Even if it it is sampled in discrete time steps, it is still analog - an analog sampled system.
The definition of digital is quite clear - representation of a value by numbers.
That is only the case for a class D amp that accepts something like S/PDIF, which is a digital system: numbers representing a value. Otherwise it always is analog - time, volts, current, that sort of (analog) quantities.
Jan
There are analog sampled systems also, but I don't understand how you could think everything is analog. If you have a set of discrete pulse width values, for example 100%, 66.6%, 33.3%, 0%, and nothing else, and frequency is constant (for example 500 kHz), then it is a 500 kHz, 2 bit digital signal. Do you still think it is analog?
A sign or signal can represent a number. But the opposit direction?!? Have you ever seen a number? I havent. It's not a physical object. Your clear definition is quite - false. Did you know that there are different classes of numbers? For example irrational numbers. Can you show me a digital system able to represent irrational numbers? No. Such thing doesn't exist. You may think of integer numbers instead. To be more precise a limited set of them can be represented by a digital signal. But if the signal can represent all rational numbers within a given domain, then it is analog.
I think you are more intelligent and/or open minded (yet sceptic in the right way) than some others. You had good questions and observations, so I could simply wrote instead of correcting mistakes.
I understand what you are saying, but a signal that consists of discrete pulses is not necessarily a digital system. Digital means that the digital 'word' is a number that represents something, and the value is not coded in the width and height of the pulses. A digital signal is for instance a digital audio S/PDIF signal. The music is coded in numbers that make up the samples. The height and width of the pulses in itself do not code the signal, these can change through the system, for instance due to level shifting or reclocking, without the value they represent changing.
In class D the signal is coded not in a number but in the width, the height, the frequency of the pulses, all analog values. Change any of those and the signal changes. Nothing digital here.
Once you get this clear there's no longer any discussion required because it is a very distinct and unambiguous definition.
Jan
Forgot to mention: any number can be represented by a digital word, of course. Irrational numbers - easy. Just do it like you do with pencil and paper. Just define a code, a required precision, and off you go. Just as we define a code to represent floating point numbers in a digital word (s).
Jan
Jan
Last edited:
One of the main reasons why the PWM signal may not be regarded as digital: The physical representation of a binary value is usually a voltage. This voltage has to be within a certain range to be termed as "low". It does also have to be within a certain ranbge to be called "high". And high will alsways be high as long as it is within the predefined range, same goes for "low". I.e slightly imprecise voltage does represent an exact value. In the case of the PWM amp we hav infinite resolution in terms of pulse withd and we have an output signal that is dependant on the feeding voltage which can vary as a function of the output current and PSU properties etc. But here the output voltage does not represent a diescrete value like high or low it IS an analog signal and and if it sags by 10% for instance the output voltage would also sag by 10%. Therefore most class-d amps use analog feedback to correct for these analog errors.
It would of course be possible (and it is of course done, like within TacT or the TI solutions) to use a digital class-d modulator giving discrete timing steps and an oputput stage of very high precision. Such an amp could be called "digital" if some inaccuracy in terminology is allowed. Buit in exact terms such an amp would have to be called a PWM based AD converter and it is therefore a mixed-signal device.
Regards
Charles
It would of course be possible (and it is of course done, like within TacT or the TI solutions) to use a digital class-d modulator giving discrete timing steps and an oputput stage of very high precision. Such an amp could be called "digital" if some inaccuracy in terminology is allowed. Buit in exact terms such an amp would have to be called a PWM based AD converter and it is therefore a mixed-signal device.
Regards
Charles
You make an important point. In digital, the height and width of the pulses can vary (within reason) without having an effect on the output signal, because those variations do not change the number those pulses represent.
In the class D example, changes in pulse height and/or width directly affect the output signal because it is the pulse height and width that represent the signal.
Jan
In the class D example, changes in pulse height and/or width directly affect the output signal because it is the pulse height and width that represent the signal.
Jan
Then could you please answer the question I asked? "...a 500 kHz, 2 bit digital signal. Do you still think it is analog?" Because I seriously don't understand what is your opinion.
You are right, for example because a signal is not a system. But this is irrelevant, since nobody stated "a signal that consists of discrete pulses is necessarily a digital system". You are denying a statement that nobody stated. I don't know why do you think this was relevant. Seriously I don't know. If you have any logical reason for it, please tell!
I never wrote anything suggesting that every signal is digital if it contains discrete pulses. I can show many examples where discrete pulses represent analog signal (analog PAM, analog PFM, etc...). But the particular example I've shown is a digital signal. It, and not everything else. So please don't talk about everything else instead of what I asked, if you don't have to.
So if there are no words, then is it not digital? And if it doesn't represent anything, then is it not digital?
Why not? This is completely new for me. Can you link any reference that contain such restriction? Do you really want to say that everything else can code a digital value except pulse width and height???
Have you ever heard about QAM, ASK, PSK, etc...?
https://en.wikipedia.org/wiki/Quadrature_amplitude_modulation
The digital word is unspeakably wider than you could dream about. (For example the whole universe is digital, because every physical quantity has a quantum (smallest possible changing) except for (maybe) time and length. Thinking voltage is really analog is only a technical simplification, ignoring existance of energy levels.)
OK, so you know a digital signal that is not coded by pulse width. Does this mean it is a general rule, that pulse width cannot code digital value? Obviously no. For example the well known SAE J1850 digital protocol does code digital values by Pulse Width Modulation:
http://img.it610.com/image/product/34a4c8e7510241498021b00c22889048.png
- You are comparing incomparable things. "Number vs. width" is meaningless. They are terms from 2 different discipline (math vs. physics).
- Do you really think that "voltage" in general can be characterised as "analog"? please tell you don't!
