Topping EHA-5

[themad]

17 minutes ago, bobkatz said:

I bought the Topping and got it this week, with hopes it could be a second or extra Stat amp. I was even planning on doing a "bargain stat amp" feature at my new journalistic home "Positive Feedback", where you can find my review of the Audeze CRBNs, for example. Anway, I'm not going to publish, I don't like to publish reviews of bad gear 😞
...

To see where the skeletons are buried in this amp, take a look at my measurements. The measurements to really study is a comparison of the Mjolnir KGSS HV Carbon and the Topping at nominally 90 dB equivalent SPL of IMD 19-20 kHz 1:1 ratio. I think that lurking in that measurement is at least one of the reasons why the Topping sounds so harsh and the Mjolnir sounds so pure. The primary difference tone at 1 kHz in the spectrum does not tell the story. Notice the high frequency side bands near the 19-20 kHz that begin to reveal to us the Topping's nonlinearities. There's more in the attachments. I'm sorry that I didn't do the multitone, I will some day, when I get around to it, and I'm sure it will reveal the skeleton underneath the "golden glow" of this amp. 

 Hi Bob,

Thanks for the great effort and information! I just do not see the graphs for the KGSSHV Carbon measurements. Perhaps the image upload did not work for that graph?

[bobkatz]

1 minute ago, themad said:

 Hi Bob,

Thanks for the great effort and information! I just do not see the graphs for the KGSSHV Carbon measurements. Perhaps the image upload did not work for that graph?

My pleasure! Just uploaded the Mjolnir measurement. Put the Mjolnir and Topping pictures side by side!

[bobkatz]

I'm sorry that the BBS interface does not show the names of the files. The unlabeled ones are square waves at 1 kHz, 10 kHz, 20 Hz, 100 Hz. Next time I upload I'll annotate all files in the GUI 🙂

[bobkatz]

On 7/28/2023 at 7:44 AM, bobkatz said:

I'm sorry that the BBS interface does not show the names of the files. The unlabeled ones are square waves at 1 kHz, 10 kHz, 20 Hz, 100 Hz. Next time I upload I'll annotate all files in the GUI 🙂

Another test picture shown is a polarity test signal. The amp is non-inverting, based on pin 2 going positive to +-marked pin on the Stax connector. I guess that's something. Although, as you would find from my measurements of Stax and CRBN headphones using a cheapo Mini-DSP Ears, cross checked wih a calibrated measurement microphone --- these headphones invert absolute polarity! But you know what, and here's my theory: Those who say they can hear a difference when absolute polarity is reversed made their judgments listening to moving coil loudspeakers --- not electrostatic headphones. A moving coil has an inherent linearity difference between out-going and in-coming due to differences in the spring/cushion of the coil and spider returning to the zero point. And also consider that magnets themselves are very non-linear in their pull at different distances. Not so with capacitive. And an electrostatic diaphragm in push-pull from the stators is  bilaterally-symmetrical, so there should be no distortion or strength of difference between push and pull. Personally I have never heard an audible difference between inverting and non-inverting signals with E-stats.
 

The Audeze CRBNs have their nano-material embedded equally throughout the diaphragm, while Stats with metallized diaphragms are coated on just one side (I think, am I correct?) so at least theoretically there could be a difference between push and pull but we are talking about microns of distance difference here. 

[kevin gilmore]

So, after a few hours in my favorite combination of methylene chloride, acetone, toluene, benzene, and 2 other magic ingredients (basically miller-stephenson ms111 only way cheaper)

Warning kids: don't try this at home. highly flammable, highly toxic, low flash point... Let the professional morons do this

the eha5 discrete module in the process of giving up its secrets.

every single part of the eha5 is a massive pile of crap. it takes a whole room full of highly competent engineers to produce something this fucked up.

i did take the other module, hooked it up to a bench supply and hd800 headphones and got a good listen. wow is this bad.

proof that obviously the lowest distortion possible in amir's s/n test is the one and only measurement that they think matters.

the rest of the specs, not even close.  

for those that don't know, the external power brick is a single 15v switching supply at 3 amps. internally it makes -15 from the +15. except for the fact that the -15 supply seems limited to about 750ma, and it runs 4 amps from this. So maybe if you are lucky you get about 175ma max current per amplifier. No way this is enough to drive those crap transformers.

transistors are 2L, G1 and the dual is k4r

 

Edited August 5, 2023 by kevin gilmore

[bobkatz]

There has to be some measurement that reveals why this sounds so bad. I have Jim Johnston’s Buzz signal as a multitone wav file. You look at the spectrum between 200 and 500 and if there’s distortion product there you know it’s a bad thing. If you can play a wav file in your spectrum analyzer I’ll be happy to send you the wav for testing this module. 

