Using optocouplers instead of GDT for half/full bridges? / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Using optocouplers instead of GDT for half/full bridges?

1 2 next

Move Thread

LAN_403

Plasmana

Sun Aug 22 2010, 09:54PM

Registered Member #3108 Joined: Thu Aug 12 2010, 05:37PM
Location: Worthing, England
Posts: 72

I had this random idea of using optocouplers instead of gate drive transformers while I am trying to get one to work!

So what do you guys think? And advantages or disadvantages?

Schematic for illustrative purpose only!

ScotchTapeLord

Sun Aug 22 2010, 10:45PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

This has been tried.

I believe the main advantage to optically isolated gate drives was more reliability at a greater range of frequencies, as gate drive transformer winding numbers have to be optimized for a particular frequency. It also allows for control over duty cycle and dead time.

I believe it was Steve Ward who said that the higher rise times from directly driving an IGBT or a MOSFET could possibly result in harmful transients across the device's gate.

People use GDTs for ease of construction when dealing with multiple gates- one supply feeding one transformer that could provide isolated signals to each IGBT, as opposed to making a separate supply for each one.

Arcstarter

Mon Aug 23 2010, 01:14AM

Registered Member #1225 Joined: Sat Jan 12 2008, 01:24AM
Location: Beaumont, Texas, USA
Posts: 2253

You can use optos, but you have to have an isolated power supply for each gate.

This is how alot of people drive brick IGBTs, except they use a UCC or halfbridge of N and P channel complimentary fets for more power. Or both! That is how Steve's drsstc-2 is driven:

The isolated power supply is very simple, believe it or not. Any good SMPS core, some driver (ZVS is perfect, and as Steve Conner once said, the sine wave output may help eliminate crosstalk, or even an offline supply with uc3842), and 4 windings with rectification and filtering. This is Steve Ward's fullbridge isolated power supply:

Avi

Mon Aug 23 2010, 07:26AM

Registered Member #580 Joined: Mon Mar 12 2007, 03:17PM
Location: Melbourne, Australia
Posts: 410

the top mosfet would never switch on, there has to be 12-15v between the gate and the source pin to turn a mosfet on and some current to flow for some small moment while the gate charges.
the top mosfets source pin is floating and not grounded and that wont cause the correct potential difference and current flow.

Avalanche

Mon Aug 23 2010, 09:56AM

Registered Member #103 Joined: Thu Feb 09 2006, 08:16PM
Location: Derby, UK
Posts: 845

It's a good way of driving gates, but probably not best suited to Tesla coils...

You can buy optocoupler gate drive ICs, have a look here -

I've used the HCPL3140. I just use 1 chip on each gate, even the low sides, and just add a bootstrap capacitor (something like a 10uF low impedence electrolytic to provide the supply for the high side, so that the cap charges up when the lowside device is switched on).

The obvious benefit of the optoisolated gate drive is that you can apply modulation, ie a lower frequency component that usually wouldn't be able to get through the GDT - but this isn't required on a TC because the only frequency is your high frequency, 50/50 on/off. You have to make some arrangements to prevent 'shoot through' when each device is independently driven like this, an easy way is to use resistors and diodes so that all your mosfets switch on slow, and off fast. A GDT seems to prevent shoot through by itself, probably because it's true AC gate drive.

It's worth a try though, I've considered trying it with those gate drive ICs, but all my coils are stupidly high frequency and the opto chips probably wouldn't perform very well.

Plasmana

Mon Aug 23 2010, 09:54PM

Registered Member #3108 Joined: Thu Aug 12 2010, 05:37PM
Location: Worthing, England
Posts: 72

So I see optocouplers are good for low frequency while GDT's are good for high frequency. That is interesting, Ill buy a couple of those optocouplers and try them out on my future half bridge projects.

Thanks for the answers guys!

Arcstarter

Tue Aug 24 2010, 01:36AM

Registered Member #1225 Joined: Sat Jan 12 2008, 01:24AM
Location: Beaumont, Texas, USA
Posts: 2253

Plasmana wrote ...

So I see optocouplers are good for low frequency while GDT's are good for high frequency. That is interesting, Ill buy a couple of those optocouplers and try them out on my future half bridge projects.

Thanks for the answers guys!

