I think they are very convinced that it is mass-market compatible that's at least what I expect th ebe the driving force behind it.
Regards
Charles
P.S: Are you going the the AES meeting in Hillerod in August ?
Regards
Charles
P.S: Are you going the the AES meeting in Hillerod in August ?
When I hear that "massproduction", I recollect PC, MP3&4 players, I-Phone, LCD panels or something like this. HI-FI Audio now is really tiny area, to being called massproduction. BTW, D2 DSP about $5, and still just few company use it.

Pabo said:I have never understood the reason for an "all digital" class d amp. For me it seems much more straight forward to just combine a DSP with an analog class d amp in order to get the filtering and processing features. Adding an integrator around a globally modulated class d stage gives just as good performance at a fraction of the complexity. The DSP can have a much lower processing power.
Now, I am truly analog. I do not understand all the sampling theories etc. but my gut feeling is that the best that can be achieved is close to an analog amplifier. The only real advantage I see is that the switching frequency is constant. In a self oscillating topology the drop in frequency when increasing the output signal reduces the loop gain and also lowers the possiblity for the integrator to act effectively. But still, our analog amps perform just as good as the Zetex amps and they are really cheap.
The reason is very simple. If the input signal is digital to start off with (which most audio signals are nowadays) it makes sense to do as much of the processing as possible in digital -- where errors (added noise and distortion) can be kept arbitrarily small by using long enough word lengths -- than converting to analogue using a DAC and driving the output into an analogue PWM modulator.
(incidentally, if you can find any (DAC+analogue PWM amp) combinations that do have as good performance as DDFA (e.g. NAD Master M2) I'd be interested to know of them 🙂
It also makes it possible to use much more sophisticated (higher performance) algorithms for modulation, dead time control, output limiting and fault detection than are possible with an analogue modulator. In particular, it's very difficult to build an analogue modulator with such a high-order feedback loop to get very high loop gain and PSRR over the entire audio band without running into stability problems.
The whole point of integration is that the complexity is hidden from the end user -- nobody thinks twice nowadays about the DSP+DAC complexity hidden inside a sound card or codec chip which would have been inconceivable for a few dollars 20 years ago.
This is all part of a general trend to move complexity into the digital world and simplify the analogue circuits -- for example, tone controls are much more flexible done digitally, and functions such as complex speaker equalisation can be done which are impossible in analogue.
(incidentally, if you can find any (DAC+analogue PWM amp) combinations that do have as good performance as DDFA (e.g. NAD Master M2) I'd be interested to know of them 🙂
I absolutely think I can but I have never seen the complete measurement pack of Zetex solution. One thing that would be very interesting is to see THD vs power at different frequencies. I figured out that the plots they usually show are a little twisted. The THD vs frequency plot is at 1W or so which is just when the noise and THD meet in the THD vs power plot. Therefore the THD vs frequency plot makes you think that it has constant loop gain in the audio band which I highly doubt.
I agree that making the processing in the digital domain is smart but once you want to make the power it is way simpler to make it in the analog domain and according to me the quality of the signal can be kept just as good.
But if you run digital all the way, aren't you stuck with digital attenuation for volume control as well? If full-scale digital word equals full power out, then listening with the volume 36dB from full scale means a 16 bit system becomes a 10bit system. With analog attenuator between DAC and power amp, the LSBs get smaller as the volume goes down, # of bits are preserved.
bwaslo said:But if you run digital all the way, aren't you stuck with digital attenuation for volume control as well? If full-scale digital word equals full power out, then listening with the volume 36dB from full scale means a 16 bit system becomes a 10bit system. With analog attenuator between DAC and power amp, the LSBs get smaller as the volume goes down, # of bits are preserved.
Which is why the internal DSP resolution is (I think) 35 bits...
