Electrostatic Headphone Amplifier Review & Comparison

[kevin gilmore]

But i actually did that. about 7 years ago. using the NWU owned ap. Thats how i know that the tektronix probes work. Thats how i know that correctly built attenuators also work. OK, maybe 8 years ago. can't remember.

And i already own a st100a. got tired of fixing it about 15 years ago. The keithely works just fine for current electrostatic design work. If i actually ever cared about measuring .0000x i would rent an ap. or a dscope.

my guess is that these days a current source modified stax is worth more than one in original condition. That seems to be the case in the used market at HF.

 

[bobbooo]

For reference, Amir has made comparative measurements before. Just FR though:

• SR-007 Mk1 & SRM-007t: https://www.audiosciencereview.com/forum/index.php?threads/headphone-measurements-using-brüel-kjær-5128-hats.15352/page-3#post-486254
• SR-007 Mk1 & SRM-313: https://www.audiosciencereview.com/forum/index.php?threads/headphone-measurements-using-brüel-kjær-5128-hats.15352/page-4#post-486266

Comparative distortion measurements this way would be great (including for the Koss amp and headphones). As @JohnYang1997 said, the real-world total combined amp + headphone distortion that we actually hear is what really matters, not an abstracted distortion value of any individual component in the audio reproduction chain taken in isolation, especially under (dubious) approximated conditions.

 

[milosz]

In a nutshell, you can’t swap a resistive load for a reactive one and think any valid measurements are going to come out of it.

These amps intended for electrostatic earphones will shift their operating point when loaded with a resistive load. Can't expect the THD and other specs to be valid under such conditions.

Amirm- did you look at the schematic for any of these amps and do any analysis before devising your measurement protocol? If it is to be science, measurement protocol is of prime importance.

 

[the_brunx]

Hmm, I have an SRM-353X just sitting in a cabinet unused.

Please send that in ASAP if you can. Im really excited to see more measurements of electrostatics. and the desirable electrostatics amps are manufactured with ccs anyway. so everyone will be happy.

 

[amirm]

In a nutshell, you can’t swap a resistive load for a reactive one and think any valid measurements are going to come out of it.

The load is reactive. It has a 200 pf cap across it. A reactive load will make the situation worse for the amp, not better.

 

[amirm]

These amps intended for electrostatic earphones will shift their operating point when loaded with a resistive load.

So do the regular headphone and power amplifiers. We use resistive loads because no one will agree on what a reactive load should be like. And indeed there is no industry reference to use anything but resistive loads for amplifiers. It is understood that the amp behavior may change with reactive load but it should be well behaved with the simpler resistive load.

 

[LTig]

even with surgery you cannot remove the 100k resistor, you have to add an external attenuator.
to avoid ground loops there is some very tricky and patented circuitry inside the ap that compensates for all sorts of issues that would show up if the input was not fully isolated. There is nothing wrong with an external attenuator.

So for the dashboard (1kHz THD) it should be sufficient to connect the AP via a cap, e.g. a 1 uF non electrolytic. Then the load at DC is negligible even for this kind of amp.

 

[amirm]

Amirm- did you look at the schematic for any of these amps and do any analysis before devising your measurement protocol? If it is to be science, measurement protocol is of prime importance.

Say what? Schematic? Whose schematic? You are saying I can't measure DACs without looking at their schematics?

So no, I look at gear as black boxes with some exceptions. Here, instead of normal 600 ohm load I use, I tested with the highest impedance available to me which is 100 K. I noted that clearly in the review:

So knowledge of the load was important and I approximated it. If it is not good enough, then I am not your answer. Look elsewhere for the data.

I know for me the measurements highly correlated with clear audible distortion as you turn up the volume on the tube amps. So my curiosity is gone.

 

[amirm]

So for the dashboard (1kHz THD) it should be sufficient to connect the AP via a cap, e.g. a 1 uF non electrolytic. Then the load at DC is negligible even for this kind of amp.

The AP is already AC coupled for these tests. I only run down to DC for frequency response measurements (which sometimes I have to override if there is excessive DC). The capacitance used however allows flat response from 10 Hz up.

 

[Francis Vaughan]

Really. The circuit analysis isn’t rocket science. The tubes are run into 66kOhm anode resistors. These form a basic constant current source. The 100kOhm load is split across the two outputs, the mid point essentially being a virtual ground. So the load each tube sees is 50kOhm. Less than the anode resistor. This is a disaster.
Now straight up, the output tube circuit has had its gain reduced to a third by this. This gain comes off the open loop gain of the amp and represents a three times reduction in the feedback available to linearise the amp. Next, it means that the output swing the amp is capable of is reduced markedly. The output circuit is pretty much the same as you would see as a driver stage in a normal tube amp. Typically these circuits are designed to drive loads of about half a mega ohm. Not one tenth of that. The constant current operation of the anode resistor is seriously compromised. It looks more like a wobbly voltage. This means the tube will lose a huge amount of linearity. It sees a dynamical varying load and the current through the tube will not be constant, which pretty much screws all design assumptions and any chance of linearity. The load seen varies in part with the operation of the other leg of the output, which is going to add to the pain. Especially as these are single ended and inherently asymmetric, an asymmetry that is significantly exacerbated by the load.
I really don’t get why there is any degree of support for the measurements. Unless you have both the schematic in front of you and can perform some simple circuit analysis of tube circuits you don’t have any basis to claim knowledge of what is happening.
The 006t and 007t use essentially the same circuit. So if they didn’t measure the same under these circumstances it would be a surprise, not the opposite.

