Post by thread:[EE]: high resolution DAC

I am looking for a way to output a very high resolution voltage from a microcontroller. I need about 24 bits of resolution. From what I've seen, the only widely available 24 bit DACs are audio DACs. Has anybody used an audio DAC in a non-audio application? One idea I had was to somehow cascade lower resolution DACs in series. i.e. use the output of one as the ground for the next and then scale the final DAC's output for my needs. This would require lots of opto-isolation, isolation transformers, and generally sounds like a bad idea to me. Does any body have any other ideas? Is there anything about an audio DAC that would make it not suitable for use in high accuracy testing equipment? Thanks, Eric Christensen -- http://www.piclist.com#nomail Going offline? Don't AutoReply us! email spam_OUTlistservTakeThisOuTmitvma.mit.edu with SET PICList DIGEST in the body

> I am looking for a way to output a very high resolution voltage from a > microcontroller. I need about 24 bits of resolution. From what I've Cascading DACs will only work IF the LSB of the MSB DACs (in fact, all bits of all higher order DACs) have an accuracy of 1 LSB. This is most unlikely to be the case and would be exceptionally hard to achieve. A tracking DAC using a A2D and feedback may do the job. You can get 24 bit Sigma Delta A2D converters. You measure and then adjust output accordingly. Samples per second are rather slow. Do you REALLY need 24 bit accuracy. That is exceptionally fine tolerance (as you know) - 1 part in 16 million. There are very very very (add a few more) few applications that can really work to this sort of accuracy overall. Are you able to share the application or give more detail. Because 1. It sounds fascinating 2. We may well be able to suggest a less precise solution. Russell McMahon -- http://www.piclist.com hint: To leave the PICList .....piclist-unsubscribe-requestKILLspam@spam@mitvma.mit.edu

On Sat, 18 Oct 2003 17:03:18 +1300 Russell McMahon <apptechKILLspamPARADISE.NET.NZ> wrote: > Do you REALLY need 24 bit accuracy. > That is exceptionally fine tolerance (as you know) - 1 part in 16 > million. There are very very very (add a few more) few applications > that can really work to this sort of accuracy overall. > > Are you able to share the application or give more detail. > Because > 1. It sounds fascinating > 2. We may well be able to suggest a less precise solution. I'm working on a current limited power supply that must control current from 0 to 4 amps with 1 uA resolution with less than 1% error. The current design uses hardware that controls current proportional to a given voltage. The idea is to have a microcontroller set that voltage with a DAC, creating a programmable current limiting device. Since there are 4,000,000 steps at 1uA each from 0 to 4 amps, that means my DAC needs at least 22 bits of accuracy. Since 24 bit audio DACs seem common, I figured they would be a cost effective solution IF they would work for non-audio applications. I can't think of why they wouldn't but... I've never done anything with audio, so I don't really know. I am considering creating several current 'ranges' so that I would need less DAC accuracy and would then just select the correct range. This may be only feasible solution, but I want to fully investigate the high resolution DAC area before I rule it out. Eric > > > Russell McMahon > > -- > http://www.piclist.com hint: To leave the PICList > .....piclist-unsubscribe-requestKILLspam.....mitvma.mit.edu Eric Christensen Iowa State University Computer Engineering and Computer Science Home: (515) 572-7708 Cell: (612) 202-9336 EraseMEechrisspam_OUTTakeThisOuTiastate.edu -- http://www.piclist.com hint: To leave the PICList piclist-unsubscribe-requestspam_OUTmitvma.mit.edu

Eric Christensen wrote: > I'm working on a current limited power supply that must control current > from 0 to 4 amps with 1 uA resolution with less than 1% error. Hi. 1% of what ? Full scale ? Or 1% of 1uA ? 1% of 4A (full scale) is actualy 40000 uA !! Lets say you are outputing 4A, what is then the allowed "error" (expressed in Ampere !) ? Jan-Erik. -- http://www.piclist.com hint: To leave the PICList @spam@piclist-unsubscribe-requestKILLspammitvma.mit.edu

