The science behind Stax's magic

[Repentant subjectivist]

Aren't there more measurements than just FR graphs? I'm not very well versed, but I can't imagine a pair of 10$ IEMs ever producing the same sound as a 009 even if you managed to match the FR graph.

 

[maverickronin]

Aren't there more measurements than just FR graphs? I'm not very well versed, but I can't imagine a pair of 10$ IEMs ever producing the same sound as a 009 even if you managed to match the FR graph.

You could get much close than many might imagine, but it's complicated.

A headphone FR graph is basically equivalent to steady state in room acoustic measurement of a speaker system. Even if the final FR is the same, the speakers could sound different due to differences such as intensity and arrival time of reflections.

Headphones are even worse that that, because each headphone is essentially it's own set of speakers and its own room and then combines with your ears in an acoustically unnatural way to add a third element. Because of that, even if two differently constructed headphones were EQed to exactly the same target curve on a specific test rig they would end up diverging to some degree when place your head, or another test rig.

IEM to IEM is easiest since sticking a sealed transducer in one's ear canal bypasses the most of these issues. Circumaural to Circumaural seems like it may be possible in many situations as well, but opens up a whole new can of worms if you want emulate spatial qualities as well.

 

[Repentant subjectivist]

You could get much close than many might imagine, but it's complicated.

A headphone FR graph is basically equivalent to steady state in room acoustic measurement of a speaker system. Even if the final FR is the same, the speakers could sound different due to differences such as intensity and arrival time of reflections.

Headphones are even worse that that, because each headphone is essentially it's own set of speakers and its own room and then combines with your ears in an acoustically unnatural way to add a third element. Because of that, even if two differently constructed headphones were EQed to exactly the same target curve on a specific test rig they would end up diverging to some degree when place your head, or another test rig.

IEM to IEM is easiest since sticking a sealed transducer in one's ear canal bypasses the most of these issues. Circumaural to Circumaural seems like it may be possible in many situations as well, but opens up a whole new can of worms if you want emulate spatial qualities as well.

If we assume that a very cheap pair of circumaural could be EQ'd to match the exact sound of a pair of high end electrostatic headphones like the 009, what is stopping anyone from managing that and simply saving everyone a bunch of money? Is it a question of just tinkering enough with EQ software and a cheap pair of headphones?

 

[solderdude]

Take 1 headphone.
Measure them on a specific rig.
Use some program that applies the exact EQ so it measures flat in FR while leaving it on the test rig.
Marvel at the flat FR.
Now move the darn headphone and measure again.
No longer flat.
Use any other measurement device... it now isn't flat no matter how you mount it.
Will your ears do the same ?

Use another headphone.
Same procedure.

Now compare the results of the 2 'properly EQ'ed' headphones and measure them on any other HATS.
Do they measure similarly different in FR ?

How is distortion ? The one that showed 2% around 3kHz for instance is that 'cured' ?

How's the resonances ? (looking at the CSD) are they suddenly gone or similar ?

Some things in life are pointless. You can get headphones to have a very similar tonality but not the same.
For some similar-ish tonality equals sound the same so those folks will tell you you can get headphones to sound the same.

 

[MayaTlab]

If we assume that a very cheap pair of circumaural could be EQ'd to match the exact sound of a pair of high end electrostatic headphones like the 009, what is stopping anyone from managing that and simply saving everyone a bunch of money? Is it a question of just tinkering enough with EQ software and a cheap pair of headphones?

