At 04:09 PM 5/1/2013, you wrote: >Good day to all. Hi, Dwayne:- >We've just been contacted and asked if we can design an inexpensive >current source that is computer controlled. However, the specs are >considerably higher than anything that I've done in the past. > >What they want: > >Output current adjustable from 10uA to 1 Amp. Step decade ranges are >acceptable - it does not need to be able to be adjustable over a 5 >decade range in one fell swoop. > >Output voltage compliance of !! -5V !! to +10V. I don't know if they >really want this thing to be capable of a bipolar output but am in >the process of finding out. For now, assume that the output >compliance is single polarity only. > >Here's the kicker: accuracy of 0.01% of the set-point. As Russell said, that's a pretty unrealistic spec. If you look closely at real specs (eg. Keithley 24xx), the best accuracy on the 100uA scale is 0.031% + 20nA, in a limited temperature range around 23=B0C. At 10uA, that's +/- 23nA or almost +/- 0.25% , which is more than 20x worse than "0.01% of setpoint". I have received reports of certain problems with the 24xx, and I'd prefer the Agilent instruments, but they make Keithley look cheap (in cost)= .. You are correct that good accuracy is not achievable realistically in a sin= gle range... range switching typically is accomplished with latching signal relays to avoid errors due to resistance and thermal EMFs, >I don't yet have any idea of frequency response requirements but, >again, am in the process of finding out. However, I *think* that >they are doing battery coulometry, so slow response is probably acceptable= .. Why such silly tight specs for a battery load? I would think 0.1% would be overkill, and probably 1% is good enough. Are they going to measure the temperature of the battery internals to 0.01 degree C? >The customer wants the units to be computer controlled. I was just >going to use some form of USB - Serial arrangement - most likely >using USB-equipped PICs. I do plan to isolate the unit from the host >USB port with some form of isolation barrier. > >I do adjustable current sources and sinks with 8 to 10 bit accuracy >on an irregular basis. But this is the first time that I've been >asked to do something approaching 14 bits of accuracy. > >Questions: > >Is anyone aware of something with similar specs that is already >readily available? The customer is currently using a bunch of >expensive Keithley units and is looking for less expensive >alternatives. Personally, I'd rather point the customer to an >already-existing solution rather than starting this particular design cycl= e. There are plenty of electronic loads used to test batteries. You can set them to provide a constant resistance, a constant current, programmed pulsed drain etc. They won't be 0.01% unless someone is lying (as data sheets often do). >But: if I do have to design this beast, Does anyone have any favorite >A/D and D/A convertors that are 14 bit accurate or better? I've been using the AD5791, which is not cheap, but it has decent DC specs (perhaps better than what the instrument mfrs are using). It has SPI as an option. Hint:- if you need smooth dynamic response, watch out for the glitch energy characteristics. You could pick a fine resolution cheaper monotonic DAC and use a sigma-delta ADC to correct the current by adjusting the code. It should boil down to getting a very accurate (and stable) voltage reference and a very accurate and stable resistance*. You can (and probably will have to) calibrate out initial errors, so budget getting something NIST-traceable with that kind of accuracy if you are going to be building them. You'll also need at least a crude thermal chamber. You will not be able to easily buy parts to guarantee that kind of accuracy, it will have to be adjusted in (probably at a minimum of one point per range) and confirmed (over the current ranges and temperature spans). There are $$$ 0.01% references and $$ 0.01% resistors, but that already exceeds your error budget. *Getting that kind of accuracy in the current sense resistor is non-trivial- at 1A it will be dissipating at least 1W if you're to maintain decent accuracy at 10% of full scale. I don't want to be too discouraging, with a reasonable spec, similar to the commercial units, it's all quite do-able, but I'm not sure I'd recommend it as a first foray into precision analog design unless schedule and dollars are quite favorable, as subtle things gang aft agley. Best regards, --=20 http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .