one is really surprised to the amount of "latent" readers lurking
if you read suntechnik's reply thoroughly, I would see that he is obviously confusing "analog with linear". something along the line "switching is not analog so it's digital". my rather aggressive reply (yes it was aggressive) was because of the fact that someone who doesn't seem to understand the difference between analog and digital seems to be confident enough at the same time as to post arrogant replies like the one I quoted above.
Read the papers on hypex.nl, but to put it simply
- the signal on the output of the amplifier is in the analogue domain (it goes up and down in proportion to the input signal - the speaker's never being pushed fully out and fully in by the switching amplifier)
- all the feedback and control is done in the analogue domain
- so it's an analogue amplifier
Zetex (Diodes) claim that their switching amplifiers (eg NAD M2) apply their feedback in the digital domain but I've never examined the documentation closely enough (following reading Bruno's 'class D is analogue' paper) to reevaluate whether this really is the case
one is really surprised to the amount of "latent" readers lurking
if you read suntechnik's reply thoroughly, I would see that he is obviously confusing "analog with linear". something along the line "switching is not analog so it's digital". my rather aggressive reply (yes it was aggressive) was because of the fact that someone who doesn't seem to understand the difference between analog and digital seems to be confident enough at the same time as to post arrogant replies like the one I quoted above.
Then it would have been a bit more elegant to draw suntechnik's attention to the fact that he might confuse "analog with linear".
"Latent lurking"? No you are wrong (to quote you...). I closely follow this thread
why would I respond elegantly to someone who choses to be arrogant without any backup to his claims? this is a rather technical discussion board, not social networking.
Because being polite is a necessary requirement for a forum like this to function properly; it has nothing to do with social networking.
You said to admire Bruno for his politeness, I'd say follow his example.
End of discussion for me.
did I insult the guy? did I call him him names? every post of yours convinces me even more that you have some sort of agenda.
Oh dear. I was a bit disappointed to find that after my posting a link to the "all amplifiers" paper (which makes the argument in quite some detail) there was still some uncertainty left in some readers' eyes but that could be due to the fact that a lengthy paper isn't necessarily a clear answer. Presuming of course that people actually read it before piping up...
But it's not necessary that people get miffed about it on my behalf
Re ADC latency, one of the other linked papers actually shows a working amp with ADC feedback. Theoretically you can live with latency of up to a whole switching cycle, in practice you need a lot less because of the extra phase shift incurred in antialias and noise filtering. But still, some of the new tricks I've worked out over the past two years have staggering implications as to how much delay and phase shift you work around without affecting loop dynamics at all.
Re Zetex: Zetex is essentially the "PEDEC" system, in that it starts off with a "perfect PWM" signal generated in the small-signal domain (i.e. without the nonidealities of the power stage) which is then subtracted from the actual power stage output to form the error signal which is subsequently integrated and used to add variable delay to the switch transitions. The only thing Zetex does that makes it look "more digital" is to convert the integrated error signal to digital and do the variable delay digitally (block diagram pages 51-52 of http://www.hypex.nl/docs/papers/AES124BP.pdf). I've not yet worked out the reasoning behind this added complexity. Performance, though not bad, is not in line with the overhead.
But it's not necessary that people get miffed about it on my behalf
Re ADC latency, one of the other linked papers actually shows a working amp with ADC feedback. Theoretically you can live with latency of up to a whole switching cycle, in practice you need a lot less because of the extra phase shift incurred in antialias and noise filtering. But still, some of the new tricks I've worked out over the past two years have staggering implications as to how much delay and phase shift you work around without affecting loop dynamics at all.
Re Zetex: Zetex is essentially the "PEDEC" system, in that it starts off with a "perfect PWM" signal generated in the small-signal domain (i.e. without the nonidealities of the power stage) which is then subtracted from the actual power stage output to form the error signal which is subsequently integrated and used to add variable delay to the switch transitions. The only thing Zetex does that makes it look "more digital" is to convert the integrated error signal to digital and do the variable delay digitally (block diagram pages 51-52 of http://www.hypex.nl/docs/papers/AES124BP.pdf). I've not yet worked out the reasoning behind this added complexity. Performance, though not bad, is not in line with the overhead.
Last edited:
Because I was using Krell FPB monoblocks to power Magnepan speakers that is why I am so curios about NCORE which seems to be a perfect match but without TCO of a class A amplification (in case it sounds as perfect as Krell class A of course). Do you ever own HiEnd class A 300W per channel amp BTW?one is really surprised to the amount of "latent" readers lurking
OK back to switching nature of a class D amp based on a very simple example with DC (that is where PWM used the most).
Once you are feeding class A or class AB with DC all active devices stays either closed or opened to some extent and reach equilibrium.
However class D amp has to fully switch on & off output devices constantly at a switching frequency (that is 300KHz-450KHz for an audio class D IC commonly) even in case of DC input in order to mimicking DC output.
Class D is an inherently different beast from a class A or AB amp. That is why class D is so efficient.
What might be the benefit of a class D amp except power efficiency?
Since class D output devices either fully open or closed (two discreet states actually and it seems like very few readers of an audio related resources understands what "digital" means really) and switching frequency is fixed all energy "pops" should be precisely equal to each other and could be very well measured.
That leads to an idea that class D amp could generate that energy "pops" without any need for a feedback based on predefined mathematically modeled transform in digital domain.
But thanks for the links to the articles that were kindly provided in the thread it looks like class D amps not very matured enough to operate without global feedback yet. Otherwise class D amp could be easily feed with a PCM or DSD input without preliminary DA conversion.
