Forums

strange DAC

Started by Unknown December 17, 2021
I need an isolated DAC. I could just generate a PWM thing, isolate it,
and lowpass filter on the high side. But we could also generate a
delta-sigma bit stream, which has advantages. If d-s, I might want to
reclock it before lowpass filtering to get clean, fast edges.

But then I could drive several d-flops.

https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1

The analog signal now has 16 levels instead of two, sorta random but
still 16.

Is this worth doing? It's hard to think about on a Friday night after
pasta and beer.



-- 

I yam what I yam - Popeye
lørdag den 18. december 2021 kl. 03.49.47 UTC+1 skrev jla...@highlandsniptechnology.com:
> I need an isolated DAC. I could just generate a PWM thing, isolate it, > and lowpass filter on the high side. But we could also generate a > delta-sigma bit stream, which has advantages. If d-s, I might want to > reclock it before lowpass filtering to get clean, fast edges. > > But then I could drive several d-flops. > > https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 > > The analog signal now has 16 levels instead of two, sorta random but > still 16. > > Is this worth doing? It's hard to think about on a Friday night after > pasta and beer. >
it is basically a FIR filter, so it's just an extra filter before your filter if you are going to add a FF anyway why not 8, like 74hc164
On 18/12/2021 13:49, jlarkin@highlandsniptechnology.com wrote:
> I need an isolated DAC. I could just generate a PWM thing, isolate it, > and lowpass filter on the high side. But we could also generate a > delta-sigma bit stream, which has advantages. If d-s, I might want to > reclock it before lowpass filtering to get clean, fast edges. > > But then I could drive several d-flops. > > https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 > > The analog signal now has 16 levels instead of two, sorta random but > still 16. > > Is this worth doing? It's hard to think about on a Friday night after > pasta and beer.
Probably not much advantage to it. You might need slightly less filtering after it, but that isn't likely to save you much. With all of the resistors the same value (since you labelled them all R) you will only get 5 levels, not 16, since there will be 0,1,2,3 or 4 flipflops high. If you have a longer chain of flipflops (or even just for 4 of them), you would be better off not weighting them all the same, but weight the conductance (1/R) as raised cosine or gaussian or something that will filter better than a flat-top impulse response.
On Fri, 17 Dec 2021 19:09:32 -0800 (PST), Lasse Langwadt Christensen
<langwadt@fonz.dk> wrote:

>l&#2013266168;rdag den 18. december 2021 kl. 03.49.47 UTC+1 skrev jla...@highlandsniptechnology.com: >> I need an isolated DAC. I could just generate a PWM thing, isolate it, >> and lowpass filter on the high side. But we could also generate a >> delta-sigma bit stream, which has advantages. If d-s, I might want to >> reclock it before lowpass filtering to get clean, fast edges. >> >> But then I could drive several d-flops. >> >> https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 >> >> The analog signal now has 16 levels instead of two, sorta random but >> still 16. >> >> Is this worth doing? It's hard to think about on a Friday night after >> pasta and beer. >> > >it is basically a FIR filter, so it's just an extra filter before your filter > >if you are going to add a FF anyway why not 8, like 74hc164
Oh, of course; that was hiding behind two layers of fetticcini with Mo's cherry-tomato vodka sauce. The resistors would be sized for some filter response. -- I yam what I yam - Popeye
On Sat, 18 Dec 2021 23:24:42 +1100, Chris Jones
<lugnut808@spam.yahoo.com> wrote:

>On 18/12/2021 13:49, jlarkin@highlandsniptechnology.com wrote: >> I need an isolated DAC. I could just generate a PWM thing, isolate it, >> and lowpass filter on the high side. But we could also generate a >> delta-sigma bit stream, which has advantages. If d-s, I might want to >> reclock it before lowpass filtering to get clean, fast edges. >> >> But then I could drive several d-flops. >> >> https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 >> >> The analog signal now has 16 levels instead of two, sorta random but >> still 16. >> >> Is this worth doing? It's hard to think about on a Friday night after >> pasta and beer. > >Probably not much advantage to it. > >You might need slightly less filtering after it, but that isn't likely >to save you much. > >With all of the resistors the same value (since you labelled them all R) >you will only get 5 levels, not 16, since there will be 0,1,2,3 or 4 >flipflops high.
Right. The resistor values determine the filter response *and* the DAC response.
> >If you have a longer chain of flipflops (or even just for 4 of them), >you would be better off not weighting them all the same, but weight the >conductance (1/R) as raised cosine or gaussian or something that will >filter better than a flat-top impulse response. >
Yes, given an RC lowpass filter at some 10s of KHz, any rolloff from this trick would be way out there and not very helpful. A single resync flop could still be useful, clean fast edges powered by some vref. The project is to put some number of programmable isolated ac/dc dummy loads on a fixed-size board. The ultimate problem will be to dump the heat. -- I yam what I yam - Popeye
On a sunny day (Fri, 17 Dec 2021 18:49:36 -0800) it happened
jlarkin@highlandsniptechnology.com wrote in
<qfiqrghi0a53qhjqsn048smqah7ltqk895@4ax.com>:

