FET Switch Isolation using optocouplers instead of conventional gate drivers

[rtarbell]

Background - skip if you want: One way of making a push-pull SMPS is to use two N-type FET switches, one called the "low side" driver (source referenced to actual ground) and the other called the "high side" driver (source referenced to the output). Some IC FET gate driver chips come with the bootstrapping option to control the high side FET switch, which can be very convenient (all you have to do is install the bootstrapping capacitor - the rest of the gate drive is already taken care of).

==> Instead of using this kind of high-side/low-side driver method, has anyone had success in switch-driving using optocouplers with their SMPS? The advantage I am looking for due to isolation is the fact that the PWM IC itself can be referenced to regular ground, while neither FET switch has to be referenced to ground because of the isolation provided by the optocouplers. This might be usable for regulating a bipolar +/- supply, instead of a unipolar + supply.

For instance, Avago makes a few optocouplers that are designed to drive FET gates at reasonable switching frequencies (the HCPL-3120 is rated at 2A max output current at up to 250kHz switching freq). Is this possible to use in an SMPS, at say, 100kHz switching frequency?

[sawreyrw]

This should work fine, but you need to provide the isolated power supplies. See Fig. 25 of the data sheet.

[Eva]

Be careful when using optocouplers this way, false tripping (and potential power stage destruction) will arise if the safe common-mode dV/dt is exceeded. In practice, this limits most optocoupler applications to moderate speed switching. HCPL3120 is rated at 15KV/us minimum (15V/ns) which is among the best that I know and enough for most SMPS with suitable snubbers. You will discover that other models have much inferior ratings.

On the other hand, IR bootstrapped gate drivers are rated at 50V/ns which gives a large safety margin For example, at full load my high voltage class D stuff Is slewing 500V in 20ns with IR2113 without problem.

Other drawbacks of the HCPL3120 include the 300ns turn-on and turn-off typical propagation delays, the 300ns max guaranteed pulse distortion, and the +/-350ns maximum guaranteed propagation delay difference between parts, which demands a 350ns or longer dead time.

Propagation delays in IR bootstrapped drivers are matched to 10ns, 20ns or 30ns depending on model. In practice, some parts like IR2113 are very stable with temperature allowing for dead times close to 0 although this is only useful for class D, not for SMPS.

[poobah]

I have used the HP OPTO drives in situations where the PWM duty cycle can run the full excursion from 0 to 100.00%. They work... but don't be misled by the advertised frequency... it is the dV/dT that places limitations on things... like EVA says. Great for low freq. IGBT stuff.

In these situations as above... the IR chips can get "lost". Command pulses too narrow will create false states on the output. Command pulses too long and the bootstrap dies or noise corruption with dedicated supplies can upset the output state as well. The bottom of the chip talks to the top through pulses... and exactly how this works is porrly defined in the datasheet.

The IR chips are very tolerant of ground bounce and at worst a small schottky from out-ground to in-ground solves this. They are good chips... the data sheets are the WORST EVER... REALLY, the WORST!

I'll leave EVA to wonder what PWM app's go from 0 - 100%

[rtarbell]

Thank you all for your replies!

I'll go ahead and try this - I should have mentioned that I'm not using these for true PWM. I wanted to use the "opto + powerFET" switch for rail switches in a rail switching supply (a la Workhorse from the schematics he posted here awhile ago), I would call it class G (BUT I know that this class-naming thing can cause a ruckus).

==> Essentially, the optos are only switching the high voltage power supplies in when needed, and then switching them back out when they are not needed according to the supplied music signal. I would *think* that these "switching" frequencies are much lower than what modern PWM uses.

P.S. I'm trying to use one floating high voltage power supply for both the positive and negative high voltage supplies needed. I got the idea from a post I read here, although I can't remember the posting off the top of my head..... but anyways, one power supply that does the job of two, now that's sneaky!

[poobah]

Those optos would be perfect for that...

You will still need a floating supply though.

[switchmodepower]

Use the HCPL-3180. It's a faster device.
I've used it for a 3KW inverter section with no problems.