Pabo said:
I absolutely think I can but I have never seen the complete measurement pack of Zetex solution. One thing that would be very interesting is to see THD vs power at different frequencies. I figured out that the plots they usually show are a little twisted. The THD vs frequency plot is at 1W or so which is just when the noise and THD meet in the THD vs power plot. Therefore the THD vs frequency plot makes you think that it has constant loop gain in the audio band which I highly doubt.
I agree that making the processing in the digital domain is smart but once you want to make the power it is way simpler to make it in the analog domain and according to me the quality of the signal can be kept just as good.
I have private results showing more performance detail than on the Zetex website but I can't publish them for obvious reasons...
The loop gain for a high-order sigma-delta noise shaping loop (which is what the DDFA is) isn't constant over the audio band but it is much higher than is possible with a conventional low-order PWM modulator loop, so the performance (SNR, PSRR, THD) can also be better because of the higher feedback factor -- this is exactly how many high-performance single-bit sigma-delta ADCs work, where up to 8th order loop filters are common.
You can work this out from the fact that a 850kHz PWM DAC with 7b time resolution (108MHz clock rate) has 57dB in-band SNR from unshaped quantisation noise but the DDFA quantisation noise must be well below -130dB (since they can get up to 126dB real SNR), so the loop gain in the audio band must be getting on for 80dB up to 20kHz.
This very high loop gain reduces all the distortion components generated by the power switching stage without the stability and overload recovery problems of trying to get such high gain in an analogue modulator.
Basically the DDFA performance is up there with the best audio DACs (Zetex are also now promoting the modulator as a high-performance multichannel DAC), so the power amplification comes "for free" as far as performance is concerned.
Assuming the audio input signal is digital, I can't see why it would ever be better to use a standalone DAC driving an analogue modulator which can only add more signal degradation.
It all depends what you define as "simple" -- as I said, the whole point of digital processing is that many things which are difficult or impossible to do in analogue become easy or even trivial, they just need some gates or DSP code.
IVX said:When I hear that "massproduction", I recollect PC, MP3&4 players, I-Phone, LCD panels or something like this. HI-FI Audio now is really tiny area, to being called massproduction. BTW, D2 DSP about $5, and still just few company use it.![]()
As has been said before, Zetex can't be targeting just the high-end low-volume Hi-Fi markets (like NAD Master M2) -- no way would this have justified the cost of developing the chips -- but also much higher-volume markets like home cinema (and other multichannel audio applications like PC and mobile), where the volumes are in the millions.
Depends of chipset cost anyway. Can you tell me something about it for 10KU? So far we use D2(for PA), as I said.
IVX said:Depends of chipset cost anyway. Can you tell me something about it for 10KU? So far we use D2(for PA), as I said.
I can't say anything about chipset selling price, you'll have to ask Zetex and it will depend heavily on volume -- though I would expect that for 10k units they'd at least be willing to talk to you and give you a ballpark figure.
For high volume (>1M) Far East applications the DDFA chipset unit price has to be able to be fairly low, which you'd expect given the technology. For 10k units it will be significantly higher, but I would have thought not out of order for a PA application.
Ian
(not speaking as a Zetex employee, because I'm not -- but I'd like to see them to sell lots of chips for obvious reasons 🙂
Does the version with the amp offer a non amplified .1 line output for a subwoofer?
Where can i find a built version of the dac preamp version for 2.1 channels?
I could only find the nad m2, but it doesn't offer a subwoofer signal output, and it costs way too much 😀.
By the way, does dual mono (or dual stereo) improve sound quality even for this technology? Because the nad m2 seems dual stereo. A version with only one card and psu would cost less, and i wonder how worse it would sound.
Sorry if this may be noob questions, but i'm no expert.
Where can i find a built version of the dac preamp version for 2.1 channels?
I could only find the nad m2, but it doesn't offer a subwoofer signal output, and it costs way too much 😀.
By the way, does dual mono (or dual stereo) improve sound quality even for this technology? Because the nad m2 seems dual stereo. A version with only one card and psu would cost less, and i wonder how worse it would sound.