>I need an isolated DAC. I could just generate a PWM thing, isolate it, >and lowpass filter on the high side. But we could also generate a >delta-sigma bit stream, which has advantages. If d-s, I might want to >reclock it before lowpass filtering to get clean, fast edges. > >But then I could drive several d-flops. > >https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 > >The analog signal now has 16 levels instead of two, sorta random but >still 16. > >Is this worth doing? It's hard to think about on a Friday night after >pasta and beer.
No idea what your speed requirements are, but the i2c PCF8591 I have used many times. You would need to isolate sda and scl, also sda pulldown backwards. It is an 8 bit ADC + DAC in one 16 pin DIL. But then you can go 2 ways.
18.12.21 03:49, jlarkin@highlandsniptechnology.com wrote:
>I need an isolated DAC. I could just generate a PWM thing, isolate it, >and lowpass filter on the high side. But we could also generate a >delta-sigma bit stream, which has advantages. If d-s, I might want to >reclock it before lowpass filtering to get clean, fast edges. > >But then I could drive several d-flops. > >https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 > >The analog signal now has 16 levels instead of two, sorta random but >still 16. > >Is this worth doing? It's hard to think about on a Friday night after >pasta and beer. > >
You could use an Icoupler. They have 10ns ish propagation delay and comes with ability to generate supply also. Then filter the PWM and you are done -- Klaus
jlarkin@highlandsniptechnology.com wrote:
> I need an isolated DAC. I could just generate a PWM thing, isolate it, > and lowpass filter on the high side. But we could also generate a > delta-sigma bit stream, which has advantages. If d-s, I might want to > reclock it before lowpass filtering to get clean, fast edges. > > But then I could drive several d-flops. > > https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 > > The analog signal now has 16 levels instead of two, sorta random but > still 16. > > Is this worth doing? It's hard to think about on a Friday night after > pasta and beer. > > >
Sure, that works. It's a species of transversal filter. You can also make a waveform generator using one or more HC4018 counters and weighted resistors. We had a discussion with George a few years back about that. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
On Sat, 18 Dec 2021 15:56:51 GMT, Jan Panteltje
<pNaOnStPeAlMtje@yahoo.com> wrote:

>On a sunny day (Fri, 17 Dec 2021 18:49:36 -0800) it happened >jlarkin@highlandsniptechnology.com wrote in ><qfiqrghi0a53qhjqsn048smqah7ltqk895@4ax.com>: > >>I need an isolated DAC. I could just generate a PWM thing, isolate it, >>and lowpass filter on the high side. But we could also generate a >>delta-sigma bit stream, which has advantages. If d-s, I might want to >>reclock it before lowpass filtering to get clean, fast edges. >> >>But then I could drive several d-flops. >> >>https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 >> >>The analog signal now has 16 levels instead of two, sorta random but >>still 16. >> >>Is this worth doing? It's hard to think about on a Friday night after >>pasta and beer. > >No idea what your speed requirements are, but the i2c PCF8591 I have used >many times. >You would need to isolate sda and scl, also sda pulldown backwards. >It is an 8 bit ADC + DAC in one 16 pin DIL. >But then you can go 2 ways.
There is one faction here that wants to put a uP up on the isolated side and use its DAC and ADC channels and let it close loops and such. OK, but we'd have to pick a chip that won't be EOL in a few years, and it would need its own program. -- I yam what I yam - Popeye
On Sat, 18 Dec 2021 17:24:39 +0100, Klaus Kragelund
<klauskvik@hotmail.com> wrote:

>18.12.21 03:49, jlarkin@highlandsniptechnology.com wrote: >>I need an isolated DAC. I could just generate a PWM thing, isolate it, >>and lowpass filter on the high side. But we could also generate a >>delta-sigma bit stream, which has advantages. If d-s, I might want to >>reclock it before lowpass filtering to get clean, fast edges. >> >>But then I could drive several d-flops. >> >>https://www.dropbox.com/s/0d7x6l6isyoldva/ML_DS_Dac.jpg?raw=1 >> >>The analog signal now has 16 levels instead of two, sorta random but >>still 16. >> >>Is this worth doing? It's hard to think about on a Friday night after >>pasta and beer. >> >> >You could use an Icoupler. They have 10ns ish propagation delay and comes with ability to generate supply also. Then filter the PWM and you are done
If I want analog accuracy, I'd prefer to run the PWM or D-S through a fast gate or flop that's powered by some good reference, dropped down from whatever isolated supply. A resistor and an LM4040 type thing would do. -- I yam what I yam - Popeye