We've looked at some tricks for passing analog signals across a galvanically isolated barrier. Let's finish up with two more.
Though I have never tried it, it is possible to use delta-sigma modulation to drive a digital optocoupler and achieve analog isolation. Ross Fosler described this technique on EETimes, a sister site. Apparently, this is not an unknown technique. It has been implemented in devices like ADI's AD7401A and TI's AMC1203.
That brings us to the linear optocoupler. Some design ideas and app notes I have seen suggest that it's possible to use ordinary optocouplers linearly. However, they will drift with temperature and age in unpredictable ways. Also, there are variations from one device to another, and that's not good when you want more accurate results.
In fact, conventional optocouplers are used as linear optocouplers in many switchmode power supplies. However, they are inside the control loop, which compensates for the variations.
Regarding devices specifically intended for linear applications, I have had much success with the IL300. It has one emitter (LED) and two matched phototransistor detectors. One detector is used as part of the feedback path around an op-amp, thus closing the loop. The other performs in a similar fashion on the other side of the isolation barrier.
It should be noted that the optocoupled gains of the devices do vary in production — so they are produced with bin categories, grouping devices with similar gains.
Several manufacturers offer similar devices, including the LOC110 from Clare (now IXYS). There are even packages with multiple analog channels.
Avago makes a linear optocoupler in a unique package, the HCNR200/20. It has also introduced the ACPL-C870, which I believe works on the delta-sigma modulation technique described earlier. I have used it as a high voltage-isolator. It worked well, but its input voltage range was limited.
Now let's try and plug any approaches that I may have missed. Tyco has some unusual devices that seem like proximity sensors but can pass data, signals (whatever that means), and power. They seem to be a long time coming, but they are very interesting nonetheless.
Let's not forget that it is possible to isolate with fiber optics using something like the PWM, VFC, or digital approaches that we have seen. Also, wireless can provide the ultimate isolation if your project has the economic headroom, though that would also require conversion to digital before the transmission.
Have you used any other approaches?