GaN Systems Amplifier Eval Board Measurements

[PeteL]

Orchard Audio has GaN-based amps. They seem to be performing exceptionally well. He's a member here (@orchardaudio), so can probably comment

It's very good, on the same level as Hypex, but 1049 just for the mono module, so two of those, + PSU, + Chassis, connectors and wiring+ assembly time and know how... We are reaching extremely prohibitive here for a DIY thing. Better be good.

 

[Matias]

What I understand from the interview that the intellectual people podcast had with Bruno Putzeys (https://www.audiosciencereview.com/...ellectual-people-podcast-bruno-putzeys.24138/) is that Bruno Putzeys doesn't see much improvement for the Purifi Audio products by switching to Gallium Nitride, he states that the use of Gallium Nitride don't improve (much) the current product. He expects more from Silicon Carbide MosFets in the future, especially for high power amplification (he mentions 10kW).

So Purifi won't (probably) use GaN in their amplifiers, it will be interesting to see if other manufacturers will switch to GaN and if they can produce amplifiers with higher Sinad that can compete with Ncore and Purifi.

More details about silicon vs GaN by Bruno and Lars here:
https://audiophilestyle.com/ca/bits...eaders-qa-with-lars-risbo-bruno-putzeys-r815/

 

[Helicopter]

Fascinating. This GaN transistor technology makes me excited about the future of Class D as it seems to have the potential to strip away half of the heat loss while maintaining high performance standards. I can't wait to see the Orchard amp and others to come.

 

[YSC]

Manufacturer eval boards are not meant to be used this way. They are designed to showcase the technology in this case particularly the efficiency of GaN Fets which this eval board achieves.

It is always up the designers to tweak it and make it better for their purposes.

These are mass produced and will not focus on audio performance. They will not be using the best quality parts because that would make them to expensive. My amps use 0.05% and 0.1% thin film low noise resistors for example, you will never see this on an evaluation board.

And personally I think by not adding a bunch of other techniques to achieve the best performance have a better showcase ability of how much this single part can achieve in their replacement with other parts?

like only chipping the engine of a car can gain 1 sec of lap time sounds more precise in showing off that chip comparing to do tyre change and aerodynamic mods and achieved 5 sec lap time

 

[Plcamp]

Very cool to see, in more ways than one?

The reduction in heatsink requirement should allow tighter layout and counter one concern…dramatically faster current slew rates vs EMI.

Otherwise this looks great, and should yield smaller, higher reliability products with time.

 

[Francis Vaughan]

The GaN devices used are about $10 US each (YMMV). 4 per channel. There are also GaN devices used in the power supply, also about $10 each, 2 off as far as I can see. So there are already about $100 worth of GaN devices in this eval kit. That puts a pretty clear floor under the price of any product.

The efficiency gains seems very worthwhile. At least a halving in heat output.

The eval board is insanely complicated. I'm having a lot of trouble working out how the open/closed loop part of it works. The Renesas D2 sound processor chips don't intrinsically provide for any feedback mechanism. They are providing the PWM, and the DSP, so just where there is a feedback loop isn't readily apparent. But it seems there is one of some form or another. Feedback is software configured. There is a whole row of anonymous glue in between the D2 chip and the TI GaN driver chips. So this would appear to embody the feedback. Pictures are too grainy to allow identifying the chips. ADC and digital feedback into the D2? Linearising the switching stage? Dunno. Would be nice to find out more.

 

[GWolfman]

Thanks!

 

[JohnYang1997]

And personally I think by not adding a bunch of other techniques to achieve the best performance have a better showcase ability of how much this single part can achieve in their replacement with other parts?

like only chipping the engine of a car can gain 1 sec of lap time sounds more precise in showing off that chip comparing to do tyre change and aerodynamic mods and achieved 5 sec lap time

But to me this only shows the performance is mediocre at best. Without further investigation, actually improving the design, no conclusion can be made.
In reality, GaN made higher switching frequency possible, made lower DT possible and that's basically it. Using the same techniques, I can only see 6dB more performance to be extracted. And thus performance is still probably determined by the design techniques/topology.

 

[EJ3]

Very cool to see, in more ways than one?

The reduction in heatsink requirement should allow tighter layout and counter one concern…dramatically faster current slew rates vs EMI.

