Dummy Load II

[()blivion]

If the circuit is oscillating enough, then it is basically a PWM load. The PSU may be able to survive that if the filters caps are a good low ESR type and the frequency is high enough. However, we will be totally unsure of how our load circuit is really performing. My biggest worry, at a glance, is exactly where all the power ends up being dissipated. If we are essentially doing PWM, then all the power is being lost in the resistors... not the well heatsink-ed FETs. Needless to say, it was not designed for this.

Edit: Unless there is some kind of feedback mechanism internal to the FET that can cause gate oscillations, I am perfectly satisfied with just a low pass filter on the gate.

 

[ronv]

Rev 2.1.1

Here is the integrator version. Looks pretty good.

 

[jocanon]

Sweet Ron, thanks! What's the new bunch of circuits to the left?

 

[ronv]

That is a 317 voltage regulator to supply the logic with 9.5 volts from the DUT.

It replaces your external 12 volt supply.

 

[jocanon]

Ah, i see now...I had to really zoom in on it to see the 9.5v

 

[Mr RB]

No disagreement. Do you see any more holes like we had when the +12 was on before the 24?

There should be no stability issues, just that one issue that if the target 24v PSU is disconnected while the current limit is turned up, the FETs will slowly ramp up trying to maintain current and then you get that overcurrent situation when the 24v is reconnected while the FETs take time to ramp down again.

Normally that is not a problem as you don't disconnect the target PSU when everything is cranked up! You have a pot to turn things down before disconnect and a switch to ramp the FEts gates down to off too.

If the target PSU has a bad fault where the output is cutting on/off I would use an open loop driver so it just fixes the FET gates at a set voltage, the FETs just act like a massive resistor. That's how my big dummy load is set up (open loop) like a resistor which is very safe if you have a bad PSU fault but won't do the auto current trimming nicety you get from the closed loop dummy load.

That could easily be setup with another switch to select open/closed loop modes.

 

[jocanon]

So, if I were to say, test the over current shut off on a power supply by cranking the current up until it kicks in, I think this may cause problems if I am understanding Mr RB's post because the power would just instantly cut out...right or not?

 

[ronv]

No, I think it behaves well on power up and down of the DUT now because the 9.5 volt logic supply will come on after the DUT.

If the target PSU has a bad fault where the output is cutting on/off I would use an open loop driver so it just fixes the FET gates at a set voltage, the FETs just act like a massive resistor. That's how my big dummy load is set up (open loop) like a resistor which is very safe if you have a bad PSU fault but won't do the auto current trimming nicety you get from the closed loop dummy load.

This would work for BJT but not for FETs due to the large variation in threshold voltage that can't be made up for with a small emitter resistor like your big load.

 

[jocanon]

This looks pretty cool (no pun intended). So if I understand correctly, we no longer need any extra power supply other than the DUT itself. There is the small issue of the fans, but I could easily power them off the DUT as well I would think. Then on the other hand, will it even need fans if I use Duset52's suggested current sense resistors soldered to the water cooled pipe? Maybe no fans are needed for the op-amps and other logic, or would it still be a good idea to throw a cooling fan or two in there?

 

[ronv]

Right you are, no extra supply. If you check out the little ammeter/voltmeter I posted for your guy you could use that for this as well since it would run either off the 9.5 or the 24/12. You probably don't need a fan if you use the resistors on the cooling pipe, but 1 might not be a bad idea. Do the FETs feel cold to the touch?

 

[jocanon]

Nah, those suckers are hot . I can't hold me fingers on them for longer than a millisecond.

 

[ronv]

Ahh, I thought they were water temperature. I lowered the temperature sensor to 50 C. Hope it is not to low, but then again we have 50% more FETs now so they should cool down. What temperature does the little read out say when they are hot. Where is it mounted?

 

[jocanon]

It only says about 20-25c I have the temp sensor glued with thermal glue to the coper pipe right next to a FET. Maybe I should put it on the plastic casing of the FET? If so, then I think 50c would be too low, not sure exactly what 50c feels like but they were above that I think.

 

[ronv]

50C is very unconfortable. I hate to put it on one fet, but that would give a faster response than the pipe unless the fuse is blown for that one. Pretty low probability I guess. A plot of case temperature vs current would be good.

 

[jocanon]

I can put that on my to-do list

Edit:
Nevermind, I'll do it now...

 

[jocanon]

Here's the results of the temperature of the plastic casing of the FETS in amps per FET at 24v:

1A=16C
2A=28C
3A=47C
4A=72C
5A=88C

At 5 amps per FET the metal backing which is soldered to the water cooled coper pipe was about 25C.

 

[()blivion]

Woh... That's kinda crazy.

I wouldn't ever think that the plastic case could get so hot if the metal tab wasn't also just as hot. The actually chip inside the package is soldered directly to the copper tab, so the heat is being taken away from where it's made almost instantly. The only way it could get hot on the back side was if the heat output was so high that it was creating a gradient in the silicon itself, pumping heat both directions. Then again, 120 Watts is pretty high, so that is probably what's happening.

Judging from those numbers, I would say we are getting well into the red for how much power the package can handle. I think 3-4 Amps per FET is the most we can consider reliable at 24 volts for TO-220. If we want more than that per channel, we are going to have to get a better package probably.

 

[Dusey52]

RthJC (thermal resistance junction-case) for the FET is 0.5 °C/W. RthCS (case-sink) isn't stated but it's not zero and is likely about the same 0.5 °C/W. So, at 125 watts, the junction is probably running 125° C above the 25° of the water pipe or close to 150°

 

[jocanon]

By rough estimate, I have about 15 hours of run time on them running at about 5 amps per FET at 24v and the only ones that are broken are the ones I shorted out. After I build the dummy load II, I would like to see at what point they break on dummy load I.

 

[()blivion]

The problem with being on or just over the max junction temp, is that the damage being inflicted is done over time. It's not always a lights-on/lights-off kind of thing. So it may take days, month, or even years for the devices to finally break down. Of course there is a temp that will, in fact, destroy them instantly. That is probably the melting point for the softest material in the die.

In any case, that is much too hot IMO. More channels, lower current, or better part power capacity is needed. Getting parts in a different package would be easy enough to do.