The crucial issue is how the signal is represented. The word 'analogue' sums it up - something electronic (i.e. voltage, current, pulse width) is an analogue of the signal. In a digital system the signal is often represented by some combination of present or missing pulses, but the size of the pulse is irrelevant (as long as it is big/small enough to be recognised as being present/absent).Pafi said:
No. You are referring to Class D as digital when it is not. Class D does not have discrete levels just before the output filter. It has a pulse width - an analogue of the signal.
No. If there are only two controlled switches then it can't be a digital amplifier. If there are 16 controlled switches then maybe it is digital. Class D uses PWM. Some other method (e.g. 16 switched high power current sources) would not be Class D.
It is a real pity that Class D is called Class D because the D makes people think of Digital. Curiously, this effect does not happen elsewhere: a Class E RF PA does not make people think of Energy or Eric or anything else starting with E. A Class F RF PA does not make people think of Fruit or Fibre.
PWM is an analogue signal representation method. Anything using PWM is by definition analogue, even if the modulator uses digital techniques.
Internally, it is a digital amplifier, because it uses discrete amplitude. It does not sample with a fixed interval but at least two times per comparator period, its statistical time frame. With PWM/DM, this slack of statistics is just, what allows the result to have analogue amplitude. But its bandwidth is limited to comparator frequency / 2, the sampling theorem.DF96 said:
No. I just meant, that dither is a common trick for masking non-linearities at low levels. Some Class-D do not need it and are indeed technically superiour to Class-A/B. But i suspect, that TDA8920B has dither.Pafi said:
There we captured the notion of switching range, and while it may be not one established in the internet yet, it is a valuable one. It is not what i wrote -- i forgot a reciprocal--; instead it is ComparatorPeriod/2PiPowerActorSwitchingTime. Explanation: In order to provide energy for only a small impulse, the switch must be active for only a short time. But its on-time cannot be infinitesimally small, as it has inertia usually expressed in parasitic capacitance, on- and off-time and transit frequency. Hence at too lo input levels, it does not switch on but ignores input.
PWM data transfer does not use an analogue. It's uses discrete time division periods to represent any arbitrary one of a discrete predetermined set of numbers. Unlike an analogue system there is no requirement for a continuous waveform. Jumping from 3 to 15 to 5 is valid. Time division data transfer isn't 'less digital' than voltage representations.
Still agree that the overwhelming majority of Class D amps are analogue devices.
Well this is slightly more complicated than that. For a PWM amp the sampling rate is usually the same as the switching frequency for weak input signals. For large input signals the sampling rate converges towards 0.5 * switching frequency.
For very low input signals the on-time of the switches is amost equal. The result of this is that for infinitesimally small input signals the difference in on-time between the two switches is infinitesimally small but not the on-time by itself. So there is no such restriction in resolution for a classic PWM amp. At least not theoretically.
Regards
Charles
Last edited:
No. The pulse width (or mark/space ratio) is an analogue of the signal.rdf said:
No. It is not inherently discrete. There are no numbers, so they can't be predetermined. You may be confusing the output of a digital PWM modulator with the inherent characteristics of PWM as a system. The analogue output of a DAC is restricted to discrete levels, but that is not an analogue property but a restriction imposed by the limitations of the digital system which, in this case, produced the analogue signal.
Thank you! It is this kind of helpful discussion, which i like about this site.
Forget all symmetry for this moment and think of a bucket voltage dropper. In order to produce a small output, on-time must be short. Below switching range, the transistor does not fully switch on but works gradually, hopefully like a linear element. Efficiency is reduced, wasting more power than outputting. Furthermore it is not linear, and its transfer function will vary with the batch, but one can be sure, that it will tend to ignore input.
If I used a PWM data scheme to transfer the following:
"No. The pulse width (or mark/space ratio) is an analogue of the signal."
what would the pulses look like?
I asked you some questions, but "what is the crucial issue" was not among them. Your "answer" is problematic many ways, but correcting them would require many time, this because I choose not to respond in detail.
Do you really want to say that you know better what I wrote than me? If you think I don't know what I wrote, than at least try to quote! I NEVER told every ClassD was digital. All ClassD I developed are analog. I emphasized this many times before. Read much more carefully!
What is the signal? Do you really think only one signal exists in a ClassD amp?
