Deadtime, shoot-through etc / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

Deadtime, shoot-through etc

Move Thread

LAN_403

Frosty90

Thu Sept 24 2009, 04:38AM

Registered Member #1617 Joined: Fri Aug 01 2008, 07:31AM
Location: Adelaide, South Australia
Posts: 139

Hi everyone,

I've completed the h-bridge for my new TC and tested it with a lightbulb load. With a 15 volt supply, and a ~100khz drive signal (which is the start-up/exciter signal from my driver board, which uses CT feedback and a 74HC14 'detector'), the bridge draws aproximately 12mA of current under no load. In TC service, it will be operating at around 250kHz. I haven't added any dead-time, only that which is due to the rise time of the signal through the GDT and 2.2 ohm gate resistors. I've meausred the rise time to be around 350ns. Im using hgtg40n60a4 IGBTs. So is this apparent no load current (note below) due to shoot through? or something else? Is it going to be a problem at full voltage (340v) and frequency (~250kHz)?

Note: I took the 12mA reading from my DMM. The analouge ammeter on my power supply doesn't move, and the multimeter shows current even with no dc bus power, just with the gate drive going, so Im thinking that at least some of the '12mA' is due to RF interference with my meter. Also, there is quite a bit of ringing (around 10 volts), across the bridge with a lightbulb load, although the leads to the load and the power supply are quite long, which they wont be in the TC.

Cheers,
Jesse

**EDIT** (since ...'s reply)
I just realised that I actually didnt have one leg of the bridge connected to the supply +'ve rail, with it connected properly, the no-load current is around 25mA, with a 15volt supply.

...

Thu Sept 24 2009, 04:43AM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

If you are only seeing 12ma of no load current, your bridge is in great shape--I would have expected more than that just from the losses involved with charging/discharging the various capacitances of the igbt

GeordieBoy

Thu Sept 24 2009, 02:40PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Firstly make sure you have a lot of DC bus decoupling on the H-bridge right at the switching devices. Then measure the DC current at the power supply end of the cable away from where all the switching is going on.

It is typical to get figures in the low milliamp range for no-load supply currents. This is real current drawn from the supply and does represent some wasted power, but it might not be as bad as you think when you increase the supply voltage up to the hundreds of volts.

The no-load current you are measuring is made up of at least three components:

1. Shoot-through currents due to overlapping of top and bottom device's conduction times.

2. Cds charging currents. When no load is connected, each MOSFET that turns on has to charge the opposing MOSFET's output capacitance. This is necessary for the drain-source voltage of the "off" MOSFET to reach the supply voltage. The energy put into this capacitance when the device is off, is dissipated in the channel when it turns on.

3. Off-state leakage currents. (Should be very small for modern MOSFETs, and insignificant when testing at reduced voltage!)

4. Avalanche losses if any of the devices are avalanching due to voltage spikes. (Highly unlikely with no load if there isn't any shoot-through.)

The main contributors are number 1 and number 2. The balance of the idle current between number 1 and number 2 changes as you increase the supply voltage. Shoot-through time usually remains constant so current due to shoot-through will increase steadily with applied current if all of the gate-drive timing stays the same. Capacitive turn-on current does not increase so rapidly because the device capacitances actually decrease with more applied voltage.

So quite often you do a test with reduced DC bus voltage and get alarmed at the current draw. Thinking if this test was at 400VDC the current would be higher and that would be lots of watts of heat. In actual fact when you wind the voltage up slowly you find that the quiescent current barely increases because the capacitances diminish as you wind up the voltage. Likewise any voltage overshoots under no-load testing at reduced voltage often look alarming. However, they usually don't scale proportionally to applied voltage so don't look as damaging when the DC bus is cranked up to hundreds of volts.

It is certainly worth testing bridges with 10 or 20 volts DC to start with, but high-voltage MOSFETs and IGBTs can behave quite atypically with low applied voltages. So if it appears to work okay, I would just increase the voltage slowly whilst monitoring device temperatures and supply current, until you get up to the full running bus voltage. Then leave it there for some time and check for heating again after so long.

Also, it is worth noteing that voltage waveforms will often look spikey when there is no load connected to the bridge because of the device output capacitance charging that I mentioned. Once a slightly inductive load is connected, the situation changes radically. Device capacitances are now charged resonantly from the load inductance and free-wheel diodes exhibit softer recovery so everything usually looks more rosey with an inductive load connected.

A good load to test an H-bridge with is a filament lamp with a small amount of "commutating" inductance in series. This will often show very clean switching waveforms during testing. Of course the real crunch is how it performs with the final intended load though.

I hope this helps,

-Richie,

Frosty90

Fri Sept 25 2009, 02:14AM

Registered Member #1617 Joined: Fri Aug 01 2008, 07:31AM
Location: Adelaide, South Australia
Posts: 139

Wow! Thanks Richie! It all makes perfect sense (obviously) now that you mention it! I'll try the tests you mentioned and post the results!

Cheers,
Jesse

Frosty90

Fri Sept 25 2009, 08:13AM

Registered Member #1617 Joined: Fri Aug 01 2008, 07:31AM
Location: Adelaide, South Australia
Posts: 139

Results:

Ok so with full voltage applied (340v plus 220uf of decoupling ), according to my meter, the unloaded bridge was pulling around 65ma with a switching frequency of around 100kHz. After 10 minutes, the heatsink had gone from 24 to 28 degrees C, and after 10 minutes, it was up to around 36 (No fan cooling but I am using the dreaded 'sil-pads'; and my driver circuit needed fan assist). So all seems well. So it seems to be a slight amount of shoot-through. I'm also guessing at full operating frequency (around 250kHz), the losses due to this would scale aproximately linearly, which I don't mind, as I'll probably endup putting a fan on there anyway.

Unfortunately, I can't scope the thing at full voltage, as I dont have an isolating transformer or isolating scope probes, I'll have to dig up my MOTs and make one.

Cheers,
Jesse

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