Attenuate then amplify?

[throbscottle]

I'm in a bit of a quandary with the input circuit for my little bench dmm project.

I've designed it so far so that the most sensitive range, 2.5v (or 250mV with different scaling in the uC) connects the adc input directly to the input socket of the meter, using a reed (or other, more robust) relay.

The more I look at it, the more this seems a Bad Idea to me, for the simple reason that if a voltage higher than the relay's contact rating is applied to the input, it could arc over and blow everything up.

I did think about:
putting two relays in series, though it seems a severe kludge,
using an extra resistance in series with the adc input,
or having a high voltage relay (expensive).

I also thought about using the first tap on the voltage divider so there is a permanent divide by 10, changing the scaling in the uC and losing the μV capability, but where's the fun in that?

So the next most logical thing I could think of was to use the always divide by 10, but then have an ins. amp with a gain of 10, and put up with the extra possible source of errors.

Bearing in mind that I'm a hobbyist with no way to do anything particularly precisely, and a very limited budget, what does anyone think?

(While I'm on the subject, I'd like to give this project some kind of name to publish the design as - suggestions?)

 

[ronv]

Usually most a to d's can go a few tenths above and below ground and can tolerate some resistance in series with the input. So you could put say a 5k resistor in series with the input then add a reverse diode between the pin and +5 and ground so if the voltage goes to high or to low it will be clamped by the diodes.

 

[ronv]

I should have asked you what the maximum voltage is you anticipate. 5K may be to small it the voltage goes to high.
What kind of micro.

 

[throbscottle]

FS voltage for that input is 250V. There's a second input for voltages up to 2.5kV which puts an extra 90M HV resistor in series. Kludgey but easy to do, and I'm not likely to be measuring voltages which change around the 250 mark.

The micro is a PIC 16F57 - not the best choice but I came from an uninformed standpoint of knowing nothing about uC's and went with simplicity & lots of i/o.

It's a 24bit adc so a bit fussy about how it's inputs are connected. At some point I have to do maths and work out exactly what rc combination I'm supposed to be using on it's input to filter switching noise. Somewhere in the docs I found a note to put a resistor up to 15K in series with each side of the input, but I can't find it now. I don't see how that's going to help with a severe over-voltage though since the adc input resistance is so high.

I've designed in spark gaps and zeners and a fuse for protection, and the adc can tolerate ±0.3v over it's supply voltage, and it's an auto-ranging design (with range hold (done) and override (to do)). It's really just this first set of contacts I'm worried about - I don't think the zener diodes would conduct before the adc was damaged, but I can't think of a better protection apart from what's in my o/p. (and I don't mind blowing up an inst. amp, just don't want the extra source of errors!).

 

[ronv]

The 16f57 doesn't seem to have an a to d so you must have a separate one. That's the one we need to look at.
Just regular shottky diodes, not zeners.

 

[throbscottle]

Correct, it's a very basic and (I'm told) old design uC. I'm using a LTC2415 adc with a 5v reference, so my FS is ±2.5v at the adc inputs (2415 gives best accuracy when Vref = Vcc). Schottky diodes hey? That's a lot of diodes to make up the ±FS then!

 

[throbscottle]

Further googling on the subject leads me to chopper amplifiers. Hmmmmm

 

[audioguru]

My multimeters can have 120VAC appplied to their inputs without damage. The Fluke meter is auto-ranging so its idle sensitivity is 200mV full scale.

 

[ronv]

You need a buffer to get the input impedance up anyway-- take a look at the CA3160.
Here is a clamp circuit I think will work.

 

[throbscottle]

That looks like an ingenious yet brutal way to treat FET's, and resolve my entire problem, if it works as expected! I would never have thought of it.

CA3160 looks interesting. If I go with the 'always attenuate first, then amplify' option, the voltage to clamp from the voltage divider will be x10 lower, at ±0.25v - or some more convenient voltage since I can change the amp gain and scaling in the uC anyway.

I'll try your FET circuit when I have somewhere to build circuits again - ie in a couple of months if I'm lucky...

Thanks

 

[throbscottle]

Okay, I tried simulating it in LTSpice (the only simulator I can make sense of!) - clamps nicely on positive voltages to 5v, negative voltages not so good. Needs a split supply (though the sim crashed before I tried that so I didn't have time).

Actually your design got me to think about how diode clamps actually work, and I realised I've been looking at it all wrong - thanks for that, ronv I love this forum - I learn so much!

 

[ronv]

Works ok for me. Try this one.
The problem with the diodes is the leakage current. Since the input impedance is very high (10 meg.) very little current makes an error. You say not so bad I can cal it out - but it also changes with temperature. With the jfets you still need to set the 0 but the drift with temperature is low.

 

[throbscottle]

Works ok for me. Try this one.
The problem with the diodes is the leakage current. Since the input impedance is very high (10 meg.) very little current makes an error. You say not so bad I can cal it out - but it also changes with temperature. With the jfets you still need to set the 0 but the drift with temperature is low.

Oh okay. I consider myself educated! You do realise you've just made me commit to reading a stack of datasheets to compare diodes with jfet junctions? (Still needed a split supply to clamp negative voltages though )

 

[ronv]

Just +5 and ground I think. That way the input can't go much above +5 or below ground.

 

[ronv]

Actually, with such a large input resistance I'm not sure you need a clamp. The internal diodes may do it. I need to read the data sheet closer. Maybe someone else has experience with this chip. All their examples show no clamps.