Switch selection for H-bridge design

[stube40]

I'm trying to design an H-bridge that can control the output polarity of an electromagenetic coil with inductance of around 50mH and very low resistance. The source voltage is around 150V at 40A and hence the design and, particularly, the switches aren't trivial.

The timing constraints make things even tougher - the circuit needs to be able to switch within a few ms of the control signal trigger going high. One switched, the circuit will stay in the same polarity for around 100ms or so. Hence, it's far more about response time and lag than it is about actual switch frequency.

I've considered both solid state relays (IGBTs and MOSFETs) as well as electro-mechanical relays. The solid state stuff seems to have considerable power losses due to the voltage drop between emitter and collector. The electro-mechanical relays suffer from being too slow and/or occasionaly bouncing when driven hard (which is likely to blow my PSU)

Does anyone have any suggestions for a good solid state solution that has low power losses, or an electromechanical solution that is fast enough and wont bounce?

 

[Mitchekj]

Hmm. I'd think a FET (or multiples in parallel) would be the best bet.

You can get 45mOhm RDSon with a good 600V 60A+ rating... you'd still be dissipating a good 72W at 40A though... so putting two or three in parallel may be an option? Gate charge and input capacitance may be an issue, not sure. I can't think of any mechanical solution if you need to switch at that speed, though.

For instance the Infineon IPW60R045CP looks like it may meet the specs, the transition times are all pretty low.

None of these are "cheap" mind you, and would need some serious heat sinking.

EDIT: Gah, see, I'm getting rusty with the electromechanical game... I don't think you can get the P-channel FETs with the rating and low RDSon you'd need though...

 

[Resqueline]

You can use N-channels only if you use a high-side / low-side driver IC. They require some minimum switching activity though since they use charge-pumps to generate the high-side gate drive voltage.

The transistors don't need to be fast. Since the inductor is large the switch frequency can be low. I guess that a 200-300V rating could be sufficient though, and these could come cheaper and better.

The current will rise (& fall) with only 3A/ms so it'll take at least 13ms to get it up to 40A. Thereafter the H-bridge must use a switch-mode current-limit scheme.
Relays are a little slow to be used in this application it seems.

 

[Mitchekj]

I wasn't sure that there are FETs with low RDSon and enough current capacity in the 300V range. That would definitely be the way to go, if so.

edit: ah ha, IXYS makes just such a thing. I learn something every day.

 

[stube40]

Thanks for all the replies, very helpful.

It's very interesting that you mention IXYS - I used an IXYS MOSFET and dirver combination in a pulsed-laser diode PCB I made. However, that doesn't make me an expert as I based my design on the data sheet reference circuit!!

Resqueline - your post got me thinking. Timing is very important to me as I want the switch to make the circuit to within +/- 1ms of a target time. However, I will get at 100ms advance notice of when this time is and hence I could pre-fire the switch to hit the precise time as long as the lag of the switch was consistent. If so, then this might bring electro-mechanical relays back into the picture. Maybe the lag varies though? Also, might I get problems with the contact bouncing?

 

[(*steve*)]

The mosfet will be much faster than a relay.

What Resqueline was saying about it taking 13 ms to get to 40A would be just as true with a relay as with a mosfet since this is a property of the device and its rather huge inductance.

You could speed it up by applying a higher voltage to the device until the current reaches a set point and then reduce the voltage to hold the current at this value.

 

[stube40]

The mosfet will be much faster than a relay.

What Resqueline was saying about it taking 13 ms to get to 40A would be just as true with a relay as with a mosfet since this is a property of the device and its rather huge inductance.

You could speed it up by applying a higher voltage to the device until the current reaches a set point and then reduce the voltage to hold the current at this value.

Hey Steve,

Yes, I agree - it will take 13 ms to charge. This is not a problem, however there needs to be tight control over when the 13ms charge period actually begins. This was the main problem with using a relay which might take 50ms or more just to close the contact.

I was talking about predicting the beginning of the charge period and hence switching the relay 50ms in advance to negate the lag, but I think there are other issues with relays too (such as bouncing).

Clearly a mosfet could achieve the response required, provided I could live with the power dissapation - 72W for each switch normal (2x simultaneous required in normal h-bridge operation) is a real downer in the context of this experiment.

I'm guessing though that this is the final word - either live with the approx 70W power disappation of a solid state solution or put up with the timing and bouncing issues of an electromechanical solution?

 

[(*steve*)]

I think you might be surprised by the resistance of the relay contacts. They're not a perfect zero ohm connection either.

You could employ a solution with both a mosfet and a relay. Simply turn the relay and the mosfet on and off together, and place them in parallel.

The mosfet will turn on faster than the relay and it will not bounce. The relay may have a lower resistance than the mosfet (and will thus shunt most of the current away from it, reducing dissipation). To counter bounce on the opening of the relay, you could release the relay a little earlier than you turn off the mosfet.

All of this presumes that the rate of turning on and off is low enough for the relay to handle.

A modification like this will mean that the mosfet will be dissipating the max power only when it is switching and it will run much cooler. The relay, on the other hand, may well run hot.

 

[Mitchekj]

That's a good point, most relays will have a contact resistance higher than a lot of FETs... they'd be dissipating just as much, if not more, power.

40A is just a lot of current, not much you can do to curtail the loss in a switch. Of course, 70W is small relative to the 6kW supply (wow.)

 

[(*steve*)]

Yes, In fact I think this application is getting to the point where the "constant resistance" of a mosfet almost gives way to the "constant voltage drop" of an IGBT.

Have a look at the current ratings of large IGBTs! (you want 1200 amps?)

and this: http://www.irf.com/technical-info/whitepaper/choosewisely.pdf

 

[stube40]

I think you might be surprised by the resistance of the relay contacts. They're not a perfect zero ohm connection either.

You could employ a solution with both a mosfet and a relay. Simply turn the relay and the mosfet on and off together, and place them in parallel.

The mosfet will turn on faster than the relay and it will not bounce. The relay may have a lower resistance than the mosfet (and will thus shunt most of the current away from it, reducing dissipation). To counter bounce on the opening of the relay, you could release the relay a little earlier than you turn off the mosfet.

All of this presumes that the rate of turning on and off is low enough for the relay to handle.

A modification like this will mean that the mosfet will be dissipating the max power only when it is switching and it will run much cooler. The relay, on the other hand, may well run hot.

Very interesting idea indeed!! Plus the IGBT thing is something I hadn't considered at all. I assume that there are pros and cons to IGBTs vs MOSfets?? (as you can see I dont know much about them).

I wanted to have a quick look at the link you posted but it doesn't work - looks like there is a few letters missing.

CORRECTION - it works fine, I'm the idiot by trying to cut and past the text into a new browser (which didnt work obviously).