What do you think of Class D amplifiers?

I think we do now generally agree on the innards of a 'full digital' class D amp.  My point remains that there are no analogue steps in the amplifier, the input PCM signal gets turned into a PWM signal that then drives the output stages. The only point at which the signal becomes 'analogue' is after the passive low pass filter on the amplifier output.

Where we disagree is to call that 'fundamentally analogue'.  To me that looks 'fundamentally digital'.

For the electronically inept, are all Class D amps created equal?

In other words are there two types - ie. some that are basically analogue with a switchmode power supply, while others are digital right the way through till it gets to the speakers?

If the above is the case, which is likely to give the better result?.......thinking aloud, I would have thought keeping things in the digital domain as long as possible seems to make sense.

Looking at the wikipedia page start to feel a bit more unconfortable with the technology. It looks like staying digital untill the end is not a practical proposition as feedback cannot be applied using conventional analogue techniques. (perhaps digital feedback is possible though at a cost). Feedback seems to be needed in order to iron out impedance matching of the load (speakers) which can result in nonlinear response. Can't see with all the limitation this technology is displaying how class D will be succesful in a HIFI environment. Can only think they will be OK in a night club where power is the main requirement.

I'm not a particular fan of Class D, but there are interesting products out there from the likes of Nad, Bel Canto, Primare and Devialet. Due to the green issues with Class AB etc, there is more effort being put into this type of design.

I've posted this before; http://www.primare.net/assets/_managed/cms/files/PrimareUFPD2.pdf

regards

Must say have not seen this before but reading it confirms to me my original suspition that class D is a no go for HIFI.

The article claims some improvements using their technology but they don't go into detail. However, from what it implies is that they have a technique to apply higher from of feedback. I expect the techique is some analogue scheme not digital. The problem is that is known fact from class A/B designs that over application of overall feedback to minimise THD and output resistance results in poor sounding amplifiers. And unfortunatelly on class D amplifiers a lot of distorsion is generated by the technique that will require considerable feedback loop gains to remove it.

I believe the high-end US audio firm Jeff Rowland Design Group is to stop offering Class D amps and Instead return to using class AB.

As I mentioned in an earlier post, my one experience with class D, the Jeff Rowland Continuum 500 integrated, was not favorable.

Dumbledore,

It is quite possible to have feedback to compensate for power supply fluctuations, non linearities in the output filter and strange frequency dependent goings on in the speaker impedance.  Have a look at the zetex/diodes chipset I linked to earlier in the thread.  NAD have used exactly this approach in their 'direct digital' M2 amplifier, which has received very positive reviews, and not just from night club owners.

While NAD seem to be charging a small fortune for their amp, the electronics to do all this cost peanuts.

Class D is generally used to cover amps that use a switching technique (output drivers are either full-on or full-off) rather than amps which vary the conductance of the output drivers in some linear way to achieve the required output.

'Full digital' class D take a digital input signal (usually some form of PCM) apply some DSP magic and use this to generate on / off signals to drive the output stages. Other class D topologies use the same output approach, but generate the on / off signals by comparing an input analogue signal with a high frequency reference waveform (usually triangular).

Both are class D, both have the same switching output, but the way of generating the output driver control signals is different.

Andyjm,

Had a look at the DDFA spec and must say at least somebody tried to do digital feedback which is a good way to go.

Is this technology perfected. I don't think so. There are two problems I can see.

One I already said, you will need a lot of negative feedback to remove distorsion inherent in non feedback class D output levels to acceptable levels. Is known fact that high degrees of overall negative feedback is not favourable for audiophile amplifiers.

Second the negative feedback circuitry used in DDFA is in fact quite an analog process. It requires a very precise ADC (analog to digital converter) to operate and do the digital feedback (they call it Analogue feedback preocessor). DDFA spec is not saying the ADC specification but compare that with the class A/B designs that use a bit more open loop gain and two feedback resistors then I would expect DDFA technology will require some perfecting (just remeber how much pefecting went into fixing the DAC technology which is still ongoing).

With these limitations in mind related to class D amplifiers the final listening tests will be critical. If a class D can sound near as good as reference class A/B amps then they will win eventually. But in view of above limitation class D may be acceptible for HIFI use in active speaker designs used primirily for the bass and midrange units. Driving single units with no crosovers will allow the amplifier designer to tune the output stage and initial PWM drive parameters so that smaller negative feedback will be required to achieve set THD which will address limitation one above.

Thx, that's pretty much what I thought.

how come? both, "digital" and analog, class D amps use PWM signal followed by a low pass filter to modulate the driver output. the only difference lies in that a "digital" amp extracts a PWM straight from PCM whereas "analog" amp creates PWM by modulating the analog wave with a triangular signal wave.

anyway, I'm too in the camp that too much negative loop feedback is too much. a class D will not measure fine without it. for instance in that Primare blurp they claim they use 26dB of NFB! just imagine how this thing measure in it's open loop version. unfortunately, such is the nature of transistors - switching them on and off constantly will always cause them to work in nonlinear manner.

I think class D has it's specific sound. I heard the new Primare integrated and also Devialet, those driving WA Sashas, and was underwhelmed, to say the least. Those amps sounded in a way very impressive - lots of detail in HF, but not natural at all and detached as opposed to sounding coherent and liquid, to what I'm used to.

I think class D is only good to drive subwoofer drivers, where high power and low bandwidth is needed.

I've read that pdf on & off for over a year - apart from saying that Primare use their own design using high bandwidth switching FETs, they actually say very little about the inner workings such as how this 26dB of NF is applied.

What amuses me are statements from some that its open loop characteristcs are poor is like say the frequency stability of a phase lock loop is poor without the feedback - of course it will be - it designed to use feedback!

As an owner of a Primare class D power amp, all I can add is he proof is in the pudding - especially now I've dispensed with bi-wiring & those cheap & nasty bridging plates on the back of my Totems.

As for those folk who feel uncomfortable with what PWM does to analogue - just what do they think that analogue to digital conversion does to SQ? Unless they listen to entirely analogue, perhaps it's time to stop worrying about their analogue signals being "pulled apart then reassembled."