Post by thread:[EE] Simple D/A converter?

Hi all, Can any of you recommend a D-to-A converter with a single-wire digital-side interface? As small as possible -- roughly SOT23-5 to SSOP-8 would be good, 5V powered, 0-5V output (I'll actually do 0.5V to 4.5V output with it). Summary is that I need to add an analog output to an existing PIC16F883-based circuit (already working on PCB), but only have one general-purpose I/O available, and no PWM pins available. I can get tons of I2C, SPI or other D/A converters, but not finding anything with actual single-wire serial. FWIW, there's too much going on in the chip to get any decent bit-banged PWM. I can do this with a small PIC, but since I only need 25 or so of these, I thought I'd check to see if anyone here knows of a purpose-built chip for this, so that I can save coding time. Cheers, -Neil.

At 11:20 AM 4/11/2012, Neil wrote: >Hi all, > >Can any of you recommend a D-to-A converter with a single-wire >digital-side interface? As small as possible -- roughly SOT23-5 to >SSOP-8 would be good, 5V powered, 0-5V output (I'll actually do 0.5V to >4.5V output with it). > >I can do this with a small PIC, but since I only need 25 or so of these, >I thought I'd check to see if anyone here knows of a purpose-built chip >for this, so that I can save coding time. If you *do* wind up having to do it yourself with a PIC, look at the PIC10f322 family. They have native PWM hardware built-in, which may dramatically reduce your coding time. No USART or I2C hardware, unfortunately. The SOT23-6 package is easy to work with. BTW - I've been purchasing SOT23-6 zero insertion force sockets (for programming) from <http://www.wvshare.com/product/499-P44-10-REV-B.htm> US $24.99 each in 3-lot. I then made up little adapter boards that map the pins to a 6-pin header to match a PICkit 2 or 3. dwayne -- Dwayne Reid <spam_OUTdwaynerTakeThisOuTplanet.eon.net> Trinity Electronics Systems Ltd Edmonton, AB, CANADA (780) 489-3199 voice (780) 487-6397 fax http://www.trinity-electronics.com Custom Electronics Design and Manufacturing

On Wed, Apr 11, 2012 at 07:38:17PM +0200, Wouter van Ooijen wrote: > On 11/4/2012 7:20 PM, Neil wrote: > > Hi all, > > > > Can any of you recommend a D-to-A converter with a single-wire > > digital-side interface? As small as possible -- roughly SOT23-5 to > > SSOP-8 would be good, 5V powered, 0-5V output (I'll actually do 0.5V to > > 4.5V output with it). > > I'd opt for a PIC. UART in, PWM out. Why not do software PWM on the existing I/O pin then? All that is needed is a low pass filter. BAJ {Quote hidden}> > -- > > Wouter van Ooijen > > -- ------------------------------------------- > Van Ooijen Technische Informatica: http://www.voti.nl > consultancy, development, PICmicro products > docent Hogeschool van Utrecht: http://www.voti.nl/hvu > C++ on uC blog: http://www.voti.nl/erblog > > -

Reading is actually fundamental. That's what I get for reading the reply and not the original post. BAJ On Wed, Apr 11, 2012 at 08:53:25PM +0200, Wouter van Ooijen wrote: {Quote hidden}> > Why not do software PWM on the existing I/O pin then? All that is needed is > > a low pass filter. > > Quote Neil: FWIW, there's too much going on in the chip > to get any decent bit-banged PWM. > > > > > -- > > Wouter van Ooijen > > -- ------------------------------------------- > Van Ooijen Technische Informatica: http://www.voti.nl > consultancy, development, PICmicro products > docent Hogeschool van Utrecht: http://www.voti.nl/hvu > C++ on uC blog: http://www.voti.nl/erblog > > -

That seems to be the popular choice, which is not unexpected on the piclist :) I'm using internal oscillators, so it's probably better for me to use something other than UART. I'll probably just reuse code from a varying duty-cycle protocol I've used a few times before, where 33% & 67% represent 0 & 1 respectively. It's very forgiving of timing variances. Cheers, -Neil On 4/11/2012 1:38 PM, Wouter van Ooijen wrote: > On 11/4/2012 7:20 PM, Neil wrote: >> Hi all, >> >> Can any of you recommend a D-to-A converter with a single-wire >> digital-side interface? As small as possible -- roughly SOT23-5 to >> SSOP-8 would be good, 5V powered, 0-5V output (I'll actually do 0.5V to >> 4.5V output with it). > I'd opt for a PIC. UART in, PWM out. >

