuse Field vs Harman Curve on my HD600 and there is a massive & obvious difference between them, so it comes down to how different the Target Curves are from each other. It would be harder to recognise the difference between the smaller changes that have been made to the Harman Curve over the years, so that would be harder to tell apart....I've not tried comparing them yet. It's easy enough to create different EQ profiles in Equaliser APO that you can flip between with a couple of quick mouse clicks....so it's easy to compare them. I'd start by choosing a preference on my HD600 (open back) which is my favourite headphone, then I'd see if my preference still stuck when using my NAD HP50 (closed back). Yeah, and I've got my own set of "reference tracks" that I know & love from different genres of music, so I would try them out on a variety of music to guage their suitability......most of those reference tracks are tracks that I think are high quality recordings that work very well on my Harman Curve & Room EQ'd JBL 308pMkii speakers which are also set up at perfect equidistant triangle, so I try to use that as a reference to help choose the right tracks.

I'm quite interested to play with creating different Target Curves using Mad_Economist's spreadsheet which is the Chris Struck Method due to it's ability to mimic any speaker in any room.....so I'm looking forward to trying that out if I can get the HRTF files for GRAS from Oratory1990 for instance (which I would plug into the spreadsheet along with any speaker of my choice that Amir has measured). But that's a project for this weekend or whenever I can get those HRTF files.

 

[Mad_Economist]

I would assume it is very easy to generate automatically from a measured impedance plot.

Ah I see, you mean like an output from a spreadsheet or other calculator based on the Zin of the headphone? That'd make sense - heck, if someone felt ambitious, that shouldn't be too hard for a web calculator I'd imagine.

B.t.w. as a side note... why doesn't it say 'technical expert' below your avatar ?

To be fully honest, I feel like giving me the same flair as Floyd Toole or j_j really undercredits them...

Wanna badge, you pay the Watchman.

This will pose a problem, as due to the omission of lower torso articulation, head and torso simulators are incapable of bowing to anyone...

 

[Jimbob54]

Ah I see, you mean like an output from a spreadsheet or other calculator based on the Zin of the headphone? That'd make sense - heck, if someone felt ambitious, that shouldn't be too hard for a web calculator I'd imagine.


To be fully honest, I feel like giving me the same flair as Floyd Toole or j_j really undercredits them...


This will pose a problem, as due to the omission of lower torso articulation, head and torso simulators are incapable of bowing to anyone...

You got a purdy mouth though...

 

[Mad_Economist]

All the more reason to include it then One of the great things about ASR is its measurements can reveal deficiencies and faults in devices that are not immediately obvious yet can impact performance, which in turn could influence manufacturers to improve on these deficiencies. A headphone may measure great under ideal conditions with a ~0 ohm output impedance source, but if the headphone itself has wildly varying or ridiculously low impedance (like the Campfire Andromeda), real-world usage will be audibly compromised. The fact is many multi-driver balanced-armature IEMs do have wildly varying impedance, so it would be useful to quantify the impact of this on measured headphones' frequency response for users. Similarly, many (probably the majority) of stereos and AVRs have ridiculously high headphone out impedances which will affect most headphones. If these are common problems that impact end users, then I don't think them being bad design is a good reason to ignore them, rather I would say the opposite.

To be clear, I definitely think electrical Z should be measured, it's just a borderline pass-fail type criteria for me - there's very few cases where I see a single-digit nominal impedance and wild swings and think "this seems like a product suited to actual use". For the designs with higher but still varying impedance, it's less likely to pose a very large issue, however, which is why I was relegating it to the "technical interest" side of things. I think we generally agree though

 

[solderdude]

To be fully honest, I feel like giving me the same flair as Floyd Toole or j_j really undercredits them...

That's the same reason I didn't want the 'title' as well.

 

[bobbooo]

Re: Zout and headphone impedance, I always liked Rinchoi's visualization of that:
View attachment 78991
Presenting both the delta and the compensated responses is IMO helpful for newer people/folks who ain't about eyeballing minutia.

Reference Audio Analyzer do the same:

They even have a tool that calculates the frequency response of any measured headphone paired with any measured source (delta also available):

And if they haven't measured the source you're using, you can enter its output impedance into this tool to calculate the frequency response. All their tools, calculators and comparators can be found here, they're pretty useful.

 

[bobbooo]

What's the typical output impedance of a smartphone?

There isn't really a typical value. Although they do generally seem to be getting lower, there have still been flagship phones released within the past year that have high output impedances e.g. the Xiaomi Mi 9T Pro with a value of 22.7 ohms. This is despite Xiaomi's older Mi A1 mid-ranger having a low output impedance of just 1.3 ohms. The only way to know for sure for any particular model is to have measurements of it.

 

[bobbooo]

And LG G7 at 1.6 ohm.

Interestingly, Reference Audio Analyzer measured the G7's output impedance to be even lower with the Quad DAC on, at 0.8 ohms, yet in agreement with your measurement at 1.7 ohms with the Quad DAC off. Another (albeit slight) advantage of the Quad DAC it seems.