> I'm working on a current limited power supply that must control current > from 0 to 4 amps with 1 uA resolution with less than 1% error. There's a lot of difference between 1uA in 4A and 1% :-) Can you explain that in more detail. eg if I wanted 3.456789 amps what would be acceptable. At 1% of reading would it be +/- 0.034567 amps (easily achieved) or 1% of full;l scale (0.04A) or ... ? Again, why do you need 1 uA resolution in the current limit. Inquiring minds want to know :-) (FWIW - the origin of that phrase is explained here http://www.yaelf.com/aueFAQ/mifnqrngmndswnttkn.shtml) Russell McMahon -- http://www.piclist.com hint: To leave the PICList RemoveMEpiclist-unsubscribe-requestTakeThisOuTmitvma.mit.edu

Sorry... I wasn't very clear on that. I need to be able to set the current within 1% of 1 uA. So 3 amps must really be 2.99999999 to 3.00000001 amps. The hardware to control the current this accuratly is being developed by someone else. The initial plan, however, calls for the current be controlled by a proportional voltage. When I starting punching in the numbers for this, I was pretty much stumped. A 24 bit DAC seems to be about the only thing precise enough. Otherwise we will need to implement the multiple range setup. Eric On Sat, 18 Oct 2003 11:54:17 +0200 "Jan-Erik Soderholm XA (TN/PAC)" <spamBeGonejan-erik.xa.soderholmspamBeGoneERICSSON.COM> wrote: {Quote hidden}> Eric Christensen wrote: > > > I'm working on a current limited power supply that must control > > current from 0 to 4 amps with 1 uA resolution with less than 1% > > error. > > > Hi. > > 1% of what ? Full scale ? Or 1% of 1uA ? > > 1% of 4A (full scale) is actualy 40000 uA !! > > Lets say you are outputing 4A, what is then the > allowed "error" (expressed in Ampere !) ? > > Jan-Erik. > > -- > http://www.piclist.com hint: To leave the PICList > TakeThisOuTpiclist-unsubscribe-requestEraseMEspam_OUTmitvma.mit.edu Eric Christensen Iowa State University Computer Engineering and Computer Science Home: (515) 572-7708 Cell: (612) 202-9336 RemoveMEechrisTakeThisOuTiastate.edu -- http://www.piclist.com hint: To leave the PICList piclist-unsubscribe-requestEraseME.....mitvma.mit.edu

> Sorry... I wasn't very clear on that. > > I need to be able to set the current within 1% of 1 uA. > > So 3 amps must really be 2.99999999 to 3.00000001 amps. The hardware to > control the current this accuratly is being developed by someone else. > The initial plan, however, calls for the current be controlled by a > proportional voltage. When I starting punching in the numbers for this, > I was pretty much stumped. A 24 bit DAC seems to be about the only > thing precise enough. Otherwise we will need to implement the multiple > range setup. So you really need 28 bits of accuracy (1:400,000,000) If you go in 1 uA steps that's 1:4,000,000 = 22 bits control accuracy. I still find it very hard to believe that such accuracy is either attainable or necessary in any device that deals in normal reality. I'm happy to be shown to be wrong of course :-). Unless you are dealing in some utterly leading edge research or measurement such precision is almost unheard of. What, are we allowed to ask, is this device ??? Russell McMahon -- http://www.piclist.com hint: To leave the PICList RemoveMEpiclist-unsubscribe-requestEraseMEEraseMEmitvma.mit.edu

Eric Christensen wrote... >I need to be able to set the current within 1% of 1 uA. > >So 3 amps must really be 2.99999999 to 3.00000001 amps. The hardware to >control the current this accuratly is being developed by someone else. >The initial plan, however, calls for the current be controlled by a >proportional voltage. When I starting punching in the numbers for this, >I was pretty much stumped. A 24 bit DAC seems to be about the only >thing precise enough. Otherwise we will need to implement the multiple >range setup. You expect to be able to set a 0 - 4 ampere current to within an absolute accuracy of +/- 10 nanoamps? This is ridiculous; what you're being asked to achieve, and what the poor guy who's designing the hardware to control the current is being asked to achieve, is far, **FAR** beyond the outer limits of what is realizable in practice. If you were talking about being able to adjust the current with a resolution of 1 part per million or so (note I'm talking about resolution, NOT accuracy), you might be able to come close with some extremely careful design. One example of such a monster is shown in Linear Technology, Inc. application note AN-86, "A Standards Lab Grade 20-Bit DAC with 0.1ppm/0 Drift", at http://www.linear-tech.com/pub/document.html?pub_type=app&document=92 But what you're asking for has not only four hundred times that resolution, it must have an accuracy of one part in 400 million, as well. Sorry, but you simply can't get there from here, period. Whatever it is you're trying to do, needs to be re-thought from the ground up. What ARE you trying to do, anyhow? (just curious) Dave D. -- http://www.piclist.com hint: To leave the PICList RemoveMEpiclist-unsubscribe-requestspam_OUTKILLspammitvma.mit.edu