10F320/2 looks really good. Only thing I'm wondering is that it seems the PWM voltage output is based on the supply voltage, so for higher accuracy, should I run the analog output back into an A/D for some type of dynamic closed-loop adjustment? If so, I'd go for something with >8-bit A/D. Alternatively, I could monitor the power-supply voltage and adjust the PWM value accordingly. But here too a 10-bit A/D would be nice. The 12F1501 looks promising. Cheers, -Neil. On 4/11/2012 2:03 PM, Dwayne Reid wrote: > If you *do* wind up having to do it yourself with a PIC, look at the > PIC10f322 family. They have native PWM hardware built-in, which may > dramatically reduce your coding time. No USART or I2C hardware, > unfortunately. The SOT23-6 package is easy to work with. BTW - I've > been purchasing SOT23-6 zero insertion force sockets (for programming) > from <http://www.wvshare.com/product/499-P44-10-REV-B.htm> US $24.99 > each in 3-lot. I then made up little adapter boards that map the pins > to a 6-pin header to match a PICkit 2 or 3. dwayne

> On Apr 11, 2012 2:54 PM, "Wouter van Ooijen" <EraseMEwouterspam_OUTTakeThisOuTvoti.nl> wrote: >> >> > Why not do software PWM on the existing I/O pin then? All that is > needed is >> > a low pass filter. >> >> Quote Neil: FWIW, there's too much going on in the chip >> to get any decent bit-banged PWM. >> > I remember someone here describing a D to A output based on a random > number. It wasn't timing sensitive. From my memory: if your random number > is below your target value turn the output on, else output is off. I don't > know if it's viable with a pseudo random generator. > I also did some PWM with "sample and hold." When not PWMing the output, I'd tristate it. The C of the series RC LPF acted as a hold capacitor. Depending on required update time, accuracy, etc., it might be possible to output a single pulse with a certain period high, then a certain period low, then tristate until you come around again. I did this to drive 48 analog outputs with a single PWM output of a PIC (demultiplexing the single PWM pin into a bunch of hold capacitors). Harold -- FCC Rules Updated Daily at http://www.hallikainen.com - Advertising opportunities available! Not sent from an iPhone

> On Apr 11, 2012 2:54 PM, "Wouter van Ooijen" <wouterspam_OUTvoti.nl> wrote: >> >> > Why not do software PWM on the existing I/O pin then? All that is > needed is >> > a low pass filter. >> >> Quote Neil: FWIW, there's too much going on in the chip >> to get any decent bit-banged PWM. >> > I remember someone here describing a D to A output based on a random > number. It wasn't timing sensitive. From my memory: if your random number > is below your target value turn the output on, else output is off. I don't > know if it's viable with a pseudo random generator. > I also did some PWM with "sample and hold." When not PWMing the output, I'd tristate it. The C of the series RC LPF acted as a hold capacitor. Depending on required update time, accuracy, etc., it might be possible to output a single pulse with a certain period high, then a certain period low, then tristate until you come around again. I did this to drive 48 analog outputs with a single PWM output of a PIC (demultiplexing the single PWM pin into a bunch of hold capacitors). Harold -- FCC Rules Updated Daily at http://www.hallikainen.com - Advertising opportunities available! Not sent from an iPhone

On 12 April 2012 14:42, Neil <@spam@picdude3KILLspamnarwani.org> wrote: > 10F320/2 looks really good. Only thing I'm wondering is that it seems > the PWM voltage output is based on the supply voltage, so for higher > accuracy, should I run the analog output back into an A/D for some type > of dynamic closed-loop adjustment? If so, I'd go for something with > >8-bit A/D. > > Alternatively, I could monitor the power-supply voltage and adjust the > PWM value accordingly. But here too a 10-bit A/D would be nice. > > The 12F1501 looks promising. > > Cheers, > -Neil. > > How about clamping the output PWM at a level a bit below the supply voltage? That would reduce the dependency. Just watch out for temperature variation! RP

