Seeing Dave & Dave talking about designs for current probes reminded me that I was going to write about a little project that I designed recently. I had been wanting a differential voltage probe for a while, but not enough to justify shelling out for a prebuilt one. I saw Jim Rowe's design in Silicon Chip, and thought about building one of these. This is quite an impressive design - 80MHz bandwidth - but has the unfortunate disadvantage of being AC coupled. It is really designed for high frequency, small signal work, while I was more after a device for general purpose work where safety isolation is desirable.
After looking at various design options (a proper differential design like Jim's, digitising in front of an isolation barrier, linear optocouplers (basically everything the Daves talked about)), I settled on the HCPL-7520-300E optoisolator. This has the ADC and DAC all built in, and basically only requires a power supply on each side. Although the bandwidth is only 100kHz, this should still be adequate for quite a few situations, and it is not too expensive at ~ AUD $7
The power supply for the module presented a few issues. I was initially going to use batteries for both power supplies, but couldn't fit them it the case I wanted to use. Then, I decided to try Jim's idea of using USB power from the scope, with an isolated DC-DC converter for the input side. However, I ran into trouble finding a converter with a suitable insulation rating. Eventually, I decided on a combination of the two approaches, using a battery for the input side, and external power for the output. A second optocoupler switches power to the input circuit when the output circuit is powered up, so the input is shut off automatically when the attached oscilloscope is powered down, preserving the battery.
Unfortunately, the module ended up being a bit more complex than I had first hoped - a negative rail generator and some op amps were needed for span and offset adjustment, and I also decided that a clipping indicator was necessary. But it all seems to work OK. I don't have any plans to manufacture these in quantity at this stage, but I can probably publish some schematics and board files if people are interested in building one up.