Just thought of something else... what about suitable protection diodes for the MOSFETs? I have no idea about how to select some for this...

Cheers
Shaun

 

That's a thought. I've been going by ST Microelectronics' datasheet for your MOSFETS; their part number is STP55NF06, their datasheet indicates it has built-in Zener diodes that would protect against both overvoltage and reverse EMF.

I've been going on the assumption that these are industry standard components.

Who is the manufacturer of the components you're actually using?

[eta]
The reason for specifically using lamps is that they have a non-linear resistance; if the current flowing through them is low, they will act almost like a dead short. The higher the current through them gets, the more resistance they have. MOSFETS exhibit a similar characteristic on a much smaller scale; the hotter they get, the greater their resistance. This is completely opposite to most other semiconductors which generally decrease in resistance with increased temperatures. This can easily lead to the dreaded "thermal runaway" condition and meltdown of the components involved.

Meanwhile, I understand that space constraints are tight. However, using small aluminium (preferred UK spelling) angles will do next to nothing for dissipating MOSFET heat. The tab on your MOSFETS are most likely made of copper that is tin plated. Copper is one of the better metals for conduction of heat. Aluminium is roughly half as good at conducting heat as copper, but it's cheap. If you really are in a crunch for space, you might have to consider water cooling. Of course, this means everything will have to be water-tight.

I think you posted the link of your train project before - but post it again, would you? I've lost it.

 

That's a thought. I've been going by ST Microelectronics' datasheet for your MOSFETS; their part number is STP55NF06, their datasheet indicates it has built-in Zener diodes that would protect against both overvoltage and reverse EMF.

I've been going on the assumption that these are industry standard components.

Who is the manufacturer of the components you're actually using?

They are indeed ST Microelectronics... on the MOSFETs themsleves it says:

P55 NF06 @
GK 2W4 V6
CHN 752

[eta]
The reason for specifically using lamps is that they have a non-linear resistance; if the current flowing through them is low, they will act almost like a dead short. The higher the current through them gets, the more resistance they have. MOSFETS exhibit a similar characteristic on a much smaller scale; the hotter they get, the greater their resistance. This is completely opposite to most other semiconductors which generally decrease in resistance with increased temperatures. This can easily lead to the dreaded "thermal runaway" condition and meltdown of the components involved.

Right, so lamps it has to be then... I hope this is just for testing purposes? I can't quite see me having 16 headlamp bulbs inside the train!

Meanwhile, I understand that space constraints are tight. However, using small aluminium (preferred UK spelling) angles will do next to nothing for dissipating MOSFET heat. The tab on your MOSFETS are most likely made of copper that is tin plated. Copper is one of the better metals for conduction of heat. Aluminium is roughly half as good at conducting heat as copper, but it's cheap. If you really are in a crunch for space, you might have to consider water cooling. Of course, this means everything will have to be water-tight.

Ah, I could use copper... or I could use water cooling for a pc... in fact I can put a radiator in the front of the bonnet... I think about this.

I think you posted the link of your train project before - but post it again, would you? I've lost it.

Sure its http://manners.homelinux.com it hasn't been updated in quite some time! I'll try and get round to doing that this week.

Cheers
Shaun

 

They are indeed ST Microelectronics... on the MOSFETs themselves it says:

P55 NF06 @
GK 2W4 V6
CHN 752

OK, good deal - we are indeed on the same page then!

Right, so lamps it has to be then... I hope this is just for testing purposes? I can't quite see me having 16 headlamp bulbs inside the train!

Yes indeed, it's just for testing. The whole idea is to protect your MOSFETS and alternator-turned-motor should current draw get out of hand.

Ah, I could use copper... or I could use water cooling for a pc... in fact I can put a radiator in the front of the bonnet... I think about this.

Copper is roughly 1.7 times better at conducting heat than pure aluminium. Water is roughly 1.5 times better at conducting heat than copper is!
Nickel is far worse of a heat conductor than aluminium, worse than that is iron, and even worse yet is tin. For that reason, I suggest that instead of trying to solder copper tubing to a copper heatsink, that you fabricate your individual heatsinks from fairly thick flat or bar copper stock, and drill the water passage edgewise right through the sink. You might bore out either side a bit, and sweat-solder something like barb fittings in for the interconnecting plumbing. You wouldn't want to try to tap threads into it, as copper is so soft it will tear easily. Besides, you wouldn't want water leaks over your circuit board.