- Do you really think if changing some parameters of a signal changes the output means it is analog? please tell you don't!
It is not just ambiguous, but full of serious mistakes. So many that I left some of them unmentioned.
Even if you don't want to learn transmittion technology, you could consider maybe the designers in TI know what they are talking about, and doesn't develop product families that are not exist.
I asked you to show a digital system (...), not to give me ambiguous commands.
Precision? I you talk about precision, then you don't understand what irrational numbers are. If you write a number with tolerance different from exactly 0, then it is not an irrational number. You can specify irrational numbers by giving them names (pi, e, sqrt2, etc...), or writing an INFINITE method to calculate the Infinite number of DIGITS of the Decimal representation.
So making a digital system that codes some of the irrational numbers is possible, but unusable for any reasonable mathematical signal procession.
To distingwish between any of the irrational numbers you need infinite amount of bits. (The same is true for rational numbers also.)
But you missed the important point completely: the word "number" alone doesn't tell anything about quantization, what is the main question. Class and definition domain of the number set is what tells if it is digital or not.
In the case of Analog PWM we have infinite resolution. But this is not universal. In case of Digital PWM the resolution is finite. For example the old TAS5001 http://www.ti.com/lit/ds/symlink/tas5001.pdf
And also every real 1 bit Sigma-Delta A/D has an output whose pulse width is quantized.
Yes, as you said these errors are analog. Not the whole amp. Yes, the supply voltage is analog, it is not constant, and modulates the output, but this is not the operation principle, but the deviation from the principle. The principle is fully quantized (both in time and voltage), therefore digital.
And should we characterize the whole amp by the type of some of the errors in implementation? Will it be free of quantization error just because some other errors can be described as analog? No.
No uncertainity, it is called a digital amplifier. Should we ignore quantization error by calling it analog? I don't think so, neither any other engineer designing such amplifiers. Distinction from traditional (Analog) PWM amplifiers is a must for many reasons including technical and marketing reasons.
I respect your demand for a more precise term. This is reasonable. However I must say I find this demand a change in the topic of the disputation, and would like to treat as a new, different question (or set of questions).
1: Should we call any other name a product family that is already named?
It can be beneficial, if the new name can imply more of the characteristics. But no matter what name you choose, it's always confusing when you call something a different name than others.
2: What should be the name? "PWM based AD Converter"? Please be a little (a lot) more careful in name giving because you already mixed the direction of the conversion. There are other problems with it, for example it exclude any other pattern generation methods like sigma-delta modulation. Should we create many different names for the different sub-types? The more serious problem is it doesn't tell anything about the application: high power signal generation.
3: In what category should we list it? Mixed signal device is fine for me, however you also can call a Power device, or Audio device... For me a logic gate is also a mixed signal device.
Last edited:
"In digital" is a syntax error. "In" must be followed by an object, not a characterisation. And with this error you avoided the possibility to think about the object, this way you didn't have chance to realize what makes a digital transmittion immune to errors: the reconstruction. But reconstruction can be realised only in the receiving device. If you have perturbation on the output of a digital amplifier, then no receiver device can reconstruct it, since the amplifier is the last digital device in the chain, the remaining (speaker, air, ear) are analog. So what you really have found is: you are analog.
With reconstruction after a small perturbation a ClassD amp is also immune (there are examples for this also), but without reconstruction a digital transmittion is also sensitive to perturbations (for example with "pixel accurate reclocking" you can transmit a HDMI signal for very high distance, without it the signal became full of errors).
My concern with class D, any other variation of "switchmode" processing, and digitizing in general, is that it can generate energy at very high frequencies that can cause slewing related distortions in down stream opamps. Obviously such circuits have some kind of low-pass filter on their outputs, but these probably don't get rid of Rf energy completely. Plus I've read that some of the class D amps are unstable driving the reactive speaker loads, and have blown up, taking the speakers with them. I don't know how to tell which class D amps are reliable and safe for my expensive speakers. I've designed switchmode power supplies, which are very similar to a class D power amp, and I know there are many ways to screw up the design.
However I tried my best to make clear every time I posted that there are both analog and digital PWM, it is clear that I failed.
I already wrote that quality of an analog PWM is limited by switching/modulating freq ratio, but I wouldn't call it resolution.
I don't know what do you mean by 1 bit PCM. If you mean listening to only Most Significant Bit of a 16 bit system, then I can say I tried also, and it is awful: you may be able to recognize the song, but not enjoyable at all. But a correctly dithered 1 bit 10 MSPS system is just awesome.
BTW: "quartz" watches with "7 melodies" and games in the 80s had quite enjoyable sound, however they all were 1 bit.
EMI. The problem is even more serious in case you want to listen to radio in the system. A badly designed ClassD amp (or SMPS) can disturb a weak station badly. And this is also a quick method to check if a ClassD device is good or not. Some of them are awful, some are almost perfect.
If an FM receiver works well next to it, then problem in an OPA is very unlikely. Sidenote: not only output can radiate RF, but supply rails, chokes and heat sinks also.
Sensibility for reactive load depends on the feedback. UcD is the best. Every other may have some problems, but can also be perfect, and I don't know easy way to tell without trying it with various loads and waveforms. But a reactive dummy load is much cheaper than a speaker.
This is true for many different systems. In these cases clock frequency and dithering determines quality and depending on these quality can vary from painful to perfect.
I can't. I don't collect such links, and I write on mobile with insanely limited bandwidth. You can search easier. TAS5001 I linked is Digital Amplifier, but if it's PWM or other method - I don't know. Not much info in datasheet.
There is PWM module in many uC. It is definitely a digital PWM, but I dont know if it helps to you or not. There are application notes on them.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.