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

That will be 1am my time, but we are just putting in a movie so I should still be up.

OK. got the thing built. I'll be testing it over the next 30 mins....

Awesome. Keep us posted!

OK, so... Tried my version with a small wall adapter. Did not work so well, which is not untypical for a new build. One side got very very hot quite quickly. Looking at the build for mistakes now. But I am also wondering if one of the parts was not bad from the start (The price I pay for using salvage parts so often). It's most likely an Op-Amp as I beat on them pretty good from time to time. Yes... these were salvaged parts AND have been used for prototyping more than once. Hey! if anyone has complaints... give me some more

I'll be back on the ball tomorrow.
-()blivion

Good morning world. It's ~9:30 AM here and I have good-ish news. The circuit works, But there are still some things that need to be worked out. I'm not 100% sure yet, but I'm pretty certain I have at least identified the problem.

The problem I was having last night was not one of dysfunctional parts, it was one of power... not enough to be exact. It appears that if the Op-Amps are powered directly off the power coming in, and the power supply can not supply enough current to keep the voltage > 3~7 volts (depends on the Op-Amp) the Op-Amps will not work properly to hold the lines low enough to prevent the load from drawing massive current. This of course causes feedback, with the Op-Amps thus getting less power and still not holding the lines low enough, and thus causing more current to drain lowering the power to the Op-Amps even farther. At some point, the op-amps just pass what ever voltage they are getting straight on to the gates, and this is about where things sit at and cook for a while.

So, if at anytime the voltage to the Op-Amps drops below this point, the before mentioned feedback will kick in and keep it stuck like this. And since this condition is seen every time during start up, as well as possibly popping up randomly at anytime. The whole problem is not good juju and needs to be worked out.

For now, I have the problem fixed in mine by powering the Op-Amps with an independent wall adapter. I am currently investigating other ways we can fix this that should not require such an "unsatisfactory" design. Things such as...

1) Pull down resistors on all the gates/Op-Amp outputs.
2) Above resistors + capacitors to make the load adjustment changes slow.
3) Under voltage lockout circuit, which keeps the gates low until enough voltage is seen.
4) Boost regulator circuit that will make a constant voltage out of extremely low voltages.

And if none of this works good, the option will always be there to attach an external independent supply as I have with mine.

Bottom line... the circuit will effectively load a power source in a controlled manner from what I'm seeing. You *COULD* order parts now if you want. Powering off of a separate supply is not really that big of a hassle. Plus we could always throw in a 9v battery.
-()blivion

P.S. Soldering all the FET's to a flattened copper tube, then running water through it does work wonders for cooling them.

That's great news! I agree it will be more convenient to not have to power the op amps separately, but in the grand scheme of things really not a big deal if it does have to be powered separately.

()

I'm curious.... If you have some spare time could you measure the gate voltage on each fet? It would be interesting to know what a typical spread in the voltage is. It would also be interesting to know how good the water heatsink works. Temp rise / power disappation. Might come in handy someday.

By my measurements and (terrible) maths I'd say 1~3 mV spread. And the water block heat sink worked very well. Two PSU's ago I was up to 120~180 Watts. The copper was cool to the touch up to ~100 Watts. And that's only with 2.5 gallons of standing water, no rad or nothing.

I'll get you some exact numbers in a bit. But I need better crappy test PSU's now. I'm 0 and 3 today...


Edit: OOPS! I'm retarded.... Those numbers are for sense resistor voltages. Disregard them. Correct numbers in newer post.

()blivion, are your PSUs all braking on you?

No.... *I'm* breaking all my PSU's on me.

Victims Broke because..
[SUB]PSU #1) apparently the case is on there to keep things like pieces of scrap wire off of the circuit board, Who knew.[/SUB] (Scratch that, I fixed this one)
PSU #2) I thought 18 Amps was the same thing as 20 Amps. The PSU strongly disagreed.
PSU #3) Was TKO because it had a snake inside it... or was overheating. I blame snakes. (Not technically "broken" I guess)

The circuit works fine.

Ah, I see

OK. got some pictures of my setup, as well as some measurements.

Voltages (Running @ 4.53V ~ 10.38A)
MOSFET------GATE VOLTAGE------SENSE VOLTAGE
1______________2.1495___________0.173
2______________2.1490___________0.174
3______________2.2765___________0.174
4______________2.1540___________0.175
5______________2.1610___________0.174
6______________2.1470___________0.173

Temperatures: 1 Hour run time.
Room Temp = 27c
Water Temp = 28c (2.5 gallons)
Sense Resist = 44c
Copper Pipe = 30c

Copper pipe temp after ~10 seconds without pump 48c



And now some pictures... (some are similar, but not duplicates)

View attachment 65972View attachment 65973View attachment 65974View attachment 65975

View attachment 65976View attachment 65977View attachment 65978View attachment 65979

Looks like it works. How did you flatten the copper, hammer? I may end up going the pump route instead of a garden hose, any recommendations on where to pick something like the up?

(),

That is a thing of beauty! Some very interesting data. The FETs only varied .13 volts for their turn on voltage - not anywhere near the worst case spec of 2 volts and 2- Only .04 degree C per watt temperature rise. That says with 1200 watts it will still only be 48 + 27 or 75C. I never thought it would be that good!

ronv said:

(),

That is a thing of beauty! Some very interesting data. The FETs only varied .13 volts for their turn on voltage - not anywhere near the worst case spec of 2 volts and 2- Only .04 degree C per watt temperature rise. That says with 1200 watts it will still only be 48 + 27 or 75C. I never thought it would be that good!