Something else - have you been using heat sink compound? I don't know where you get it in the UK, perhaps Maplin or Ferrel (sp), but it's available at local Radio Shack stores here in the States. Comes in a small tube, looks like a ghostly-white paste. It makes the thermal bond between the semiconductor and heat sink much better.

Sure its http://manners.homelinux.com it hasn't been updated in quite some time! I'll try and get round to doing that this week.

Good deal!

I did some poking around on there - happened to notice the PC heatsink that you machined down on the bottom so that it was flat. While the flatness may add to the asthetic appeal, I'm afraid that you diminished the efficiency of the heat sink by removing a fair portion of the "heat distribution channel", if you will - the bottom of the heat sink needs to be fairly thick to efficiently conduct the heat towards the radiating fins.

Oh, one other thing that hasn't been talked about - cooling for the alternator-turned motor. You'll need quite a bit of air moving through it to keep it from turning into a molten blob. Had you considered some type of a blower? I'm afraid that a standard PC fan just won't do. Perhaps a squirrelcage-type auto heater blower from the auto wrecking yard would do the trick. While you're in the heater section, might as well get a heater core too - you could use that for the radiator for the water-cooled electronic components

 

Yes indeed, it's just for testing. The whole idea is to protect your MOSFETS and alternator-turned-motor should current draw get out of hand.

Excellent I went to a car breakers yard yesterday and spent some time removing headlamp bulbs from a huge collection of old headlamps!
One thing I thought of, they have two elements in them, and on the bulbs it says 55/60W so, can I not use both elements thus reducing the number of required bulbs to 8?

Copper is roughly 1.7 times better at conducting heat than pure aluminium. Water is roughly 1.5 times better at conducting heat than copper is!
Nickel is far worse of a heat conductor than aluminium, worse than that is iron, and even worse yet is tin. For that reason, I suggest that instead of trying to solder copper tubing to a copper heatsink, that you fabricate your individual heatsinks from fairly thick flat or bar copper stock, and drill the water passage edgewise right through the sink. You might bore out either side a bit, and sweat-solder something like barb fittings in for the interconnecting plumbing. You wouldn't want to try to tap threads into it, as copper is so soft it will tear easily. Besides, you wouldn't want water leaks over your circuit board.

I kinda knew copper was better as thats why the best PC heatsinks are copper... but I used the ally as that was what I had to hand...

Just to make sure we are on the same page with this too... I am wanting to start the petrol engine using the alternator, which means it should only be switched on for a short time, hopefully in the order of seconds. However I don;t know how much current it will draw when trying to turnt he engine over.

Thats an interesting suggestion about the water cooling... I'll have to look into this more. I had already found a large radiator used for PC watercooling that I thought would fit perfectly in the grill at the front of the bonnet.. so this would be an idea use for it. I might look into buying the copper heatsink parts designed of computers as it would save me time, and also I'm not sure I;d trust myself making something water-tight!

Something else - have you been using heat sink compound? I don't know where you get it in the UK, perhaps Maplin or Ferrel (sp), but it's available at local Radio Shack stores here in the States. Comes in a small tube, looks like a ghostly-white paste. It makes the thermal bond between the semiconductor and heat sink much better.

Perhaps you mean Farnell? yes they do it, as do number of other places.. in fact I have some from when I built my pc, its called ArcticSilver.

Good deal!

I did some poking around on there - happened to notice the PC heatsink that you machined down on the bottom so that it was flat. While the flatness may add to the asthetic appeal, I'm afraid that you diminished the efficiency of the heat sink by removing a fair portion of the "heat distribution channel", if you will - the bottom of the heat sink needs to be fairly thick to efficiently conduct the heat towards the radiating fins.

oops
Maybe water cooling here too...

Oh, one other thing that hasn't been talked about - cooling for the alternator-turned motor. You'll need quite a bit of air moving through it to keep it from turning into a molten blob. Had you considered some type of a blower? I'm afraid that a standard PC fan just won't do. Perhaps a squirrelcage-type auto heater blower from the auto wrecking yard would do the trick. While you're in the heater section, might as well get a heater core too - you could use that for the radiator for the water-cooled electronic components

Yes... well the fins on the top of the engine will keep air moving, and the alternator has a fan ont he shaft which will be underneith the train blowing cold air up inside. Not sure if this would be enough.


Cheers
Shaun

 

Hi SgtWookie,

I have setup the bulbs on a stand and will wire them up soon (I need to buy some more crimps that fit the spades on the back of the bulbs) ... and I also need to purchase some sunglasses

Once I have done that, whats next?