 

[dense]

If you are referring to the tube amps, I would hold my breath. The current numbers do not hold water from a basic scientific point of view. A trivial circuit analysis is quite clear on this point. The output stages are being run way outside of their design parameters when loaded by 100kΩ. Vigorous hand-waving not withstanding.

Whether it's the tube or not, I am not against any personal favors. But from the results I'm seeing, I can't agree with the statement that this product is the reference of a headphone.

 

[Francis Vaughan]

Whether it's the tube or not, I am not against any personal favors. But from the results I'm seeing, I can't agree with the statement that this product is the reference of a headphone.

Well the headphones have not been tested. Just the amplifiers. End to end you may well see a very different result.

 

[amirm]

Typically these circuits are designed to drive loads of about half a mega ohm. Not one tenth of that.

Come again? Where do you get these assumptions? Here are the specs for Stax SR-007A

170K is not 500K as you state.

Here is another:

145K is not 500k.

By your assumptions the amps would perform like hell with real loads as well!

As I said, the best verification of your theories are measurements. If you don't have them, then you are just guessing and conjecturing, not adding anything to the discussion but FUD.

 

[Francis Vaughan]

I was referring to the general style of circuit used as a driver. Which is why I said typically.

 

[Francis Vaughan]

As I said, the best verification of your theories are measurements. If you don't have them, then you are just guessing and conjecturing, not adding anything to the discussion but FUD

Sorry but no. You started the measurements with the assumption that the measuring regime was correct for purpose. That it was the best you could do and not otherwise justified is the start of the problem. You got results you didn’t expect. You are now justifying those results as being good and denying that it might be the measurement setup that is at fault. This is bad science.
If you don’t have the background to understand the circuit you should probably stop digging the hole any deeper.
The impedance you quote is the complex impedance at 10kHz. That is quite simply not the same thing as the resistive load placed on the anode. Spend a little time doing the work to understand the circuit. Tube circuits are not hard but they don’t behave like solid state.

 

[amirm]

The impedance you quote is the complex impedance at 10kHz.

Correct. They don't specify anything else so you can't make the assumptions about half a meg load and such. You are pulling that out of thin air while building a case that impedance is so important.

And what to do about varying impedance between the two samples I showed you? Don't you think the amp needs to be tolerant of good bit of variation in impedance given that they manufacture headphones with different impedances?

And remember, as I noted, distortion measurements on AP are AC coupled. So any argument you are building on DC resistance, doesn't hold.

 

[amirm]

If you don’t have the background to understand the circuit you should probably stop digging the hole any deeper.

I have full knowledge of the load and that is the only thing that matters here: how the instrumentation loads the amp. How an amp works is not my business. It is the designer's. You on the other hand, didn't bother to do a 1 minute search on the impedance of the load and ran off with super faulty assumption. Learn what is important, and what is not.

 

[amirm]

Really. The circuit analysis isn’t rocket science. The tubes are run into 66kOhm anode resistors. These form a basic constant current source. The 100kOhm load is split across the two outputs, the mid point essentially being a virtual ground. So the load each tube sees is 50kOhm. Less than the anode resistor. This is a disaster.

So to go back to this, I showed that the impedance at 10 kHz is 145K ohm for one of the Stax headphones. If you are right, then the performance at 10 kHz should be a near disaster as well. Correct? And worse above that frequency. Correct?

 

[Francis Vaughan]

So to go back to this, I showed that the impedance at 10 kHz is 145K ohm for one of the Stax headphones. If you are right, then the performance at 10 kHz should be a near disaster as well. Correct? And worse above that frequency. Correct?

Harmonic distortion at 10kHz may well be poor. Given even the second harmonic is at 20kHz, this may not have worried the designers. All harmonics above 10kHz are inaudible.
It would be really interesting to know the exact input circuit of the AP here. A more rigorous analysis and comparison of expectations would follow. The AP I assume defines that 100kOhm flat across the audio spectrum. That is what makes it different to what is basically a capacitive load. A 100kOhm resistor with a large DC blocker capacitor is significantly different to a few hundred pF capacitor. I would hope it is 100kOhm at 20Hz. The headphones OTOH will not be.

 

[JohnYang1997]

So please
Someone measure the impedance across 20khz range of typical modern electrostatic headphones like L300.
Someone measure the amps with DC block and without to find out how much difference this makes to the operation point and performance.
Also someone measure the amps with headphones. Compare the input voltages when clipping, relative SPL(power) outputs.