By the way, any chance we could get all the same suite of APx555 measurements that you made of the LG G7 but of your Samsung Galaxy S8+ as well? Those S8+ measurements would be relevant to all S8, S9, S10, Note8 and Note9 (plus variants) users, as they all have the same audio chip. If you do end up measuring headphones, I think the only consumer audio device you won't have comprehensively covered would be smartphones And considering a large proportion of headphone users still use their phone to play music, and headphone jacks are making a comeback, I think smartphone audio performance would be of interest to and attract a lot of readers.

 

[Tks]

I thought I'd drop this here to ask: what do you want from headphone measurements? What do you go in looking for?

Very simple, an averaged FR, and THD at various SPL's (96dB ideally if it had to just be one).

Oh and very much would want to see channel variance (which is driver matching, or how closely they are matched). Aside from these two things, all is well otherwise

Oh and for fun - perhaps measurements of things like foam pads or dust filters/grills on or off? Lots of people say this makes a massive difference. When I try it on some of my headphones, it's not really much of a big deal.

 

[outerspace]

I want to suggest Amir or to other measurement enthusiasts one thing that made his graphs most useful and comparable. There is a problem for most of measurements - they all are frequency response depended. For example CSD in almost all cases just reflects FR (link 1, link 2). Even THD and other distortions are FR dependent in many ways and don't shows real behavior and have bad comparability. FR affect magnitude of signal witch is directly affect overall amount of distortions and also affect specific parts of distortions such as high-order harmonics.

For good comparability measurements should be made with frequency response equalized to one target (preferrably to harman target curve or at least to just flat) with minimum phase equalizer. There is other side effect of this - step response, square wave, impulse response, tone bursts etc will be looks the same in most cases.

 

[Dreyfus]

If the project would continue I would welcome a tonal balance chart that shows the pressure distribution of a headphone. I like the charts at headphonecheck.com ("simple view" showing the emphasis of bass, fundamental, mids, upper mids and highs). Too bad they measure with a falsly calibrated KU 100. I think their charts are very easy to interpret by beginners, though.

 

[outerspace]

I like Tyll's square plots because in his graphs the headphones that were able to better reproduce those squares are mostly planars, like Audeze/Fostex/Hifiman and the bass coming out from these is different, in a better way, than dynamic cans.

Accurate square response of the diaphragm do tells me how fast and precise the bass will sound.

Designing a diaphram able to quickly rise and quickly fall is very difficult and I tend to appreciate the folks who design headphones that measure better with square signals, although I'm aware that only this kind of measurement doesn't tells much about the sound profile or audible THD.

There is old topic were guy emulate Tyll's real measurements (square plots, step response and impulse response) just by FR: head-fi topic and archived innerfidelty post. Unfortunately graphs are lost. But he emulate real measurements precisely.

 

[trl]

Thanks for sharing the link, although I can only see one pic there (I'll retry later).

I do like and I fully agree with the explain here: https://web.archive.org/web/20160411225758/http://www.headphone.com/pages/evaluating-headphones, at least this is how I see it too:

"[...] the ability for the headphones to create a straight line at the top and bottom, even if it's tilted will indicate coherent performance in the lows".

 

[Mad_Economist]

Thanks for sharing the link, although I can only see one pic there (I'll retry later).

I do like and I fully agree with the explain here: https://web.archive.org/web/20160411225758/http://www.headphone.com/pages/evaluating-headphones, at least this is how I see it too:

"[...] the ability for the headphones to create a straight line at the top and bottom, even if it's tilted will indicate coherent performance in the lows".

The thing is, this characterization is at minimum misleading. The "tilt" of a square wave reflects the low frequency phase shift; this in turn reflects the effective highpass frequency. Here's a square wave looped back into a consumer sound card (reflecting its lowpass filter, thus the ringing at the edge):

And here's that same square wave with a 10hz highpass on the soundcard output - no change in the relative level of the square wave's harmonics, but the phase shift makes itself known:

Tyll did a ton of great things for headphones as a community regarding measurements - his characterization of square waves wasn't one of them. At the most charitable, they're a reasonable proxy for some frequency response relationships that have associations with what he attributed to them - at worst, they're just not a good tool to observe headphone behavior.

 

[solderdude]

At the most charitable, they're a reasonable proxy for some frequency response relationships that have associations with what he attributed to them

Indeed. I also found that when using HATS, even after correction for FR, 'ringing' is shown in plots that doesn't appear to come from the driver but rather seems to be a resonance of the ear canal.
Of course, the 'signal' at a real eardrum will also 'resonate' but perhaps at a slightly different frequency.

As a stand-alone measurement they do tell me whether bass is rolled-off, has a boost at certain frequencies, indicates a headphone sounds 'dull' or sharp. To me square waves and needle pulses thus do add some info (time related) when one evaluates a suite of measurements. Or should I say confirms some measurements.

In short.. I do find some value in these type of measurements but are not essential.