Edit: I have now bought the connectors, so I can wire it up now.

Cheers
Shaun

 

Hi SgtWookie,

All wired up! here is a schematic of exactly what I have now... where do I go from here?

Thanks for your advice as always, much appreciated.

Cheers
Shaun

 

OK, before you power it up again...have you done something to improve the cooling on your power MOSFETS?

While the headlamps are capable of absorbing a decent amount of the power, the rest will be split between the alternator and your MOSFETs. Without adequate heat sinking, I'm afraid that even WITH the headlamps in the circuit, the MOSFETs may still have to dissipate sufficient power that they may overheat, and self-destruct.

The MOSFETs SHOULD be able to handle the power, IF they have proper heat sinks, including using heat sink compound, and that you have adequate water flow through the new sinks to carry the heat away.

I really don't want for you to have to go to the bother and expense of replacing another set of those MOSFETs, and I know you don't either.

I suggest the most rapid path to success is carefully going over all of the options prior to flipping the switch on; as that switch may yet again trigger the release of the magic smoke.

 

OK, before you power it up again...have you done something to improve the cooling on your power MOSFETS?

No I haven't...

I really don't want for you to have to go to the bother and expense of replacing another set of those MOSFETs, and I know you don't either.

I suggest the most rapid path to success is carefully going over all of the options prior to flipping the switch on; as that switch may yet again trigger the release of the magic smoke.

Whilst specatacular and quite fun... no I don't really want to do that again!

So, water cooling then.. hmm there is a slight problem... the high side FETs will be ok, but when it comes to the low side, I cannot connect the heatsinks together as it will short out the three output phases as the case is connected to the drain... I could buy some insulating thingies.. but I assume the heat transfer wouldn't be anywhere near as good as having it directly in contact with the heatsink (and thermal paste)...


Going back to the frequency/inductance of the windings... what minimum frequency would you suggest? I am looking at about 400Hz would this be enough do you think?

Cheers
Shaun

 

No I haven't...
(snip)

So, water cooling then.. hmm there is a slight problem... the high side FETs will be ok, but when it comes to the low side, I cannot connect the heatsinks together as it will short out the three output phases as the case is connected to the drain... I could buy some insulating thingies.. but I assume the heat transfer wouldn't be anywhere near as good as having it directly in contact with the heatsink (and thermal paste)...

No, you need direct contact with the copper heatsink.

I suggest that you need FOUR heatsinks; one long one for the three high-sided FETs, and three short ones for the low-side MOSFETs. You could do the plumbing using surgical tubing. Barbed fittings should make a water-tight seal that doesn't require clamps.

If you can keep the water flow rate high enough, you may be able to get away with plumbing up to three in series.
To avoid sharp bends in the tubing, consider staggering the positions of the low side MOSFETs. That would give you room to do a loop-the-loop with the surgical tubing.

Going back to the frequency/inductance of the windings... what minimum frequency would you suggest? I am looking at about 400Hz would this be enough do you think?

400Hz wouldn't be a bad choice. You'd be able to hear it, and it may prove tiresome after a while. If you were up around 25KHz or so, it would be out of the audible range. However, switching that fast may keep the MOSFETS in the linear (transition) region for too long. This is dependent upon your driver circuit, and the MOSFETs themselves.

 

No, you need direct contact with the copper heatsink.

Thats what I thought you'd say

I suggest that you need FOUR heatsinks; one long one for the three high-sided FETs, and three short ones for the low-side MOSFETs. You could do the plumbing using surgical tubing. Barbed fittings should make a water-tight seal that doesn't require clamps.

I have some 3/8" plastic tubing... so I'll be able to use that...
I have ordered a radiator, pump and one heatsink from a computer water cooling place. The heatsink is for mosfets on a computers motherboard, I ordered one just out of interest to see what it was like. It should all be here early next week.
I still have yet to figure out how to do the 3 smaller heatsinks... as I said before I'm not sure I'd trust myself to be able to make it water-tight.. I've had enough magic smoke already!

If you can keep the water flow rate high enough, you may be able to get away with plumbing up to three in series.
To avoid sharp bends in the tubing, consider staggering the positions of the low side MOSFETs. That would give you room to do a loop-the-loop with the surgical tubing.

The pump can move 400l/h so hopefully that'd be ok... with the radiator at the front of the bonnet with perhaps a fan...
Thanks for the ideas for positioning them... I'll let you know how I get on.

