Controller No5 With Protection - Looking for Explosions

[Peterae]

Here's a circuit of the next controller, i have implemented Poynt's TVS protection and also the pulses width limiting circuit.
The TVS are 500Watt rated.

[poynt99]

Let's hope the circuit functionality isn't affected too much Peter and your effect is still there

Is the TVS on the gate a bi-directional type?

.99

[Peterae]

Hi Darren
Yes the gate is Bidirectional and the drain are unidirectional, sorry couldn't find a symbol for them in my cad package so really i guess i should of created one  :

I think the TVS should be good, not yet sure how the pulse width limiter will work but i can always cut them out if they are causing me problems.

Bit's are on order right now, and i will etch the board over the next couple of days.

[BEP]

Are these TVS's selenium or silicon? One was faster than the other.
My phone sucks or I would look it up.

[Peterae]

Fast Response time: Typicaly less than 1ps from 0 volts to VBR for unidirectional and 5nS for bi directional

[poynt99]

Peter,

Just in case, I want to mention that the leads for the TVS diodes should be clipped as short as possible, and soldered directly to the MOSFET leads if possible. You want to eliminate as much stray inductance as is practical.

.99

[Peterae]

OK Darren
Cheers just etched the board, not so easy with a big heat sink, i guess i could tag them on underneath directly across the terminals, i have used really thick tracks though and am using the same bigger FET's i used for Ds board instead of irf840 this time around.

I will go with tagging on bottom then thx.

[BEP]

Just my experience folks but the TVS should have minimum conductor length between the TVS and the source of the spike. You want more inductance after the TVS than before it.

Always place the protection nearest the source of the damaging energy.

[Peterae]

Not quiet ready for testing but almost there.

Controller 5

[darkspeed]

Looks good!

[Peterae]

Thanks Ds
Initial tests show it works well, i have 2 scope shoots taken across each FET gate/source

During this initial test i have nothing connected to the drain's

gate5_1 is taken @ 1Mhz

gate5_2 is taken @ 100Hz

The only worry is over 1Mhz the driver chip starts getting hot and the gate drive is about 8 Volts instead of 18 Volts.

[Ken]

Peter,
You could use liquid cooling. Its easier than most people think.
Of course this would change the length of the leads.

[Peterae]

Thanks for the advice Ken 'a bucket full of oil'  

These new fet's do not seem that good, i have just powered them up and i cannot even see a switch off at my maximum setting of 255nS and on the scope it looks like when i set for a 32nS pulse the pulse is over 400nS DOH so it looks like it's back to irf840A's

IXFH12N100Q
http://www.farnell.com/datasheets/42399.pdf

Here's 2 pictures 1 with 0 phase delay and the other with 255nS delay.

[Peterae]

OK swapped to IRF840A's now and that's much better

3 Picture's
1 Across Gate/Source no load to Drain
2 Across load coil with 0 delay phase
3 Across load coil with 255 delay phase - Notice the pulse 

[poynt99]

That's good news Peter...you're back!

Hopefully the rig is a little more robust as well. Have you pushed it yet?

.99

[Peterae]

Hi Darren
Just pushed it up to 70volts and i am not getting my usual crashing spots, not sure what to make of it yet.

First thing i will say is that i need a delay near 250nS to get the pulse to appear so it would seem that the gate drive has slowed down something.

[Peterae]

OK i have run with 90Volts for the coil supply and am not getting any noise that i have been getting, the coil sings at the fundamental frequency my Sig Gen is set to with no White Noise as i get with re-triggering, which is because of the lack of Noise being injected.

2 things maybe happening ,either the Gate drive has slowed the fet switching down too much or the EM bursts are a result of very fast pulses way above the TVS rating and can only occur without the TVS.

I guess first i need to remove the Gate drive cap and resistor and see if i can get the Noise bursts back.

The other problem i have is the IRF fets have metal tab's and these dam heat-sinks require a screw so i have no way of insulation without getting some screw isolators, i already have the flat heat washer.

[poynt99]

Sometimes with ruggedness built in, performance can suffer.

You may want to try increasing the input cap value, or bypassing it to see how detrimental (if at all) it is to the gate drive. Keep the PW low just in case. You may also want to remove the 20 Ohm resistor temporarily as well if possible. Perhaps with some luck you can eliminate both, and just leave the TVS diodes in place.

.99

[Peterae]

Yep no problem Darren

I am going to swap over the heat-sinks from a previous board that have clips so i do not have the Drain connected to the heat-sinks.

First i will try without gate cap and resistor and see what happens.

[Peterae]

OK swapped my heat-sinks over and so now my FET's are isolated from the heat-sink metal.

Next up was to test the board without the gate drive resistor and capacitor

See scope shot across Gate/Source with no load below.

Next i started winding the volts to the coil supply up, 70volts no problem, 80 no problem, 90 was running well until the pulse showed up and then my psu shut down due to over current, the Non delayed fet had gone short, the TVS diodes read ok.

There was no sign of the pic crashes and re-triggering i had before to cause the FET to turn on more than it should, so what do you make of that, these TVS diodes should react in a pico seconds or so 

[Peterae]

Interestingly i cant see much difference between the gate drive with or without the protection circuit in place, a bit higher in volts without, but interesting that it didn't blow with the protection.

It is now looking like the TVS that is stopping the crashing of my pic by stopping something, off course to prove this i need to take the TVS off and see if the pic crashing comes back.

[Peterae]

Now i have taken the TVS off the Drain and run some tests, i am still unable to get the EM noise that crashes my pic, i went up to 80 volts, i dare not go any higher because there is only one likely hood when the noise starts.

The only thing i have not tried is taking the TVS diode off the gate.

I dont know why some controllers are better than others for causing this noise, the difference with this one is i did not use any SM components, my gate track is 3 times longer than usual because of the larger heat sinks i used and the places for the cap and resistor in the gate control circuit.

It maybe time to switch back to controller 4 and try the TVS diodes on that controller.
It seemed that the pulse across the coil was shorter without the TVS diodes but i didn't grab a shot to prove it.

I am not sure why the TVS diodes didn't protect the fet in the previous test above, someone mentioned not to run over half the fet's rated voltage to be safe

[BEP]

Insert a current limiting resistor between the PSU and the coil. Scope across that and give it the juice. If the voltage drop across the resistor doesn't show an actual over-current condition things might become interesting.

[poynt99]

When the FET blew, was it possibly because of an extended time in an over-current condition?

If so, the TVS diodes can't (and aren't meant to) protect against that.

.99

[BEP]

Or against the intrinsic diode being slower than the FET junction