Sorry if this may be noob questions, but i'm no expert.
Zetex alternatives
Anyone come up with other vendors who have similar closed-loop architecture and excellent performance?
Anyone come up with other vendors who have similar closed-loop architecture and excellent performance?
The digital circuits then predistort the input to the PWM modulator which drives the actual power output stage to compensate for the timing and amplitude errors in it, such that it has the same performance as the reference DAC and the error signal tends to zero.
Digital predistortion, hah?
This way one can achieve all the -120dB figures from virtually anything at very one well defined load/temperature/moon phase condition.
Guys, learn yourselves some analog stuff...
Digital predistortion, hah?
This way one can achieve all the -120dB figures from virtually anything at very one well defined load/temperature/moon phase condition.
Guys, learn yourselves some analog stuff...
Methinks you don't understand how the circuit operates...
The digital predistortion cancels out the distortion generated in the output switching stages, and is adaptive i.e. it varies with time to track variations and null out the errors. And it's inside the overall feedback loop, so you can think of it as linearising the open-loop transfer function before overall negative feedback is applied.
The overall distortion performance is then the same regardless of time, temperature, voltage, and unit-to-unit differences (e.g. component variation) -- which is more than can be said for most analogue circuits 🙂
The output switching stage distortion is extremely dependent on the load involved and changes drastically with real-life reactive loads. This is not a fixed voltage-to-voltage dependency, in which case the pre-distortion would actually work. The temperature (through power delivery) also modulates the load resistance by in excess of 10% if it is a, say, for example speaker in a, say, reality.
eheh all want to know how real work but will be never in public domain...sure we can trust Zetex and Nad plus forum review....very positive for the M2!
if is dsp ,with 4channel amp can be done digital xover...will be real gem !!!
Och no need one vinyl/tube/FR-OB for me...
if is dsp ,with 4channel amp can be done digital xover...will be real gem !!!
Och no need one vinyl/tube/FR-OB for me...
Last edited:
The output switching stage distortion is extremely dependent on the load involved and changes drastically with real-life reactive loads. This is not a fixed voltage-to-voltage dependency, in which case the pre-distortion would actually work. The temperature (through power delivery) also modulates the load resistance by in excess of 10% if it is a, say, for example speaker in a, say, reality.
Without giving anything away, the modulator knows both instantaneous output voltage and current (and temperature) and so it can calculate the static and dynamic load impedance -- the predistortion isn't fixed, it varies with load current.
Anyway this is just added linearisation inside the overall feedback loop, it improves linearity on top of what the negative feedback does.
The key point is that the entire circuit has almost identical performance to the reference DAC alone, so you could say that the output stage (power amp) is effectively almost distortion-free.
Last edited:
That starts to sound interesting... either that's impressive or just too good.
What's the current sensor? Can it sense DC...switching frequency?
Also I'd be interested to know the overall loop feedback factor above 10kHz.
I suspect we're talking the whole audio range here not 100Hz, aren't we?
What's the current sensor? Can it sense DC...switching frequency?
Also I'd be interested to know the overall loop feedback factor above 10kHz.
I suspect we're talking the whole audio range here not 100Hz, aren't we?
That starts to sound interesting... either that's impressive or just too good.
What's the current sensor? Can it sense DC...switching frequency?
Also I'd be interested to know the overall loop feedback factor above 10kHz.
I suspect we're talking the whole audio range here not 100Hz, aren't we?
The current is sensed by using the Vds of the FETs, there were a couple of AES papers about this. The amp switches at about 850kHz and has got multiple poles, so the loop gain even at 20kHz is still pretty high -- you can guess this by the fact that the distortion is still low (the gain needs to be high anyway for the noise shaping to work).
Of course many of the details of the processing were not disclosed... ;-)
- Status
- Not open for further replies.
- Home
- Amplifiers
- Class D
- Zetex DDFA