400Hz wouldn't be a bad choice. You'd be able to hear it, and it may prove tiresome after a while. If you were up around 25KHz or so, it would be out of the audible range. However, switching that fast may keep the MOSFETS in the linear (transition) region for too long. This is dependent upon your driver circuit, and the MOSFETs themselves.

So it would just create a hum at 400Hz... I'm not sure I'd mind that as I'm not going to be keeping it on.. it'll just be on for a matter of seconds to get the petrol engine turning over.

Cheers
Shaun

 

Plain plastic tubing may not be a good choice. I suggested surgical tubing because it remains very flexible over a broad range of temperatures, and resists "kinking" if you avoid making bends too tight. Plastic tubing tends to kink, which would restrict the flow of water.

I suggest that you experiment with a piece of the tubing and a pan of water warmed up on the stove, just to see what it might do.

As far as the three smaller heatsinks - it really shouldn't be too difficult. Simply drill one hole edge-to-edge for the water passage, enlarge the passage a bit at both ends for the barbed fittings, and offset the hole for the MOSFET mounting screw. Solder in barbed fittings on either side, make sure the solder didn't coat the water passage bore, and you're done. See the attached.

 

Hi SgtWookie and all,

First of all sorry for dropping off the face of the planet for a while!

I have managed to try out the water cooling and everything... but I had the funny feeling that those MOSFETs just weren't up to the job...

I tried watercooling an IRAM20UP60A and got the headlight bulbs to light nicely and got a nice slightly flat A hum out of them! Hooked up the alternator and it turned for about 1/8 of a turn and then it died. before it died though the current went over 40A...

Anyway, this made me think perhaps I should go back to making my own bridge out of individual devices. The first thing I thought about was the fact that it was at least 40A @ 24V... not far off 1kW... and the MOSFETs I was using were only rated at 30W... I have no idea how to work out how much power would be going through each device..

Anyway, I tried the driver that S_lannan suggested... in fact it turned out to be the same driver thats in the IRAM20UP60... and I got some IGBTs for the bridge...

I connected the headlamps to the alternator and it worked! at least a little bit... the alternator turned and as the bulbs got brighter it stopped... it didn't have any kind of power there. At this point it was drawing about 9A.

I then tried connecting the alternator directly.. and one of the IGBTs failed.. so I was wondering about doubling them up? would the driver be able to drive 12 IGBTs? also, in the circuit diagram in the driver datasheet it shows a resistor between the output and the gate... what is the purpose of this? I put in a 100ohm as a pure guess, but have no idea if that is a good thing or not. I have googled around but cant seem to find anything, or at least anything that I understand about this resistor and how to choose a value for it.. or if it is in fact needed at all...

Is it worth me drawing up a schematic for you?

Thanks for all your help so far and as always its much appreciated.

Cheers
Shaun

 

an update...

I tried doubling up the IGBTs... and again it worked fine with the headlight bulbs... only this time when I connected the alternator in series with the bulbs the driver IC died again... it was fine when the alternator was in series before...

none of the IGBTs have died.. I have tested them and they still work perfectly.. its just the driver that keeps going...

is it because the driver can't supply enough current to charge the gate when they are passing a lot of current? this would seem to be what is happening, the big problem with my theory is that IGBTs aren't supposed to need much current at all... that being one of their advantages...

any suggestions?
do I need to build my own drive circuit instead?

Cheers
Shaun

 

Shaun,
Haven't forgotten about you, I just haven't had the time to read through all the datasheets and compare them to your observations. I have never used an IGBT, so this will be new territory for me.

 

Shaun,
Haven't forgotten about you, I just haven't had the time to read through all the datasheets and compare them to your observations. I have never used an IGBT, so this will be new territory for me.

Thanks for the message It's very good of you to give up your time for everybody on here... so if/when you find the time it is very much appreciated

As far as I can tell IGBTs are just an FET and a BJT in a darlington pair... so the gate drive requirements shouldn;t be that much different from my MOSFETS..

I have just found this document that looks like it might be helpful to me.. so I'm going to have a read of that...

Cheers
Shaun

 

First of all sorry for dropping off the face of the planet for a while!

A Yank expression for that is "Stuff happens"

I have managed to try out the water cooling and everything... but I had the funny feeling that those MOSFETs just weren't up to the job...

I tried watercooling an IRAM20UP60A and got the headlight bulbs to light nicely and got a nice slightly flat A hum out of them! Hooked up the alternator and it turned for about 1/8 of a turn and then it died. before it died though the current went over 40A...

