Help with PSU (Temp control fan, load bank, & PWM circuit)

[ronv]

OK, Here is a bit on how it works:
R4, R3, and U4 form a voltage divider from +12 and ground. The wiper on the potentiometer (U4) can be set to .65 volts. This signal is applied to the positive terminal of the op amp making it's output go positive. This turns on the FET. The FET turns on until the voltage drop across the .13 ohm resistor also = .65 volts. This represents a current of 5 amps (.65 volts/.13 ohms = 5 amps. The circuit is what is called closed loop so the current will remain at 5 amps even if the +24 drops.
U5 is just an amplifier with a gain of 7.66 times. It amplifies the voltage across the .13 ohm resistor. So .65 volts X 7.66 = 4.98 volts (call it 5). This displays 5 volts on the meter that = 5 amps per FET or 50 amps.
Same deal with U6 except the gain is 10X. The output of the temperature sensor is 0.01 volts per degree C or 1 volt per 100 C. So at 88C the output will be 8.8 volts to the meter.
U1 is a comparator. The output of the comparator will go to ground if the temperature sensor goes above .88 volts (88C). If this happens the output will ground the bottom of R4 making the + side of the op amp go to ground thus "asking" for 0 volts (no current) across the .13 ohm resistor. It will also sound the alarm and light the LED.
Because U1 has positive feedback the temperature must go belo 76C before things will start over.

We can make it go higher - how about 60 amps? This will be pushing the temperature limits when running the 24 volt supply but you will probably have more water than () had in his test.
Hope this helps.

 

[jocanon]

Awesome ronv! Thanks for the explanation. Let's make it capable of 60 amps. I am not sure I will go that high, but would not mind having the option.

 

[ronv]

OK. R3 now 68.1k.

 

[jocanon]

I feel like I should owe you guys something. You have been so helpful!

 

[()blivion]

Remember, this is *OUR* hobby... plane crasher. We do this for our selves, you owe us nothing.

 

[ronv]

Just post a few pictures of the planes when you get a chance.

 

[jocanon]

Remember, this is *OUR* hobby... plane crasher. We do this for our selves, you owe us nothing.

Lol. I like how I go from RC stunt pilot to plane crasher . Although its true, if you can't afford to crash them you can't afford to fly them!

 

[()blivion]

Yep, I have three. One is broken, one is never going to be finished, and the other is very very small. OH! And I just got a indoor lighter than air unit I have no gas for.

 

[jocanon]

Just post a few pictures of the planes when you get a chance.

This was right after I finished building it. It flies amazing! The throw on the elevator is almost 90 degrees so it will do some crazy stunts! If you want to see some truly inspirational video of this plane in action check it out here, this was me flying the other day...hahaha I wish I was that good, no this is a guy named Daniel Holman, he is a very good pilot:

https://www.youtube.com/watch?v=0HqN0YRTkt4

 

[ronv]

Way cool!

 

[()blivion]

Yup, really cool. At that point, it's basically a helicopter that looks like a plane. lol.

Back on Topic, sort of /)
I'm currently looking into a P-Channel version of this circuit.

The Idea
I think I said this before (PM?) but the idea is that by nature of P-Channel function, a P-Channel version of this circuit would have it's tab soldered to ground instead of the positive supply. Thus allowing for power supply's to be tested that have more than one positive rail, but a single ground. This is normally the case for most multi-rail power supply's, including computer power supplies, and as such would be very useful to design for. But it's looking very difficult to bring from theory into practice.

For the circuit we have going now, (electrical) current comes in through the pipe, into the drain of the MOSFET, then out the source pin, and into our sense resistor, then ground. We tap in between the sense resistor and MOSFET and feed that voltage back to the negative of our op-amp, and a negative feed back controls the current like we want. Electrically, the N-MOS version of the circuit would work 90% the same for a P-MOS version. The only difference is what would be the drain, would now be the source and vice-verse, and then we would flip the op-amp. Or in other words, in both circuits the drain is soldered to the pipe, but the drain on an N-channel MOSFET needs to be more positive than the drain on a P-channel and thus, voltages on the gate have opposite effects.

Here is a schematic to demonstrate that they are virtually the same circuit. Notice that the only difference is that the source and drains are flipped, as well as the op-amps inputs. And that they are even fed with the same control.

View attachment 66086

The Problem
This seams like it would work great.... there is a catch. On the left side, the sense resistor is between the drain and ground. THIS CAN'T WORK because then we will not be able to solder the backs to the pipe anymore, thus loosing our massive cooling advantage. The "supper cooling" of this project is it's key point, and must be maintained at all cost. I would go as far as to say, putting the N-Channel's on multiple pipes, separated by insulating tubes, while running insulating cooling, would be better than putting the sense resistors between the MOSFET's and the cooling pipe. And that is far far to complex to do either.

So, what can be done? I was thinking about configuring the sense resistor in a high side way. But then we need to reference the voltage drop off of the positive rail. I don't see a way to do this really without some complex configuration. Something like optical isolators and so on. And I almost dislike this more than the separate tubes idea.