Eric, One additional note about audio DACs, since you asked: I think you're going to find that when an audio DAC is advertised as "24 bit", that's an expression of its signal-to-noise ratio, not its DC accuracy. I'm not very familiar with these devices but I would expect, given their intended application, that their DC accuracy would be rather poor. Dave D. -- http://www.piclist.com hint: To leave the PICList RemoveMEpiclist-unsubscribe-requestTakeThisOuTspammitvma.mit.edu

Eric Christensen wrote: > I'm working on a current limited power supply that must control current > from 0 to 4 amps with 1 uA resolution with less than 1% error. So you're saying you want to be able to set the current to, for example, 3.500007 amps plus or minus .035 amps?? Sounds a bit silly. This is just like saying you insist someone measure the distance from Boston to Seattle in multiples of 5 feet, but they are allowed to be off by 30 miles. What's the point? Even if you could get such a reference signal with zero error, how are you going to design the analog section with a noise level where 1uA resolution is meaningful? ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: To leave the PICList EraseMEpiclist-unsubscribe-requestspamspamBeGonemitvma.mit.edu

Eric Christensen wrote: > I need to be able to set the current within 1% of 1 uA. > > So 3 amps must really be 2.99999999 to 3.00000001 amps. > The hardware to control the current this accuracy is > being developed by someone else. There was a post stating that when choosing a college one should prefer the college with the best physics labs, not EE courses. And, by the way, "someone else" have to be Nobel Prize Winner, I think, to maintain 0.00000001 % precision current sensing resistor. Mike. -- http://www.piclist.com hint: To leave the PICList RemoveMEpiclist-unsubscribe-requestKILLspammitvma.mit.edu

Eric Christensen wrote: > Sorry... I wasn't very clear on that. > > I need to be able to set the current within 1% of 1 uA. > > So 3 amps must really be 2.99999999 to 3.00000001 amps. You're seriously asking for a current source that can deliver 3A +-10nA!!!? That's an error of 3.3 parts per *billion*, or .0033PPM. I doubt that would be within the capabilities of NIST even under optimal laboratory conditions and lots of tax dollars to spend, let alone someone who has to ask for help from the PIClist. Just to point out one of MANY problems, let's be generous and say you can get the sum of all your proportional temperature dependencies to be no more than 10ppm/degC. That means a .00033 degree C change in temperature blows your error budget. Even if you could get the temperature dependency down to 1ppm/degC, you'd still need to hold everything constant to within 1/300 of a degree C. I think you have misstated the true problem somewhere and latched onto an unworkable solution. In any case, you're waaaaaay over your head. Go home. Do something else. Forget about this. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: To leave the PICList piclist-unsubscribe-requestSTOPspamspam_OUTmitvma.mit.edu

Eric, Don't be discouraged by the negative comments you are getting here. However, DO note that almost everything that people are saying technically is true. - your spec is so stunningly tight it is almost unbelievable - the difficulty in achieving this is utter leading edge or beyond. - Very probably you don't actually need anything like this accuracy for several good reasons. So far you have given us only general details about what you are trying to do. It may be that the requirement must be kept secret and can't be shared in full detail. HOWEVER it would be a very very very good idea if you could describe what you are trying to achieve in as much detail as possible. NOT how you think you are going to do it but the actual aim. Then people may well be able to provide some good advice. In the absence of more detail. all people can say is that it's not practical. Is this a consumer or industrial product or a research one off? What is the budget? More detail is essential I'm afraid. Russell McMahon -- http://www.piclist.com hint: To leave the PICList spamBeGonepiclist-unsubscribe-requestSTOPspamEraseMEmitvma.mit.edu

