Digital amplifiers

[Dave2002]

On another thread mention was made of digital amplifiers, and Class D amplifiers - http://www.for3.org/forums/showthrea...725#post314725

To avoid complicating things further I though it would be interesting/useful to have a discussion here.

Gordon hints that some digital amps may not use feedback to control linearity problems - I hadn't thought of this specifically.
Class D amplifiers are seemingly not the only kind of digital amp. It was mentioned that there are others, E,F ... I think going up to H. I know hardly anything about these - I come from Barcelona.

So-called Class T amps are a development of some amplifiers based on audio chips made by a now defunct company called Tripath.

A big feature of some of these amps is that they need fairly low power to drive them, as because of the switching in series with the load (the speaker) they are either open circuit - no current flows, or completely closed - very low impedance, so all the current flows through the load. Thus the amp itself is very efficient and effective at driving some speaker loads.

A lot of AV amps use this kind of technology.

It might also be of interest to have a new thread on active speakers.

[Stephen Smith]

Dave
I tried to PM you yesterday, but the system says that it is not possible to send because your inbox has reached its limit.

(Apologies I can't discuss the various amps - I have no electronic engineering capability.....)

[David-G]

It would be enlightening, and aid my comprehension of this discussion, if someone could briefly outline what a digital amplifier is! Thanks.

[Dave2002]

It would be enlightening, and aid my comprehension of this discussion, if someone could briefly outline what a digital amplifier is! Thanks.

David-G

There are two obvious "definitions":

1. An amplifier which allows direct digital inputs, such as USB, SPDIF etc. Such amplifiers could use normal analogue circuitry, but have integrated DACs at the front, so that everything can go in one box.

2. An amplifier which uses some form of digital processing in order to achieve the amplification. Class D amps do this, as they take the relatively low frequency signals - which could simply be analogue input signals - and then raise the signal to a very high frequency by some modulation process. It is then possible to switch power transistors on or off to achieve the amplification, and as Gordon has pointed out there can be a lot of crud in the process which needs to be filtered out, though perhaps this might happen naturally as most transducers are not going to respond at the high frequencies used in this process.

Note that a Class D amplifier does not have to "digital" in the sense of 1 - it may not have a DAC, and may not be able to accept any form of digital inputs directly. There are other forms of amplifier which can be considered digital - as I mentioned before. I'd like to know more about them.

Clearly it should also be possible to do the following:

3. An amplifier which combines the ability to input digital signals and also uses digital processing techniques in order to achieve the amplification to drive the speakers.

There is also the possibility of having digital preamps or mixers. These could take in multiple digital sources, and then perform combining operations, adding, blending and filtering on each channel. I suspect that many professional mixers are now digital, but there probably aren't many in the domestic market. For example, if one wanted a treble filter, this could be done in the digital domain, without using any analogue circuitry. I know bass, treble filters, and equalisers are not always recommended for high quality sound, but they are useful in some situations, and they can be implemented using digital processing.

[Gordon]

As I said in the other thread, "digital" amplifiers are nothing special at all, there is no magic. Amplifiers are man made devices with man made defficiencies. The various letters simply denote the different ways "amplification" occurs and to what degree they are effective at what they are intended do do. EG Class C is almost entirely used in RF work.

So what do we mean by Amplification anyway? Amplification in general is a very simple affair; all you want is a precise replica of a given low level signal but at a much higher power. Simples!

Usually in audio we mean amplification of the voltage - which inherently embodies the signal fidelity - from a low level to a higher one. In a pre-amp this is almost certainly done in Class A. Further along the audio chain amplification is needed from perhaps a few hundred millivolts into high impedance [say 10 KOhm or more, leading to a pitifully small power - V squared over R] to enough volts to drive a loudspeaker, typically many Watts into a very low load. In the other thread we explained at length why this was fraught. Class A can be usd but it wastes a lot of power, B and AB simply share the load more efficently among several output devices - a distributed amplifier model.

At low levels digital amplification is trivially easy - take the binary numbers [assume PCM for now] and shift them in significance so that say the given LSB is moved to be the next one up and replace the old LSB with a random 0 or 1 or, better still, compute it using the gain factor you want, also expressed in binary.

Example: take an audio 16 bit sample: 0110 0110 1010 1101. Increasing its amplitude [as seen by a DAC] by 2 [let's keep it simple] gives: 1100 1101 0101 101x where x is the new LSB which is not known from the original 16 bit expression so it has to be estimated. This is all done by a simple binary point shift.

To reduce the gain by 2 shift the bits to the right instead. In decimal arithmetic, which you all do all the time, you do the same thing - to multiply by 10, the base of the number system, shift the decimal point.

