Note to Readers: (added on the 31st, Oct 2022)
Please note, this thread should be considered as a discussion only. As it evolved it cannot give working ideas as to how to implement completely digital drives for Class-D.
--------------------------------------------
I do not know whether someone has already done this, but logically and technically, it seems possible.
Suppost a digital stream similar to that used to drive DACs is fed to an algorithm running on a microcontroller or small processor. The algorithm would use extrapolation to calculate the value of the instantaneous voltage at points which are not covered by the data from the stream. These calculated values/data would then be used to calculate the pulse width to feed the driver of the output stage. This would still result in an audio output, but without a DAC.
Please note, this thread should be considered as a discussion only. As it evolved it cannot give working ideas as to how to implement completely digital drives for Class-D.
--------------------------------------------
I do not know whether someone has already done this, but logically and technically, it seems possible.
Suppost a digital stream similar to that used to drive DACs is fed to an algorithm running on a microcontroller or small processor. The algorithm would use extrapolation to calculate the value of the instantaneous voltage at points which are not covered by the data from the stream. These calculated values/data would then be used to calculate the pulse width to feed the driver of the output stage. This would still result in an audio output, but without a DAC.
Last edited:
Yes this has been done and even sold commercially both in the form of finished amplifiers and in the form of IC's from companies like TI. See:
https://www.hypex.nl/p/application-notes-white-papers/ (A true One-Bit Power D/A Converter)
https://www.ti.com/lit/an/slaa117a/slaa117a.pdf
In general 'power DACs' have issues with power supply rejection, complexity of design, poor performance metrics in comparison to self oscillating designs at the same average switching frequency.
https://www.hypex.nl/p/application-notes-white-papers/ (A true One-Bit Power D/A Converter)
https://www.ti.com/lit/an/slaa117a/slaa117a.pdf
In general 'power DACs' have issues with power supply rejection, complexity of design, poor performance metrics in comparison to self oscillating designs at the same average switching frequency.
The running algorithm should use the digital data, instantaneous rail voltages and the set volume as independent variables. That way, it would calculate the required pulse width according to the audio signal and rail voltages. This should provide power supply rejection.
The obvious disadvantages are added complexity and the requirement of an ADC working at the switching frequency.
Last edited:
or you could just do feedback... however this is non trivial post filter. Furthemore you have the delay and non-idealities of an ADC to deal with. I do remember some NAD model that implemented pre-filter feed forward correction of pulse area that would compensate for power supply variation without too much complexity.
You need a high-resolution low-latency ADC if you want to close the feedback loop. Such designs exist, but are unusual. See
https://www.diyaudio.com/community/threads/pcm-to-pwm-conversion-101.205496/page-3#post-6902928
https://www.diyaudio.com/community/threads/pcm-to-pwm-conversion-101.205496/page-5#post-6904813
https://www.diyaudio.com/community/threads/pcm-to-pwm-conversion-101.205496/page-3#post-6902928
https://www.diyaudio.com/community/threads/pcm-to-pwm-conversion-101.205496/page-5#post-6904813
If DSD is effectively a PWM signal, can this not be fed directly to the output stage without any processing?
Yes this works with the following issues:
1) switching frequency is too high so this can only be done inefficiently using quite a specialist low power output stage
2) no feedback so PSU variation, non ideal switching, filter non linear behavior all affect the output
... 3) no volume control if not made with a PSU manipulation.
//
//
ECDesigns does this more or less like you describe in a quite innovative way with the PowerDAC-S. I am interested in this device but think the 120W power consumption for 2 x 4W is a drawback. One has to read a few times what the device exactly is as it is neither a DAC nor an amplifier! Difficult to describe so one better reads what John says about it.
https://www.ecdesigns.nl/
Played with a number of cheap and middle class PowerDACs/FDAs of various kinds (Nuforce DDA-100/120, Wadia 151 PowerDAC Mini etc.) the last 10 years and I think they are the future. Only because my favorite source has analog outputs I went back to analog inputs amplification. IMO the Denon PMA-60 is an outstanding example of the Full Digital Amplifier (FDA) technology, I was very fond of it.
FDA has success but in the area of smart TV sets. For the classic conservative audiophile they are maybe too modern so one reads negative comments about the technology but trying out is a better way to get to know them. It also ends the 2 box approach DAC/amplifier and when combined with an internal streamer it also ends the need for the usual interfaces (hooray!) and the need for expensive analog volume control, source selection and cabling.
You see the name giving to have some quirks. There are FDA with SPDIF/USB and there are FDA/streamers that have USB/SPDIF and (wireless) LAN.
https://www.ecdesigns.nl/
Played with a number of cheap and middle class PowerDACs/FDAs of various kinds (Nuforce DDA-100/120, Wadia 151 PowerDAC Mini etc.) the last 10 years and I think they are the future. Only because my favorite source has analog outputs I went back to analog inputs amplification. IMO the Denon PMA-60 is an outstanding example of the Full Digital Amplifier (FDA) technology, I was very fond of it.
FDA has success but in the area of smart TV sets. For the classic conservative audiophile they are maybe too modern so one reads negative comments about the technology but trying out is a better way to get to know them. It also ends the 2 box approach DAC/amplifier and when combined with an internal streamer it also ends the need for the usual interfaces (hooray!) and the need for expensive analog volume control, source selection and cabling.
You see the name giving to have some quirks. There are FDA with SPDIF/USB and there are FDA/streamers that have USB/SPDIF and (wireless) LAN.
