I meant the other way around actually, switching frequency an integer multiple of the ADC sample rate, or of n + 1 times the ADC sample rate when you use an n bit SAR that is clocked at n + 1 times its sample rate. It has the advantage that any switching frequency ripple on the ADC reference gives only a gain error and no mixing products at sum and difference frequencies.
Back on topic: this is the Axiom/Teledyne Dalsa patent for a low-latency ADC that is used in the feedback of the class-D amplifier of post #56: US9602126B2, https://patents.google.com/patent/US9602126B2/en
Back on topic: this is the Axiom/Teledyne Dalsa patent for a low-latency ADC that is used in the feedback of the class-D amplifier of post #56: US9602126B2, https://patents.google.com/patent/US9602126B2/en
Since all the code resides inside the ISR (which is very quickly executed), all ADC values (load current, bus voltage etc.) used within the ISR code may be thought of as sampled at the switching frequency itself, even though more samples are available from the ADC that is clocked much higher, but are not used.
So, you could, in a way, say that the switching freq is an integral multiple of the effective A/D sampling rate, as they are the same frequency. Also, note that any higher is just not useful, as changes can be made only once in a PWM switching cycle anyway. So, there you go, the answer is essentially yes.
So, you could, in a way, say that the switching freq is an integral multiple of the effective A/D sampling rate, as they are the same frequency. Also, note that any higher is just not useful, as changes can be made only once in a PWM switching cycle anyway. So, there you go, the answer is essentially yes.
Though you don't need the opamp filters when you're DSP, it's highly unlikely that they would do any harm to the DAC's sound signature, but really can't say the same about crossover networks and tubes.
I'm on the cheap side as well. But, I like DAC+linear amp for domestic power, will consider D-amp for subwoofer/lows only. I would like 16 assignable channels for an 8ch surround system and active crossovers, all cheap, but with cost reduction brought about by DIY action and not the semiconductor economy of scale as in PCM-PWM chips.
Marcel, the low latency A/D patent you cite is interesting. I'm not an expert in z-domain stuff, but it looks like the author is splitting the loop gain between the forward and feedback paths in a different and clever way, so that the net loop gain seems to remain more or less the same as before. I guess there are no drawbacks for this type of implementation. Figure 6A&B also show our current topic as a potential application for the method.
However, if I'm not wrong, it's going to be a while before anyone could freely use this method, as the patent was issued quite recently, in 2014.
However, if I'm not wrong, it's going to be a while before anyone could freely use this method, as the patent was issued quite recently, in 2014.
You can't legally make your own ADC based on this principle, that's for sure. Apparently you can buy modules that use their chip. On another thread, someone found this:
https://truebluebox.com/portfolio-item/power-amplifier-da-input/
I don't know if you can buy the chip itself in small quantities.
https://truebluebox.com/portfolio-item/power-amplifier-da-input/
I don't know if you can buy the chip itself in small quantities.
That "True Blue" website seems pretty commercial, like a commercial on TV... They tout DIY, but offer no system architecture diagram showing how their particular amplifier works. They just say things like "It sounds better than equipment many times its price!" OK.. "You can make a Bluetooth speaker with it!" Yes, as a DIYer I know you can, but how many DIYers would for 1k$ - by the time you add their optional volume control board? I'd be wondering how the analog gets into it, via a mixer with the output feedback signal? What happens between the USB and the speaker terminals? How's the clocks done? What power supply would I need?
The latency of the feedback ADC is a thing only when you're trying not to oscillate. What if one tries to pick up the oscillations instead, like Bruno Putzeys did with the UcD ?
This way, regular parts (unlike the low latency ADC) can be used to get a high loop gain, post-filter feedback. However, the associated delay would need to be predictable in order to design the switching frequency.
Parabolic hysteresis is another method that looks promising.
This way, regular parts (unlike the low latency ADC) can be used to get a high loop gain, post-filter feedback. However, the associated delay would need to be predictable in order to design the switching frequency.
Parabolic hysteresis is another method that looks promising.
Not required, it may be possible for the skilled person to modify the module's power stage (alone) to his requirement (read power level), while maintaining similar gain and feedback structure to get more or less similar performance.