Well, it has nothing to do with frequency, it is all about the current the opto can supply. The higher the freq, the faster the gate has to charge and discharge, hence more power.

Also i see that schematic showed no ground, and that is what Avi was talking about. That is why the isolated supply is needed, so you can ground the 0v of the supply to each mosfet without shorting stuff out.

teravolt

Tue Aug 24 2010, 05:09AM

Registered Member #195 Joined: Fri Feb 17 2006, 08:27PM
Location: Berkeley, ca.
Posts: 1111

the thing people don't like opto drive is the delay of the opto wich can be around 100 to 200 ns and that ens up with bridge that is delayed from the OCD wich may be compinsated with phase delay now . also you have to mess with 4 isolated power supplies. but I have a inerest in this as well insted of optos I use plastic fiber optics for galvonic isolation. I still neeed 4 suppies. the fiber optics feed a inverting buffer that drives 2 ixdd414pi in series with about .5ohm with the gates of my cm600's. the max F i got out of the bridge is about 100Khz with dead time added. they may have opto's with drivers bult in to them now in one IC

Wolfram

Tue Aug 24 2010, 07:39AM

Registered Member #33 Joined: Sat Feb 04 2006, 01:31PM
Location: Norway
Posts: 971

Arcstarter wrote ...

Well, it has nothing to do with frequency, it is all about the current the opto can supply. The higher the freq, the faster the gate has to charge and discharge, hence more power.

Yes it does. Except for special high-speed ones, optocouplers tend to have a lot of propagation delay, usually in the hundreds of nanoseconds to a few microseconds. This is very significant at higher frequencies.

To answer the original question, yes it is possible as many others have said, but it's not exactly easy. Firstly you will either need some pretty fast optocouplers with gate driver outputs or some pretty fast optocouplers and external gate driver circuits. The first option is not really a good one, as the gate driving optocouplers I've seen tend to be quite weak current-wise, so the best option seems to be the fast optocoupler + gate driver one.

This means you'll need four fast optocouplers and four gate driver ICs. Expect to pay around a dollar each for the optocouplers, and around twice that for the gate drive ICs if you're making a full-bridge. Then you need to provide isolated power for each of the gate drive sections. You can use isolated DC-DC converters, but that tends to get expensive, or you can make something yourself, but that might be a project in itself. You could do it the easy way and bootstrap the upper drivers, and get away with a single isolated supply for the whole bridge. You need to be careful if you do that, as the voltage across the switching devices tends to swing negative when the reverse diodes inside the devices conduct, you need to protect the gate drivers from this negative voltage.

You also need to be careful with the driver circuit, to use the right amount of dead-time and avoid cross-conduction at all costs. In mild cases, cross conduction leads to a lot of EMI and voltage spikes, and in more extreme cases, exploding IGBTs. Everything also needs to be well shielded, to avoid the disatrous situation where noise causes all the gate drivers to turn on at the same time. One more thing that you need to be careful with is how the circuit behaves when it loses power. I know Steve Ward had some trouble with one of his earlier gate-driver designs turning on all the IGBTs in his bridge when the driver circuit lost power.

Or you could just get a 20-cent toroidal ferrite core and wind it with some 6-10 turns (for typical SSTC frequencies, 200-300kHz, more turns for lower frequencies) of wires from a CAT5-cable and be done with it. You don't even need to buy a core, you probably already have some of the common gray noise suppression toroids . They are not ideal, but they tend to work well enough in my experience. The only common mistake is to use the wrong kind of core, i.e. powdered iron. Powdered iron cores are usually painted yellow and are most commonly found in the output filters of PC PSUs. Do not use these to make GDTs.

Anders M.

Conundrum

Tue Aug 24 2010, 06:12PM

Registered Member #96 Joined: Thu Feb 09 2006, 05:37PM
Location: CI, Earth
Posts: 4061

strange.. what causes this delay, the conversion of light into carriers within the phototransistor presumably?

A while back I "discovered" that LEDs can be used as emitters and sensors, so a worthwhile experiment might be to use a red LED + parallel 1K dampening resistor as the emitter and infrared LED as the detector, to see if the propagation delay is shorter...

-A

"Bother" said Pooh, as the ZPM sprouted Replicator claws...

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob