class D - anyone?

[petercom]

Just to add fuel to the thread, as it has gone quiet again, what do we actually call a Class D amplifier? Class D does not necessarily mean digital, it just happened to be the next letter in the alphabet after A, B, C. Class D refers to using a Pulse Width Modulated (PWM) signal where the envelope of the width of the pulses follows the signal to be amplified. The high frequency PWM output is then filtered leaving the required signal to go to the loudspeakers.

There is nothing to say that Class D has to have a digital input or that it is really digital at all. All that it means is that the amplifying devices switch between one voltage state and another and stay on, or off, for a time that varies according to the desired signal. As such you can apply analogue feedback from output to input or, as is more common these days, digital feedback where the output is re-sampled, DSP applied, and then the input corrected accordingly. Either way the feedback needed is considerable.

If you want a small, highly efficient, low heat amplifier then PWM can be combined with a switching power supply. The problem then is that you have two sources of high frequency interference to contend with - the muck coming out of the switching power supply and the remains of the pulse waveform coming out of the outputs.

To get an idea of what this looks like check the following link:
http://stereophile.com/solidpoweramp...ro/index4.html

This is one of the better PWM amps where the designer has sought the goal, as epitomised in his analogue designs, of 'vanishingly low' distortion. Whether low distortion is a better goal than high frequency interference I'll leave you to decide.

The interesting comment in the listening notes in this review is the emphasis on 'flat' two dimensional sound. To that I would add 'musically uninvolving' too. In fact these type of amps sound to me exactly like good, high feedback, ss amps.

How far removed these are from the valve amps that this site promotes. The distortion of a 300B SE is palpable in terms of both measurement and audibility. Yet I prefer the harmonically related distortion prone signal from a 300B SE anyday.

What we have to consider IMO is that the level of distortion, and its harmonic relation to the music, is still low enough to be hardly detectable. We have become so caught up in the 'lower is better' spec war that we have forgotten that it is really difficult to hear harmonically related distortion at levels of around 1%. Our forebears knew this in the '40s and '50s but, before subjective reviewing, hi-fi mags went by measurement as a measure of goodness.

Funnily enough there is an article on this in the same issue of Stereophile
http://www.stereophile.com/reference...ard/index.html

Whilst I think Keith misses the point, the message is clear that a relatively high level of harmonically related distortion where the harmonics are naturally graded from the second harmonic downwards is preferable to distortion where odd harmonics predominate. And what it doesn't point out is that non-harmonically related distortion, such as intermodulation, has a pretty nasty effect on the quality of the listening experience.

[vitalstates]

Thanks for that Peter, I was wondering where or why the word 'digital' had sneaked into the discussion.

Is it easy to summarise the differences between this and the class T offerings?

Regards

Ed

I thought that "Class T" was the name used for the same stuff using 'TriPath' chips.

Likewise "Class Z" for stuff with 'Zetex' in it.

[petercom]

Class T is a term coined by Tripath in an attempt (probably successful considering the number of manufacturers that have adopted it) to differentiate their Class D amplifiers from others of the ilk. What they did differently is to use DSP to map the characteristics of the output devices in order to overcome non-linearities. There are other niceties, such as operating at very fast switching frequencies which was novel in its day, but others now do this too.

Tripath make a good, cheap, efficient Class D amplifier on a chip for other manufacturers to implement low heat, low consumption, high power products.

[acorn]

Hi all


I have been following this post with interest and I tend to agree with Peter.

Digital is a nasty compramise and personaly should be left where there is no other alternatives like ie computers, you have so many stages in digital in the converstion proccess which in turn corrupts the signal like quantisation then there is the non- monotonic conversion which is most destructive at low signal levels so they use digital dither to mask this distortion into noise which then results in that cd gritty sound and not forgetting the accuracy of the dac itself an other source of jitter.

what a mess, in theory it may seem all right but in practice its an engineering nightmare, we should scrap it and start from fresh.


Regards Acorn

[vitalstates]

Hi Acorn

forgive me but which are you talking about, digital or class D??

Regards

Ed

[acorn]

Hi ED

I was talking about Digital, but taliking about class D every type of class of amp as its pros and cones, correct me if I am wrong but there as never really been a good class D valve amp it may be possible using silicon to obtain better results but your back to the problem of using to many stages and introducing distortion into the system, dont these class D chips work on this digital principal I am not quite sure ?


Acorn

[pre65]

Hi-could you build a class D (or T) valve amp ?

would anyone want to ?

Philip

[petercom]

As I pointed out above, Class D describes amplification using a Pulse Width Modulated waveform. This is at much higher frequency than the audio waveform and acts as a carrier for the latter. As it is only the Width of each pulse that 'encodes' the audio waveform, the pulse height is immaterial and can therefore be switched between two states, on and off. It can, therefore, be said to be 'digital' in nature.

However most of us think of digital as a method of encoding a signal using a digital 'word' where the signal is sampled and the resulting data encoded in digital form. This type of digital signal has a repetitive timing frequency which allows it to be reconstructed if damaged by transmission. You can also use the digital 'word' to incorporate error correction, further reducing the possibilities of data loss due to waveform damage. This is the basis for the data that is used in computers, Red Book CDs, DVDs and MP3s.

PWM, however, is a halfway house between digital and analogue. Because the width of each pulse varies the pulse repetition is difficult to sense by any error correction method. Correct reproduction of the data thus relies on accurate timing throughout the circuit and freedom from waveform distortion by the amplifying devices. This wasn't possible in early Class D amps which is why they sounded horrible and why Tripath scored an early hit when they worked out how to solve the amplification problems by mapping the characteristics of the amplification devices within the digital feedback loop.

But in modern Class D amplifiers, because you have to sample the signal waveform anyway, the analogue waveform goes through an A to D converter and usually incorporates DSP to handle filtering and feedback before another conversion to PWM. Even if the input is digital it will still be converted to the digital format used in the amplifier. The output of the amplifier is also re-converted back to the input digital format to handle feedback in most cases.

Now, if you think digital is a good thing and a the data in a digital waveform is indestructible then all this is wonderful. However if, like acorn, you are deeply suspicious of all this digital manipulation of the precious data that is a music signal, then why bother?

[NealG]

Very well put Peter. One small point, Red Book CD 'data' has no form of error correction.