Class D popping up everywhere. Innovation?

[fdeck]

Well take any 3rd year linear systems class in when you get your BSEE and you would understand (if you passed ). Actually a smoke signal (smoke digitally modulated with a blanket) or even Morse code (got these from wiki, sorry) are considered digital. Look up the strick definition. A digital system has discrete time points when things happen and the mathematics of continous linear systems do not apply. Period.

Do you mean a "sampled" system? There are digital systems that have a continuous time axis. Consider a simple combinatorial circuit consisting of a bunch of gates, where the propagation delay time of each connection is affected by jitter due to voltage noise. Any engineer would call it a digital system.

A discrete time axis is not a necessary feature of a digital system.

 

[fdeck]

With an o-scope you would see what looks like a 1KHz sine wave.

Done that. At the output of a Class-D amp, you typically would see the 1 kHz sine wave plus some residual switching hash that is quite effectively filtered by any speaker, but not by your scope.

 

[agedhorse]

Well take any 3rd year linear systems class in when you get your BSEE and you would understand (if you passed ). Actually a smoke signal (smoke digitally modulated with a blanket) or even Morse code (got these from wiki, sorry) are considered digital. Look up the strick definition. A digital system has discrete time points when things happen and the mathematics of continous linear systems do not apply. Period. You can not apply linear equations or use continous time math ( for example Laplace transforms, etc) or integrals. There are digital analogies you must use instead. These are use for the simulations. The simple fact that the digital switching speed is an issue and/or the artifacts are filtered-out should be a huge clue. A digital system may have any number of discrete states. Binary is the term related to two digtal states. Also Bi-stable means digital. There is no stable condition between states. And yes, a class D is a "mixed signal" design and it implies both digital and analog circuits exist in the same system, however usually it is used to refer to an integrated circuit that uses both analog and digital circuits. PWM is a digital modulation technique. A pulse train of a variable rate modulates a binary switch (transistor).
Oh and BTW anything we actually implement today has both digital and analog properties since the switching times are greater than zero. However even though a digital system has analog properties, it can not be considered analog for analysis.

Do what would you consider the residual crossover pulses for a steady state sine wave in a class B amplifier to be... an analog or digital signal? It's a series of time spaced pulses that certainly resemble a digital signal but are created as an analog artifact.

 

[gaffster787]

Do what would you consider the residual crossover pulses for a steady state sine wave in a class B amplifier to be... an analog or digital signal? It's a series of time spaced pulses that certainly resemble a digital signal but are created as an analog artifact.

well , yes exactly... i'd call that cross-over distortion a non-linear distortion caused by devices acting like digital switches because of biasing issues. The AB class with helper caps and diodes improves it; but not completely. As you know, just as digital amps have analog properties, analog amps can have digital properties. When a real audio signal is applied, the spectrum of this non-linear distortion will spread and be less of an issue because of the bell-curve frequency content of real audio that drives the amplitude of the distortion down compared to a sine. Sine waves are not always the best way to characterize an audio system, but at least it is apples to apples. The biggest problem I have with my very light Carvin 500W digital is that I call pull it right off my cab if I walk too far!

 

[agedhorse]

Taking this a step farther, how about the marketing folks getting hold of this and calling an analog amp "digital" bacause it generates a "digital signal". You know those marketing types

 

[Passinwind]

Taking this a step farther, how about the marketing folks getting hold of this and calling an analog amp "digital" bacause it generates a "digital signal". You know those marketing types

I thought an on-off switch was enough grounds to call it digital?

You can pay me later...

 

[fdeck]

I thought an on-off switch was enough grounds to call it digital?

You can pay me later...

Only if you operate it with one finger.

 

[fdeck]

Taking this a step farther, how about the marketing folks getting hold of this and calling an analog amp "digital" bacause it generates a "digital signal". You know those marketing types

I remember when CD's first came out, a number of hi-fi makers came out with new amplifiers and speakers, labeled "digital ready."

 

[Jerrold Tiers]

As far as what you see with a scope........

BEFORE the filter, you see a series of square pulses of varying length. With some imagination, you can see the general sine wave character of the signal (if you put a sine wave in)

AFTER the filter, you see a sine wave (if you put one in) with "wiggles" that are the residual of the square pulses.