Slew rates: Something that I haven't seen many people pay attention to lately. Size: rack mountable dimensions are fine & let the air go through to cool. Smaller dimensions are cool too; but too many are doing them too small for good airflow, screwing things up for people who run their equipment hard. Please make the dimensions large enough so that they can be run at full throttle without overheating and without needing cooling fans.

 

[EJ3]

But to me this only shows the performance is mediocre at best. Without further investigation, actually improving the design, no conclusion can be made.
In reality, GaN made higher switching frequency possible, made lower DT possible and that's basically it. Using the same techniques, I can only see 6dB more performance to be extracted. And thus performance is still probably determined by the design techniques/topology.

If it is optimized to gain the 6dB AND the cooling is also optimized THEN there is a win. Either, by itself, is NOT a win & I'll stick with my 2200's.

 

[JohnYang1997]

If it is optimized to gain the 6dB AND the cooling is also optimized THEN there is a win. Either, by itself, is NOT a win & I'll stick with my 2200's.

It's very likely the whole amplifier only has less than 20W to dissipate, and less than 10W in normal conditions. So the size can be made very small without heated up.
6dB is too small of an improvement for the whole range of amplifiers. But 6dB is very much a lot in the sota amplifiers.

 

[YSC]

It's very likely the whole amplifier only has less than 20W to dissipate, and less than 10W in normal conditions. So the size can be made very small without heated up.
6dB is too small of an improvement for the whole range of amplifiers. But 6dB is very much a lot in the sota amplifiers.

that's why I feel this evaluation board did what it intended to, showing off the potential on a as simple as possible design, so to show the improvement is solely from the new component, and I think 6db is a lot for a single component choice? kind of like say, switching a chip/caps can produce it then it will be amazing, it's like don't need to bother for hair picking for 6db in topology, and possibly getting even better SOTA amps or similar to current amps with cheaper manufacturing cost?

 

[JohnYang1997]

that's why I feel this evaluation board did what it intended to, showing off the potential on a as simple as possible design, so to show the improvement is solely from the new component, and I think 6db is a lot for a single component choice? kind of like say, switching a chip/caps can produce it then it will be amazing, it's like don't need to bother for hair picking for 6db in topology, and possibly getting even better SOTA amps or similar to current amps with cheaper manufacturing cost?

GaN fets are expensive. One GaN fet used here can buy you a chip amp with 2 BTL outputs and it will outperform this evaluation board. So...

 

[levimax]

that's why I feel this evaluation board did what it intended to, showing off the potential on a as simple as possible design, so to show the improvement is solely from the new component, and I think 6db is a lot for a single component choice? kind of like say, switching a chip/caps can produce it then it will be amazing, it's like don't need to bother for hair picking for 6db in topology, and possibly getting even better SOTA amps or similar to current amps with cheaper manufacturing cost?

This looks quite complicated for a "as simple as possible design"

 

[Billy Budapest]

THD+N looks much better than most amplifiers out there. If only we could have great SINAD as well, GaN is definitely gonna stomp the market.

SINAD is the inverse of THD+N so I don’t understand your comment.

 

[orchardaudio]

Starkrimson Ultra seems to be a great application of GaN with 500W at 4 ohms.
https://www.audiosciencereview.com/...-gan-class-d-modules-starkrimson-ultra.18644/

Full details and additional test results here:
https://www.audiosciencereview.com/...amplifier-modules-250-500w-into-8-4ohm.23518/

 

[pozz]

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

 

[Vini darko]

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

Probably low feedback design.

 

[EJ3]

Wow, those NAD amps look really nice! Great power and excellent measured performance. I can see why you feel no need to switch to a class D design.

Unfortunately I failed to mention to Amir to go ahead and test them in the 2 ohm mode (I run one of them in 2 ohm stereo for my 2 subs [each sub has dual 4 ohm voice coils & is wired as a 2 ohm unit]) and in bridged mono 4 ohm mode (as that is the way I run the other 2, one for the left channel and one for the right channel). There is no shortage of clean power in my system. Perhaps some can hear what noise there is but I cannot.

 

[kipman725]

At the prices purifi charge they should be using GaN. In hard switched applications they are a dramatic improvement over Si. One reason however may be that they require more design effort. Its easy to break them without knowing why if you don't have the right probing techniques due to things like low margin between turn on Vgs and damage Vgs.