Class D or AB

[Bob Lee (QSC)]

The output voltage is compared to the input voltage (using
some circuitry for scaling the voltage) to determine which is
greater.

That's not correct. The usual method for class D PWM involves comparing the signal voltage with a very precise triangle wave. This converts the input voltage into a time analog, which controls the switching of the MOSFETs, which alternate between the positive and negative rail voltages, apportioning these voltages by time instead of by controlling a push-pull voltage divider (as would be the case with a class AB amp). A duty cycle of 50% positive and 50% negative would result in the integration filter putting out 0 volts. A duty cycle of 75% positive and 25% negative results in the
integration filter putting out 0.5 × the positive rail. And so on.

This could be done digitally, that is, with the time apportionment done by a digital process, but it usually isn't. I recall one class D amp, made by TacT, that had a digital audio input and a processor that computed the pulse widths from the incoming data streams. That was an entirely numerical process, so it would be proper to call that class D amp "digital." Ironically, the volume control was done through apparently analog means: by raising or lowering the ± rail voltages (i.e., varying the regulator circuitry).

 

[agedhorse]

not to hijack, but I took a shuttle 6.0 to Iraq for 12 months. I pushed it to the limit in 120°F ambient temperature. I had it cranked in duststorms. I even had it cranked on dirty power and this little shuttle 6.0 never quit on me. The only way we lost one of our 2 shuttles was due to user error and turning it on set to 110V with 220V juice; and even then GB fixed the sucker for free! that was 2009. I moved to Japan and I was able to move with my shuttle where she produces the same great sound that she made on day 1.

My Class D is battle effin tested durable!

P.S. The same wikipedia that warnergt cites also states specifically that Class D Amplifiers are not digital.

Yes sir, you have discovered first hand that not all amps are designed and manufacturerd to the same reliability standards. I have supported quite a few of our players who active duty servicemembers, while on their deployments overseas. Since our products are targeted primarily for the more professional, or touring oriented player and low cost is not the primary concern, there are liberties that we can take to improve performance and reliability (we use some mil spec. materials and assembly methods) to where they are just as reliable and any other amp regardless of class. This is much the same approach mil. qualified equipment compares to it's civilian counterparts. Military equipment is designed and manufactured to be used under adverse conditions with high reliability and low cost is not even considered. I think your experience highlights that there does not have to be a difference between different classes of amps, just between different designs and deployments.

 

[agedhorse]

Regarding the analog versus digital discussion, I think this is a much more complicated subject than it might appear on the surface.

IMO, the most accurate way to describe a class D amplifier is an analog amplifier that uses some digital-like design elements.

There are not really logic states, there are "voltage versus time bursts" where the voltage is generated by semiconductor devices (they don't have to be MOS-FETs) operating in either cutoff or stauration. The voltage between "states" doesn't matter except that the maximum output power of the amp is proportional to the square ofthe voltage.

In many of these amps, the sampling frequency of the PWM is not fixed as in the truely digital world. The sampling rate can change with power level and frequency and in some designs this is an essential piece of the performance equation.

The varying width or duty cycle of the pulse train does not generally fit into the digital world as it's not traditional to perform any computations on the pulse stream, nor can the data be stored or manipulated by traditional digital methods.

There is generally continuous proportional feedback between the output and input for (nonlinear) error correction, the same analog issues of stability (phase/gain margin) apply. This is not generally a digital technique.

BUT, the signal is not truely analog either as there is no (or little) desired operation within the linear portion between cutoff and saturation.

IMO, class D is better described as a non-linear analog amplifier. It's unfortunate that the letter D was used, but that's what came after C.

 

[warnergt]

I agree with that. There are steps in the process that are digital. I would have to ask the same question, though:
What digital code is representing the signal at any particular point in time?

Keeping in mind that the output is analog and unquantized, how is this represented as a unique set of logic levels?

I don't think I would call it a digital "code" but it is represented
by a digital binary signal. There are only two logic levels:
on and off, 1 and 0, true and false, positive and negative,
whatever you want to call it.

 

[warnergt]

Discrete physical states are not what distinguish a digital system.

That contradicts wikipedia. I'm not interested in taking up
this argument. Perhaps you should post your definition on
wikipedia and see how long it lasts.

In fact, a digital circuit has voltage levels that vary all over the place. In a 5-Volt logic system, "0" might be represented by voltages from 0 to 1 V, and "1" by voltages from 3 to 5 V. And the voltages swing between these levels during transitions.

In that circuit, the discrete levels would be '0' and '1'. The
voltage is arbitrary because it's digital.

 

[warnergt]

There is no digital, i.e. numerical, representation of anything in a class D amp.

Doesn't matter. Numerical representation is not a necessary
condition to be digital. Go back to my open door sensor
example.

 

[warnergt]

Read the Bruno Putzeys AES paper, link above. He's also an audio engineer with patents in amplifier technology, and he disagrees with you.

How about a link?

Class D amps are largely analog but there is point in the signal
path where it is digital. And that was my point when it was
asked how a class D amp could possibly be considered digital.

 

[warnergt]

Quick jog off topic. In the presence of one of Wiki's founders an interviewer said they were going to add an incorrect definition for Elephant. He shouted, "don't do that!" it would take them weeks to find the error.
It can be a good source. However don't stop thinking for yourself. For more info on how Dictionaries, encyclopedia etc. are assembled read "The Professor and the Mad Man".

I agree. Wikipedia certainly does contain inaccuracies and
many points of contention. On the other hand, it is also subject
to a great deal of scrutiny. If you believe something is inaccurate,
you can go in and change it or challenge it. Topics on science
and engineering generally get sorted out pretty quickly on
Wikipedia because it is usually not that difficult to arrive at
a concensus.

