GaN Systems Amplifier Eval Board Measurements

[JohnYang1997]

The total power needed to drive a MOSFET gate is:
View attachment 136726

Driving a MOSFET a typical gate drive voltage will be 10 to 12V, while driving a GaN device the gate driver voltage will be 5 to 6V.

Another thing to factor in is that the total gate charge for GaN transistors will be lower than that of an equivalent MOSFET.

In summary, even though the switching frequency goes up by a factor of two the power needed to drive the gate will still be lower than for a MOSFET at a lower frequency.

Here is a very good app note on the subject:
https://www.ti.com/lit/an/slyt664/slyt664.pdf

For my design the total gate drive power needed is about 65mW per GaN FET x 4 = 260mW. I have verified this by measurement.

I was talking about 1.7MHz and up. Different GaN Fets have different total gate charge. Also the driver itself has increasing dissipate on itself with higher switching frequency, although it's also dependent to what gate driver it is. 600-800KHz is of course not that much of an issue. But the next time you double the frequency it will make some differences. Like 450KHz to 900KHz to 1.8MHz. And this is in the context AM frequency range needs to be avoided. If avoiding AM is out of the question, then the context of that comment breaks down.

 

[David_M]

Can someone please explain the sidebands around the switching frequency and its harmonics?

If I'm not mistaken, class D works by varying the duty cycle of the switching waveform to match the amplitude of the input music signal. That mush of sideband noise around the main switching signal represents the amplitude variations of the input signal.

 

[pozz]

If I'm not mistaken, class D works by varying the duty cycle of the switching waveform to match the amplitude of the input music signal. That mush of sideband noise around the main switching signal represents the amplitude variations of the input signal.

Seems cogent. Those sidebands are very consistent.

 

[dronepunk]

Is there an Orchard Audio GaN test in the works?
It seems that this US designer knows how to implement this tech well.

 

[pma]

If I'm not mistaken, class D works by varying the duty cycle of the switching waveform to match the amplitude of the input music signal. That mush of sideband noise around the main switching signal represents the amplitude variations of the input signal.

No, the frequency remains the same in case of PWM modulator with fixed frequency. I can show other class D amplifiers spectra without such effect. Something modulates frequency and finally looks like jitter.
dF here is 53kHz, so it could not be a result of 1kHz input signal.

 

[orchardaudio]

Is there an Orchard Audio GaN test in the works?
It seems that this US designer knows how to implement this tech well.

Yes. It will be tested by @SIY for AudioXpress, just like my previous one here:
https://audioxpress.com/article/fre...d-audio-bosc-and-purifi-audio-eigentakt-eval1

I also posted all of my own measurements here:
https://www.audiosciencereview.com/...amplifier-modules-250-500w-into-8-4ohm.23518/

 

[orchardaudio]

No, the frequency remains the same in case of PWM modulator with fixed frequency. I can show other class D amplifiers spectra without such effect. Something modulates frequency and finally looks like jitter.

I agree for PWM the switching frequency stays the same (that is assuming fixed frequency PWM). My initial thought is that it is jitter as well, but it's too hard to tell without knowing exactly what kind of modulation they are doing.

 

[pma]

This is the spectrum measured at AIYIMA A07 from 500kHz to 700kHz with
1) no input signal
2) 10kHz input signal (2Vp output at 10kHz)

no input signal

and with 10kHz sine input

Side lobes here are at 20kHz (2 x 10kHz) distance from the carrier frequency. As always, more detailed measurements are needed than a general picture.

Spectrum of PWM signals

Now let's feed the A07 with 20kHz sine wave and we get textbook spectrum PWM signal as per (b) above, except for some digital artifacts:

So there is no explanation for the plot that @amirm has measured, in the PWM principle itself.

 

[David_M]

No, the frequency remains the same in case of PWM modulator with fixed frequency. I can show other class D amplifiers spectra without such effect. Something modulates frequency and finally looks like jitter.
dF here is 53kHz, so it could not be a result of 1kHz input signal.

BTW, I never said the frequency changes, only the duty cycle of the (constant frequency) PWM signal is changed as the signal amplitude changes, much like a DSD signal.

