Mosfet Heatsinks and capacitor voltage

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
i have constructed the circuit below to control a halogen bulb with two MOSFETS.

previously i had been having issues trying to derive a dc supply voltage for my gate drive opto.
Though the circuit above seems to work and the uneven power dissipation seen prior appears to have been solved.

Onto the capacitor question: when i vary my duty cycle i can see the voltage of my cap (C1) drop from 12V to around 8.5V. This isnt much of an issue for the fets that i'm using as Rds(ON) reaches a minimum around 8V.
Though i was wondering what steps i could take to keep the capacitor voltage higher?
The cap i have in place is a 330uf bog standard electrolytic, though i do have a few 35V 220uf low ESR caps i can butcher from fried buck converters.
I guess i could use caps with a higher rating?


Also, i am wondering if there is any recommended MOSFET heat-sink types? (my fets are TO-247AC)
The Fets get above 100 Celsius at the moment.
This is breadboarded currently but i will be using Vero-board when i am confident with the circuit.

Any guidance is appreciated, as always.
Thanks!
 
Last edited:

wayneh

Joined Sep 9, 2010
17,496
Why not just use a bigger cap for C1? And the 680Ω resistor - it should probably be a lower value if your opto is drawing enough current to take the zener out of regulation. Don't go too low though, you need to protect the zener from overcurrent.
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
Why not just use a bigger cap for C1? And the 680Ω resistor - it should probably be a lower value if your opto is drawing enough current to take the zener out of regulation. Don't go too low though, you need to protect the zener from overcurrent.
That's that question answered, ill do both when i get the cash - thanks.
i have been limited to the 680 ohm resistor as its the only 3W resistor i have, all my others are 1/4 watt, though i don't have any caps or larger capacity either, so ill likely purchase both.
 

ronv

Joined Nov 12, 2008
3,770
The opto should only draw 6ma.. Do you have anything else attached to the top of the cap.
Does the dc value change or do you see short pulses down to 8.5 volts.
What is the PN of the zener?
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
The opto should only draw 6ma.. Do you have anything else attached to the top of the cap.
Does the dc value change or do you see short pulses down to 8.5 volts.
What is the PN of the zener?
There isn't anything else connected to the top of the cap aside from the zener, opto-driver and bypass capacitor.
I am monitoring the DC value with my multi set appropriately, i see the dc level reduce when i PWM the gate (@25kHz).
The forward voltage drop of my zener is 1.2V
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
You need a filter on the rectifier output.
Add a 100μF or larger capacitor from the diode cathode to common.
i did have another capacitor on there previously, though i found that the presence of that capacitor caused uneven heating of my fets stemming from diode leakage - i think :S
 

crutschow

Joined Mar 14, 2008
34,281
What diode are you using?
You may need to go to a Schottky diode since a standard diode can have a lot of recovery time reverse current at the high PWM frequency.
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
What diode are you using?
You may need to go to a Schottky diode since a standard diode can have a lot of recovery time reverse current at the high PWM frequency.
I just checked with the added capacitor again, it is indeed the cause of the uneven heating.
The type of diode i am using at the moment is an STTH3010D - Ultrafast recovery high voltage diode.

Thanks for the help folks, i thought i had managed to escape my diode problem haha
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
Just tried with a High-Voltage Schottky Rectifier Diode.
The lower fet (drain connected to the load) gets very hot very fast.

Also, when i raise the duty - the voltage of the capacitor jumps right up to 70+V. It is rated for 50V.... eeeek.
I have a 400V cap i can sub in, but the diode still appears to leak and heat M4 as a result.
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
Firstly, let me apologise for the inconsistent labels between the image in the first post and this simulation below.
i have been playing with LTspice:
i can see that during times when the MOSFETS are open, the capacitor (C2) charges through the diode (D3) and the current flows though the open M2 lower fet.
During times that the MOSFETS are not open, the body diode of M2 (D2) conducts to charge the capacitor, causing uneven dissipation as the inrush from the capacitor is passed through the lower fet when the PWM output is low.

hotfetwaveform.PNG
When i remove C2 (as per the original post above) the current through D2 peaks around 50mA - which i correlate to little perceptible difference in temperature between the two fets when i test on my breadboard without C2.
With C2 in place, the current through the body diode D2 surpasses an amp.