Interesting. It's been so long since I've visited the thread, and so many other projects in between, I'll have to read back and take a look at why that might've happened.

Anyway, this made me think perhaps I should go back to making my own bridge out of individual devices. The first thing I thought about was the fact that it was at least 40A @ 24V... not far off 1kW... and the MOSFETs I was using were only rated at 30W... I have no idea how to work out how much power would be going through each device..

Power dissipated in the device = Current through the device x Voltage drop across the device.

I just had a thought (yes, 'twas painful ) that perhaps you are using the same power source to control the gates as you are using for the power to the load. As the current increases through the wires from the power source (or light bulbs), the voltage drop across them also increases.

If the voltage on the gates caused the MOSFETS to enter the linear region under such a heavy load, power dissipation (meaning heat) in the MOSFETs would increase a great deal.

MOSFETS have a positive temperature coefficient, meaning the hotter they get, the more resistance they have. This usually means they are much easier to parallel then "normal" BJTs are.

However, if the temperature rise was quite rapid, it may burn out immediately. Once one MOSFET burned out, the remaining MOSFETS would "pop" very quickly. This would be difficult to diagnose unless you had a storage 'scope that was capable of measuring/storing gate voltages and their transitions from the start of the test to the failure of the component.

Anyway, I tried the driver that S_lannan suggested... in fact it turned out to be the same driver thats in the IRAM20UP60... and I got some IGBTs for the bridge...

I connected the headlamps to the alternator and it worked! at least a little bit... the alternator turned and as the bulbs got brighter it stopped... it didn't have any kind of power there. At this point it was drawing about 9A.

Rather discouraging when things start blowing up. Well, we're in rather unexplored territory at the moment.

I then tried connecting the alternator directly.. and one of the IGBTs failed.. so I was wondering about doubling them up? would the driver be able to drive 12 IGBTs? also, in the circuit diagram in the driver datasheet it shows a resistor between the output and the gate... what is the purpose of this? I put in a 100ohm as a pure guess, but have no idea if that is a good thing or not. I have googled around but cant seem to find anything, or at least anything that I understand about this resistor and how to choose a value for it.. or if it is in fact needed at all...

Is it worth me drawing up a schematic for you?

Thanks for all your help so far and as always its much appreciated.

Schematics are always good to put up. You don't have to draw in all of the IGBT's; merely indicate which sections that are repeated "n" times.

 

A Yank expression for that is "Stuff happens"
Interesting. It's been so long since I've visited the thread, and so many other projects in between, I'll have to read back and take a look at why that might've happened.

I think that IGBT failed because of a bent pin on the driver... so that would be one thing solved, and would explain why none of the others went, nor have gone since.

Power dissipated in the device = Current through the device x Voltage drop across the device.

Yes that makes sense... what I meant though was - I know the whole bulb array/alternator setup was drawing x current... but how much of that total current was going through each device... I think it is 2/3? so if it was drawing 9amps in total, each device would have 6amps going through it... am I along the right lines?

I just had a thought (yes, 'twas painful ) that perhaps you are using the same power source to control the gates as you are using for the power to the load. As the current increases through the wires from the power source (or light bulbs), the voltage drop across them also increases.

If the voltage on the gates caused the MOSFETS to enter the linear region under such a heavy load, power dissipation (meaning heat) in the MOSFETs would increase a great deal.

MOSFETS have a positive temperature coefficient, meaning the hotter they get, the more resistance they have. This usually means they are much easier to parallel then "normal" BJTs are.

However, if the temperature rise was quite rapid, it may burn out immediately. Once one MOSFET burned out, the remaining MOSFETS would "pop" very quickly. This would be difficult to diagnose unless you had a storage 'scope that was capable of measuring/storing gate voltages and their transitions from the start of the test to the failure of the component.

I'm not surprised that that thought was painful

I am indeed using the same source... it is two 110Ah 12v boat batteries in series... however, the driver circuit is running off a 15v regulator (3amp variable version)

So I'm not sure what you describe is in fact the problem... I'm probably wrong !

Rather discouraging when things start blowing up. Well, we're in rather unexplored territory at the moment.

Yes quite... and well I don't mind stuff blowing up.. I look at it as R&D however, I am pleased that its not the actual bridge devices that are blowing up now...

I feel I am getting there... I now have the PIC controller sussed and the output from that is fine and I can easily change it, so thats one part of the problem solved... it would appear that I have a bridge setup that doesn't explode when I connect things up.. always good it just seems this bridge driver is the weak link now...

