Stax Tube Amplifier Distortion vs Solid State

[Francis Vaughan]

Stax aficionados will bring in the war chest now

I don't see why. Overall the new measurements are pretty much what one would expect. The tube amp is getting into trouble a bit earlier than the SS amp, and clipping a bit more softly. Otherwise they are very similar. Nothing like the original measurements that had the 007t in trouble vastly earlier.

0.2% distortion is where both amps start to dominate over HP distortion. That is -54dB. Usefully the SR-303 has a sensitivity of 100dB at 100V. Given we have a load of 100kΩ*, 100V is 10mW. The HP distortion graph for the 313 goes into clipping at about 114dB = 500V. The amplifier graph for the 313 shows it reaching -54dB distortion at a hair over 100mW, or closer to 320V. So the numbers are not wildly off. Either way it is pretty much at the limits of the power rails, whihc is what we expect. The 313 is in the usual SS near vertical take off of distortion, so there is wiggle room for picking your points.
The HP driven by either of the two amplifiers reaches about the same distortion at 3% aka -30dB. This doesn't match anything on the amplifier only distortion, and I would suspect we can safely assume we are past useful measurements in the clipping area. We may be clipping the HP.
If we look at the lower levels, we can see at the LHS, that both amps are driving the HP at about 0.05% distortion: -66dB and at 104dB SPL, which would be a hair past 10mW on the original amplifier graph. At this point the 007t originally measured at about -44dB distortion, which is something of a gap to the -66dB measured driving a HP. Indeed the 007t never measured -66dB distortion at any drive level in the first measurements. The best it ever did was -58dB. So the gap is minimally 8dB, and from the numbers, about 22dB.

The new measurements might not please the tube fanatics, but they don't show that the 007t amp in much worse a light than one might reasonably expect.

* The load is a bit ambiguous. The AP actually presents 100kΩ per leg of its differential input to ground, plus capacitance. So it may be more accurate to take 200kΩ. Capacitance doesn't make much difference at 1kHz, but becomes the dominant load near the top of the range. This is all very ball park, so I'm not worrying right now. If the graph was generated assuming 100kΩ the derived voltages will still be correct even if the power delivery isn't.

 

[EdW]

From these reviews it seems that Stax’s strength is in designing electrostatic headphones. Their circuit design skills are not at the same level and these ‘phones all need bespoke drive circuits. The current circuits don‘t do justice to them.

Perhaps they should strengthen their analog electronics capability?

 

[F1308]

How loud is +112 dBSPL?

It will depend on the frequency. Very different if 16 Hz than 3500 Hz.

https://en.m.wikipedia.org/wiki/Equal-loudness_contour

But bearing in mind 85dB should be the limit while being subjected to it along 8 full hours, you are to halve that for every 3dB step upwards.
85...8 hours.
88...4 hours.
91...2 hours.
94...1 hour.
97...30 minutes.
100...15 minutes.
103...7,5 minutes.
106...3,75 minutes.
109...1,875 minutes.
112...0.9375 minutes.

Now, after those 56 seconds at 112 dB you are supposed to cover your ears and hear NOTHING for the next 24 hours if you want to keep your hearing undamaged, since you have reach the cumulative dose.

For loudness we have to talk FONOS, not dB.

 

[pma]

So, in fact, the distortion at acoustical side below 112dB SPL is in fact lower with the tube preamp (maybe because its different Zout), or am I missing something ??

 

[the_brunx]

How loud is a typical rock concert or a night club? The sound engineers there seem to set the reference for how loud it needs to be to sound the best/most engaging.

 

[F1308]

How loud is a typical rock concert or a night club? The sound engineers there seem to set the reference for how loud it needs to be to sound the best/most engaging.

Without considering the consequences...?

 

[Maki]

From these reviews it seems that Stax’s strength is in designing electrostatic headphones. Their circuit design skills are not at the same level and these ‘phones all need bespoke drive circuits. The current circuits don‘t do justice to them.

Perhaps they should strengthen their analog electronics capability?

Their circuits are "good enough". The 313 shows little distortion up to a hair's breadth away from nwavguy's recommended 115dB target. Sure it may still introduce audible distortion but when you look at the system as a whole including the headphones it still compares favorably to most of the competition. The market for stats is small and opamps are mostly untenable for the voltages involved making everything discrete and harder to design, though I've taken a look at some of the APEX opamps and they might be promising. They are priced quite high though, a hundred up to a thousand bucks for a single opamp. And you'd need a few at least for a NFCA type design.