If i put a resistor in series with the additional capacitor C2, i can limit this current but also inhibit the charging of the capacitor.
Is there any other solution to this problem? not that i am even fully sure what the problem is.
I could reduce the resistance in series with C1 as stated above, or increase the value of the capacitor used for C1.
Or i could try limit the current through the body diode of the lower fet and use the additional capacitor C2.

But is the diode leaking? or is the capacitor charging through D2?

I am not fully certain if my assessment of the situation is correct, nor (if it is) what the best route to take is.

My threads and questions always seem to be a pain in the ass when it comes to trouble shooting.
 

Attachments

Last edited:

ronv

Joined Nov 12, 2008
3,770
It may be a problem with where you are measuring from. Everything is now referenced to the source of the FETs, not ground. So in spice you can go to view, then set probe reference and put the ground probe on the source connections to see the supply voltage and gate drive.
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
yeah my application of the simulations is seeming to confuse my comprehension.

On the breadboard, switching the connections of my drains switches which one gets toasty, but in simulation i cant see this. (Scratch that, i hand drew the circuit and changes to the load im making on the breadboard, i can see that the FET that is conducting (body diode) for the capacitor changes when i reposition my load).



I am going to see about procuring a lower value resistor to put in series with the zener as per the top post, hopefully that will keep the zener/capacitor voltage up.
I cant really say with conviction why my FETs are heating unevenly with the addition of the extra cap, so i'm not sure on how to prevent the problem.

I believe the issue to be the current through the body diodes, but am unsure what the best solution is.
This 9+ days wastage of time is going to look rather silly in my project lab book. :(
 
Last edited:

crutschow

Joined Mar 14, 2008
34,281
I think we are going in circles. :confused: I believe you need to make a fundamental change in your circuit as I mentioned earlier.
The high peak capacitor current is from the current through the zener.
If you remove R3, C1, and D4, and use a series regulator,such as an LM317 or LM7812 to generate the 12V, than the current will only be whats required to drive the opto and the gate drive current.
That should be much smaller than the zener current with the 680Ω resistor and thus the capacitor current will be much smaller also.
You can also add a small resistor (1-10Ω) in series with the diode to limit the peak current.

Also why are you using such a high PWM frequency?
 

ronv

Joined Nov 12, 2008
3,770
I think you could also return the cap after the rectifier diode to ground instead of to the source of the 2 FETs. That will keep that current out of the bottom FET.
Tell us more about what you are doing and where you are measuring when you see the 12 volts go to 8.
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
I think we are going in circles. :confused: I believe you need to make a fundamental change in your circuit as I mentioned earlier.
The high peak capacitor current is from the current through the zener.
If you remove R3, C1, and D4, and use a series regulator,such as an LM317 or LM7812 to generate the 12V, than the current will only be whats required to drive the opto and the gate drive current.
That should be much smaller than the zener current with the 680Ω resistor and thus the capacitor current will be much smaller also.
You can also add a small resistor (1-10Ω) in series with the diode to limit the peak current.

Also why are you using such a high PWM frequency?
definitely cycling in circles with this drive circuit, though it wasn't my intention when i created this thread.
I can remove the zener diode and the problem persists.
I clarified this by using a isolated dc supply for my gate drive opto and putting D3 and C2 into the circuit. The Fets heated unevenly with C2 and D3 from source to drain.
I did have some lm2596's i was using to drive the gate, but they kept smoking when the voltage at the capacitor i was using to supply the DC in jumped above 40V when switching.
The PWM frequency is at that level to ensure inaudible operation, using a lower frequency produces quite the buzz.
I think you could also return the cap after the rectifier diode to ground instead of to the source of the 2 FETs. That will keep that current out of the bottom FET.
Tell us more about what you are doing and where you are measuring when you see the 12 volts go to 8.
I reckon it is a good idea to return the cap to ground than to the source to avoid the fet body diode.
Ill look into that today.