If you want a concrete illustration of what Solderdude just wrote, on my own head, with Oratory1990's Harman presets for the HD560S and HD650, and only looking at their relative differences below 5khz (above it's way worse) :
https://www.head-fi.org/threads/the...-at-a-breakthrough-value.943107/post-16300055

Lots of potential difficulties when it comes to matching two headphones to reach the exact same FR curve, below threshold of audibility, at your own eardrum, across the whole range of audible frequencies. Among many others :
- Sealing issues because of perfectible design (in general below 250-300hz).
- Sealing issues, sample to sample and left / right variance because of bad manufacturing.
- Variance because of how the headphones interact with your own anatomy (increasingly important past 1khz or so).
- As already said, variance because of exactly how the headphones are positioned over your head.
- Impossibility to effectively EQ a pair of headphones because of some intrinsic problems (sharp high Q null for example). If the headphones are properly engineered these should be minimal.
- Difficulties to effectively headphones because of a jagged response all throughout the FR spectrum which will require a lot of filters and fine-tuning and enculage de mouches.
- The sheer difficulty to characterise the discrepancies in FR curve between two headphones on your own head - your own ears are unlikely to be good enough and third party measurements can't be used (for perfect matching, they're useful for bringing headphones to a near point). Microphones that can be positioned in the concha or at the entrance of the ear canal are of limited use past 1khz and have their own tolerance and positioning issues. Perhaps only probe microphones located near one's own eardrum could successfully do so ?
- Insufficient software capacity in some circumstances (depends on device or software) : lack of PEQ, too small number of PEQ bands to fully address the discrepancies (may be a problem if you're trying to EQ headphones that present quite a few high Q peaks and dips in the trebles, which already are difficult and perhaps sometimes undesirable to effectively EQ in and of itself), too small range of available Q values, etc.

Some ANC headphones are capable to equalise themselves to a somewhat precise target below 1khz or so in real time, to compensate for sealing variation. Very theoretically, if that target was the same, two different ANC headphones of that kind would be able to reach it on your head (very theoretical as the algorithms are likely to be a at least a bit different). The next frontier in headphones design as far as SQ is concerned likely is to get past that 1khz barrier (both to solve for undesirable variations because of individuals' anatomy - some degree of variations may actually be a good thing unless I'm mistaken -, and positional variations) but it's a lot, lot more complicated.

 

[preload]

I used to own an SR-009 with maxed out Woo Audio WES. Til I picked up a pair of the MrSpeakers Ether Flow Open, a planar magnetic design. The Ethers sounded better. Sold my entire Stax rig.

 

[maverickronin]

If we assume that a very cheap pair of circumaural could be EQ'd to match the exact sound of a pair of high end electrostatic headphones like the 009, what is stopping anyone from managing that and simply saving everyone a bunch of money? Is it a question of just tinkering enough with EQ software and a cheap pair of headphones?

Basically what @solderdude said.

You can't actually get 100% there. To use another speaker analogy it's like trying to make up for a lack of room treatment with EQ.

If you want to get as close as possible you'd at least need individualized measurements of both headphones on your own head along with impulse response convolution. Something like Impulcifer or a Smyth Realiser. (These are meant to simulate speakers over headphones, but the same concept applies.)

 

[preload]

Basically what @solderdude said.

You can't actually get 100% there. To use another speaker analogy it's like trying to make up for a lack of room treatment with EQ.

If you want to get as close as possible you'd at least need individualized measurements of both headphones on your own head along with impulse response convolution. Something like Impulcifer or a Smyth Realiser. (These are meant to simulate speakers over headphones, but the same concept applies.)

Agree. The Harman group took measurements of $100 Sony headphones and eq'd their test AKG headphones to match the response curve. When they did this, listeners ranked the Sony-eq'd AKG phones as very highly preferred. But the actual Sony headphones that the measurements were taken from sound terrible.

Also unlike speakers, calculations thst consider smoothness and tracking of a target frequency response curve only correlate somewhat with listener preference when it comes to headphones.

 

[John_M]

Isn't it supposed to be about very fast "attack" and decay speed, and very low distortion? If so, I don't think this could be replicated by EQing cheaper headphones.

The SR-009 is supposed to have under 1% distortion at 20hz, although there's no review from Amir to confirm this.

 

[solderdude]

Amir did measure his 007 on the 5128 but haven't seen him measure it on the new test rig.

He might have distortion measurements but given the +10dB bump pointing towards leaky seal may also skew distortion when measured.

 

[preload]

The SR-009 is supposed to have under 1% distortion at 20hz, although there's no review from Amir to confirm this.

