Well of course. I never said anything contrary.
"Digital" means "expressed in numbers".
It does not have to be binary, but of course it is rechnically convenient and thus has been accepted as standard.
But i guess my comment about semantic precision has nothing to do with this discussion (and whether class D may be called "digital" just because even if wrong it has been accepted as handy marketing name by the industry).
You confuse switching with digital. Each modern wall wart adapter contains a switch mode supply. That does not make it a digital device. Same applies to a class-D amp.
A flawed design !
Except some special cases it would be better to use a triangular wave.
The most common misconception is that the representation is digital because of the on-off nature of the "switches". But there is a big difference that no one mentioned so far: A digital (i.e. logic) signal is presented by two voltage ranges (and a third forbidden one to be complete). When the level is within one of the defined RANGES the value is interpreted as a clear High or Low or 1 or 0.
Within the class-d amp we switch between the rail voltages. If a logic signal drops by say 5 % it is still a logical 1. Within the class-d amp it is a5 % lower output voltage. So it can't be called digital at all. It is a duty-cycle modulated analog signal. When it comes to the timing the duty-cycle of a "normal" class-d amp can vary with a theoretically infinite resolution. If i were digital is would be discrete time and discrete value.
The reason why a post-filter feedback amp can perform better than a pre-filter one is because it can compensate the disadvantages of the output filter.
Charles
Yes, a class-d amp could actually be called a voltage controlled switchmode regulator fed by an audio signal.
Regards
Charles
So I've heard it said that all signals are analog. SPDIF and TOSLINK are carrying square wave analog signals. Within those analog signals there could be digital information encoded. It just needs to be interpreted by an appropriate device. If it's not being interpreted by a decoding device, it's just another pure analog signal. Is this the point that's trying to be made? Someone mentioned a digital sawtooth wave. What is a digital wave? Do we ever find these in nature?
Thanks for clarifying. It's confusing not just because it switches, but because it does seem to involve clear ranges where it's on or off. I looked at how the saw tooth interacts with the input wave and where they intersect sets the limit. The saw tooth frequency acts as a clock. The output filter is going to be appropriate for that sawtooth frequency. Is that right? If it's too high for the clock you're going to hear conversion noise, whether that conversion process is pure analog or not. It's the same kind of problem you have with really real digital converters. The thrid forbidden range sounds intriguing!
Perhaps in an "abnormal" class-d amp, like the TAS5825m sporting a "DSP" based PCM code to PWM duty cycle converter, each PWM value /is er, I assume to be theoretically discrete as that directly corresponds to each PCM word.
Given that one "could" AM modulate the output of such a device by time-varying the power supply voltage; that's just introducing an analog process atop the - again what this poor soul sees - an essentially digital one, consisting of those "on/off states" and those "discrete time intervals". In the TAS5825 and the like, anyway.
If you look at the diagram for a TAS5828 - just 3 ticks up in its designation - it clearly shows a functional block "Analog to PWM converter". I would agree that PWM output could have "theoretically infinite resolution".
Where's Mr Risbo when you need him? Probably spending his time doing something really great for Audio. I'd accept his word as a final say on this debate, as I believe he had a significant hand in designing that TAS5825 PCM to PWM converter. I simply assume it delivers discrete PWM states, as that seems logical considering what it is.
I wonder with manufacturers such as Hypex or Purifi have not jumped on the GaN bandwagon. AFAIK they have no products with GaN. Maybe I am wrong.
The two different processes both create a need for filtering - it docent make them "digital"... or the "same"....
//
If you watch the video posted in this thread with the interview with Putseys you will find that he answer this question.
//
I've said it before, once the design criteria has been set for a circuit design, with multi layered circuit board also, and manufacturing procedures have been put inplace, it is very very expensive to scrap and re-do the whole thing again. (maybe in the future they will)
Here is just part of the Hypex manufacturing floor, you can imagine once a design criteria is set to do something, to change it becomes a massive undertaking.
https://www.hypex.nl/media/1d/5a/b1/1664454679/Seetek-factory-10.jpg
BTW: Infineon Tech who has 56,200 employees worldwide and generated revenue of about €14.2 billion ($54,852,200,000 usd) in the 2022 fiscal year. Has just sealed the purchase of GaN Technologies.
https://www.infineon.com/cms/en/about-infineon/press/press-releases/2023/INFXX202310-014.html
Cheers George
Last edited:
Because GaN is not reliable when used for audio. If you look at the FETs used by Hypex, their target application is automotive diesel injector drive. So tough stuff.
But before this, when Purifi for a couple of years ago started the design, they made the choice. GAN has been around for a long time. The answer is in the video. If it had any advantages that really mattered, they would have chosen it. But there isn't any. Not for this application (audio). I trust Purifis knowledge here. Others use it in order to have a sales pitch as they cant design amps on the same level as purify - GAN or not...
//
There have been plenty of extremely knowledgeable people on this forum here explaining exactly why.
I have also made and developed my fair bit of Class-D amplifiers, also explaining why.
But I guess such explanations don't do much when people believe that everything that's for sale is true.
For some reason it also needs to come from some kind of special guru person to make people want to understand it.
Diesel engine? GaN is for F1: https://podcasters.spotify.com/pod/...the-1st-GaN-system-for-fuel-injection-e1qsiuf
(Trying to be funny here)
The problem with audio i guess is that FETs are already fast enough.
A digital sawtooth is often easy to create from a (low cost) MCU. If, for example, the MCU already excists in the solution and the ADCs is used for other functions.
The analog comparator, driver and FETs can be low cost. Analog in. So low cost class D design driving a low cost speaker. Flawed but working. Is the sound analog or digital? Lets say 8 bit sawtooth.
(The class D amp is analog with analog input)
Last edited:
- Status
- Not open for further replies.