Cooking eggs on a mosfet

[mindoverflow]

Hi,

I have a circuit that looks like this schematic, with good heat sinks (no air flow) that heats too much in about a minute or so. The H bridge is PWM driven at varying frequencies.



I'm attaching an LTSpice file (I don't know if it will function for you as it contains the custom IRF540 and 9540 models) and the simulated plots (at 10khz pulses). What do you think about it ?

The IRF540 gate voltage is clipped at 3V due to the 670Ω resistor (used to limit the current through the 15V zener) acting as a voltage divider with the pull up resistor. Do you have a solution for this ?

Do you have any suggestions to improve the graphs ?

I don't have an oscilloscope to confirm that the IRF models are being correctly simulated.

And back to the main issue, why is it heating that much ? (using a 25V power supply, I haven't tried under 15V I wonder if it would make any difference 12-15V)

Thank you.

 

[KeepItSimpleStupid]

Post a schematic image as well, please.

 

[mindoverflow]

The first attachment is a schematic, just click on the thumbnail image.

 

[ronv]

The lm339 can drive very little current so the FETs will switch very slowly and will probably have a lot of shoot thru current where the top and bottom one are on at the same time. Post your .asc file or have a look at the current thru the fets as they switch.

 

[mindoverflow]

Yes, you seem to be right, I see current shoots for a considerable amount of time as in this current thru mosfets simulation:

But seriously, is there a problem with the attachments? because KeepItSimpleStupid asked for the schematic while it was there and you are asking for the asc file while it is in the first post 7 Kb

 

[ronv]

With me it's hard to tell. But I have it now.
Take a look at this. Might work for you. I didn't have your models so I used some with the same gate charge.

 

[jjw]

The LM339 can't source current and the 4.7k pullup resistor may be too big to switch the FET on fast enough.
Also the voltage divider at the gate may prevent to switch off the FET.
Could you use some H-bridge driver IC ?

 

[MikeMl]

MindOver, you might get some useful info by reading this thread.

 

[alec_t]

The IRF540 gate voltage is clipped at 3V due to the 670Ω resistor (used to limit the current through the 15V zener) acting as a voltage divider with the pull up resistor. Do you have a solution for this ?

Remove the 670Ω resistor. It's not needed, as the LM339 has open-collector outputs.

 

[mindoverflow]

ronv thank you, your schematic generates almost perfect square wave outputs, and I think it is clever to switch off the mosfets throu the diode by the way, would a 1n4148 make the job or should it be a schottky diode ?
It looks perfect, I'll try to find a place for the extra parts on the board or add it as a module and would report the test results. Thanks mate.

For those who can't check the schematic posted by ronv (or won't) here is almost the same:


@jjv Yes, an H-bridge driver IC or mosfet gate drivers would be the best solution, but I didn't find any on the local market and I can't order online, then I went for the comparator solution, and now I ended up with 2 useless circuit boards (well at least for more than a couple minuts) then I'm trying to find a fix for this.

MikeMl , Thank you.

alec_t , well, I think an open collector doesn't prevent a short between the zener and the comparator which would lead to an excess of the zener power rating and/or the mili amps of the comparator output. You correct me if I'm wrong.

 

[alec_t]

You correct me if I'm wrong.

The '339 open-collector output can only sink current, i.e. pull the output towards ground. It has no path for current to flow from the positive rail into the zener in your original circuit. That is why the 339 output needs a pull-up resistor (4.7k). In your original circuit the zener current comes from the positive rail via this pull-up resistor only, so is limited to ~2mA. Your post #10 circuit does need the resistor to limit the zener current, because Q3 can pull up to the positive rail.

 

[mindoverflow]

alec_t Sorry, I haven't got your idea yet. I may have a wrong understanding of the thing, but what I think of is that an open collector doesn't source current but in this case current is sourced from V+ through the zener diode when the breakdown voltage is reached, then I see both the zener diode and the comparator output at risk.
I tried a simulation like this one:

And I see about 2 amps of current flowing through the components. I would appreciate that someone tell me where the mistake of my understanding lays.

 

[jjw]

The zener should be connected from comparator output to ground, not from +25V to comparator output.
The 4.7k resistor limits the zener current.

 

[mindoverflow]

The zener should be connected from comparator output to ground, not from +25V to comparator output.
The 4.7k resistor limits the zener current.

For what purpose ? That would give a 0 V across the diode at switching on cycles and clipping the Vgs to -15V during the off cycles. That's nonsense (frying the mosfet or keeping it on all the time) then better not using the zener at all.

Did I misunderstand your post ?

 

[jjw]

Yes, you misunderstood.
The purpose of the zener is to limit the gate voltage to 15V.
If the schematics in post #12 is intended to simulate the gate drive of NFet, when comparator output is 0, nothing limits the current through zener and output transistor of the comparator and both will fry.
When output is high ( the ouput transistor is off), the gate gets 25V from 4.7k resistor which is too much for the NFet.

 

[mindoverflow]

jjw Ok, I see now. But the post #12 is in reply to alec_t who says that the 680Ω resistor isn't needed and that the open collector doesn't source current, then I made that test circuit and a simulation to explain that the current is sourced from the V+.
And about your statement, the test circuit is suitable for the PFet switching (plus a resistor between the comparator and zener) just like my circuit on post #1, but ronv posted (#6) a solution that looks to be perfect.

 

[alec_t]

Try this sim and convince yourself the zener current is limited:

 

[mindoverflow]

alec_t Ok, in your circuit , the current is being limited by the pull up resistor as the zener is connected to the ground and comparator output but in mine it was between V+ and comparator output (parallel to the pull up). But now I see what you were talking about you were looking at the NFet part of the circuit and I was looking at the PFet, then we were talking about two different things but yes, the PFET's zener too could be moved to the ground (you didn't mention that) maybe that's what jjw was talking about, I didn't get his idea at first.

 

[mindoverflow]

Well, just to report the results, I managed to build the circuit on a perfboard and it works like a charm well, at least for the main issue, there is no traces of heat at all (continuous 1 amp or so and peaks of 3 amps). Then only problem is that the uC is being reset at every reversing (powered from 2 different power supplies and sharing the same ground - not optocoupled) then I'm still investigating if it's the backEMF / Flyback that's causing this (I'm using flyback diodes but not shottcky) or a problem in the perfboard.

Then thank you ronv and everybody else.

 

[alec_t]

Pushing 3A through perfboard traces might well cause problems. Do you have a star ground arrangement?