Maybe it's for a warp field generator? You need something pretty precise to keep those dilithium crystals properly aligned. Talk about your overspec'd power supply... I know that some NMR stuff is on this order of resolution (bias coils in spectroscopy). R "Jan-Erik Soderholm XA (TN/PAC)" wrote: {Quote hidden}> > Eric Christensen wrote: > > > I need to be able to set the current within 1% of 1 uA. > > > > So 3 amps must really be 2.99999999 to 3.00000001 amps. > > So, about 0.003 ppm then. > Well, I'm out of ideas... > > What on earth is this device to be used for ? -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Russell McMahon wrote: > ...So far you have given us only general details about what > you are trying to do. It may be that the requirement must > be kept secret and can't be shared in full detail... No secrets, just for operating air to air shoulder launched inter-planet rockets for Saddam. (See "[OT;] Shoulder launched rockets missing in IRAQ") Mike. -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

> From: Dave Dilatush[SMTP:KILLspamdavedilatushspamBeGoneCOMCAST.NET] > Sent: Saturday, October 18, 2003 10:24 AM > To: EraseMEPICLISTEraseMEMITVMA.MIT.EDU > Subject: Re: [EE]: high resolution DAC > Eric, > One additional note about audio DACs, since you asked: I think > you're going to find that when an audio DAC is advertised as "24 > bit", that's an expression of its signal-to-noise ratio, not its > DC accuracy. I'm not very familiar with these devices but I > would expect, given their intended application, that their DC > accuracy would be rather poor. That's right. A DAC meant for DC applications is referred to as a "precision" DAC, which means that each step is within 1/2 LSB of its absolute voltage. The 24 bit rating of an audio DAC means that it has 2 ^ 24 discrete output steps. This provides for that "granularity" over small ranges of output voltage. In reproducing audio, that is all that matters; you don't want to hear the steps in the output when the peak-to-peak signal is small. There is no guarantee that the cumulative accuracy is particularly good. In other words, audio DACs are designed for AC use, not DC. John Power -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Greetings, Instead of purchasing a DAC just make an R2R ladder network DAC. We build one in our EE course in second year, it was 16bit accuracy and worked rather well. You can just add more resistors to achieve the higher resolutions like 24bit or whatever you require. I'm not sure how good they perform as far as distortion and non-linearity are concerned but at least it's an option. Cheers Damien At 10:40 PM 17/10/2003 -0500, you wrote: {Quote hidden}>I am looking for a way to output a very high resolution voltage from a >microcontroller. I need about 24 bits of resolution. From what I've >seen, the only widely available 24 bit DACs are audio DACs. Has anybody >used an audio DAC in a non-audio application? > >One idea I had was to somehow cascade lower resolution DACs in series. >i.e. use the output of one as the ground for the next and then scale the >final DAC's output for my needs. This would require lots of >opto-isolation, isolation transformers, and generally sounds like a bad >idea to me. > >Does any body have any other ideas? Is there anything about an audio >DAC that would make it not suitable for use in high accuracy testing >equipment? > >Thanks, >Eric Christensen > >-- >http://www.piclist.com#nomail Going offline? Don't AutoReply us! >email @spam@listserv@spam@spam_OUTmitvma.mit.edu with SET PICList DIGEST in the body -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

I don't think it's possible to get the *requested* precision that way. Even 16 bits of precision it not easy, but of course it depends on how you define "rather well"... Jan-Erik. Damien Cahill write: > Instead of purchasing a DAC just make an R2R ladder network DAC. We build > one in our EE course in second year, it was 16bit accuracy and worked > rather well. You can just add more resistors to achieve the higher > resolutions like 24bit or whatever you require. I'm not sure how good they > perform as far as distortion and non-linearity are concerned but at least > it's an option. -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

