Hi,
What clock rate did the ADC run at? I suppose you took advantage of the 1.5MHz PWM rate to run the modulator and ADC at somewhat more relaxed rates.
Cheers,
B.
PS. I should add that I know of at least one other person who has been working on ADC feedback. He worked for Crest Audio until management struck and closed down the R&D lab. He's out of a job now. Nice world to live in if even class D designers can be jobless
What clock rate did the ADC run at? I suppose you took advantage of the 1.5MHz PWM rate to run the modulator and ADC at somewhat more relaxed rates.
Cheers,
B.
PS. I should add that I know of at least one other person who has been working on ADC feedback. He worked for Crest Audio until management struck and closed down the R&D lab. He's out of a job now. Nice world to live in if even class D designers can be jobless
In early development we tried just about every combination of PWM / Modulator / ADC clock rate – and the ACTEL FPGA's where not ROM Based!!! Only needing to change the external Intergrator TC / Gain...
Now have many old Actel devices with the cryptic “V3, V3.1 etc etc etc”
The good thing with the old Actel was its not easy to copy the code - an issue when you leave boards with other "3rd Parties"
John
Hey - nobody pays me
Now have many old Actel devices with the cryptic “V3, V3.1 etc etc etc”
The good thing with the old Actel was its not easy to copy the code - an issue when you leave boards with other "3rd Parties"
John
Hey - nobody pays me
An interesting illustration of our different styles of working... I calculate the whole smash once and stick to it. Usually that gets me the optimum straight away. But then again just trying out various things in practice can bring up some interesting new ideas. Like making level translator coils on PCB... very adventurous to my standards.JohnW said:
There there now, next you'll be telling me you're poor.JohnW said:Hey - nobody pays me
Bruno Putzeys said:
Yes, I didn't mean to imply that there wasn't a DAC. I just wanted to say that there was not a sine wave re-created anywhere.
I find this question of DAC choice to be quite fascinating. Is it better to buy a conventional DAC, so that I have a sine wave going into the modulator, or to build a one-bit DAC, so a PWM-type signal is going into the modulator?? (you already answered this question with regards to pre or post-filter feedback)
In the former case, I would want to have the output of the DAC be full scale and then do analog volume control, tone control, etc. so that I run at maximum SNR. In the latter case, I don't really care if the output of the 1-bit DAC is full-scale or not, because I have no crossover distortion in my Class D stage. I could buy a PWM controller which does all the attenuation, tone, balance, etc in the digital domain and spend all my time on the 1-bit DAC and Class D design.
It makes me wonder about the possibility of a modulator which works with either a sine wave input or an analog-domain PWM input. This way I can use my amp with a digital source, like my DVD player, or an analog source, like my television.
You can build a DAC with over 120dB of SNR. With current chips it's not even that hard. If you do that you can dispense with the volume control.dutch_hokie said:
Off the shelf all-in-one PWM controllers do not have that kind of performance. If you want good PWM-DAC performance, you will need to build your own digital PWM controller.
From a practical point of view, the former is the easiest to do. The latter only makes sense if you want to improve THD at high modulation indexes at the expense of output filter correction.
Why you expect that cross-over distortion (or the lack of it) only comes into play when using digital PWM is something of a mystery to me
. Apart from the high-level modulation distortion issue and the output filter thing there are no such crucial differences between the two approaches.
To be digital or not
Hi Bruno, JohnW, others
To return to the original topic:
Consider the power switching signal of an AD modulated switching amp which switches between two states. If it's a digitally modulated signal the waveform will have a modulation that changes time domain parameters, eg pulse width or pulse runlength, in a discrete manner. If you're in marketing or PR you can safely put on your 'digital hat' and say that you have a 'Pure Digital Amplifier'. -1, +1, -1, +1. . . all that matters
But where is the information conveyed in this signal? To answer this we must don our 'engineer's hats'. Is information conveyed in the two states? Yes, but not exclusively. It's also conveyed in the peak to peak amplitude of the signal, in a linear fashion; otherwise your supply voltage wouldn't make any difference to the volume. And information is also conveyed in the nature of the transition between the two states in both a linear and a non-linear manner, otherwise you could employ huge dead times with no effect on distortion. All these things matter, and they matter greatly. Because what we have here is actually an analog waveform!
The quality of the final analog output is inextricably associated with the switching waveform, and all so-called 'Pure Digital Amplifiers' employ on one or more of such waveforms.
I suppose it's a bit late now - the three of us have each having made several 'Pure Digital Amplifiers' - to be told that strictly speaking, there's no such thing!
