Anything that gets a review by John Darko is a hard pass in my book. 🤣 I listened to several of the products he reviewed and gave high praise to and they were unlistenable.
Correct, or maybe even if it's .04%, as you can't hear the difference when listening through speakers that are in whole percent figures.
It only becomes obsessive if you stare incessantly at a THD Analyzer and Osilloscopes etc, and don't listen to music.
Remember what gods like Nelson Pass, DanD'Agostino, John Curl etc etc all say, "global feedback is a necessary evil" in many poor designs. The "the best feedback is no feedback" if it can be done still keeping reasonable THD figures.
Cheers George
This is where I thank the laws of nature that evolution decided that mammals needed huge quantities of negative feedback, both local and global.
So, the low global feeback cult has been around what 50+ years now, so exactly what is special about the peachtree amplifier again that makes it somehow groundbreaking? Or is just another design appealing to people who 'believe' in a certain approach?
So, the low global feeback cult has been around what 50+ years now, so exactly what is special about the peachtree amplifier again that makes it somehow groundbreaking? Or is just another design appealing to people who 'believe' in a certain approach?
Yeah, just like the "gods" said the earth was the center of the universe. As it turns out, the best feedback for class d is lots of feedback. One can live in the past in denial or accept the advances of technology and move forward. The "feedback is bad" paradigm has been proven to be false.
As I understood, the design philosophy during that time was "do a good job with no feedback, then add a little to make it even better".
Yes, and it still is but more with BJT's which are more linear than Mosfets.
But now the Peachtree GaN-1 and Carina GaN with Class-D can do it without global feedback also because of half the amount of "dead time" that GaN-fet have over Mos-fet, and that equals much less hf distortion. (they may use some form of pre distortion though)
GaN Tech: Optimum dead-time range depends on load current, but with the GaN FETs, this range is about 50% that of MOSFET. (The 4 A to 20 A GaN FET dynamic optimum range is 5 ns – 20 ns vs. 9 ns – 36 ns for MOSFETs)
Cheers George
Last edited:
I strongly disagree, you are oversimplifying the problem.
Yes, not even look at that number, and IMD that is likely much more important, listen to the thing and see if you like it. Look at the number after the fact and see if there is any correlation.
I am 50/50 on the so-called 'gods'. They generally tend to tell you half the story, and keep the other very important half hidden for job security. So what happens is that a good design gets put in the wrong application and sounds worse than the 'poor design'. You really need to do your job, spend years figuring out what is really is going on.
I believe I have seen evidence where global feedback is put with class D, in the so-called "post" output filter arrangement. It's interesting that a place like TI, rather than putting all the chip efforts in their fully digital class D units, have developed out these analog feedback designs. I assume they sound good. Perhaps better than their fully digital models, which have a different kind of feedback, if any.
Perhaps you just cant get away from it, for the "a little bit better" part of the 50 year old design philosophy.
Yes, and it still is but more with BJT's which are more linear than Mosfets.
But now the Peachtree GaN-1 and Carina GaN with Class-D can do it without global feedback also because of half the amount of "dead time" that GaN-fet have over Mos-fet, and that equals much less hf distortion. (they may use some form of pre distortion though)
GaN Tech: Optimum dead-time range depends on load current, but with the GaN FETs, this range is about 50% that of MOSFET. (The 4 A to 20 A GaN FET dynamic optimum range is 5 ns – 20 ns vs. 9 ns – 36 ns for MOSFETs)
Cheers George
And this is where self oscillating class D has a bit of magic. You use the phase shift of the output filter to help drive the oscillation and in the process get the non linearity of a large inductor included in the feedback loop. So the clever way to approach the problem is to understand the UcD approach. Applying a class A-B amplifier philosophy to a switching design can only be due to not understanding what you are doing OR trying to sell to people who believe what the magazines have been writing without thinking about it.
disclaimer: sellers are free to sell sub-optimal approaches and buyers are free to prefer that approach. Ain't nowt as queer as folk after all.
Perhaps it's now time for the thread starter to summarise and answer the question posted in the title - after all, some feedbacks and hints has been posted. What are your thoughts now, George?
//
//
The Mod crew believes this is a good time to step away from the keyboard and enjoy other things in life. Check back in a few hours.
I haven't posted here in about 20 years, but this got me to, lol.
Earlier in this thread someone asked "why doesn't anyone do feed forward" - i.e., have the amp "know" the impedance of the speaker ahead of time and then adjust the FR of the signal ahead of time, making feedback less important.
What I'm wondering, though, is if the Technics "digital amps" aren't already doing exactly that via their "LAPC" feature. The new SU-GX70 integrated amp costs almost exactly the same as the GAN-1 + Node stack (about $2000). As far as I can tell it doesn't have feedback, only feedforward via "LAPC". It sends test tones to the speaker during setup:
https://us.technics.com/products/network-audio-amplifier-su-gx70
Indeed, as far as I can tell only their $10,000 flagship has feedback, via something they call ADCT. https://us.technics.com/products/stereo-integrated-amplifier-su-r1000
Am I wrong about the Technics amps?
EDIT: Here is a video of the LAPC system playing test tones:
Earlier in this thread someone asked "why doesn't anyone do feed forward" - i.e., have the amp "know" the impedance of the speaker ahead of time and then adjust the FR of the signal ahead of time, making feedback less important.
What I'm wondering, though, is if the Technics "digital amps" aren't already doing exactly that via their "LAPC" feature. The new SU-GX70 integrated amp costs almost exactly the same as the GAN-1 + Node stack (about $2000). As far as I can tell it doesn't have feedback, only feedforward via "LAPC". It sends test tones to the speaker during setup:
https://us.technics.com/products/network-audio-amplifier-su-gx70
Indeed, as far as I can tell only their $10,000 flagship has feedback, via something they call ADCT. https://us.technics.com/products/stereo-integrated-amplifier-su-r1000
Am I wrong about the Technics amps?
EDIT: Here is a video of the LAPC system playing test tones:
Last edited:
Yay! wer'e back on line, hope all the agro has now chilled out.
I thought their SE-R1 (around $20k) was their flagship GaN equipped amp that used linear power supplies as well!, it also used LAPC. Maybe it also has no global feedback. But the real advantage of the SE-R1 was that it was said to also use 3 x higher switching frequency of 1.5mhz instead (because with GaN HF stability it can be done) rather the norm of around 600khz with Mosfets.Doing this allows the low order switching noise passive output filter to be set far higher freuency also by 3 times, and therefore keeping any phase shift from it well away from the audio band.
Cheers George
I thought their SE-R1 (around $20k) was their flagship GaN equipped amp that used linear power supplies as well!, it also used LAPC. Maybe it also has no global feedback. But the real advantage of the SE-R1 was that it was said to also use 3 x higher switching frequency of 1.5mhz instead (because with GaN HF stability it can be done) rather the norm of around 600khz with Mosfets.Doing this allows the low order switching noise passive output filter to be set far higher freuency also by 3 times, and therefore keeping any phase shift from it well away from the audio band.
Cheers George
Last edited:
- Status
- Not open for further replies.