I have been experimenting this evening, and doing a fair bit of reading, and the more I read, the more I think that the problem is the amount of current the driver can source. I cant find the rating anywhere on the datasheet, but on the product details on the farnell website it says it can source 120mA... I think its possible for one IGBT to have a 150mA peak... 15v and a 100ohm resistor...

The outputs from the duff driver aren't completely dead, just not reaching full voltage and its not a very clean signal. so maybe its just slightly killing it? hmm

I could increase the resistor value I guess... but thats just going to slow down the switching time... then again maybe that doesn't matter at my frequency range.

Also, the lower the resistor value the better in terms of the miller effect (which I don't fully understand)...

Schematics are always good to put up. You don't have to draw in all of the IGBT's; merely indicate which sections that are repeated "n" times.

ok.. I have found some other half bridge drivers which I want to give a go (based on my theory) if that doesn't work then I'll draw up the schematic and post it on here

Cheers
Shaun

 

OK, current in series is like voltage in parallel.

If, for instance, you have a 10v supply with a lightbulb connected to the + side of that supply, then an inductor, and then the drain of an N-ch MOSFET, and the source of that N-ch MOSFET connected to the - side of the supply - when you turn on the MOSFET, let's say there is 10A going through the MOSFET. Well, that same 10A is going through the inductor and the light bulb. Since I=E/R, and your E=10, and your I=10, the R must also be 10.

Let's talk about your PIC program for a bit. You say you're switching the MOSFETs on and off at 400Hz (or thereabouts.) Are you actually swapping between phases at that rate? Or are you just turning them all on and off?

If you're trying to spin the phases at a 400Hz rate without the rotor moving first, it's not likely that the rotor will ever get going - there is simply too much rotor inertia.

You'll still need to turn the bridge on and off to limit current flow to acceptable levels, but you also need to start the phases in the field windings switching polarity between each other at a slow rate when the power is first applied, then ramp up the rate of change.

I hope this is making sense to you.

 

OK, current in series is like voltage in parallel.

If, for instance, you have a 10v supply with a lightbulb connected to the + side of that supply, then an inductor, and then the drain of an N-ch MOSFET, and the source of that N-ch MOSFET connected to the - side of the supply - when you turn on the MOSFET, let's say there is 10A going through the MOSFET. Well, that same 10A is going through the inductor and the light bulb. Since I=E/R, and your E=10, and your I=10, the R must also be 10.

Makes sense, but wouldn't R be 1ohm?

Let's talk about your PIC program for a bit. You say you're switching the MOSFETs on and off at 400Hz (or thereabouts.) Are you actually swapping between phases at that rate? Or are you just turning them all on and off?

its the frequency of each phase... and there are 6 steps per cycle:

010
110
100
101
001
011

1 = high side on
0 = low side on

If you're trying to spin the phases at a 400Hz rate without the rotor moving first, it's not likely that the rotor will ever get going - there is simply too much rotor inertia.

You'll still need to turn the bridge on and off to limit current flow to acceptable levels, but you also need to start the phases in the field windings switching polarity between each other at a slow rate when the power is first applied, then ramp up the rate of change.

I hope this is making sense to you.

yes it does the thing is, the alternator started to turn... but within a few seconds, the headlamps got brighter and it stopped turning... my guess was that as the headlaps got hot, their resistance increased, am I wrong in that assumption?

I have done a new schematic... I know there are value missing, I will update that tomorrow when I have the parts to hand.

However, this worked as I described above.. when I removed the headlamps and connected the alternator directly.. the driver (IR21365) died... the outputs weren't perfect square waves.. I know it wasn't a fault which shut it down, as that turns on an LED (not shown in the schematic).

The thing is... the IGBTs are fine.. and the MCU and opto isolators are fine.. it is just the driver IC thats going. One thing I have noticed is that it can supply 120mA... and I have a theoretical max of 150mA(15v/100ohm). I was thinking about increasing the resistance to say 147ohm (adding a 47 in series)...

the other thing I have discovered in reading lots of articles etc, is that it is reccomended that you clamp the gate to stop its voltage going too high due to the miller effect (at least thats how I've interpreted it)... so I am planning on adding a 16v zener across the gate/emitter pins (this is one way described in an article I found.. although that was back to back zeners, I believe this is because of the posibility of using a negative gate voltage to hold it low)

the bottom line is.. I have had it working.. but when more current goes through the IGBTs, the driver dies for some reason... so other than the ideas I have had above.. I'm not sure what to do next

Cheers
Shaun