Damien Cahill wrote: > Instead of purchasing a DAC just make an R2R ladder network DAC. We > build one in our EE course in second year, it was 16bit accuracy and > worked rather well. I can imagine it "worked rather well" if all you did was run audio thru it. However, I seriously doubt your hand built DAC was monotonic between all counts. That would require resistors with relative accuracies of better than .002%. Even if you could by some miracle trim your resistors to the required accuracy, a tiny temperature drift would get them all out of whack again. > You can just add more resistors to achieve the > higher resolutions like 24bit or whatever you require. No, you can't. Having 24 bits of meaningful resolution implies monotonicity. The required resistor accuracy to achieve that with a R/2R ladder from discrete parts is unattainable. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: PICList Posts must start with ONE topic: [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads

You are right! Are there any details on the web? george ----- Original Message ----- From: "Stef Mientki" <spamBeGones.mientkiKILLspamMAILBOX.KUN.NL> To: <.....PICLISTspam_OUTMITVMA.MIT.EDU> Sent: Sunday, October 19, 2003 11:57 PM Subject: Re: [EE]: high resolution DAC {Quote hidden}> My employer has an 128 channel, 32 bits AD-converter. > But I don't think there's anyone in this newsgroup, > who will believe it ;-) > > Stef Mientki > > -- > http://www.piclist.com hint: PICList Posts must start with ONE topic: > [PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads > -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Denny Esterline wrote: > Given today's culture of paranoia, you might do well to more > clearly indicate when joking about something like this. :-) > > > No secrets, just for operating air to air shoulder > > launched inter-planet rockets for Saddam. Hmm, you are right, setting precisely current through coils you can compensate Earth's field. With vacuum zero field detector you can thus obtain the vector of the Earth's field... Hmm, the same but unmanned September jets. Mike. -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

> From: Damien Cahill[SMTP:TakeThisOuTdamiencahill.....TakeThisOuTAUSTARNET.COM.AU] > Sent: Sunday, October 19, 2003 5:27 PM > To: TakeThisOuTPICLISTKILLspamspamMITVMA.MIT.EDU > Subject: Re: [EE]: high resolution DAC > Instead of purchasing a DAC just make an R2R ladder network DAC. We build > one in our EE course in second year, it was 16bit accuracy and worked > rather well. You can just add more resistors to achieve the higher > resolutions like 24bit or whatever you require. I'm not sure how good they > perform as far as distortion and non-linearity are concerned but at least > it's an option. How did you manage that? Even if you use 0.1% resistors, you can only get 1 part in 1000 accuracy. Sixteen bits needs 1 in 65536 which is 15 parts per million. Are there resistors with that accuracy? John Power -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

John N. Power wrote: > > From: Damien Cahill[SMTP:.....damiencahillRemoveMEAUSTARNET.COM.AU] > > Sent: Sunday, October 19, 2003 5:27 PM > > To: RemoveMEPICLISTspamBeGoneMITVMA.MIT.EDU > > Subject: Re: [EE]: high resolution DAC > > > Instead of purchasing a DAC just make an R2R ladder network DAC. We build > > one in our EE course in second year, it was 16bit accuracy and worked > > rather well... > > How did you manage that? Even if you use 0.1% resistors, you can only > get 1 part in 1000 accuracy. Sixteen bits needs 1 in 65536 which is 15 > parts per million. Are there resistors with that accuracy? The best "standard" resistors at Farnell are 0.01%, 0.6 ppm/DegC (SMD). That's still 100 ppm, so a 16-bit DAC is still "hard" even with these... And they are $15 - $40 *each* in single quantity... :-) Jan-Erik. -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

>> > Instead of purchasing a DAC just make an R2R ladder network DAC. We >> build one in our EE course in second year, it was 16bit accuracy and >> worked rather well... >> >> How did you manage that? Even if you use 0.1% resistors, you can only >> get 1 part in 1000 accuracy. Sixteen bits needs 1 in 65536 which is 15 >> parts per million. Are there resistors with that accuracy? > > > The best "standard" resistors at Farnell are 0.01%, > 0.6 ppm/DegC (SMD). That's still 100 ppm, so a 16-bit > DAC is still "hard" even with these... > > And they are $15 - $40 *each* in single quantity... :-) Wait a minute, you need to read up a bit on R2R ladders if you think you need .01% resistors to get 16 bit performance. It isn't a walk in the park, but the tolerance of the resistors in the ladder does not have to be very good. -Bob -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Bob Blick said: > Wait a minute, you need to read up a bit on R2R ladders if you think you > need .01% resistors to get 16 bit performance. It isn't a walk in the > park, but the tolerance of the resistors in the ladder does not have to > be very good. Following up on that, I am not suggesting you do it ;-) Buy a DAC if you want accuracy, don't build one. -Bob -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

