Fast, wide bandwidth high side isolated current sensing down to DC

[dmills]

I have been sketching out a fairly high frequency switcher and noodling around with an idea for getting the output from a inductor DCR or high side resistive current sense setup out in a way that does not suffer too horribly with ground bounce, massive dV/dt and all the other nasties while retaining good CMRR and bandwidth, and think I just might have something.....

Basically the idea is to feed the differential voltage into a floating IF input on a double balanced mixer, with the RF input driven at some suitable RF frequency, then just sync demodulate the result once you are back on the control board away from the roar of the switching stuff. Obviously pick a RF drive frequency that is not the third harmonic of the switching rate (VHF drive is not out of the question so a few tens of MHz of bandwidth may well be reasonable, or square wave drive actually at the switching frequency as long as the transformers LF response is not too rotten, there will be a glitch at switching point but that is probably uninteresting anyway?).

Annoyingly the usual suspects (MCL) do not seem to have diode ring mixers that isolate the IF connections from the RF and LO ports, but there is no fundamental issue that I can see. Transformer isolation with response down to DC whats not to love?

The nice thing about this is that there is no need for active devices in the noisy domain, and the transformers only need a response down to whatever frequency the RF drive is, so they can be tiny little VHF things, it may even be possible to build both the required transformers as trifiliar windings on a single small pigs nose ferrite, in which case you are down to something like two four diodes and the transformer assembly in the power domain, all the rf drive and sync demod can be on the control board and is in any case trivial.

Thoughts?

I might need to order some diodes and small ferrites to have a play.

73 Dan.

[ejeffrey]

You can do this.  There is no need for an isolated mixer ring, you can just use an external transformer on the RF and LO ports.

Analog devices does something similar with the ADuM4190, converting an analog signal to PWM and coupling that via transformer.  That makes the coupled signal digital, and avoids the need to send an LO across the isolation boundary, or use good mixers.  The (very expensive) AD215 works almost exactly as you describe but in addition uses the ~400 kHz LO to transmit power across the boundary.  The receive side powers the transmit side.  One nice feature of this is that the isolator also can provide bootstrap/standby power to the output of an isolated power supply.

[jbb]

I’ve wondered about this in the past. I expect it could work.

There are also the feedback linearised optocouplers (HCPL 315 or something), but i’m not sure they’d have the bandwidth you need (and they need a power supply).

[dmills]

The linear optos are unimpressive bandwidth wise, and also can have issues with high dv/dt, they have their place, but it is not here.

Looks like the balanced mixer and sync demod is a somewhat more standard thing then I was expecting, I am still going to whistle up a few MCL mixers and some ferrites and see what might be possible.

73, Dan.

[T3sl4co1l]

So, a chopper amp.

You will find CMRR issues are still present.  A transformer is a transmission line component, so you have equivalent capacitance between windings, which injects CM into your diff mode at high frequencies (peaking at 1/4 wave = winding length or thereabouts).

You have to use balanced windings and shielding to maintain high CMRR.

Ethernet transformers do all of this internally, and are quite cheap.  May be very attractive.  Downside: only functional 1.5kV isolation.

"RF" of low MHz will suffice for Ethernet transformers (1-30MHz say).

Tim

[David Hess]

Transformer based isolation amplifiers have been around for a long time.  Chop the input, move it across an isolation transformer, and synchronously demodulate it.  Analog Devices, Burr-Brown, and others made a lot of money with products and solutions for this.  They have been largely replaced by linear optocouplers which are simpler to use but lower accuracy and previously voltage-to-frequency converters and now ADCs connected via a digital isolation barrier where high accuracy is required.

If the current is discontinuous, then just a current transformer is sufficient and I do not see why the continuous current at the switching node would be important enough to bother measuring except during development where high bandwidth AC/DC current probes can be used.

Still, it would be interesting to see what kind of performance a home grown isolation amplifier using an Ethernet or similar transformer could reach in this application.  I would adjust the chopping to be out of phase at the switching frequency so that the chopping noise is during the smooth uninteresting portion of the current waveform.  This would preserve the current transitions as much as possible.