> Can any of you recommend a D-to-A converter with a single-wire digital-side interface? Any SOT23-5 PIC. Scott Datallo's Sigma Delta ADC routines. Software PWM if needs be. Can interleave between ADC isochronous actions. If IRQ available PWM can be IRQ driven AND simplify Sigma Delta isochronous timing need. Not totally trivial but reasonably simple. For extra points you can almost certainly do the ADC input and digital output on the same pin :-). Russell McMahon On 12 April 2012 20:41, Richard Prosser <KILLspamrhprosserKILLspamgmail.com> wrote: {Quote hidden}> On 12 April 2012 14:42, Neil <RemoveMEpicdude3TakeThisOuTnarwani.org> wrote: >> 10F320/2 looks really good. Only thing I'm wondering is that it seems >> the PWM voltage output is based on the supply voltage, so for higher >> accuracy, should I run the analog output back into an A/D for some type >> of dynamic closed-loop adjustment? If so, I'd go for something with >> >8-bit A/D. >> >> Alternatively, I could monitor the power-supply voltage and adjust the >> PWM value accordingly. But here too a 10-bit A/D would be nice. >> >> The 12F1501 looks promising. >> >> Cheers, >> -Neil. >> >> > > How about clamping the output PWM at a level a bit below the supply > voltage? That would reduce the dependency. Just watch out for > temperature variation! > > RP > >

Hi Neil, On 13 April 2012 00:26, Neil <spamBeGonepicdude3spamBeGonenarwani.org> wrote: > What exactly do you mean by "clamping the output"? I actually need 0.5V > to 4.5V output, and Vdd will be 5V. But I was looking for the A/D > feedback,etc for accuracy, since the PWM voltage output will be subject > to variations in Vdd. > > Cheers, > -Neil. > Basically I was thinking about a system where the output voltage was limited to, say, 4.5V by means of a series resistor & precision zener / TL431 etc. This would then be independent of supply voltage but subject to any temperature drift in the clamping device. The minimum voltage would be close to zero - if you need it to be 0.5V, then you'd need to add it in - use an opamp circuit to provide the clamping function with a 0.5V offset.? RP

If you use the PIC to do the ADC it will be susceptible to the same voltage fluctuations as the DtoA output. You would need to use a voltage reference chip for accuracy so why not just power the PIC from a voltage reference and that would govern the AtoD accuracy. On 12 April 2012 22:25, Richard Prosser <TakeThisOuTrhprosserEraseMEspam_OUTgmail.com> wrote: {Quote hidden}> Hi Neil, > > On 13 April 2012 00:26, Neil <RemoveMEpicdude3TakeThisOuTnarwani.org> wrote: >> What exactly do you mean by "clamping the output"? I actually need 0.5V >> to 4.5V output, and Vdd will be 5V. But I was looking for the A/D >> feedback,etc for accuracy, since the PWM voltage output will be subject >> to variations in Vdd. >> >> Cheers, >> -Neil. >> > > Basically I was thinking about a system where the output voltage was > limited to, say, 4.5V by means of a series resistor & precision zener > / TL431 etc. This would then be independent of supply voltage but > subject to any temperature drift in the clamping device. The minimum > voltage would be close to zero - if you need it to be 0.5V, then you'd > need to add it in - use an opamp circuit to provide the clamping > function with a 0.5V offset.? > > RP > >

Correction, just power the PIC from a voltage reference and that would govern the DtoA accuracy. On 13 April 2012 00:21, Chris Roper <caroperEraseME.....gmail.com> wrote: {Quote hidden}> If you use the PIC to do the ADC it will be susceptible to the same > voltage fluctuations as the DtoA output. > You would need to use a voltage reference chip for accuracy so why not > just power the PIC from a voltage reference and that would govern the > AtoD accuracy. > > On 12 April 2012 22:25, Richard Prosser <EraseMErhprossergmail.com> wrote: >> Hi Neil, >> >> On 13 April 2012 00:26, Neil <RemoveMEpicdude3EraseMEEraseMEnarwani.org> wrote: >>> What exactly do you mean by "clamping the output"? I actually need 0..5V >>> to 4.5V output, and Vdd will be 5V. But I was looking for the A/D >>> feedback,etc for accuracy, since the PWM voltage output will be subject >>> to variations in Vdd. >>> >>> Cheers, >>> -Neil. >>> >> >> Basically I was thinking about a system where the output voltage was >> limited to, say, 4.5V by means of a series resistor & precision zener >> / TL431 etc. This would then be independent of supply voltage but >> subject to any temperature drift in the clamping device. The minimum >> voltage would be close to zero - if you need it to be 0.5V, then you'd >> need to add it in - use an opamp circuit to provide the clamping >> function with a 0.5V offset.? >> >> RP >> >> -