Edited August 5, 2023 by bobkatz

[spritzer]

Anybody who measures for lowest distortion and thinks that is some sort of a golden benchmark, knows absolutely fuck all about amplifier design let alone the history of hi-fi.  I have nothing against meaningful measurements but when something sounds like crap and measures ok... you are doing something wrong.  Same goes for awful measurements, if you sit down and it sounds good then something is wrong.  Like when Amir showed the world he's fucking retarded by not understanding how electrostatic amps work.  A real Schrödinger's cat quantum phenomenon, the device sitting in two states at the same time. 

The issue for me here is twofold, it's a low power amp which is asked to do far more than it can deliver and then we have the whole can of worms of using transformers to step up the signal with all of their limitations.  Anybody who's taken a SRD-7 and some cheap chip amp off ebay knows if can swing a lot of voltage (hence why Stax put a clamping circuit into the later ones) but it doesn't sound all that great. 

 

 

[bobkatz]

@spritzer  I fully agree that if it sounds bad it is bad, regardless of how it measures. And, for example, plain THD is very deceiving at best. I have an expert friend who knows how to interpret differences in the thousandths of a percent. But that’s beside my point.

Given the sensitivity of modern test gear that is able to measure noise floors below our perception, there ought to be a test signal that exercises the demon that we audibly perceive. And we should  not cease to seek those test signals, because they can lend objective support to our subjectivity. 

[kevin gilmore]

so you say you should be able to measure why it sounds like it does...

first 2 are 1khz into 50 ohm non-inductive load. notice that the rise time and fall time are not identical. also notice severe slew rate limiting in both directions and kink on the positive slope.

3rd picture is 10khz.

kind of what you would expect from an amplifier with an open loop gain of around 120db and a group delay of 10us or more.

i don't have the transformers, but i'll bet it looks even worse driving that transformer.

pretty sure the a90d will perform exactly the same.

edit: see pictures in next post.  definitely same as a90d

Edited August 6, 2023 by kevin gilmore

[kevin gilmore]

success soon

yep, 39 transistors including 2 dual matched npn.

 

Edited August 6, 2023 by kevin gilmore

[spritzer]

Fuck this is dire, sure they were just on the cusp of making the best electrostatic amp there ever has been.   

That is a lot of transistors, any joy with piecing it all together? 

[bobkatz]

Why so many transistors?
 

anyway, It’s real ironic. Seeking a good, economical way to help “stat-o-philes” to get into our hobby and with an amp that has more drive capability than a typical Stax amp.
 

I already own an expensive KGSS, and it sounds fantastic. Thank you Kevin and Spritzer  

 I thought it would be “simpler” to build a lower voltage amp driving a great transformer than to build a direct drive high voltage amp. 

But the step up transformer way has its own challenges more than the obvious problems of the non-linearity of the transformer. The problem is an extremely low impedance primary that’s very difficult to drive. I know, I just measured impedance and phase shift of a great toroid. Holy shit that’s tough to drive!
There’s no such thing as a free lunch 😞

We shall see if I can come up with anything. And still keep the power supply to +- 24 volts. That’s my intentional design constraint

 

best wishes to all  

 

Bob

 

[kevin gilmore]

sure, +/-24v

at 5 amps each rail

balanced drive

large heatsinks

custom hand wound bifilar transformer with a really large cross section to keep the flux as low as possible and magic hybrid core ferrite materials.

$$$

 

 

Edited August 7, 2023 by kevin gilmore

[spritzer]

Yeah, there are no shortcuts here.  You could run a simple Class A amp off +/-24V rails but I'd stay away from the old Hood design.  It has an output capacitor (as it is a single rail design) and that is just asking for trouble with this kind of load plus limited voltage swing. 

[Kung]

great job, only here in the world can see such wonderful reviews of electrostatic amp 

[kevin gilmore]

its easier to do the board, then let the software turn it back into a schematic

update: should be final schematic.

that is a 3 pole (18db) low pass filter at 18khz. part of schematic now.

 

 

 

 

nfca3.PDF

Edited August 22, 2023 by kevin gilmore

[thebrunx]

On 7/28/2023 at 1:09 PM, bobkatz said:

I bought the Topping and got it this week, with hopes it could be a second or extra Stat amp. I was even planning on doing a "bargain stat amp" feature at my new journalistic home "Positive Feedback", where you can find my review of the Audeze CRBNs, for example. Anway, I'm not going to publish, I don't like to publish reviews of bad gear 😞

What a disappointment this amp turned out to be. My disappointment came in stages. The first stage was an optimistic listen: "Hey, not too bad". At that time I was listening to some simple classical and folk arrangements: with solo voices and guitar, the weaknesses of the Topping circuit are not immediately obvious. But as soon as I played more complex music, full orchestra, rock band, you name it. Where it starts to get real harsh, bloated, lose dimension is with complex music, full orchestra, rock band, you name it. The more I listened, the more I started to hate this amp. 