And "warm sounding" HiEnd class A tubes amps operate easily without the need for a global feedback for decades already that what class D amps claimed to be as good sounding as - right?
In the late 1990's there was a whole horde of people who thought likewise: that power FETs switched with infinite speed and had zero resistance, that power supplies are infinitely clean and low-impedance and that the output filter could be ignored. None of these actually hold in reality. But even if they did, it would only mean that the conversion from digital to analogue could be done perfectly using power FETs. It would not imply that the power fets were doing anything digital. It would only mean they'd be performing their analogue function very well. Since then, people have realised that they don't and they have put on their engineering hat "ok these are the physical limitations of real life circuitry, what do we do?" Engineering is first understanding the science of the real universe, and then trying to get something done based on that understanding.
Indeed, many people think that 2 states means digital. Hahaha, funny, isn't it (poke poke). Since you evidently consider yourself one of the few who do, you know of course how ridiculous that is. Digital means that whoever receives the signal next will interpret the signal as a string of discrete symbols. An ADSL signal is a continuous time warble but it is digital in that the end result after reception is the same bytes that were put in at the transmission end. A PWM signal is analogue because the recipient just takes the physical waveform. Since you're not one of the many who think of digital as something that happens magically as soon as a square wave turns up, you know all of this.
Could you supply a reason why a car would be more mature if it could operate without feedback (i.e. a driver or cruise control)? Could you explain why an airplane autopilot would be more mature if it kept the thing flying straight without feedback? Or is it only amplifiers whose maturity depends on their ability to operate without feedback? It just so happens that cars with no mechanism to detect whether it's still on the road are obviously impossible, as is an airplane that stays aloft with the controls stuck in one position, and amplifiers can be made that do produce a recognizeable output signal without feedback (well apart from emitter followers in the output stage of course, which are just single-component transconductance amps with feedback around them).
Is feedbacklessness an engineering concept at all? Suppose I gave you a black box containing an amplifier, and I charged you to discover whether it used feedback or not? How would you do it? Would you measure it? Would you hear it? If you can only determine if an amplifier is "mature" by opening it and looking at the schematic, doesn't that mean that by necessity you've failed to do so by testing how well the amplifier amplifies signals?
I would think that the mature amplifier is the one who manages to replicate the input signal with the greatest accuracy. Ncore does so 30 times better than the best zero-feedback amplifiers ever made. In addition I would think that it is a sign of maturity for the audio engineer not to cling on to culturally transmitted childhood fears of the feedback bogeyman. To avoid feedback just because it's the done thing to speak badly of it is not very mature at all.
At the peril of quoting myself too much, have a look at http://www.linearaudio.net/userfiles/file/letters/Volume_1_BP.pdf. If after reading this you still think feedback is inherently bad, you've gone religious. In that case you might have accidentally stumbled onto the wrong thread. There are enough feedback thrashing threads to go and find fellowship in.
Last edited:
I know a fair bit of drivers I would like to remove from the feedback loop...
Great paper! Guess audio people should read more Norbert Wiener instead of The Absurdist Sound... And it is not often that you see people who can put Otala in a rational perspective.
I actually forgot to trot out the really good argument. Noise shapers. You can't make "digital" class D amplifiers without a noise shaper. As it happens, these are essentially control loops with lots of feedback around a quantizer. The most extreme case being a 1-bit sigma-delta modulator. It continues to amaze me how zero pheedback phools can enjoy listening to SACD's which by definition require an extremely heavy-handed feedback loop in order to get the audio band free from rubbish, whilst at the same time avoiding it elsewhere in their system. I can only surmise they have never seriously looked at noise shaping or sigma-delta modulation. For instance just today someone was clumsily describing an amplifier that used "pops" (pulses) that cancel eachother out in the long run. So how exactly does one create that signal in the first place if not with an integrating feedback loop that keeps the tally? A "digital" class D amp that first makes a noise shaped PWM signal and then recklessly drives a pair of FETs with it is still a feedback amp like just any other, except that the feedback signal is taken entirely at the wrong circuit node.
Oh well. It's all the better when people decide that they are henceforth audio experts. After all, imagine them declaring themselves to be expert surgeons instead.
Oh well. It's all the better when people decide that they are henceforth audio experts. After all, imagine them declaring themselves to be expert surgeons instead.
So are you saying that you with all your stuffy textbooks somehow know better than I do what my ears hear? That's almost like the people who claim that Feng Shui is "unscientific" despite the fact that I *know* it works!
[ok, what's the emoticon for irony here...?]
perhaps? I tried looking for emoticons saying "somewhat resembling iron" but that got me nowhere.I tried that once, saying that "constant loop gain across the audio band makes the amplifier sound like it has no feedback". But in the long run my embarrassment at such shpiel got the better of me.Bruno the BEST way to do feedback is to use the proper amount of it then HIDE it from the end user's knowledge
Before:
After:
MODS! We need an Irony emoticon!
iron-y
Ah, but *everybody* knows double-blind testing is worthless. Nobody listens to music like that, so it is not a realistic test. And in a double-blind situation there is so much performance anxiety and negative vibrations around that there is no way you will be able to listen properly. Not to mention the people sneering at my crystals and destroying the aura.
Ah, but *everybody* knows double-blind testing is worthless. Nobody listens to music like that, so it is not a realistic test. And in a double-blind situation there is so much performance anxiety and negative vibrations around that there is no way you will be able to listen properly. Not to mention the people sneering at my crystals and destroying the aura.
- Status
- Not open for further replies.