On 4/12/2012 4:25 PM, Richard Prosser wrote: > Basically I was thinking about a system where the output voltage was > limited to, say, 4.5V by means of a series resistor & precision zener > / TL431 etc. This would then be independent of supply voltage but > subject to any temperature drift in the clamping device. The minimum > voltage would be close to zero - if you need it to be 0.5V, then you'd > need to add it in - use an opamp circuit to provide the clamping > function with a 0.5V offset.? RP Gotcha now. But space is super-tight, so op-amp is probably out. I can just send the appropriate digital signal range (103 to 922) to get 0.5V to 4.5V, with the only side effect of reducing resolution by 20%.

Nice idea, if I can find an appropriate reference. It would need to be >4.5V (when supplied from down to 4.8V), Only thing I find in this range is 4.5V with a MAX6107, but I'll avoid Maxim if possible. I do intend to add a 2.5V reference though, and do some math to compare the output and compensate it accordingly. Cheers, -Neil. On 4/12/2012 6:21 PM, Chris Roper wrote: {Quote hidden}> If you use the PIC to do the ADC it will be susceptible to the same > voltage fluctuations as the DtoA output. > You would need to use a voltage reference chip for accuracy so why not > just power the PIC from a voltage reference and that would govern the > AtoD accuracy. > > On 12 April 2012 22:25, Richard Prosser<RemoveMErhprosserspam_OUTKILLspamgmail.com> wrote: >> Hi Neil, >> >> On 13 April 2012 00:26, Neil<RemoveMEpicdude3TakeThisOuTspamnarwani.org> wrote: >>> What exactly do you mean by "clamping the output"? I actually need 0.5V >>> to 4.5V output, and Vdd will be 5V. But I was looking for the A/D >>> feedback,etc for accuracy, since the PWM voltage output will be subject >>> to variations in Vdd. >>> >>> Cheers, >>> -Neil. >>> >> Basically I was thinking about a system where the output voltage was >> limited to, say, 4.5V by means of a series resistor& precision zener >> / TL431 etc. This would then be independent of supply voltage but >> subject to any temperature drift in the clamping device. The minimum >> voltage would be close to zero - if you need it to be 0.5V, then you'd >> need to add it in - use an opamp circuit to provide the clamping >> function with a 0.5V offset.? >> >> RP >> >> -

Resurrecting this thread from a month ago. I had issues fitting a piggy-back pc-board into the original product. However, I have been slowly migrating the parent of this product from a 16F to an 18F23k22 and accelerated that effort as it would've possibly let me use the D/A technique below... specifically because I'm using the internal oscillator, and the 18F will let me run that much faster. Net result is that it works nicely -- at 32Mhz. And since I'm using C now for my 18F apps, it got implemented very rapidly. Thanks for informing me of this technique. Now though, I have to ask about this method -- why exactly the need for a random number? I am simply sampling as much as possible, and toggling the output bit depending on if the sampled value is greater or less than the measured value. That measured value being the filtered output bit. Cheers, -Neil. On 4/12/2012 10:34 AM, Adam Field wrote: >> I remember someone here describing a D to A output based on a random number. >> It wasn't timing sensitive. From my memory: if your random number is below >> your target value turn the output on, else output is off. I don't know if >> it's viable with a pseudo random generator. > To expand on this, I found the original thread: > > http://www.piclist.com/techref/postbot.asp?id=/piclist/2001/02/04/004250a&key=random%20analog > > I haven't tried it yet, but it looks almost too interesting to not try.

{Quote hidden}> Resurrecting this thread from a month ago. > > I had issues fitting a piggy-back pc-board into the original product. > However, I have been slowly migrating the parent of this product from a > 16F to an 18F23k22 and accelerated that effort as it would've possibly > let me use the D/A technique below... specifically because I'm using the > internal oscillator, and the 18F will let me run that much faster. > > Net result is that it works nicely -- at 32Mhz. And since I'm using C > now for my 18F apps, it got implemented very rapidly. Thanks for > informing me of this technique. > > Now though, I have to ask about this method -- why exactly the need for > a random number? I am simply sampling as much as possible, and toggling > the output bit depending on if the sampled value is greater or less than > the measured value. That measured value being the filtered output bit. > > Cheers, > -Neil. >From your description, it sounds like you're using an adc to measure the output voltage and then setting the output bit high or low as appropriate. This is similar to delta modulation. Is this what you're doing? I think the random number approach did not use an ADC. Instead, the desired output is compared to random numbers and the pin set high if the desired output is above the random number and low if it's below. If the random number were not random, but were, instead, a steadily increasing number, you'd get standard PWM. But, you would have to increment the number at fixed intervals to get the required duty cycle. As I read this random number method, you can call run it at irregular intervals and the output will still be correct due to the use of the random numbers. It's a really interesting idea. Harold -- FCC Rules Updated Daily at http://www.hallikainen.com - Advertising opportunities available! Not sent from an iPhone

