How much single pulse power MOSFETs can handle?

[Ekke]

Hi all!

I'm thinking to build a capacitor discharge welder, but there seems to be a problem with MOSFETs, I can't figure out how many of them I will need.

I'm thinking to use these MOSFETs:
IRLS3036-7PPbF
https://www.electro-tech-online.com/custompdfs/2009/10/irls3036-7ppbf.pdf

Main reasons:

• Voltage (enough, not too high)
• Low R_on (really important with high currents)
• High pulse current (?)
• Package (I like reflow)

I would like to use up to 48V, but my capacitors will be 50V (65V surge). Is 48V too high? But the main problem is this: Those capacitors can store ~600J. Single pulse width will be something like <15-20ms. Fig. 8 in datasheet says "Single pulse" in the corner, but those means continuous pulses and are limited by junction temperature ("Repetitive rating; pulse width limited by max. junction temperature.")?

I found this info from some random source: "If you weld nickel, you can use a constant value of about 0.003Ω.". So, if I have 48V/0.003Ω = 16000A.

But, if you count ESR of capacitors (many parallel: 0.64mΩ), resistance of MOSFETs (10 parallel: 0.15mΩ) you get 48V/3,79mΩ = 12665A. It's still quite a lot...

There will be some wiring etc. so resistance goes up a little, but the maximum current will still be >10000A? So, can 10 of those MOSFETs handle it (10x1000A repetitive peak? but not @ 48V) or do I have to use something else? 600Ws [J] doesn't sound so much, so why it seems that I will need 10000 MOSFETs?

If I counted right after 1ms pulse @ 10000A there will be ~24V/160Ws left in the capacitors, so the maximum pulse will be quite short. Any info and calculations are welcome..

 

[Ekke]

So after all this is a hard question?

 

[Sceadwian]

You don't use a mosfet for that kind of thing, they can't switch high currents like that, the reasons are quiet complex. Simply charge the capacitor tap the electrode and it will discharge. If you want to get to a set distance discharge a secondary high voltage capacitor through the circuit as an initiator, then the lower voltage high current pulse can follow that ionized trail to intiate the primary welding current which is low voltage.

You need to factor in inductance from every part of your circuit at this point as at those pulse currents it will be the primary delay factor, significantly more than you seem to think as you didn't even mention inductance in your post. At those currents even short straight pieces of wire have significant inductance.

 

[Ekke]

You don't use a mosfet for that kind of thing, they can't switch high currents like that, the reasons are quiet complex.

The funny thing is that I know at least 4 CD welders that uses mosfets. Fets surely need a protection circuit because of the induction peak, I'm not sure how delays will be a problem..

 

[Sceadwian]

Odd, I'd figure they'd use SCRs, for a capacitive welder SCR's handle peaks much better I thought.

 

[smanches]

The funny thing is that I know at least 4 CD welders that uses mosfets. Fets surely need a protection circuit because of the induction peak, I'm not sure how delays will be a problem..

Are these commercial products ir DIY projects?

 

[gerty]

Where I used to work we had several brands and all of them were scr fired.

 

[Sceadwian]

Now that I think of it more I don't see how it possibly could have been a Mosfet in a CD welder, Mosfets have a DV/DT limit that can't be broken or the parasitic BJT latches in some manner as to cause device destruction.

 

[Ekke]

Are these commercial products ir DIY projects?

DIY. Haven't seen any internals of commercial products.

Here is a one:

YouTube - Spot Welder / Discharge Cutter finally completed!

"Thanks for the feedback! Good question regarding weld current

I use a "1 Farad" capacitor, switching is controlled by 4 x IRFP2907 MOSFETs in parallel. Now, the IRFP2907s have a pulse current rating of 816Amps limited by junction temperature, according to the datasheet

I have no idea what that means. But if I use less than three of those MOSFETs, they literally explode into tiny fragments at weld times over 7mS (and man they're expensive!)

So the answer is... One billion?"