The varying length pulses are in fact an analog representation of the signal.

You could think of them as a pulse train of 1s and zeros, with varying numbers of 1s or zeros in each pulse, I suppose. But there do not exist any circuits in a natural sampling modulator to actually count up and assign those varying numbers of 1s etc.

The modulation process is an analog one, with a comparison of the analog signal to an analog ramp voltage. That comparison converts the analog amplitude to an analog time duration. it is a "modulation" process, in which a higher frequency (the switching frequency) is "pulse width modulated" by the signal. To "demodulate" it, a filter strips out the higher frequency by an analog averaging process (filtering) to reconstruct the analog output voltage.

it isn't much different from an old "bucket brigade" flanger as I mentioned. Those 'sample" an analog input, giving a sample that is an analog voltage level, not a digital word. This is passed along and finally output through a simple filter which removes the sampling frequency and leaves the signal.

'digital" implies a conversion to a representation of an analog signal by discrete states having a limited number of possible values (much less than the analog system has). The digital system is a "quantized' discrete time system, where the analog is not "quantized", although it may be somewhat 'discrete time".

AM radio might be considered discrete time (each half-cycle of the radio frequency is a 'sample"), but it is really an analog system. Ditto for the flanger above.

is this anything more than just some fun discussing it? Well after 5 or 6 pages, probably not.....

But the distinction of analog vs digital is somewhat important marketing-wise for the folks who get upset with 'digital" stuff.... if they think it is digital they may decide up front that they don't want to mess with it without even a playing test. Personally, I am of the opinion that Class-D is analog enough that they shouldn't worry, at least if it is done right.

Of course there are those who might not like it UNLESS it was digital. They are welcome to consider class-D to be digital!

 

[gaffster787]

I thought an on-off switch was enough grounds to call it digital?

You can pay me later...

I like it. you could use your power switch to digitally signal the guitarist to turn down. the singer probably would'nt even notice. you make a very interesting observation because power on and power off "transient conditions" are major concerns with amp design. Usually there is some form of relay that disconnects the output from the speakers to protect them from the POP !!!! -until things settle out. All kinds of soft-start issues with power supplies as well, because there is an almost infinite demand for curren at first. And when powering down you must be sure that the residual voltages that linger because of capacitance do not prevent a good reset to digital circuits when power is reapplied. These are the issues that come from years of hard lessons in between the layoffs and outsourcing of our jobs.

 

[gaffster787]

There is alot of truth to Jerrold's statements if you assume digital is a subset of discrete time systems, however that is not the standard definition. Maybe it has become the common usage of the term that digital has to be more than one bit and quantization is involved. However there are many 1-bit Digital to Analog (or A-to-D sigma-delta) chips that sorta go against that idea. Also you can characterize an old school AM radio with a y(t) = blah(t) equation. A baseband signal is mixed with a carrier continously with an analog mixer.
If there in a Y = blah(nT) equation were n = 0, 1, 2,... well that is no longer analog. Sampling is when you mix (or convolve) an analog signal with a digital signal. The rules change, and a class D amp is by definition a DIGITAL amp, as advertised. But I will agree it is a minimally digital system compared to many others. ok, i swear i'm done now but may appear again. does a brewer drink beer on his day off?

 

[lindseyp]

FYI, A digital system uses discrete (discontinuous) values. ON or OFF. Any switching system is considered digital. ...

For some reason I thought that made it "binary," not necessarily digital.

 

[Handyman]

I remember when CD's first came out, a number of hi-fi makers came out with new amplifiers and speakers, labeled "digital ready."

You've got to love the marketing folks. I saw Avatar at an Imax last week, and the little gee-whiz mareting piece at the beginning referred to the theatre's "state of the art digital speakers".

Silly me, still using electro mechanical speakers.

 

[fdeck]

it isn't much different from an old "bucket brigade" flanger as I mentioned. Those 'sample" an analog input, giving a sample that is an analog voltage level, not a digital word. This is passed along and finally output through a simple filter which removes the sampling frequency and leaves the signal.

LOL, the same thing, but with photo-sensitive buckets, just won its inventor a Nobel Prize.

 

[Passinwind]

I like it. you could use your power switch to digitally signal the guitarist to turn down.