I WILL BET THAT THERE ARE MEASUREMENTS THAT WILL REVEAL THIS AMP'S WEAKNESSES. You just have to know how to find out where the bodies are buried. I did find some that clearly show its weaknesses. Attached below. Frequency response (with a 100 pf load), Output level just before clipping, THD at various frequencies at normal levels and just before clipping (spectrum and %), IMD (19-20 kHz) (spectrum and %). The one thing I should have measured is a special multitone signal I use, I bet that would show where the bodies are buried, and tell us why the amp falls totally flat the more complex the music you put into it. But I'm so disenheartened and disappointed by the sound of it that I don't have the energy to bring it back to the test bench. 

I like to display an amplifier's measurements in equivalent SPL, based on Stax 007 and CRBN nominal sensitivity of 100 volts RMS = nominally 100 dB SPL. It helps bring a real-world perspective to amplifier measurements. 

Topping Factory spec for max is 700 volts RMS (I suppose the factory took this a cat's hair below clipping and only at 1 kHz). I measured maximum output at 1 dB below clipping at 630 volts RMS at 1 kHz, equivalent to 116 dB SPL. But When performing the 20 Hz THD test, the amp's DC protection circuit kicked in and the most I could get out of it at 20 Hz with a continuous sine wave was the equivalent of 105 dB SPL, 177 volts RMS before the amp shut down. So the lower the frequency, the worst the amp's headroom for transients. Plus, the THD at low frequencies is pretty bad, the transformer saturates very strongly.

To see where the skeletons are buried in this amp, take a look at my measurements. The measurements to really study is a comparison of the Mjolnir KGSS HV Carbon and the Topping at nominally 90 dB equivalent SPL of IMD 19-20 kHz 1:1 ratio. I think that lurking in that measurement is at least one of the reasons why the Topping sounds so harsh and the Mjolnir sounds so pure. The primary difference tone at 1 kHz in the spectrum does not tell the story. Notice the high frequency side bands near the 19-20 kHz that begin to reveal to us the Topping's nonlinearities. There's more in the attachments. I'm sorry that I didn't do the multitone, I will some day, when I get around to it, and I'm sure it will reveal the skeleton underneath the "golden glow" of this amp. 
 

The date in the Fluke scope pictures is wrong... it's actually yesterday, 7/28/23. I have to fix the date in my Fluke Scope. Every time the battery dies, it loses the date 😞

 

Post the equivalent measurements of the Carbon for comparison please. 

[kevin gilmore]

So a person accused me of faking my scope picture, because it had to be my signal generator. And that the amp performed better at the higher gain setting.  I completely disagree.

Q15 has no emitter resistor and the resulting gain of just this transistor is 30db. which makes the miller capacitance absolutely huge.

which is why you see two different slew rates. the first bit is due to the output amp (everything left of r12 which is a diamond buffer driving a pair of current mirrors driving another diamond buffer). The second bit is due to Q15.

edit: shenzen audio is now selling these things brand new in the box on ebay for $50 off. soon it will be $100 off.

Edited August 27, 2023 by kevin gilmore

[Juansan2]

25 minutes ago, kevin gilmore said:

So a person accused me of faking my scope picture

Sorry but what the absolute f*ck is wrong with some people?

[Fitz]

26 minutes ago, kevin gilmore said:

So a person accused me of faking my scope picture

 

Do.. do they not know who you are?

[kevin gilmore]

well if i used one of my digital scopes, then did an edit on the computer and then played it back on the scope, its probably fairly easy. but on a real analog scope, it would be some kind of impossible.

[Fitz]

The question is less about the possibility of faking it, than it is your track record for this kinda stuff. It's not like you're some random yahoo showing up with "proof"

[bobkatz]

On 8/25/2023 at 12:54 PM, thebrunx said:

Post the equivalent measurements of the Carbon for comparison please. 

I think I did put up a Carbon measurement in this thread. If it's not here I'll have to dig up the measurements and show you all. 

[Juansan2]

21 hours ago, Fitz said:

The question is less about the possibility of faking it, than it is your track record for this kinda stuff. It's not like you're some random yahoo showing up with "proof"

Exactly!

[spritzer]

For some reason, people really drank the Cool-Aid on this stuff so anything said about is fake or an attack.  It can't be that it is just a crap product...