Any thoughts?

(Note: This is not really related directly to your project jocanon, though it builds off of it and will make it more useful for others)

 

[ronv]

If you google high side current sense you should get some hits. It does make it more complicated.
Why can't you do separate NFET pipes?

 

[()blivion]

Why can't you do separate NFET pipes?

I figure these things are going to be the main problems, in no order.
Less (physically) optimal, less (physically) stable, more places for leaks, electrolysis will occur, less isolated channels, slightly worse cooling.

 

[()blivion]

LOL... My stupidity is to the levels of full on autism apparently. I was so hell bent on it being hard to do that I overlooked the most obvious thing... Flip the control voltage divider.

HERE is a simulation of how it's done with P-MOS. And here is a schematic for those who can't see the simulator.

View attachment 66109

Notes:
You can divide up the FET + op amp combos at any number to make numerous "channels" for different power rails. You need only add another control line. And all the FET's can be soldered to the same pipe as is desired for best cooling with out complexity.

1K resistors should be applied to all of the feedback paths and in series with the control divider output. This is just a precaution, because this configuration has a somewhat high common mode voltage.

Notice that the simulated FET's all have different voltage thresholds and the circuit still shares the total current equally, just as the one before did.

 

[ronv]

Now I get it. You want the pipe at ground.

So I guess were ready to build this?

() I have a bunch of big ol PFETs if you decide to build it I can slip some in an envelope.

 

[()blivion]

So I guess were ready to build this?

Probably, but the N-FET version. I'm only doing the P-FET version for my own thing... QED.

Now I get it. You want the pipe at ground.

Yeah, FAR FAR FAR more useful for other things if the pipe was the ground. Then one can do multiple rails, referenced to the same ground and still have all the backs soldered to the same cooling pipe (as I keep repeating over and over again). Then one can divide up all the MOSFET's for different rails and different current adjustments. Such as doing a 12v/20A rail, a 5v/35A rail, and a 3.3v/50A rail.

Such a dummy load would be more useful for fully loading a computer PSU during tests. Especially some of the higher end PSU's that put out > 1000 Watts and are $$$. Repairing and "burn in" testing these PSU could be useful. And selling such dummy loads, fully built for doing just that could also be useful (read: profitable).

() I have a bunch of big ol PFETs if you decide to build it I can slip some in an envelope.

Really? I would like that... I would like that a lot. Whats the part number? How many are you going to give me? I'll PM you my address, unless you would prefer a email?

I want to make a fully featured computer PSU tester eventually. I was planing on getting all the parts to build a few units and sell them on ebay, see what they auction for. I need to prove the circuit is functional and economical before I move to "production". I have about $500 I could drop into parts if it looks like a promising venture. As it appears now, just from a glance you can't find a good 1+ hour full load power supply tester for under $500. I think this circuit + physical components can be built for far less than that. It's a thought.

I'm still waiting for the other shoe to drop of course.

-()blivion

 

[jocanon]

()blivion, so is the difference between the P-FET version you are talking about now and the N-FET version that you can test multiple rails simultaneously?

 

[()blivion]

()blivion, so is the difference between the P-FET version you are talking about now, and the N-FET version [you had before, is] that you can test multiple rails simultaneously?

Yeah you have it right. (Edit: I misinterpreted what you were saying at first)

It's a little thing that bytes you in the end. For the most part, the two circuits work the same. The difference is that the former circuit has all the FET's tabs at positive. That is, the pipe you solder the FET's onto connects to the positive of the DUT. On the P-FET version, the pipe becomes the ground. This is important... almost every single power supply in existence has one single shared ground, and multiple other rails. THESE RAILS WILL SHORT TO EACH OTHER IF CONNECTED TOGETHER!!! (not trying to yell) Think about it, if you tried to say... hook up the 5v rail AND the 12v rail of your normal PSU to two different FET's with separate grounds.... Well.... you couldn't... won't work. This is because the positive of all the FET's have to be soldered to the pipe (for cooling) which connects them. So you can't separate the positives. And no matter what you do, the ground for both rails is going to be connected internally. So the N-FET version is really only useful for one single voltage. This is NOT optimal for testing computer PSU's, it won't load them enough.

The P-FET version on the other hand is the other way round. The backs are still soldered to the pipe, BUT THEY ARE NOW GROUND!!! allowing you to attach the one ground to all the FET's. Then you can separate FET's and different current control knobs for the different voltages. And they will all work happily together.

Bottom line
As far as it applies to you and what your doing. P-FET's are more expensive per Watt. So for a given amount of budget, you can make a higher wattage dummy load with N-FET's. Thus it is more beneficial for YOU to buy and build for the N-FET version that we already have figured out. The P-FET version is better for loading more than one voltage/rail at a time. you don't need that.

 

[jocanon]

Okay, so it sounds like it is time to pull the trigger on the N-FET version.

 

[()blivion]

Yeah, we should go for it. I'm sure we can make it work.

Whats the BOM again though? I just want to check it and make sure it's right.
(Edit: NVM, I still have the Email ronv sent me, I assume your still using that?)