If the gain factor needed isn't an easy form [ie a power of 2, like 4, 8 , 16 etc] but is itself a longer, fractional binary number then a shift won't do; but some simple logic will produce the right answer. The fact that the new LSBs are estimates illustrates the erosion of precision that results from the starting point of 16 bit - you can't create information you never had any more than in analogue you can't increase gain without also increasing the noise that comes with it. The same applies at 24 bits, a calculation loses some of the precision.

So where is the mystique about "digital" amplification? Doing the amplitude increase is also possible trivially in a "power" amp by making the size of the voltage steps produced at the output of the DAC much larger so that each bit changes the output by a lot more [again assuming a PCM scenario].

The issue comes in getting lots of watts out and that means lots of current as well as volts. A D Class amp uses a crude output stage that switches large current - almost willy-nilly - into the load without worrying about the effect on the precision of the resultant output voltage. A feedback loop and/or a suitable filter of some kind [eg Quad Current Dumpers] or this Class T device we've been told about, see patent here:

http://www.google.com/patents?vid=USPAT5777512

uses correction loops to get the voltage right again. The patent refers to "noise shaping" by which they mean removing crud you don't want including the nasty distortion caused by the D class. Getting this to work at high precision is hard as is getting to work at high power.

Simples!!

[Dave2002]

This article from the IEEE might interest and amuse - http://spectrum.ieee.org/semiconduct...hook-the-world

[Gordon]

This article from the IEEE might interest and amuse - http://spectrum.ieee.org/semiconduct...hook-the-world

Hmm, I remember a lot of these devices when they first came out!

[OldTechie]

Some info about the modules PMC use in their DS001 amplifier:

http://www.hypex.nl/downloads/white-papers.html and http://www.hypex.nl/docs/papers/ncore%20wp.pdf

I particularly like the penultimate slide in the Powerpoint:

The Road To Heaven
• Specify the performance and accept the design
The Road To Hell
• Specify the design and accept the performance

[Gordon]

Some info about the modules PMC use in their DS001 amplifier:

http://www.hypex.nl/downloads/white-papers.html and http://www.hypex.nl/docs/papers/ncore%20wp.pdf

I particularly like the penultimate slide in the Powerpoint:

The Road To Heaven
• Specify the performance and accept the design
The Road To Hell
• Specify the design and accept the performance

Nice one OT!!

[Nevalti]

Some info about the modules PMC use in their DS001 amplifier:

I was told that PMC no longer use Hypex modules, nor the 'Flying Mole' amps but there is obviously room for 'badge engineering' confusion - so you may well be absolutely right. A couple of years ago, I asked PMC where they got the amps from and they claimed to make them themselves - hmmm. I tried to get hold of some PMC modules for testing but a helpful chap at PMC, who wanted to remain anonymous, told me not to bother as my Bryston amp was, in his opinion, much better.

The Americans seem to have adopted Class D with more enthusiasm than we have. I read so many good reports that I tried a few to see if I could 'go green' and save a lot of heat and space but they were all a big disappointment when played at realistic volumes (with my speakers). At low to medium volume they can be very impressive but when driven hard(ish), all the ones I heard became rather unnatural. If low level only is required, I feel that something like a Bantam Gold is probably satisfactory for most of us. I have one for 'drifting off' music in my bedroom and it never annoys or intrudes at that sort of volume even though it is driving my lovely old TDL Studio 1s..

[OldTechie]

I was told that PMC no longer use Hypex modules, nor the 'Flying Mole' amps but there is obviously room for 'badge engineering' confusion - so you may well be absolutely right. A couple of years ago, I asked PMC where they got the amps from and they claimed to make them themselves - hmmm. I tried to get hold of some PMC modules for testing but a helpful chap at PMC, who wanted to remain anonymous, told me not to bother as my Bryston amp was, in his opinion, much better.

They have certainly dropped the Flying Mole. That was clearly a horrid device with published frequency response curves that were highly dependent on the attached load. I was surprised that PMC ever used them, but maybe it worked ok with the DB1 which is, I think, the only speaker they sold with them.

If it is the Hypex Ucd400OEM as reported in lots of places, it's not really as high power as is implied. Hypex rate it at 400W - but that is into 4 ohms, so PMC sensibly says 200W, the Hypex 8ohm rating. PMC specifies the distortion only at half power, claiming .05% at 100W. Maybe 100W should be the rating. Bryston are much more conservative with their ratings - typically expecting .01% or less at the full rated output. However, you can buy a lot of CDs for the price difference between a DS-001 and a PowerPac 60.