Sample & hold error

What's the acquisition time requirement? It could be done, but it would need a bigish storage cap.

What are the other specs? Voltage range, available supplies, budget, etc?

John

Bigish indeed. Perhaps a 10-turn pot and a motor drive? With a brake?

Voltage range means a _lot_. If that's 27uV drift out of a 100uV range, then the application is easy-peasy. If it's 27uV drift out of a 100MV range, it may be a teeny bit harder.

I had several nF in mind, with 10s of fA of leakage current. Not unreasonable, not really difficult. But a digital solution would likely be cheaper.

John

e:

Slow, bandwidth can be as low as a couple hundred hertz.

None of those are decided yet, except the Vin range, which can be anything you like at least upto yet-to-be-fixed-but-7-15v.

ke?

Its for a few off only, not a production run.

cheers

NT

So, an op amp buffer on a hold capacitor of 2 uF size will have to pull less than 27 E-6 V/sec * 2E-6 F = 50 pA ; yep, there's ways to do that. Your PC board layout will have to have some guard rings (and drive them at the node voltage), and if you don't trust the switch that charges the capacitor, backing it up with a small reed relay would be prudent.

If your tracking of the input has to slew too fast for a big hold capacitor, you'll have to use a fast-sample capacitor and transfer to the long-time-hold one as a separate operation.

Digitizing would be pretty easy- particularly if you have some time to acquire the data. Could be a one or two-chip deal for three or four channels (not many micros have DACs built in), and you won't have to worry about PCB leakage and that sort of thing.

In general, I've found real-world time constants in the 30 minutes and above range to be troublesome (back when PID temperature controllers were analog we _had_ to do that), and you're talking more like a four hour time constant.

A digital alternative to using a micro would be a simple tracking ADC (up/down counter feeding a R-2R network and a comparator). Just gate the clock to freeze the output. But a micro would be simpler, cheap and would just work.

If he doesn't need startling accuracy (say better than 1%) a DAC made from discrete resistors will probably get him what he wants. 1 chip plus some R's - certainly cheaper than am all analogue solution. If there is a question regards to programming than a PICAXE solution is the simplest and cheapest. If he already has a programmer then SWORDFISH Basic is free, easy and powerful.

Using a CD4017 and some resistors will get rid of all harmonics up to the ninth, so a fixed LPF will make a nice sine wave.

Cheers

Phil Hobbs

Wrong thread!

PH

But the teacher doesn't want them to digitize.

We should set up a standard rate for doing the kids' homework for them, like #300/hr or so.

Cheers! Rich

Can't you put a digital S/H inside an old capacitor case?

I think this is where you need to go to the digital world.

You can use a simply AVR or PIC, a very low pin count that has a ADC on it. The output you can do with a PWM scheme to get the value. It's a matter of simply holding it in memory as long as you wish!

If this has to be able to capture a very high speed transition, then use a sample and hold as the front end so that the processor will have enough time to get the reading. Jamie

I think I can use an LM324 or similar to make a simple 4 bit DAC with pfb Rs for the hold function. This would meet all the specs as far as I can see. An R chain to set the comparator levels, 4 trs to disconnect the input, resulting in an indefinite hold.

Any gotchas I should beware of with this? The only one I'm thinking of is Vin range being exceeded. Fairchild's datasheet says its good for the full supply range though on input pins.

NT

You might need a higher input impedance op amp - something with JFET or MOS inputs plus you will need a guard rail around your capacitor which will also have to be large and very low leakage - if it can be done at all.

l.

There is no capacitor, that's the point. The 4 digital channels keep their state by using pfb Rs

NT

didn't someone say digital was out?

What I proposed IS in the digital world.

NT

I've done a quick sketch of the cct diagram

Points marked S connect together, 0v gives hold, V+ gives tracking of input.

If looking to expand it to 8bits, addressing the errors would become an issue.

Have I missed anything in this 30 cent ADC?

NT

Sorry, replied to wrong post!

NT