Cheers
John H
Hi Bruno, JohnW, others
To return to the original topic:
Consider the power switching signal of an AD modulated switching amp which switches between two states. If it's a digitally modulated signal the waveform will have a modulation that changes time domain parameters, eg pulse width or pulse runlength, in a discrete manner. If you're in marketing or PR you can safely put on your 'digital hat' and say that you have a 'Pure Digital Amplifier'. -1, +1, -1, +1. . . all that matters
But where is the information conveyed in this signal? To answer this we must don our 'engineer's hats'. Is information conveyed in the two states? Yes, but not exclusively. It's also conveyed in the peak to peak amplitude of the signal, in a linear fashion; otherwise your supply voltage wouldn't make any difference to the volume. And information is also conveyed in the nature of the transition between the two states in both a linear and a non-linear manner, otherwise you could employ huge dead times with no effect on distortion. All these things matter, and they matter greatly. Because what we have here is actually an analog waveform!
The quality of the final analog output is inextricably associated with the switching waveform, and all so-called 'Pure Digital Amplifiers' employ on one or more of such waveforms.
I suppose it's a bit late now - the three of us have each having made several 'Pure Digital Amplifiers' - to be told that strictly speaking, there's no such thing!
Cheers
John H
Re: To be digital or not
Where "opinion" comes in is precisely the non-engineering bit (ie the salesbabble). Does one try to root out the d-word and explain to everyone that "switching amplifier" or "class D amplifier" is right and "digital amplifier" is wrong? Or does one go with the flow and sell one's product by whatever name that people find fanciest?
I'm an idealist in very much everything I do, so I firmly choose the first option. OTOH I shouldn't blame others for opting for the second.
IMO it's never too late to set the records straight.John Hope said:
Where "opinion" comes in is precisely the non-engineering bit (ie the salesbabble). Does one try to root out the d-word and explain to everyone that "switching amplifier" or "class D amplifier" is right and "digital amplifier" is wrong? Or does one go with the flow and sell one's product by whatever name that people find fanciest?
I'm an idealist in very much everything I do, so I firmly choose the first option. OTOH I shouldn't blame others for opting for the second.
Bruno Putzeys said:
Hi Bruno,
I cannot fully agree with you on this issue - I've seen with a client UCD based product "aliasing" products between the RF leakage of 8Fs DAC and the UCD modules operating frequency. These products where -80db to -100db down Ref. full scale.
At lower power levels (where the modulator clock frequency seems "stable") these where at a "fixed" frequency- however as power output increased these aliasing products started to vary in frequency. Measurements of the UCD modulator confirms that at lower power levels the output looks more like fixed frequency PWM where at higher power (say 10W - 20W) the PWM reaches "full" modulation, and then starts to vary its switching frequency.
This issue arose when on some pieces of music that contains high Bass energy (a Fat Boy Slim track I recall), birdie - tweeting like sounds where heard....
But maybe FAT BOY SLIM sounds better with birds singing in the background - I take my hat off to you...
However this confirms my greatest fear of Class D (non synchronous and synchronous) - the danger of aliasing products, however the danger of the UCD approach is that these get modulated with the music - never a good thing audibly.
Along time ago I encountered similar issues with Class D and my solution was to use a truly spread spectrum clock. Even on "Fixed Frequency" Class D, this dramatically improved measurement & audio quality.
Audio quality was easy to understand - randomize any aliasing products - including any non-linear demodulation products in the Analogue front end - however the measurement improvement was a surprise (maybe one and the same thing) it seems integrators / noise shapers like be "doing something" and not sitting around in an "Idle State"
I did not spend much time on investigating the UCD issue - but I could not understand how such "Low levels" -80dB to -100dB could be so obvious audible - I suspect however that connecting the output of the audio analyzer attenuated the RF leakage level (who knows); I should have listened to the unit with the analyzer connected.
I also tested the RF output of the CD player by directly connecting the audio output to spectrum analyzer, and found nothing I would be concerned about / out of the norm…..
These issue aside (reduced by adding extra filtering on the modules input), UCD is an incredible product – far better then the "Hi-End" CD player used as the front end source I suspect. But I would really worry about using these with the current “fashion” of filter-less DAC’s!
John
Just for the recorded, I consider a Class D amplifier “Digital” as it conveys the audio information in a stream of 1’s and 0’s – this is surly the most basic definition of digital – “Expressed in numerical form”.
My dictionary states: -
“Digital - in which the data is represented by combinations of discrete pulses usually denoted by Os and Is" - now that is clearly a Class D amplifier, you cannot argue with that!
The argument about which element is truly the DAC is also besides the point, DAC means Digital to Analogue converter – the signal at the Reference node is Digital 1’s & 0’s – the ONLY place where the waveform truly resembles the analogue signal is at the output of the LPF on the HBridge – there is no other node where without extra filtering you can recover the audio – unlike a conventional DAC + Analogue input Class D.
John
My dictionary states: -
“Digital - in which the data is represented by combinations of discrete pulses usually denoted by Os and Is" - now that is clearly a Class D amplifier, you cannot argue with that!
The argument about which element is truly the DAC is also besides the point, DAC means Digital to Analogue converter – the signal at the Reference node is Digital 1’s & 0’s – the ONLY place where the waveform truly resembles the analogue signal is at the output of the LPF on the HBridge – there is no other node where without extra filtering you can recover the audio – unlike a conventional DAC + Analogue input Class D.
John
Bruno, JohnW et al,
Don't you sometimes long for the '70s and '80s product brochures that weren't afraid to get technical? I remember the first CD player Philips lauched that came with a whole booklet explaning how it worked. Philips Matchline components were also usually accompanied by simple schematics explaining why an amp our set of loudspeakers was superior over other designs.
Ever since marketing has become all about image and having cool features this has been excluded from the brochures. This is even more apparant in the PC industry where I'm faced with prbabble all the time which is nothing more but a string of meaningless buzzwords tied together. Especially the Taiwanese are inventive with thinking up hollow phrases and features which aren't a feature at all.
Sadly this is all done to convince the customer that they're buying something cool with hip new features. These days buzzwords sell products, never really the actual technology or product specifications.
That motivated me to start my website to offer just that; a different perspective, usually just the cold hard facts backed by my engineering expertise and background. We're currently venturing into home theatre and audio as well, as unlike with PC components, which can be benchmarked, audio components are never really put to the test, but we aim to do just that.
Btw. I tip my hat to both Bruno and JohnW, both of you stay true to your design philosophy which often is a far cry from what marketing specialists, or upper management, would like to see. I think Bruno said it best when he said 'It sucks, but it does so cheaply' which accurately represents the mindset of the Chinese/Taiwanese and their approach to product development.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Don't you sometimes long for the '70s and '80s product brochures that weren't afraid to get technical? I remember the first CD player Philips lauched that came with a whole booklet explaning how it worked. Philips Matchline components were also usually accompanied by simple schematics explaining why an amp our set of loudspeakers was superior over other designs.
Ever since marketing has become all about image and having cool features this has been excluded from the brochures. This is even more apparant in the PC industry where I'm faced with prbabble all the time which is nothing more but a string of meaningless buzzwords tied together. Especially the Taiwanese are inventive with thinking up hollow phrases and features which aren't a feature at all.
Sadly this is all done to convince the customer that they're buying something cool with hip new features. These days buzzwords sell products, never really the actual technology or product specifications.
That motivated me to start my website to offer just that; a different perspective, usually just the cold hard facts backed by my engineering expertise and background. We're currently venturing into home theatre and audio as well, as unlike with PC components, which can be benchmarked, audio components are never really put to the test, but we aim to do just that.
Btw. I tip my hat to both Bruno and JohnW, both of you stay true to your design philosophy which often is a far cry from what marketing specialists, or upper management, would like to see. I think Bruno said it best when he said 'It sucks, but it does so cheaply' which accurately represents the mindset of the Chinese/Taiwanese and their approach to product development.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
That's highly suspect I'd say. As you know the zero order hold function will produce a zero around 8fs, which is where the bass energy will be. Therefore there will be very little or no 8fs present under that stimulus, and therefore not cause demodulation.JohnW said:
You may have to look for some other possible causes, including baseband rubbish in the CD player or steady HF tones higher up the spectrum.
JohnW said:Just for the recorded, I consider a Class D amplifier “Digital” as it conveys the audio information in a stream of 1’s and 0’s – this is surly the most basic definition of digital – “Expressed in numerical form”.
My dictionary states: -
“Digital - in which the data is represented by combinations of discrete pulses usually denoted by Os and Is" - now that is clearly a Class D amplifier, you cannot argue with that!
The argument about which element is truly the DAC is also besides the point, DAC means Digital to Analogue converter – the signal at the Reference node is Digital 1’s & 0’s – the ONLY place where the waveform truly resembles the analogue signal is at the output of the LPF on the HBridge – there is no other node where without extra filtering you can recover the audio – unlike a conventional DAC + Analogue input Class D.
John
Form and content. What you're telling me is that by the looks of a signal you can tell whether it is analogue or digital. I've already explained at length this is not so.
Since it matters whether your signal is 1's and 0's and not 0.9973...'s and -0.002...'s so to speak, the signal is not digital, because not the encoded bits themselves matter, but the voltage and the timing. You are also suggesting that it is the filter that turns the signal into analogue. Not so. It removes the HF from a squarewave, leaving the LF. I've yet to see a filter that takes a digital input signal and produces an analogue output. It takes a squarewave and produces a smooth signal but that's quite a different affair.
Your dictionary is entirely right. Since you cannot write down your reference node signal other than in a very precise oscillogram (because the difference between 0.9973... and "1" matters), your signal is analogue.
It's so simple once you get your head round it.
Maybe - however it's normally only LF energy that has any power content to "Push" the modulator into "Frequency" modulation - it's at this point that any "static" aliasing products start to modulate...
Yes but you are correct it could be any RF products that introduce the spurie - but "birdies" are a result of modulation - as apposed to "whining" which is a fixed tone.... begs the point about input LPF's.... which was the reason of the posting.
Synchronous Class D helps to resolves most aliasing issues (care taken with out of band NS energy structure)...
John
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.