 

[georgehifi]

his focus was to keep the level of feedback low to show the advantage of the GaN transistors not needing much. He was not aiming to produce the best performance possible. Personally I wish he had targeted that given that is what we like to see here.

Any chance of doing the couple of measurements again that didn't impress like the others did, but with the feedback set up higher where you think it should be, as this may well flip your conclusion around??.

Cheers George

 

[amirm]

That would require modifying the design which is outside of the scope of what I can do. Too much feedback can cause oscillation so you can't just vary it. It would also require reverse engineering the design which would be non-trivial.

 

[georgehifi]

Too much feedback can cause oscillation

They give the computer program on the data sheet to do it I believe, via the SCAMP connector, so you'd think it wouldn't oscillate, and remain stable.
https://gansystems.com/wp-content/uploads/2020/05/GS-EVB-AUD-xxx1-GS_Technical-Manual_Rev-200526.pdf

Cheer George

 

[georgehifi]

@amirm

https://gansystems.com/webinar-playback-class-d/
44 minute onward shows you how to use the feedback control in real time, so listening comparisons can be made. and as I said the Scamp connection also gives DSP control values to be had.

Cheers George

 

[amirm]

@amirm

https://gansystems.com/webinar-playback-class-d/
44 minute onward shows you how to use the feedback control in real time, so listening comparisons can be made. and as I said the Scamp connection also gives DSP control values to be had.

Cheers George

I haven't watched but I know it has an on off for feedback. Is this what you are saying? I spoke to the designer prior to review and he didn't mention anything about increasing feedback beyond what it provides.

 

[georgehifi]

Hi Amirm
You said you measured it with low setting of the feedback as the designer gave it to you,
"his focus was to keep the level of feedback low to show the advantage of the GaN transistors not needing much"

Not measured at normal/high feedback.
Is there anyway you can re-test this, as the couple of figures you weren't impressed with with the higher feedback could then be very good.
And I believe the DSP access with the laptop program "Audio Canvas III" can do a number of thing also.

Cheers George

 

[ousi]

Looks like there are some consumer amps of reasonable price started using GaNFETs.
https://www.peachtreeaudio.com/products/gan400

Would like to see that getting measured

Previously Merrill was the few others that produce consumer amps using GaNFETs, but even the cheapest one costs 16K !?

 

[mocenigo]

These GaN devices could also be used to push up the switching frequency of classD amps by some factor 2, thus allowing a better (power) bandwidth...and come closer to good classAB designs...next gen Purify???

"come closer to good class AB designs" how?
Unless... why do you want to make the good class D designs worse?

 

[onix]

"come closer to good class AB designs" how?
Unless... why do you want to make the good class D designs worse?

I dunno that Class D should hold class AB as reference designs. IMHO, this forum is about scientific-based true reproducibility (neutrality and no color added) of recorded sound.

What people seem to like about those "other" amps is in fact the color that they add... the color that comes from higher level harmonics rooted in nonlinear distortion, e.g. an organ sounds much better in a church than in your home, due to multiple resonances, and damping both linear and nonlinear.

 

[EJ3]

I dunno that Class D should hold class AB as reference designs. IMHO, this forum is about scientific-based true reproducibility (neutrality and no color added) of recorded sound.

What people seem to like about those "other" amps is in fact the color that they add... the color that comes from higher level harmonics rooted in nonlinear distortion, e.g. an organ sounds much better in a church than in your home, due to multiple resonances, and damping both linear and nonlinear.

You are right to be going for neutrality (just as I am). BUT NONE of the CLASS D's have COME CLOSE to the Benchmark Media units. The CLASS D's that I have heard (unfortunately none are of the units that AMIRM has tested, as I have never even seen one of those class D's) have obvious (to me) non-linearities. Also I may be wrong, but I think that my NAD 2200's hold on to neutrality pretty well (see Amirm's review on this site).

 

[SIY]

The CLASS D's that I have heard (unfortunately none are of the units that AMIRM has tested, as I have never even seen one of those class D's) have obvious (to me) non-linearities.

Which ones would they be? I've had a lot of amps run through here for comparisons, and the good Class D (Purifi, Orchard, nCore...) and good Class A/AB didn't sound any different.