I already have a digital TASER for that.

Interesting discussion, thanks all. The first time I looked at a Class D amp's output into a standard resistive dummy load I did quite a double take, even though I knew it might look like it did.

 

[fdeck]

I already have a digital TASER for that.

Interesting discussion, thanks all. The first time I looked at a Class D amp's output into a standard resistive dummy load I did quite a double take, even though I knew it might look like it did.

When looking at clipping waveforms on my scope, to make sense of what was going on, it helped me to add a simple RC filter with a corner of around 50 kHz, well above the audio but well below the switching frequency. Or, if you have a digital scope, turn on the waveform averaging function.

 

[Jerrold Tiers]

There is alot of truth to Jerrold's statements if you assume digital is a subset of discrete time systems, however that is not the standard definition. Maybe it has become the common usage of the term that digital has to be more than one bit and quantization is involved.

The rules change, and a class D amp is by definition a DIGITAL amp, as advertised. But I will agree it is a minimally digital system compared to many others. ok, i swear i'm done now but may appear again. does a brewer drink beer on his day off?

This IS a bit amusing, isn't it?

the definition I am using distinguishes the 'digital" from the "analog" in terms of the signal representation........ Which I submit is a better method for making the decision as far as audio systems.

Thus, in a digital system according to my definition, and this INCLUDES a sigma-delta A-D*, the ultimate signal representation is as a numeric digital word. Of course that directly implies discrete time, and it also has "intentional quantization" into discrete levels. Processing of those words must treat them as numeric mathematical abstractions of the actual signal, and must take into account a periodicity which is not strictly inherent in the signal representation. The values may be directly stored in their numeric form and later output at any arbitrary rate.

Now, an ANALOG system must, in my definition, always represent the signal as an analog value of some sort. It may be a voltage, or it may be, as in the case of class-D, a time period. But the signal, even if sampled into a discrete time signal, must continue to be represented by a "continuous" analog value. This is true of the class-D, and of the bucket-brigade delay, for instance. In the case of the class-D, the periodicity is inherent in the representation, and the values may not be directly stored without further conversion, only the "signal" may be stored in its analog form, even in the bucket brigade chip..

This retains the "sense" of the system, whether or not it meets the "standard definition".

* A sigma-delta modulator is very close to a class-D, and might be approximated by differentiating the output of a class-D to obtain an inverse pulse density modulation. But in the sigma-delta A-D, the pulses are counted and made into a discrete-value digital word representing the sample value. Therefore the result in that case is not "analog", even though part of the A-D is analog and similar to a class-D.

Almost any A-D has parts which can be compared to the class-D, but the results are eventually "digitally integrated" into a number, whereas in the class-D, the result is "analog integrated" to an analog output value.

 

[agedhorse]

Here's another more analog way of looking at a class D output signal. Each infinitesmal time slice represents a bit of energy per unit time. That would be units of power. The instantaneous power varies along the sine wave. Now integrate these instantaneous power slices with respect to time and you have energy.

So, the "area" of each of these discrete and finite time samples represents energy delivered for that time slice and if stacked edge to edge represents the area under the power curve. There's nothing digital about it... except that it looks digital because ofthe time slices and samples and reconstruction but that's the gray area between digital and analog IMO. It's neither yet both at the same time.

A D/A converter does much of the same thing, though with numbers or a mathamatical series of bits.

This could be an interesting AES or NAMM discussion for sure.

 

[agedhorse]

Only if you operate it with one finger.

You don't know how close I came to spraying soda on my monitor...

 

[rust_preacher]

This is also interesting: Comparing a bucket-brigade delay circuit with a digital delay - a digital delay memory fundamentally has digital "words" stored, these can be broken down to ones and zeros, that makes it fully digital. Bucket brigade circuits are considered analog, as a "continuously varying" voltage can be read out of any tap on the bucket brigade delay line. HOWEVER, fundamentally, the "buckets" are charge storage semiconductor wells, and they store charge as electrons. There, one electron could be interpreted as the fundamental digital unit of charge! In practice, you would have to cool the bucket brigade, or CCD circuit to liquid nitrogen temperature to discern charge differences at the one electron level, so for practical audio purposes, the bucket brigade circuits are fully analog.