Sure but at what volume?

 

[John_M]

Sure but at what volume?

I'm sure I've read that the SR-009 distorts less at 100dB than at 90dB, although I can't remember where and you would be entitled to take that with a pinch of salt.

If only we had someone who could test/ measure these claims.

 

[seyl3r]

I'm sure I've read that the SR-009 distorts less at 100dB than at 90dB, although I can't remember where and you would be entitled to take that with a pinch of salt.

If only we had someone who could test/ measure these claims.

https://www.stereophile.com/content...09-sr-007-sr-507-sr-404-ltd-sr-404-and-sr-003

 

[solderdude]

I'm sure I've read that the SR-009 distorts less at 100dB than at 90dB, although I can't remember where and you would be entitled to take that with a pinch of salt.

If only we had someone who could test/ measure these claims.

Tyll already did many years ago.

Distortion at 100dB is low but not lower even though it appears that way. It appears to be higher at 90dB SPL because the S/background noise ratio is higher. And he even had a special box to keep external sounds out.

 

[John_M]

That's probably what I'm remembering. There seems to be something off with the left speaker here. The older SR-009 reviewed on the same page doesn't show this and the reviewer queries whether it's a measurement error.

 

[preload]

I'm sure I've read that the SR-009 distorts less at 100dB than at 90dB, although I can't remember where and you would be entitled to take that with a pinch of salt.

If only we had someone who could test/ measure these claims.

Wow that IS magic!
I remember trying to boost the bass on my sr-009 with eq and it didn't sound very clean. I guess it wasn't due to distortion. Or perhaps it was related to my tube amp.

 

[John_M]

If you're in California, you could volunteer your SR-009 set for testing.

Here is the other SR-009 % THD+ noise graph. Looks very pretty, including at 100 dB, but it would be interesting to see a like for like comparison on the same equipment with e.g. the Focal Utopia.

 

[solderdude]

but it would be interesting to see a like for like comparison on the same equipment with e.g. the Focal Utopia.

here you are .. Utopia.

Ignore the spikes. It is range switching artifacts from the used analyzer.

 

[John_M]

Thank you. These measurements look to be consistent with Amir's.

Isn't this at least part of the answer to the original question? The SR-009 clearly distorts less than even the Utopia with its super beryllium drivers, which I think was one of the lowest (if not the lowest) distortion headphones measured by Amir. Just look at those 100 dB lines tracking along neatly at 0.1% - even at 20hz and even when you crank the volume up. Clearly not something that could be replicated by just EQ'ing some cheap headphones.

 

[MayaTlab]

Isn't this at least part of the answer to the original question? The SR-009 clearly distorts less than even the Utopia with its super beryllium drivers, which I think was one of the lowest (if not the lowest) distortion headphones measured by Amir. Just look at those 100 dB lines tracking along neatly at 0.1% - even at 20hz and even when you crank the volume up. Clearly not something that could be replicated by just EQ'ing some cheap headphones.

I don't know exactly what the threshold of audibility for THD is at various frequencies, and I'm not even sure that there is enough accumulated research to determine that exactly, but so far all I've read on the subject suggest that both headphones' THD at lower frequencies already is largely below that threshold.
I'm typing this while listening to music on a modified and PEQed HD650 which sub-bass response I consider more satisfying than an unEQed K371 or an AirPods Max for example - even though it has more THD at lower frequencies than them (and the AirPods Max a pretty neat feature common to ANC headphones whereby it compensate the FR curve in real time to deliver a constant FR below 1khz or so, even with some breach of seal). Possibly because it's still low enough at the volume I'm listening that I can't hear it.
BTW if you want to see low THD, the AirPods Max is a good contender : https://www.head-fi.org/threads/apple-airpods-max-measurements-brüel-kjær-5128.951184/
That doesn't necessarily make it perfect sounding (but I wouldn't be surprised to learn that it helps with maintaining ANC performance even when seal is lightly breached).
I believe that Tyll also measured a lot of Audeze as having quite low THD.