Last edited:
Addition: as if it is not hard enough to describe stuff in detail in a foreign language... There are streamers with built in storage and FDA so the the chain is only one box, no external vintage multiplexed audio interfaces needed, digital stream till the power amplifier. Operation is often via an app but some exist that have displays and knobs. Imagine what will gained by having such a device and what will be lost. FWIW such setups have highest WAF, low(est) weight, small size and low energy consumption too 🙂
Now the quest which one is the better one.
Now the quest which one is the better one.
I think, a system as described makes sense if the audio stream/s is/are to be digitally processed. The latter can include equalisation, mixing and the addition of effects. Without the latter, it makes little sense to add so much complexity when a simpler and yet more reliable system can be used instead.
The same aim can be met if digital processing is done before it is fed into a normal DAC which adds simplicity, well tested reliability and more functionality.
My conclusion reading replies is that using a DAC to feed an amplifier is the simpler way which provides better results without too much complexity.
The same aim can be met if digital processing is done before it is fed into a normal DAC which adds simplicity, well tested reliability and more functionality.
My conclusion reading replies is that using a DAC to feed an amplifier is the simpler way which provides better results without too much complexity.
It is not that complex, these devices are smaller, simpler to operate, have less conversion steps, no DAC chip with its peculiarities, no jittery classic interfaces and more functionality. One nice aspect is the total silence between tracks. I don't know where the reliability subject comes from but FWIW I haven't had any defective streaming FDA device in 10 years. As music is digital already it makes sense to go the digital only route in a way. Well, it is the mainstream way of working in foreseeable time anyway.
If complex would be the bottleneck then we wouldn't be communicating like we are now.
They do suffer from prejudice though, that adds jitter 🙂 It could be a nice change to just try it out.
If complex would be the bottleneck then we wouldn't be communicating like we are now.
They do suffer from prejudice though, that adds jitter 🙂 It could be a nice change to just try it out.
Last edited:
If you have a PCM stream of numbers; 1003, 4, 18, 10000778 etc (i.e. PCM samples) and would like to get to and audible signal the only way is by a DAC. A class-D amplifier is an analog contraption - i.e. its impossible to do it without a DAC. I had hoped for a long time that there was some sort of short-cut, but learned that there was no such thing.
DSD is not possible to manipulate in terms of volume or EQ, hence the need for conversion to PCM in order to do that - the back to DSD for making DSD records. Kinda' sad.
Check this (they claim no D/A but... well...):
https://hifimediy.com/product/hifime-udp80-usb-and-spdif-true-digital-power-amplifier/
//
DSD is not possible to manipulate in terms of volume or EQ, hence the need for conversion to PCM in order to do that - the back to DSD for making DSD records. Kinda' sad.
Check this (they claim no D/A but... well...):
https://hifimediy.com/product/hifime-udp80-usb-and-spdif-true-digital-power-amplifier/
//
FDA do without a DAC chip for a long time now and convert a datastream in the digital domain to a class D amplifier (with digital source selection and volume control). So a class D amplifier with digital inputs which tells it is very possible. As said, I have had many of these and tried enough of these out to know they work OK. I never found a DAC chip in them. One does find ADCs for classic analog sources in them. Don't know about DSD, never bothered with that and OP did not mention it. I tried to have a subsection here on FDA but this was disapproved. Please do not deny what has been developed many years ago. The DDX technology in the device you linked was one of the first with this technology. There is D to A conversion but at the power amplifier, not by converting PCM to analog and amplify it the analog way.
Please read details of the product PowerDAC-S by ECdesigns as that goes a step further.
Even the current darling of ours the MA12070 exists in a digital only input version, the MA12070P.
Please read details of the product PowerDAC-S by ECdesigns as that goes a step further.
Even the current darling of ours the MA12070 exists in a digital only input version, the MA12070P.
Last edited:
There is no logical contradiction or fallacy in claiming a digital audio stream intended for a DAC can be processed using software algorithms to add features like frequency equalisation, the addition of a volume control parameter/variable and diverse source mixing. Such a stream, once it is transformed into pulse widths, can be applied to drive the driver & output of a Class-D amplifier.
A DAC receives a string of digital values at a sufficiently high frequency. These values are the instantaneous voltage values that a DAC is expected to output. Using logic circuits like counters, a pulse can be timed so that it starts when counting is initiated and end when the output binary number equals the value that is normally applied to a DAC.
To add other functionality frequency equalisation can be applied to the binary stream. Other binary streams can be mixed together in various ways. Finally, also effects can be applied. All this can be done using algorithms or better programmatic functions/subroutines.
This is conceptually sound.
A DAC receives a string of digital values at a sufficiently high frequency. These values are the instantaneous voltage values that a DAC is expected to output. Using logic circuits like counters, a pulse can be timed so that it starts when counting is initiated and end when the output binary number equals the value that is normally applied to a DAC.
To add other functionality frequency equalisation can be applied to the binary stream. Other binary streams can be mixed together in various ways. Finally, also effects can be applied. All this can be done using algorithms or better programmatic functions/subroutines.
This is conceptually sound.
Indeed, that's precisely what is done in a digital mixing desk. The digital format preferably has to be some sort of PCM.
Indeed, but when you do it open loop, the amplifier's supply is effectively the DAC reference. It is hard enough to keep the DAC reference sufficiently clean when the load is a few milliamperes, with amperes it will be proportionally more difficult.
You need clock frequencies of the order of a gigahertz or more to get a decent resolution with that approach, but you can use lower clock frequencies when you combine it with noise shaping.