ClassD is a class of circuits, not a circuit. Saying ClassD has levels is nonsense. A particular type can have signal levels. Some types have analog signal level some others have digital. I don't know how many times I have to repeat this to different persons, and how many times I have to show examples of digital PWM.
Because...??? I'm pretty sure you have no idea what kind of circuit I referred, however I linked a datasheet before. It is explicitely called a digital amp by Texas Instruments, it is ClassD, and the simplest application has 2 switches/channel on output. Be so kind to specify what feature is thought impossible by you in the case of that IC!
I dont know why do you think 16 is more digital than 2.
.
Or pulse frequency modulation. Or phase modulation. Or no modulation at all. Or variable digital pattern. Or multi-level modulations. Or the combinations of these. And obviously every analog modulation has a digital version. Yes, literally every single one. PWM also.
Nobody talked about this kind of circuits before you. You should check the subject of the discussion before correcting others.
Fortunately nobody in this topic showed the littlest sign being this stupid. However very strong attempts were made.
SAE J1850. Do you think it was analog?
Last edited:
You maybe inherited 1 of the 3 mistakes in one sentence of Grasso:
comparator freq is bad term, switching freq is the right term, and there can be an fsw/2 limit, but not due sampling, and that limit is not in bandwidth, but in loop bandwidth of a symmetrical, ideal, naturaly sampled PWM amplifier with I type control feedback.
Sampling rate is 2 times the switching freq, but not equidistant in case of naturally sampled PWM, this is why effectively it drops to 1 times fsw at the limits of modulation index. In case of uniform sampled PWM the sampling rate is already decreased to 1*fsw, but it remains the same at any modulation index.
Here every statements are true(ish), only the logic chain and some premises are missing. I know what mistake you wanted to correct, but without the circumstances the reasoning is floating.
Grasso! Audio amplifiers have a symmetrical input range. Idle state is halfway between minimal and maximal signal. If you ignore this, and assume operation point is on the edge, than you obviously get a distortion, but this is not how ClassD audio amps work. No. Symmetry is not maintained in a way of joining together 2 asymmetrical PWM blocks at the limits. You'd better look at and analise at least 1 actual classd audio amp before trying to make general rules about all of them.
"...analogue of a signal..."
Analog (spelled analogue by some people) is an adjective. Analogy is a noun. They are different words with different meaning. Using nouns and adjectives interchangeably makes conversation painfully ineffective.
Analog (spelled analogue by some people) is an adjective. Analogy is a noun. They are different words with different meaning. Using nouns and adjectives interchangeably makes conversation painfully ineffective.
New people, old mistakes repeated. Even if amplitude is discrete, the information is carried by Pulse Width in the classical analog ClassD amp. And in this case it is analog. In some other cases pulse width is also discrete, therefore digital.
The question appears to be meaningless. PWM is used to transfer a value, not a sentence.rdf said:
The music. Originally a time-varying voltage, but in PWM it is a time-varying pulse width.Pafi said:
Never heard of it. What is it? I know little of commercial amps. This is DIY audio.
The meaning is clear. Using an adjective as a noun is a tool sometimes used in English; I believe it is called a 'nominalized adjective'. When overdone it can lead to bad prose. If I meant to say "analogy" I would have done so, but as you say that means something different.
Last edited:
Do you know how a computer works?
All electronics transfer 'values', the point is PWM is not limited to analogs. If the position now is that PWM is incapable of carrying arbitrary data a lot of IR remotes are going to stop working.
I worked with computers for about 23 years, so yes I think I know how a computer works. I was developing, supporting and teaching about real-time SCADA systrm and applications software.rdf said:
If I implied that PWM is limited to analogues then I am sorry. Like all analogue techniques, PWM can be used for digital purposes.
I don't know why people find it so hard to grasp that pulse width is an analogue quantity. Class D uses pulse width (or pulse repetition rate - another analogue quanitity). Class D does not use PWM to transfer numbers or coded sentences; instead there is a relationship between an arbitrary (but bounded) pulse width and an arbitrary (but bounded) signal voltage. Hence Class D is analogue.
All voltages are in the physical world are analogue quantities. The interpretation of that voltage is digital or analogue. Again, I agree that the overwhelming bulk of Class D amplifiers are analogue devices. Many here, you excluded to your credit, can't seem to grasp that PWM can also carry data.
FWIW, I'm embarrassed to admit my computer history begins with the IBM 1130.
- Status
- Not open for further replies.