Forward error correction of PWM amps

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Available PWM full bridge amps are dirt cheap DIY makes no sense.
There are 2 channel amps 250 W output per channel , 500 kHz sampling, and switching power supply for less than 150 UK pound available.
However one should be clear that PWM amps work in time domain thus there are 2 sources of very tricky distortion.
First quantization error. At very low input noise causes quantization error and this appears as distortion in frequency domain. Sounds like crossover distortion but far more nasty.
Second timing errors. Timing error cannot be avoided in any way because switching speed depends exponentially on crystal temperature ( carrier mobility). Temperature , crystal defects and other physical effects affect rise time and fall time differently.
But there is a cheap way out of it, and only recently I read about the first commercially available PWM amp that employs forward error correction.
There are several methods of forward error correction in terms of error pickup
but eventually they come up with a very fast very linear class A amp capable of about 1 watts power that cancels every distortion at the output
common to the PWM and the class A.
Given this the old topics BJT or MOSFET class A or AB super ding dong powersupply are over and out.
 
Because the situation has changed as a product utilizing feedforward correction is now available. It is no longer a design idea. I do not have access to the schematics yet but I think sooner or later the schematics will be leaked.
As far as the rumors go the device overcomes also some of the drawbacks of PWMs such as sensitivity to load changes vs frequency which is typical for passive crossover speakers.
 
Are you talking about the processor Zetex produces ?
It is used in the NAD M2 amp.

"Diodes Zetex Ltd., which developed a novel feedback topology in which the output pulses are continuously compared with a reference to produce an error signal. This error signal is integrated, digitized (at 108MHz), and fed back, with noiseshaping, to the PWM modulator. The signal is also monitored at the output low-pass filter, to give a low output impedance. The Zetex team refers to their topology as a Direct Digital Feedback Amplifier, and the NAD M2 is the first commercial product to feature DDFA.
"

These processors are not for sale.
 
No the briefing intro said it is a feedforward corrected standard PWM module
( possbly the widely used Bang&Olufsen) using a 1 watt ultra linear class A amp
that cancels errors.
meanwhile those PWM amps sound good but not excellent. Imho this is due to a
rarely discussed issue, the so called fine dynamics, that is how are minimal differences of volume reproduced. This is an issue where fast class A amps excel ( provided the speaker excels, too). The issue is "resolution" . This is why those old vinyls , recorded up to 1960 and some even mono, sound more lively than any CD, despite the much lower dynamic range, and high noise floor.
This is basically an effect of human audition psychology but that is what matters.
 
The amp is provided by French firm devialet and works apparently according Hawksford/P.Walker error correction. It contains a 0.5 Watt very fast and linear class A which supplies the voltage to the load and a class D ( possibly a Bang&Olufen 250 Watt module) that supplies the current thus that the voltage remains constant. This is just "current dumping". It comes in an aluminum case and looks more like a model flying saucer than an amp, but very elegant. It supports streaming but accepts also analog inputs. No cheap at all 10.000 Euros. Very enthusiastic comments from first listening tests such as analog is finally out.
It appears to have three 250 watt modules one can be configured as subwoofer amp, has lots of digital filters.
 
Available PWM full bridge amps are dirt cheap DIY makes no sense.
There are 2 channel amps 250 W output per channel , 500 kHz sampling, and switching power supply for less than 150 UK pound available.

Is money the key criteria for DIY?

First quantization error. At very low input noise causes quantization error and this appears as distortion in frequency domain. Sounds like crossover distortion but far more nasty.
I disagree on mixing up noise with quantization.
Analogue PWM modulators allow any duty cycle without quantization.
Their noise mechanism is depending massively on the jitter mechanisms of the used comparator. For sure a very different animal vs resistor noise, but not identical to quantization (predefined discrete values).


I'll give details once I have received their brochure.
If interested in relevant and correct information - better go for a beer with the design engineer.
Most likely the brochure is just there in order to manipulate your emotions and spend 10k money.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.