Differential Probe based on AD830

[leonerd]

A recent video by Marco Reps () involved a differential probe, the schematic for which he has supplied - https://drive.google.com/file/d/0B6S_PcWWM1YlVDhreUN4NDJXN3c/view (also attached).

The design seems based around the AD830 difference amplifier, which seems ideally suited to the task. I think I understand the operation of most of the circuit, but the front-end parts confuse me a little. The resistor chains R1-R4 and R5-R7+P1 are the main input dividers to bring the input down to lower level, IC1 (the AD830) takes away the common-mode and amplifies the difference by two, and IC2 (the AD844) is an output buffer with switchable gain between x1 and x10.

A question on it though: Why the capacitors C1-C8 around the input section? Are they doing bandwidth-limiting or noise filtering, or what is their purpose? C4 and C8 are trimmer caps, so presumably they need to be trimmed into some condition. What might that be?

[baltersice]

The design is based on an old Elektor article (May 1994), there you can read all about it. The caps are for frequency compensation (like those you find on passive probes too).

[leonerd]

The design is based on an old Elektor article (May 1994), there you can read all about it.

I did see that yes, but it appears to only be in German (and also behind a paywall, but I'd be happier with that if it was in English).

The caps are for frequency compensation (like those you find on passive probes too).

Ah, I see.

I've been considering making something like this, only without the huge divide-down ratio it has. I don't need to be reading several-hundreds of maybe-mains volts, but I would like very small differentials like across shunt resistors or FETs, or maybe RS-485 lines and similar. So I was thinking of making a probe that had a much smaller divider, maybe even just 2:1, and then compensate for that in the amplifier so the output value was 1:1 in relation to the input difference.. or maybe even higher for really small measurements.

Do you think this design would work as well for that?

[baltersice]

I think it would be ok for that, but the AD830 is pretty old by now ...
Maybe you are up for a challenge? There are more ambitious projects out there ...

[leonerd]

I think it would be ok for that, but the AD830 is pretty old by now ...

Is there anything inherently bad with it? Some of the parts I end up using are at least 1990s if not 1980s - those seem fine. Mature and stable. The AD830 in particular is still a current chip, Farnell and Mouser both stock it; it seems quite cheap for the specs.

Maybe you are up for a challenge? There are more ambitious projects out there ...

That's up at 1GHz of input bandwidth - not sure I'd really be able to make use of that in any case, having only a 50MHz bandwidth scope. A more ambitious one might be more experimental. This design seems tried and tested.

[Kleinstein]

The AD830 has a rather high noise level, both voltage noise and current noise. This could be important for looking at small signals - due to the hard to avoid divider at the input, small signals are not that unusual. The old style discrete solutions tend to be lower in noise by something like a factor of 3, maybe 5.

[exe]

Is there any other design for a diffprobe  (for small signals)? I didn't find any except those mentioned here. I'm a bit hesitating to work with such sensitive FETs due to ESD. I think it's very easy to kill input fets.

[oldway]

AD830 is indeed an interesting device. But his high quiescent current (about 15 mA) make it difficult to use it for battery powered differential probe.

[Kleinstein]

JFETs are usually not that sensitive to ESD. In the final circuit the 1 M or similar resistor at the input helps a lot to make it ESD safe.  A not so pleasant property of FETs is there rather scattering threshold voltage - so offsets and drift can be significant or at least adjustments needed.

15 mA of supply current is also bad for battery operation. Also the 2.3 V amplitude limit for the input is not that high - a higher maximum amplitude would allow for less division.

[NorthGuy]

You don't have to use discrete JFETs. You may use a JFET-based op amp instead, such as OPA659.