On 5/13/2012 11:13 PM, Harold Hallikainen wrote: > > From your description, it sounds like you're using an adc to measure the > output voltage and then setting the output bit high or low as appropriate.. > This is similar to delta modulation. Is this what you're doing? > > I think the random number approach did not use an ADC. Instead, the > desired output is compared to random numbers and the pin set high if the > desired output is above the random number and low if it's below. If the > random number were not random, but were, instead, a steadily increasing > number, you'd get standard PWM. But, you would have to increment the > number at fixed intervals to get the required duty cycle. As I read this > random number method, you can call run it at irregular intervals and the > output will still be correct due to the use of the random numbers. > > It's a really interesting idea. > > Harold Yes, that's pretty much what I'm doing... I think I'm missing part of a discussion somewhere where that random number technique was explained as the technique I used, and after thinking and thinking I could not figure out the purpose for random numbers. I'll have to find that other discussion with the technique I used. Curious though -- what's being used to *compare* the random number to the desired output

Neil wrote: {Quote hidden}> On 5/13/2012 11:13 PM, Harold Hallikainen wrote: > > > From your description, it sounds like you're using an adc to measure the > > output voltage and then setting the output bit high or low as appropriate. > > This is similar to delta modulation. Is this what you're doing? > > > > I think the random number approach did not use an ADC. Instead, the > > desired output is compared to random numbers and the pin set high if the > > desired output is above the random number and low if it's below. If the > > random number were not random, but were, instead, a steadily increasing > > number, you'd get standard PWM. But, you would have to increment the > > number at fixed intervals to get the required duty cycle. As I read this > > random number method, you can call run it at irregular intervals and the > > output will still be correct due to the use of the random numbers. > > > > It's a really interesting idea. > > > > Harold > > Yes, that's pretty much what I'm doing... I think I'm missing part of a > discussion somewhere where that random number technique was explained as > the technique I used, and after thinking and thinking I could not figure > out the purpose for random numbers. I'll have to find that other > discussion with the technique I used. > > Curious though -- what's being used to *compare* the random number to > the desired output? I wrote the original random number ATOD post a few years ago. I used the random number ATOD trick to create a PWM on a few projects including the tone generator for a DTMF dialler.. To understand the random ATOD start by reviewing how a normal timer based PWM works. The output voltage is equal to the time on divided by time off multiplied by voltage swing between 1 and 0 logical levels plus logic 0 offset. A random number ATOD also creates a an output ratio of 1's and 0's. The ratio of output 1's to the total bits The output voltage is equal to the ratio of 1's to the total 1's and 0's multiplied by the voltage swing between 1 and 0 logical levels plus logic 0 offset. They will produce the same output voltage. To create a Random number PWM. 1) Decide the range of the random number RandRange typically 2^8 or 2^16 . 2) The output voltage when output is logical 1 is Vout1 and output voltage when output is logical 0 is Vout0; 3) Desired Output voltage is Vout somewhere between Vout0 and Vout1 4) Calculate a random threshold number. The ratio of RandomThreshold / RandomRange is equal to the desired output Vout / (Vout1 - Vout0) 5) RandomThreshold = (Vout - Vout0) / (Vout1 -Vout0) * RandRange . 6) Generate a random number as often as possible and compare it to the RandomThreshold. If the RandomThreshold is higher or equal to the random number output a 1 else output a 0 Thats it. I typically used a 8 bit random number for fast moving data like the DTMF and 16 bit random numbers for slow moving data. The resolution is very similar to a timed PWM. Advantages of random PWM, the noise is easier to filter because it has higher frequency components than the timed PWM. Disadvantages to the random PWM. It is computationally more intensive than a timed PWM. Regards, Walter.. -- Walter Banks Byte Craft Limited http://www.bytecraft.com