 

[JohnYang1997]

Nice.

 

[EdW]

Their circuits are "good enough". The 313 shows little distortion up to a hair's breadth away from nwavguy's recommended 115dB target. Sure it may still introduce audible distortion but when you look at the system as a whole including the headphones it still compares favorably to most of the competition. The market for stats is small and opamps are mostly untenable for the voltages involved making everything discrete and harder to design, though I've taken a look at some of the APEX opamps and they might be promising. They are priced quite high though, a hundred up to a thousand bucks for a single opamp. And you'd need a few at least for a NFCA type design.

Agree. But to make their headphones truly superb they need to improve the electronics, since this matters as much as the transducer design. The product is only as good as its weakest link.

A competent amplifier design is likely going to need discrete components to handle the voltages needed. It clearly isn’t a ’lego’ job with off the shelf amplifier ICs handling the signal swings.

 

[Degru]

From these reviews it seems that Stax’s strength is in designing electrostatic headphones. Their circuit design skills are not at the same level and these ‘phones all need bespoke drive circuits. The current circuits don‘t do justice to them.

Perhaps they should strengthen their analog electronics capability?

Estats are also very challenging loads to design direct drive amps for, so it's no wonder that only the high end is really good. Plus small niche market while there is tons of competition for normal amps. There are no easy or cheap off the shelf solutions to implement.

The easy solution is releasing a high quality transformer box to use with speaker amps. This is actually what they were doing in the beginning, a few decades ago. Then you have high parts cost for the transformers themselves but there isn't much to screw up outside of that and the hard work is left to speaker amp makers, where there is tons of competition and RnD from numerous places.

Also I am not sure that stax are really inherently exceptional at making estats as much as them being the only major player in the market. Other companies like Sennheiser or Hifiman seem to treat estat as some kind of holy tech that is worth $50k and don't release estat products normal people could buy. There is no real competition outside of small hobbyist stuff so of course stax is "best".

 

[Francis Vaughan]

So, in fact, the distortion at acoustical side below 112dB SPL is in fact lower with the tube preamp (maybe because its different Zout), or am I missing something ??

Its almost certainly dominated by the headphones, and neither amp is adding any meaningful distortion here. This is loud. The SS amp measured standalone with distortion at least 10dB better, and up to 18dB, better, than the HP amp combo is measuring as at these levels.

The tiny difference could be lots of things, generally not important. It looks for the most part as if somewhere under the HP distortion both amps eventually get to clipping and suddenly spike up past the HP. The 313, being SS does this very suddenly, whereas the 007t is likely going into a soft clip a bit earlier, and clipping earlier because of the higher voltage drop across the tube.

 

[Francis Vaughan]

The easy solution is releasing a high quality transformer box to use with speaker amps. This is actually what they were doing in the beginning, a few decades ago.

I own a pair of the original Lambdas. (They are on long term loan to a colleague, otherwise I would get them out and do a bit of testing. I really need to get them back.) When I first bought them I got the transformer energiser. I then bought the SRM-1 mk-2. It was a huge improvement.
The SRM-1 has more than a fleeting resemblance to the 007t. They both share basically the same input stage, and the 007t's tubes replace the output transistors in what is otherwise a nearly identical topology. It isn't a particularly advanced design. Back in the early 80's they were doing the best they could.
Transformers have all manner of difficult problems to manage. To make at a sane price and get the quality isn't trivial. Doable, but not trivial.
If you look at Kevin Gilmore's high end designs you get near perfect amplification, but the level of complexity is jaw dropping. He seems to have thrown the kitchen sink at the designs. Commercial realisations are not cheap. Then again the current Stax amplifiers are not either.

 

[Frank Dernie]

And oh, this is evidence that electronic distortion can be worse than our transducers (speakers and headphones)

One of the great mysteries of the science of hifi, to me, is why and by what mechanism that could possibly be.

 

[Maki]

Estats are also very challenging loads to design direct drive amps for, so it's no wonder that only the high end is really good. Plus small niche market while there is tons of competition for normal amps. There are no easy or cheap off the shelf solutions to implement.

The easy solution is releasing a high quality transformer box to use with speaker amps. This is actually what they were doing in the beginning, a few decades ago. Then you have high parts cost for the transformers themselves but there isn't much to screw up outside of that and the hard work is left to speaker amp makers, where there is tons of competition and RnD from numerous places.