When i connect my multimeter to the anode and cathode of the zener diode in the circuit assembled on my breadboard, i can observe its drop from 12.2V to 8.4V.
The voltage drops as a i raise the PWM duty and the gate capacitance is charged and discharged more often per period of time.

Thanks for the help folks. i need to try put this silly issue to bed.
 

crutschow

Joined Mar 14, 2008
34,281
..........................
I can remove the zener diode and the problem persists.
I clarified this by using a isolated dc supply for my gate drive opto and putting D3 and C2 into the circuit. The Fets heated unevenly with C2 and D3 from source to drain.
.........................
If all you have is D3 and C2 in series then there should be no steady-state current through C2 since the diode will cause the cap to charge to its maximum peak voltage and then the current will stop.
The only path for the current is through the reverse biased diode, either reverse leakage, or reverse recovery circuit.
Standard silicon rectifiers can have a significant reverse recovery current at 20kHz.
For example my simulations with a 1N4001 showed a significant peak reverse current (many amps) at each reverse-bias transition.
A 1N4148 switching diode showed only about 4mA peak.
 

Thread Starter

urb-nurd

Joined Jul 9, 2014
269
If all you have is D3 and C2 in series then there should be no steady-state current through C2 since the diode will cause the cap to charge to its maximum peak voltage and then the current will stop.
The only path for the current is through the reverse biased diode, either reverse leakage, or reverse recovery circuit.
Standard silicon rectifiers can have a significant reverse recovery current at 20kHz.
For example my simulations with a 1N4001 showed a significant peak reverse current (many amps) at each reverse-bias transition.
A 1N4148 switching diode showed only about 4mA peak.
I had a bleeder resistor across the capacitor that ended up being blackened by the power dissipation before i noticed. So that would of caused a steady state current through the cap causing it to charge (slightly i presume) every cycle.
Though, that being said - i can visibly see the DC level creeping back on my scope.
As i am doing this project in-between a-lot of school work and assignments, my tracking of confounding variables becomes fragmented with every change made to the circuit. I need to work on that.
I will read through this PDF today to inform myself about reverse recovery of diodes.

I have drawn up a schematic this morning to avoid using the body diodes to charge the caps, though i will still likely omit the lm2596's i have laying around, as i don't have the components to regulate the dc into them properly.
Well, i guess i do have 12V zener diodes and an adequately rated resistor i could use to ensure 12V into my buck converter, but i might as well use a rectifier and a cap charged to the supply peak then use the zener and another cap to maintain a 12V regulated dc supply for my gate drive opto.

My drawing had a floating source, so i guess ill be using my 12V dc out from my zener to power a isolated dc-dc buck converter. Using my lm2596 board here would short out the bottom fet.
One day i will be proficient in reducing my repetitive errors....... one day!
I think the cap that i place across my supply will likely be charged to a relatively high voltage when the supply voltage spikes, so i will need to ensure i don't blow any caps there.
I'm also hopeful that if the diode rectifier leaks, the transformer will handle the short circuit current without any deleterious heating.

So with the gate drive issue on hold until i procure a suitable converter, can anyone indicate any tips on how to efficiently heat-sink the MOSFETS?
Should i just get a clip on heat-sink and hope for the best?
 
Last edited:

crutschow

Joined Mar 14, 2008
34,281
Experience is being able to recognize the same mistake when you make it again. :)

If the resistor got black from heat then it was likely discharging the capacitor significantly between the 50Hz peaks.

So did you try the circuit without the resistor to see if it still heats?

If the circuit works properly you shouldn't need a heat sink, but you could try some of the clip-on ones to start.
 

ronv

Joined Nov 12, 2008
3,770
Actually it was me who led you astray by cutting it to close with the 680 ohm resistor. But don't give up on it yet.
Forget about what I said in post 15 and either add the cap Carl suggested from the output of the rectifier diode D3 to the sources. Or make the resistor R3 smaller by about 1/2. A 1k in parallel with the 680 would probably do it.
The power in the body diode is very small.
When you run your big lamp you may need a heat sink of some size.
 
Top