GaN-based Class D power amps


The use of GaN-based power transistor tech is now emerging for Class D audio power amplifiers. Seems appropriate to devote a forum thread to this topic. At least 3 companies have commercial class D amps in their books:

Merrill Audio, with their model Element 118 ($36k per monoblock, 400 W into 8 ohms, 800W into 4 ohms), Element 116 ($22k per monoblock, 300 W into 8 ohms, 600W into 4 ohms) and Element 114 (coming soon). 
https://merrillaudio.net

Review of Element 118 at this link:
http://www.enjoythemusic.com/superioraudio/equipment/1018/Merrill_Audio_Element_118_Monoblock_Amplif...

ADG Productions, with their Vivace Class D amp ($15k per monoblock pair, 100W into 4 ohms). (The designer emailed me indicating he has another product in the pipeline.)
http://agdproduction.com
Review of the Vivace Class D moniblocks at this link (warning: link might not work (1/11/2019)):
https://positive-feedback.com/reviews/hardware-reviews/agd-production-vivace-gantube-monoblock-ampli...

Technics SE-R1 Class D stereo amp ($17k per stereo amp, 150WPC into 8 ohms, 300WPC into 4 ohms) 
https://www.technics.com/us/products/r1/se-r1.html
Preliminary review of the Technics SE-R1 at this link:
https://www.stereophile.com/content/technics-se-r1-digital-amplifier
Technics also has a lower priced GaN-based class D integrated amp in their catalog:
https://www.technics.com/us/products/grand-class/stereo-integrated-amplifier-su-g700.html

Anyone listened to or own any of these amps?


Previewcelander

Showing 18 responses by roberttdid

Positive Feedback is .... IMHO a rag. I don't really pull much of any value out of their reviews. Recently read a review of an "exotic" cable and there was virtually no verification of claims made by the company  (some of the claims literally were Wikipedia extracts and not even applicable to their specific material usage). It was error after error. Is it par for the course? Maybe. That was just my latest exposure, and can't say I have ever come away with a "Positive" impression of an unbiased analytical review.
georgehifi,

If you believe this is the actual phase shift of the output of an IcePower based amplifier, can you explain this near perfect 1KHz square wave?

https://www.stereophile.com/images/archivesart/555Belfig2.jpg

If you have that much phase shift at 1KHz and changing considerably over the harmonics of a 1KHz square wave, then there is no way the wave would look square.
Perhaps there are holes in your knowledge that prevents your understanding of how these things work ..... 

1) @atmasphere , this would only be true if there was a difference in phase-shift between the left and right channels:

The ear uses phase information for echo location, so it can alter the soundstage too

2) @georgehifi , I am still waiting for you to explain how this 1KHz square wave for an ICE_Power module with "theoretically" large phase shift is really quite a nice square wave. If there was a large phase shift between all the harmonics that build up this square wave, then the square wave would be a mess. Hint .... the graph doesn’t show what you think it does.
If you believe this is the actual phase shift of the output of an IcePower based amplifier, can you explain this near perfect 1KHz square wave?

https://www.stereophile.com/images/archivesart/555Belfig2.jpg

3) @georgehifi , EPDR only impacts devices operating in the linear region. Class-D devices switch hard on/off, hence EPDR is e meaningless term for Class D.

4) @georgehifi , at 500Khz, for dynamic speakers, a typical tweeter may be around 100-200 ohms. Given a magnetic storage element, but no LC output filter, the harmonic content at 500Khz, given the high impedance of the speaker at that frequency is not going to damage the tweeter (and that’s not even taking into account inductive properties of crossover capacitors). Other speaker technologies may have issues obviously.

5) @georgehifi , the filter does not need to "absorb a lot of wattage". The elements are reactive and not used in a manner where the energy is stored, then dissipated (like a logic gate). Losses would be due to parasitic resistances which would be kept to a minimum.
I would agree. If the roll-off is intentional, closely matched components can also provide well matched phase between channels.

To avoid this both channels have to have enough bandwidth that this is off the table.

For the most part you are probably not lost djones51, which is why I brought up the square wave question, which once again was ignored. If your square wave is "square", then you don’t have phase "issues". There is dogma, and then there is science.

atmasphere’s comment is related to a comment I made w.r.t. consistent phase shift of both channels. He is posing, which is true, but also implementation dependent, that having a high bandwidth ensures reduced phase-shift in the audio band, which ensures there is not much differential phase shift between the channels. However, dependent on the reason for the phase shift, the shift may be by design consistent between the channels without requirement of extended bandwidth. A difference in phase shift between two channels would not be distortion. If the difference is linear in phase, it is the same as moving one speaker a small amount. The the phase shift is predominantly not linear, it could contribute to "smearing" of the sound stage, as positions would vary based on frequency. Keep in mind, most of that processing of position happens at <10KHz which is evidence by measuring timing ability of subjects with reduced hearing bandwidth (mainly from age). Timing being differential timing between ears, not any absolute timing.

Back to linear phase, which is just a time shift (and great thing about using digital filters is you can do complex filtering and maintain linear phase (or not)). Of course, there is nothing stopping a competent analog designer from designing a front end that intentionally adds more phase shift at low frequency versus high frequency to compensate for the amplifiers characteristic phase-shift. In an all digital amplifier, i.e. with digital input, phase shift of the amplification section is near meaningless as you just run a digital filter that compensates for the shift and makes it linear in phase.
1) No, that is a pulse train with infinite bandwidth and perfect phase. It's not reality.

2) That is a 10Khz square wave and actually looks quite good. You don't know for certain that's a phase shift issues without looking at the phase shift (in time!) at harmonics at 30Khz, 50Khz, 70Khz, and given the module bandwidth that is going to cause slew issues.