Also I am not sure that stax are really inherently exceptional at making estats as much as them being the only major player in the market. Other companies like Sennheiser or Hifiman seem to treat estat as some kind of holy tech that is worth $50k and don't release estat products normal people could buy. There is no real competition outside of small hobbyist stuff so of course stax is "best".

And what of transformer core saturation issues? I'd really like to know if transformers are the way to go, but all the measurements and data I've seen regarding output transformers shows them being less competent as opposed to direct drive and mostly still exist as a hold over from tube days where we didn't have solid state electronics for driving speakers. They introduce their own distortion and issues, with saturation being the primary (no pun intended). That being said most of the data available is relevant to tube amps driving speakers which use the transformers the other way around.

 

[pma]

Its almost certainly dominated by the headphones, and neither amp is adding any meaningful distortion here. This is loud. The SS amp measured standalone with distortion at least 10dB better, and up to 18dB, better, than the HP amp combo is measuring as at these levels.

The tiny difference could be lots of things, generally not important.

I agree, however IME even the amp with higher intrinsic distortion can make lower distortion at acoustical side of the transducers, mainly because of interaction of output impedance with electro-mechanical transducer non-linearities. At least this is true with dynamic transducers and current drive instead of voltage drive. I am not an expert in electrostatic transducers, so I do not know there.

Just for fun, I have arranged a quick measurement of HD598 distortion taken simultaneously at acoustic output and electric input, at 1kHz.

 

[the_brunx]

The best way to find out is to measure the sr303 on a srd7 mk2 driven by a capable amp like the benchmark thx ahb2. That would be a dream.

and then with a Sennheiser HE60 which I believe will have better distortion figures.

 

[solderdude]

I agree, however IME even the amp with higher intrinsic distortion can make lower distortion at acoustical side of the transducers, mainly because of interaction of output impedance with electro-mechanical transducer non-linearities. At least this is true with dynamic transducers and current drive instead of voltage drive. I am not an expert in electrostatic transducers, so I do not know there.

Just for fun, I have arranged a quick measurement of HD598 distortion taken simultaneously at acoustic output and electric input, at 1kHz.

That coincides with my measurement (of HD599) at 1kHz at 90dB SPL. I can't measure much higher.
1% at 106dB SPL is audible but 106dB @ 1kHz would mean about 115dB SPL in the bass.

 

[Francis Vaughan]

And what of transformer core saturation issues?

Make the transformer big enough that the core doesn't saturate. Saturation isn't the problem for a headphone, there is essentially no power being transformed. But the other evil second order issues need addressing. The big one is hysteresis. That may need exotic core materials. Then you need to cope with the vagaries of things like inter-winding capacitance and leakage inductance, that will mess with response. These are all in competition with one another. The transformer will not operate in isolation from the electrical load of the headphones, which makes things generally harder. Hum pickup might matter too for the ultimate performance. You might even be chasing weird effects like magnetostriction. And you still need to provide a bias voltage. So you didn't avoid the pain of a power supply of some kind.

This isn't to say that you can't make a very good transformer, but to compete with a competent SS solution may be cost ineffective. Companies like Lundhal will probably be pleased to quote.

The best way to find out is to measure the sr303 on a srd7 mk2 driven by a capable amp like the benchmark thx. That would be a dream.

The SRD-7 was in my experience a pretty second rate answer. But indeed, it would be a pretty interesting experiment.

 

[the_brunx]

[QUOTE="


The SRD-7 was in my experience a pretty second rate answer. But indeed, it would be a pretty interesting experiment.[/QUOTE]

which amp did you use? and what exactly improved when you used srm1?

 

[Maki]

Make the transformer big enough that the core doesn't saturate. Saturation isn't the problem for a headphone, there is essentially no power being transformed. But the other evil second order issues need addressing. The big one is hysteresis. That may need exotic core materials. Then you need to cope with the vagaries of things like inter-winding capacitance and leakage inductance, that will mess with response. These are all in competition with one another. The transformer will not operate in isolation from the electrical load of the headphones, which makes things generally harder. Hum pickup might matter too for the ultimate performance. You might even be chasing weird effects like magnetostriction. And you still need to provide a bias voltage. So you didn't avoid the pain of a power supply of some kind.

This isn't to say that you can't make a very good transformer, but to compete with a competent SS solution may be cost ineffective. Companies like Lundhal will probably be pleased to quote.


The SRD-7 was in my experience a pretty second rate answer. But indeed, it would be a pretty interesting experiment.

That's fair, I was mostly thinking about the SRD-7 and similar sized boxes which have some pretty small looking transformers, much smaller than even the ones in tube amps.