> I don't think it's possible to get the *requested* precision > that way. Even 16 bits of precision it not easy, but > of course it depends on how you define "rather well"... 16 bits is plenty hard enough. It requires that the R/2R ratio for the MSB be better than 1/32768 accurate at least afaik. 20 bit DACs are very common, made with PWM. Such are the DACs used until recently in the tuning of nearly all TV sets (now they are mostly synthesized). A long term stable 24 bit adc however is laboratory stuff. Olin L. pointed out why. A slow output PWM dac could be done easily with a pic (maybe not 24 bit). Peter -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Bob Blick wrote: > Wait a minute, you need to read up a bit on R2R ladders if you think you > need .01% resistors to get 16 bit performance. It isn't a walk in the > park, but the tolerance of the resistors in the ladder does not have to > be very good. The absolute tolerance isn't that important, but the 2:1 ratio needs to be as good as the number of bits in the ladder. If this 2:1 ratio is off, then the output is no longer monotonic, which generally makes the DAC useless for the specified number of bits. For example, consider a 16 bit DAC at the crossover between 7FFFh and 8000h. The two voltages will be produced by different sets of resistors that need to match in ratio well enough to produce the expected 1/65535 of full scale step size. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Peter L. Peres wrote: > 20 bit DACs are very common, made with PWM. Such are the DACs used until > recently in the tuning of nearly all TV sets (now they are mostly > synthesized). A long term stable 24 bit adc however is laboratory stuff. > Olin L. pointed out why. A slow output PWM dac could be done easily > with a pic (maybe not 24 bit). Yes, you can get arbitrarily high resolution using PWM if you're willing to wait long enough. However, the problem then becomes the reference supply. It would be very hard to make the reference supply stable enough so that the noise (even the low frequency noise that makes it thru the filter) doesn't swamp a single LSB change. A very small temperature change, for example, will do that at 24 bits as was pointed out earlier. ***************************************************************** Embed Inc, embedded system specialists in Littleton Massachusetts (978) 742-9014, http://www.embedinc.com -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Peter L. Peres wrote: >>I don't think it's possible to get the *requested* precision >>that way. Even 16 bits of precision it not easy, but >>of course it depends on how you define "rather well"... > > > 16 bits is plenty hard enough. It requires that the R/2R ratio for the MSB > be better than 1/32768 accurate at least afaik. > TI has announced fast (1.25 Mc) 18 bit DACs (ADS1625 / ADS1626) Stef {Quote hidden}> 20 bit DACs are very common, made with PWM. Such are the DACs used until > recently in the tuning of nearly all TV sets (now they are mostly > synthesized). A long term stable 24 bit adc however is laboratory stuff. > Olin L. pointed out why. A slow output PWM dac could be done easily with a > pic (maybe not 24 bit). > > Peter > > -- > http://www.piclist.com hint: The list server can filter out subtopics > (like ads or off topics) for you. See http://www.piclist.com/#topics > -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics

Stef Mientki wrote: > > > > 16 bits is plenty hard enough. It requires that the R/2R ratio for the MSB > > be better than 1/32768 accurate at least afaik. > > > TI has announced fast (1.25 Mc) 18 bit DACs (ADS1625 / ADS1626) > Stef Both ADS1625 and ADS1626 are *ADC*'s ! focus.ti.com/docs/prod/folders/print/ads1625.html http://focus.ti.com/docs/prod/folders/print/ads1626.html (And the ADS1625 is $56.40 at Digikey in single quantity.) And about : > here are the specs: > > http://www.vsmmedtech.com/ctf/techlib/MEG_Specs.asp it says : "The dynamic range for MEG is 32 bits. Linearity is better than 10**-6" So it's not a monotonic 32 bit range, if I'm reading that correctly. More like aprox 20 bits, but I might be missinterpretting the "linearity" value above... Still, an impressing piece of equipment :-) Jan-Erik. -- http://www.piclist.com hint: The PICList is archived three different ways. See http://www.piclist.com/#archives for details.