I have designed quite a few audio products, commercial acoustic spaces, have patents in the field, published papers and have a few related degrees. I have shown I understand the tech, can talk in details without hand waving, and don't need to quote others words.

... And, I am still waiting for your justification for why the square wave I posted, which is a very very good 1Khz square wave, is possible if there are such big phase issues?
Why ricevs? To prevent people from being mislead.
On the other hand @georgehifi won't address the points I make as they have nothing to do with snake oil and they are correcy. That's why he won't address that near perfect 1Khz square wave (but did try to deflect with a potentially slew limited 10Khz square wave that still looked pretty good).
This does not mean that EPC "Invented" GaN. They had the first commercially available/successful enhancement GaN devices, but enhancement GaN devices were made in the late 1990's. EPCs claim to fame was they were the first to successfully grow GaN on silicon allowing them to reliably make low cost cost, high performance devices. They were not the first to do this either, just the first to do it successfully for mass production.
georgehifi10 posts07-02-2020 8:19pm

In June 2009, EPC delivered the first commercial enhancement-mode GaN (eGaN®) transistors.

 


Hey @georgehifi, still waiting for you to explain the perfect 1KHz square wave for a Class-D amplifier, even though you claimed that amplifier had really bad phase shift at low frequency.

Also waiting for your proof that the slope on the 10KHz was due to phase shift, and not bandwidth induced slew limiting. It was still a pretty good looking 10KHz square wave at that.
GaN System is not a subsidiary of EPC, and EPC is not the inventor of GAN FETs, though they had the first commercial enhancement mode GaN.


Ricevs is right, and Georgehifi is right. The boards are the GaN eval boards, but they were developed by Elegant Audio Solutions, a small but reputable consulting company in Class-D audio who did some of the early D2Audio work (not Renesas). So in that aspect Georgehifi is right.

Ricevs is also right in that these are evaluation platforms. They are not designed to get the most of out the hardware, just like any other evaluation board from any other semiconductor vendor. They demonstrate the functionality and they have good (but not great) performance.
Actually it was one of @georgehifi's posts that was removed. Can't say both, but for sure 1.


Not a manufacturer of anything Class-D, and certainly not a troll. Just the guy pointing out that @georgehifi still has not justified how a Class-D amplifier with "awful" phase shift (his claim) is able to deliver a near perfect 1KHz square way. Not to mention why his only retort was a pretty good 10Khz square wave with sloped edges that he claimed was due to phase issues, though was far more likely slew limiting given 10KHz square wave harmonics are all-odd, and would be bandwidth impacted. Phase issues usually show are a more distorted shape by the way.
I am sure we can pull it back ron1264, but it will likely take judicious use of the report button. Unfortunately, some people have very dogmatic views that are not based on a deep knowledge of a topic, but a very thin knowledge. Though ricevs and I posted rather definitive information, there was no change in behavior. This is what happens when you approach a data set with limited knowledge and are unable to understand there are different approaches and different trade-offs that will yield different results. One could use a higher filter frequency and allow more switching residue in order to minimize bandwidth. One could use feed-forward techniques to reduce the impact of dead-time. One could use an all-pass analog filter to compensate for the phase-shift to achieve closer to constant group-delay. These are the things real working designers know (at least some of them), not the knowledge of arm-chair designers.   On top of that, is the knowledge of what is "audible" and what is not. A knowledge that some also lack. Put together it is dangerous combination.
Here we go again, @georgehifi without a justification for a perfect 1KHz square wave and a pretty good 10KHz square wave from a similar Class-D amplifier. I know why. Does @georgehifi?

Here we go again, withe @georgehifi claiming the only way to fix phase-shift (which is not a problem, differential group delay may be), is by increasing the frequency it is like he is totally unaware of:
  • All pass analog filters
  • Feed forward techniques for Class-D modulation
  • Accepting higher levels of switching frequency pass-through (since most speakers don't care and will filter it out themselves). 

When you refuse that others may know more about a topic, in some instances a lot more, you will never progress in your own knowledge.

That's a lot of lack of knowledge all wrapped up in one post.
You can’t fix a broken record @atmasphere , you can only throw it away. @georgehifi refuses to extend his knowledge level beyond 1 graph, for 1 amplifer (that has a similar model has a pretty good 1KHz and 10KHz square wave). It’s pointless. He is out to sewer, and because I have constantly called out his erroneous statements he won’t even respond to me any more. Is what it is.

p.s. Heinlein was my favourite author growing up.
And I quote,

A stable and load-invariant self-oscillation condition is developed for a class D amplifier employing only one single voltage feedback loop taking off after the output filter. The resulting control method is shown to EFFECTIVELY REMOVE THE OUTPUT FILTER FROM THE CLOSED LOOP RESPONSE. Practical discrete implementations of a comparator and gate-drive circuit are presented. A high-performance class D amplifier employing only 14 discrete transistors is constructed. Higher-order extensions of the control circuit are demonstrated which produce extremely low levels of distortion.


https://www.researchgate.net/publication/268383069_Simple_Self-Oscillating_Class_D_Amplifier_with_Fu...
For the slow people at the back of the room. If this technical paper is beyond your skills, just ask questions.


And I quote,

A stable and load-invariant self-oscillation condition is developed for a class D amplifier employing only one single voltage feedback loop taking off after the output filter. The resulting control method is shown to EFFECTIVELY REMOVE THE OUTPUT FILTER FROM THE CLOSED LOOP RESPONSE. Practical discrete implementations of a comparator and gate-drive circuit are presented. A high-performance class D amplifier employing only 14 discrete transistors is constructed. Higher-order extensions of the control circuit are demonstrated which produce extremely low levels of distortion.


https://www.researchgate.net/publication/268383069_Simple_Self-Oscillating_Class_D_Amplifier_with_Fu...