Current into a capacitor causes voltage to rise. That's what happens to a mosfet gate which is secretly nothing but a capacitor, as far as input current is concerned.
Of course a mosfet can be operated in the linear range or as a hard switch, just as a bipolar transistor can. It's just less convenient to do linear with a mosfet because of the gate voltage cost. Let's look at a very different mosfet.

This one is a, "logic level" type. The front page rating is 4.5 Vgs for, "on" and 10 Vgs for 3.5 milliohms. Then we go straight to the graphs to get a better feel for those numbers. Figure 3: This one is done turning on at 4 volts with a quick pulse rating of 130 amps. That is usually a useless number for us amateurs, so go to Fig. 12 and see Rds dropping like a rock between 3 volts and 4 volts. You're in the 5 to 7 milliohm range at 4 volts. If you go to 6 volts, you can promise 5 milliohms at operating temperatures. Still, the curve keeps curving!

What do you call, "completely on" when the curve never stops curving?
It depends on the amount of current you need.
Let's try 7 milliohms @ 4 Vgs and 125C and (first page) 50 C/Watt temperature gain.
Limit is 150C, room temp is 25C so 125C/50 = 2.5 watts.
P=I^2R
2.5W/.oo7 = 357
square root it for 18.9 amps
and, from E=IR we have .132 volts from drain to source.
Look at Fig. 8
We just about hit the dotted line on the left edge.

Let's go looking for trouble with Fig. 12.
Suppose your gate is rising through 3 volts and you have 20 milliohms.
2.5 watts/.02 ohms is = I^2
Current limit is 11.1 amps and Vds is .22V
Look at Fig. 8 again. We're in the middle of a safe area but we've calculated a limit.
That graph can only be true with a heat sink.
That's where you get into trouble and that's why you want to get out of the switching range as quickly as possible.

Bipolar transistor graphs are considerate enough to tell the SAO at DC.
This mosfet datasheet wants to play, "pulses only".
You can see that you really have to dig to get the information out of this data.
The best I can do is drag you through the process so you see that, "on" is related to your needs for current.

Excellent explanation, thanks.
Questions: so mosfets have no inherent saturation points as bipolars do? is a higher gate-to-source operating voltage (within spec limits, of course) always better?

 

Excellent explanation, thanks.
Questions: so mosfets have no inherent saturation points as bipolars do? is a higher gate-to-source operating voltage (within spec limits, of course) always better?

A Mosfets Drain-source resistance reaches a minimal value at a given VGS, for my FET it is 3.5 milliohm when VGS is above roughly 7V.
A higher gate source voltage is generally better for efficiency - but within limits.
The VGS voltage can not exceed the stated datasheet constraint without failure, for my FET this is +- 20V.
So i am driving my fet gate at 10V, which is plenty for my application - but more than 20 will fry the FET.

 

I have attached a schematic that i have composed as best i can at this present moment.

The lm2596 was not available in lt spice so i chose another component to give an idea of the hookup.
The DC-DC regulator is fed 34V from the bridge, with a bypass cap across its input as per the datasheet.
The output terminals are hooked up to a gate drive optocoupler that pulls the output terminal to VCC with a high PWM input, and pulls it to the -ve terminal of the DC-DC converter when low. i have omitted this as i couldn't find the spice file, also i had set my PWM to to 100% duty so the opto should of held the gate at the same potential at the output of the converter.
When i switched the duty up from 0% to 100%, the bulb lit up for less than half a second before the lm2596 appears to have failed.
The output of the converter was set to 10V, so when 10V was applied to the gate, the FET opened up, conducting a high inrush current - then the converter failed.
If this schematic sheds any light on the matter, then the last hour was not spent in vain.
If the issue is still clear as mud, ill draw something up tonight in GIMP or take a photo of my hand drawn schematic.

 

Excellent explanation, thanks.
Questions: so mosfets have no inherent saturation points as bipolars do? is a higher gate-to-source operating voltage (within spec limits, of course) always better?

Well, I'd say 0.132 volts drain to source at 18+ amps is pretty darn saturated!
But the curves on the graph show you can do better than that with more than 4 Vgs, like, 6 Vgs.
And, yes, the graph always curves the same way for every mosfet I've ever seen. They keep getting more conductive until you pop the gate.

 

My Gimp skills have corroded with time. Evidently.

This shows how i had the Dc-Dc converter set up when it fried.
As before, if anyone can spot why it failed, it would put my obsessive mind at ease.
The cap


Thanks folks!

 

Now that you have presented a schematic I can see that you are conducting 120 Hz pulsed current through the mosfet. I confirm that the gate driver needs local capacitor help. If I had a 'scope on this, I would experiment with, "How many microfarads can I install without the electrolytic getting so slow that it doesn't improve anything". We're talking about buying a low ESR and low ESL capacitor.

The gate driver would only use up about 1% of the charge in the 220uf cap you provided, but you need another one on the output side of the LM2596. That's where your electrolytic, low ESR cap goes, but snuggle it up to the gate driver and KEEP the .1 uf ceramic cap in there.

With your switcher operating at 150 KHz it can get in 3150 pulses during the 21 mililiseconds I recommended with a 100 ohm resistor in series with the gate. With 1/2 ohm, the switching supply can get in about 15 pulses in the switching time. That seems adequate to me, so I don't know why your switcher blew smoke unless the switcher didn't like the gate driver dragging the output voltage to nearly zero when it first switched on. I think we're down to "decoupling and bypass capacitors" save the day.

 

Now that you have presented a schematic I can see that you are conducting 120 Hz pulsed current through the mosfet. I confirm that the gate driver needs local capacitor help. If I had a 'scope on this, I would experiment with, "How many microfarads can I install without the electrolytic getting so slow that it doesn't improve anything". We're talking about buying a low ESR and low ESL capacitor.

