LSA Voyager GaN 350 Teardown (Class D Amp)

[respice finem]

I guess the real "GaN popcorn time" may come to the forum one day - this is when the first Hypex / Purifi GaN modules arrive...

 

[Doodski]

My preference runs to Made in USA

Yes, I like the SchiiT and JDS Labs stuff. It's priced fairly and closer to home.

I rather like the GaN idea, although not enough to spend $$ to find out.

I think the technology is good although it does not seem too have matured yet at the implementation level.

 

[antcollinet]

Yes, I like the SchiiT and JDS Labs stuff. It's priced fairly and closer to home.


I think the technology is good although it does not seem too have matured yet at the implementation level.

At the moment it seems to be solving the problem "how to put GAN into an amplifier", rather than actually making amplifiers work better.

It has potential for the latter, but an engineer or three somewhere need to work out how to actually do it in the real world. I'd like to see them shift the pulse frequency up as far as possible while keeping losses the same, so the output filter bandwidth can go way beyond 25KHz.

 

[Doodski]

At the moment it seems to be solving the problem "how to put GAN into an amplifier", rather than actually making amplifiers work better.

It has potential for the latter, but an engineer or three somewhere need to work out how to actually do it in the real world. I'd like to see them shift the pulse frequency up as far as possible while keeping losses the same, so the output filter bandwidth can go way beyond 25KHz.

It'll come.
Why should it go beyond 25KHz?
I want a schematic of any one of these snazzy new class D amp we keep running into. I've never seen one.

 

[antcollinet]

It'll come.
Why should it go beyond 25KHz?
...

It seems quite a few class D amps (including this one) have distorted filter FR depending on load impedance, rolling off (or up) well below 20KHz. I'm assuming a wider band filter would make it easier to avoid that. Or maybe it just needs more competent filter design.

Either way a higher switching frequency should make it possible to significantly reduce any distortion products hitting the audio band.

 

[Doodski]

It seems quite a few class D amps (including this one) have distorted filter FR depending on load impedance, rolling off (or up) well below 20KHz. I'm assuming a wider band filter would make it easier to avoid that. Or maybe it just needs more competent filter design.

Either way a higher switching frequency should make it possible to significantly reduce any distortion products hitting the audio band.

Hmmz... I see. I need to get up to speed on class D theory of operation. I just know what I've gleaned here at ASR.

 

[pma]

It seems quite a few class D amps (including this one) have distorted filter FR depending on load impedance, rolling off (or up) well below 20KHz. I'm assuming a wider band filter would make it easier to avoid that. Or maybe it just needs more competent filter design.

Either way a higher switching frequency should make it possible to significantly reduce any distortion products hitting the audio band.

All that have a principle based on hysteresis switching with output LC filter out of the loop. You cannot put it in the one global loop with this principle. So the modulation of FR is inevitable. TI PFFB concept is only a partial cure but it worsens stability. Add a nonlinearity in output inductor that is out of FB. So it is the principle regardless implementation. GaN or other cosmetic changes make no much difference. Yes maybe radical increase of switching frequency might help, still you have the inductor nonlinearity. The only advice is go discrete and make a serious design based on different than hysteresis principle, nowadays it means avoid class D chip of any kind.

BTW I wonder why @amirm does not measure output impedance vs. frequency in his reviews, he has AP and it would give us the answer immediately if the class D amp under test is worth considering.

 

[audio2design]

, nowadays it means avoid class D chip of any kind.

Do you include Purifi in that statement?

 

[pma]

Do you include Purifi in that statement?

It is not a chip amplifier based on hysteresis switching. Would you please use a complete quote next time, not to misquote. Thank you.

 

[orchardaudio]

I'm not really sure what we are looking at in the pic. There are XLR connectors inside the layout that go nowhere and the wires are a bit of a spiders web. I think it's just a pic of stuff that is not really meant to serve a purpose. It's probably just stuff assembled to make a picture and not a actual thing one can buy from what I can tell.

Too bad… I rather like the GaN idea, although not enough to spend $$ to find out. My preference runs to Made in USA, with the UK next. Benchmark DAC/Pre and amps with Quad, Harbeth and KEF (although I think the KEF are manufactured in PRC).

That picture is from a pre-production unit and the wiring has been cleaned up since, for production units.

I offer a 14-day return policy, if you are not happy you may return the units and get a full refund. You are responsible for return shipping though.

I do this to be able to get products into the hands of customers, and it has been very successful, I get very few returns.

 

[orchardaudio]

All that have a principle based on hysteresis switching with output LC filter out of the loop. You cannot put it in the one global loop with this principle. So the modulation of FR is inevitable. TI PFFB concept is only a partial cure but it worsens stability. Add a nonlinearity in output inductor that is out of FB. So it is the principle regardless implementation. GaN or other cosmetic changes make no much difference. Yes maybe radical increase of switching frequency might help, still you have the inductor nonlinearity. The only advice is go discrete and make a serious design based on different than hysteresis principle, nowadays it means avoid class D chip of any kind.

BTW I wonder why @amirm does not measure output impedance vs. frequency in his reviews, he has AP and it would give us the answer immediately if the class D amp under test is worth considering.

I am not sure what you are trying to say here, my amps have feedback from both sides of the LC filter. Here is a block diagram. They are also fully balanced from input to output.

 

[pma]

I am not sure what you are trying to say here, my amps have feedback from both sides of the LC filter. Here is a block diagram. They are also fully balanced from input to output.

Regardless your image, the frequency response into various load speaks for itself, the PFFB concept with 2 loops over hysteresis modulator is ineffective. With complex load it will be even worse.

You and everyone interested might like to read and study Bruno Putzeys's presentation at

https://www.hypex.nl/img/upload/doc/an_wp/WP_AES124BP_A_universal_grammar%20of%20class_D.pdf

to understand limitations of specific class D topologies.

 

[orchardaudio]

Ok that is for that specific amplifier. For My amps the 8ohm load and 4 ohm load response is very close.

Only 0.4dB difference in response at 20kHz between 8ohm and 4-ohm loads.

 

[artm]

AGD amps are universally praised for their sound quality. If we can measure them that would be very telling. If they measure comparably well to their SQ, then the testing methods are valid. If they do not, then the methods are inadequate or incorrect.

AGD's designer knows his stuff and is not just putting in a GaN in a typical layout. He's intimately familiar with the GaN's invention and understands RF circuitry. These put his amps in a higher class. I hope they can measure as well.

 

[ousi]

Actually, this reminds me of a LM3886 amp I ordered from Taiwan. The cables they use and the location of certain things, along with that crappy soldering job. I ended up having to return the amp because it arrived damaged, and some parts were "unsoldered" when it rocked it way from Taiwan to the US. Hopefully it's not the same case as in this amp that the transportation "unsoldered" some of those joints.