 

[warnergt]

That's not correct. The usual method for class D PWM involves comparing the signal voltage with a very precise triangle wave.

OK, I'll give you that. You are very likely correct here.
I was thinking of a circuit employing feedback. That may not
be the usual case.

 

[BurningSkies]

OK, I'll give you that. You are very likely correct here.
I was thinking of a circuit employing feedback. That may not
be the usual case.

Really? Telling Bob that he's "very likely correct?" with his professional experience and academic standing? The last time this was worked out we also had Mr. Tiers from Ampeg explaining why D isn't digital too.

 

[warnergt]

Really? Telling Bob that he's "very likely correct?" with his professional experience and academic standing? The last time this was worked out we also had Mr. Tiers from Ampeg explaining why D isn't digital too.

I have great respect for Mr. Tiers. He is welcome to challenge
my arguments.

 

[warnergt]

Oh, and I have great respect for Bob Lee, too. He's a smart guy
and a valuable contributor at TB. We just disagree on a point.

 

[fdeck]

That contradicts wikipedia. I'm not interested in taking up
this argument. Perhaps you should post your definition on
wikipedia and see how long it lasts.

From the Wikipedia article: "By contrast, non-digital (or analog) systems represent information using a continuous function." The input voltage, modulation duty cycle, and output voltage of a switchmode amplifier are all continuous functions. (Note that Wikipedia has an article describing the properties of continuous functions).

Look at the common properties of digital information given by Wikipedia. A switchmode power amp (considering for instance the self-oscillating type) has none of those properties. Look at the list of historical digital systems. Conspicuously absent from that list are PWM based and bistable analog control systems such as a bimetallic thermostat.

A switchmode amplifier converts a continuously varying input voltage to a continuously varying duty cycle, and back to a continuously varying output voltage again.

In that circuit, the discrete levels would be '0' and '1'. The
voltage is arbitrary because it's digital.

The voltage can be arbitary because a digital circuit has noise immunity. What makes it digital are the rules for what represents a '0' or a '1.'

 

[agedhorse]

One thing that's missing from a PWM system is the concept of a digital word, or byte of data that is made up of a series of bits. This data can be acted upon under the rules of digital computation (adders, multipliers, etc) to generate a new related byte.

You can't take a PWM signal and do these computations to them, nor can they be stored in conventional digital memory because of the varying time component. This time variation does not have a direct digital equiv. (at least up to the point that I lost track of the state of "current" digital technology).

It's the time varience that makes it an analog like function with digital like properties IMO. Kind of a bi-stable function of sorts. It's really quite a gray area and fascinating too those of us who are involved with this technology.

 

[seamonkey]

How about a link?

Class D amps are largely analog but there is point in the signal
path where it is digital. And that was my point when it was
asked how a class D amp could possibly be considered digital.

Invalid Link Removed

There's a good definition of "digital" and "analogue" in it. Great read.

 

[seamonkey]

BTW - Hearing works from auditory nerve impulses being sent to the brain.
"Pulses" - it's not digital. And thank gawd they're decoding some of this for cochlear implants for those who can't hear. Now they're even singing, check out youtube.

 

[megafiddle]

I don't think I would call it a digital "code" but it is represented
by a digital binary signal. There are only two logic levels:
on and off, 1 and 0, true and false, positive and negative,
whatever you want to call it.

Actually that is a digital code. It just has only bit of resolution. You can represent the state of the switching frequency with only one bit of resolution,
with a 'zero' representing the negative rail voltage and a 'one' representing the positive rail voltage. But as a digital quantity or representation, that's all
you can do. A digital 'one' is nothing more than that. It is like a digital quantum. More importantly, it has no duration. It could represent a duration, as
in a digital clock, but that is not the case here. The logic levels here represent the pos and neg supply voltages.

But here's the important point - the bass signal requires more than just those logic levels to define it. It requires the duration of the those logic levels.
And the digital representation of that duration is not present anywhere in the Class D amp.

Imagine what you would need to do if you wanted to send that "digital" information from a class D amp to a computer for recording. The data stream with
no signal would look like this:
0101010101010101010101010...

And with a signal it would look like this:
0101010101010101010101010...

Something's missing, no?

Now you could say, "ok, I'll just send more 'ones' for longer pulse widths and less 'ones' for shorter pulse widths, like this:
0000011111000011111100011111110011111111...

And that would work except you just added your own analog to digital conversion! That wasn't there in the amp, you had to generate it. Now it's digital.
Before that it wasn't.

Notice that you have also just quantized the pulse width. It's no longer continuously variable. It is now a discrete or integral number of 'ones'. And therefore the
bass signal has also been quantized. A characteristic of digital. In the amp itself, no such quantizing occurs.

 

[RedLeg]

 

[dincz]

As I asked before: Which characteristic of the input signal is represented by the amplitude of the pules?

I can think of no other answer than "none". The on/off voltages are irrelevant in terms of the analog/digital distinction.

Maybe a class D amp would be better described as a pulse width modulator - a device in use for decades and never thought of as digital.

 

[onosson]

While I am not an engineer or computer scientist, perhaps part of the problem here is that some people seem to be assuming that if an amplifier is to be classified as digital, it must necessarily be representing the input signal in a digital manner. But is this really necessary?

If some portion of the amplifier is converting something into a digital signal, even if the bass guitar waveform is represented by analog means throughout the entire signal path, is that an analog or digital amplifier?

In other words, how much digital is required for it to be a digital amplifier?