The gate driver would only use up about 1% of the charge in the 220uf cap you provided, but you need another one on the output side of the LM2596. That's where your electrolytic, low ESR cap goes, but snuggle it up to the gate driver and KEEP the .1 uf ceramic cap in there.

With your switcher operating at 150 KHz it can get in 3150 pulses during the 21 mililiseconds I recommended with a 100 ohm resistor in series with the gate. With 1/2 ohm, the switching supply can get in about 15 pulses in the switching time. That seems adequate to me, so I don't know why your switcher blew smoke unless the switcher didn't like the gate driver dragging the output voltage to nearly zero when it first switched on. I think we're down to "decoupling and bypass capacitors" save the day.

The strange thing was, the converter output was being pulled to zero when i powered up the circuit, it was when i jumped the duty from 0 to 100% that the failure happened.
It appears that it was the output being pulled up that cause the problem.
I got another 5 lm2596's on the way and another alternative rated for higher power. Hopefully, the definitive answers will come.
Thanks #12!
Much appreciated!

 

Right now I'm thinking 22 uf to 47 uf as the driver capacitor. They will empty by 10% or 5% when the driver turns on, and that seems adequate, too.

 

This is really getting interesting.
'converter output was being pulled to zero' needs some more explanation! Was the converter output zero when you powered up, and continued to be zero, and nevertheless you applied the PWM?
What is the source of your 35V pk sine wave input? Is it likely to be higher on no load?
What is the period of the PWM with which you had adjusted the duty?
I am assuming a LM2596 board when you are referring to LM2596. Right?
@#12:
Shouldnt it be 1% and 0.5%, which is even more adequate? (210nF / 22uF)

 

This is really getting interesting.
'converter output was being pulled to zero' needs some more explanation! Was the converter output zero when you powered up, and continued to be zero, and nevertheless you applied the PWM?
What is the source of your 35V pk sine wave input? Is it likely to be higher on no load?
What is the period of the PWM with which you had adjusted the duty?
I am assuming a LM2596 board when you are referring to LM2596. Right?
@#12:
Shouldnt it be 1% and 0.5%, which is even more adequate? (210nF / 22uF)

My microcontroller is powered from a a separate source, so the duty cycle of the PWM output to the gate drive optocoupler was 0.
The converter was configured to output 10V, i powered up my circuit (from my transformer).
When power was applied, the duty cycle of the PWM controlling the gate drive opto was still 0, hence the output of the opto was pulled to the -ve output terminal of the converter.
Looking at the god awful schematic above, it can be seen that if pin 6 and 7 (opto output and gate) are pulled down to pin 5 due to the lack of input at the opto, then pin 8 and the +ve output of the converter are also pulled to the -ve terminal.
So the circuit is powered up, but the mosfet gate is seeing little voltage.
I jumped the duty from 0 to 100%, pulling the output of the opto up to the +ve converter terminal potential, and the gate along with it.
This resulted in the bulb i am controlling illuminating for less than half a second, then the gate voltage appears to have disappeared and the bulb turned off.
I tested the fet and it appears in good health, but the converter appears to have failed.

My supply output is higher on no load, but negligibly so i believe. The failure happened just after loading the transformer.

The pwm frequency used was 20kHz.

When i say lm2156, i do indeed mean a generic board set up with the typical configuration.

Thanks

 

I'm no expert here, but shouldn't you protect the gate-source and drain-source with TVSs?
I've attached a file that better illustrates what I'm trying to say.

 

It's only a 10 volt supply for the gate and there are no inductors in the load end.

Besides, the mosfet didn't die, the switcher chip died.

 

It's only a 10 volt supply for the gate and there are no inductors in the load end.

Oh, I knew that!!! I'm just saying so the noobies can learn something (yeah, right...ha ha ha)

 

Oh, I knew that!!! I'm just saying so the noobies can learn something (yeah, right...ha ha ha)

Yeah, what #12 said!
resistive load and all that

 

Oh, I knew that!!! I'm just saying so the noobies can learn something (yeah, right...ha ha ha)

It's OK. Part of getting along here is being willing to say, "Oops, missed that one. "

 

Yeah, what #12 said!
resistive load and all that

Now seriously, I've used JDT's circuit in the past and it worked flawlessly, you may want to consider it. What I used a floating-type DC-DC converter for that.

 

Now seriously, I've used JDT's circuit in the past and it worked flawlessly, you may want to consider it. What I used a floating-type DC-DC converter for that.

I have had another thread discussing my attempt to implement two mosfets for AC as per the circuit linked above.
Could you elaborate on the floating type converter? Does that mean isolated?
I know that my converters are not isolated, i havent had the time to look into dc-dc converters due to other obligations.
I have a module coming up this month on them though.
Thanks

 

I have had another thread discussing my attempt to implement two mosfets for AC as per the circuit linked above.
Could you elaborate on the floating type converter? Does that mean isolated?
I know that my converters are not isolated, i havent had the time to look into dc-dc converters due to other obligations.
I have a module coming up this month on them though.
Thanks

Yes, it's completely isolated, and I used it exactly as shown in JDTs post.

 

Would it be correct to assume both minus pins on the regulator are tied together?
How big is the cap on the input to the regulator.
High frequency switchers like this can be fussy. It may not like a breadboard.

 

Would it be correct to assume both minus pins on the regulator are tied together?
How big is the cap on the input to the regulator.
High frequency switchers like this can be fussy. It may not like a breadboard.

No, the minus pins are not tied together on the regulator, that's what makes it a floating converter. The cap at the output of the regulator that I used was a 10 µF tantalum plus a 0.1 µF X7R