Click to expand...

Agreed, it does look like a thing of beauty...now let's just hope I can replicate it...I will probably have a lot of "stupid" questions along the way of putting it together, being the newbie that I am ! Lucky for me, it looks like a pretty simple circuit.

I flattened the copper pipe mostly in a vise, some with a hammer, then I finished it off with a sanding bit in a pencil style hand grinder (generic Dremel). It is not so important that the pipe be 100% flat on it's surface because the solder will wet and fill most spaces. More importantly is that it is CLEAN and shinny so the solder will wet the metal well. You also need a heavy iron to solder something like this, mine is 140 Watts and only got the job done after a few minutes of heating.

I recommend you take these steps for best results...
(1) Tin all your MOSFET's backs. (use lots of flux)
(2) Tin your pipe well. (use lots of flux)
(3) Clean clean clean your solder job up, then reflux all the parts lightly. (Use a file or wire brush, NOT SAND PAPER)
(4) Heat the pipe until the solder re-liquifies.
(5) Apply the MOSFET's to the pipe CAREFULLY. (You *MAY* want to put your MOSFET's on a former/holder of some kind)
(6) IMMEDIATELY remove the heat and allow it to cool. (careful not to disturb it at this critical stage, it's much like glue drying.)
DO NOT FORCE IT TO COOL. LINGERING HEAT IS REQUIRED TO MELT MOSFET BACKING SOLDER.

As for water pump, Ebay I guess. Just make sure you get the pipe and tubing that fits said pipe figured out first. then you'll know what diameter the outlet needs to be. Or you could work backwards from a pump I suppose, whichever is easier for you. I got this particular pump for free out of some ornamental fountain my mother was tossing out. And it happens to be ideal for the local hardware stores selection of clear tubing, which also happens to fit the copper pipe I have, which ALSO happens to fit some radiators I have.

I was planing on liquid cooling a PC, but then I looked at the physics of such things and saw absolutely no reason to. It won't make my PC faster, and even when one water cooling, your still actually air cooling, defeating the purpose. So I decided to look into phase change or TEC cooling instead. THAT actually achieves a real goal. Not the snake oil of liquid cooling.


Edit: @ronv: Yeah. I think it turned out quite nice myself. Thanks for your positive criticism. Means a lot.

Interesting data indeed, If I tossed out MOSFET #3 things would be even better variance wise.

()blivion said:

DigiKey BOM (Bill Of Materials):

10 x STP80NF12...................= $11.20
10 x AC07000001307JAC00.....= $7.84
5 x TL082CP........................= $1.25
1 x P160KN-0QD15B1K...........= $0.84
1 x RNF14FTD23K2................= $0.15

Total minimum cost: $21.28 + S&H

*optional*
5 x small bypass capacitors for the Op-Amp's ≈ $2.50
10 x ALSR10R1300FE12....= $17.60 (instead of 10 x AC07000001307JAC00)
1 x Better Pot that comes with a knob.
More than the minimum amount of each part, In case you break a few.

Extended Cost: Should not need to exceed $50

Schematic
Here is the schematic again, simplified for readability...

View attachment 65959

Notes
The choice of resistor will let you choose between precision and price. The $7.84 resistors are 5% tolerance units. The other units are 5 times more accurate, at just 1% tolerance, but add $10 to the price tag.

The Op-Amps used are IC's with two Op-Amp circuits per single 8 pin package. 6 of the pins are the Op-amps inputs and outputs, and are whats shown in the schematic. The remaining two are the positive and negative inputs, and should simply be connected directly to the input power. These are not shown in the schematic, but they are necessary. And it is entirely possible (probable?) that these pins will need to be bypassed with some capacitors right at the pins, to prevent oscillation.

Going farther
One can extend this basic device for a little more cost, if you want the following features...

Over current shut down.
Overheating shut down.
Current/Wattage readout and measurement.
Timer based shutdown.
Closed loop cooling.

Click to expand...

So I am about to place an order for the above materials. I just have a few questions first:
1. The original linear circuit had those 3 current sense resistors, does the new and improved circuit not have need of these?
2. What are the important specs to look for in a pot if I want to upgrade to one with a knob? Does it just need to be a 1k pot capable of 1/5watts, or is there anything else specs wise that matters?
3. What benefit do the "optional" 5 small bypass capacitors provide, trying to decide if I should get them or not?
4. Lastly, and I apologize if this was already cover, I just want to be sure, so this will work on 12 & 24 volts...i.e. if I run the node 10 X I can plug in the 24v supply and crank it up to 45-50amps, and likewise with the 12v supply to 45-50amps but half the watts?

Thanks again for all the help! I am going to go down to home depot and see what I can find as far as copper pipe...I guess I should probably pick up some PCB board at radio shack or eBay as well. I am also thinking that it might be a good idea to encase this whole thing in some sort of non-conductive case when it is all done. I have 4 little kids running around the house and would hate for one of them to get shocked

Your Questions
#1 Yes, we do not need these anymore. The new circuit takes one single point of current sense and spreads it to each and every MOSFET.

#2 Doesn't matter, get what ever you like, just needs to be 1k. Or we can change the other resistors if you would prefer.

#3 Not a benefit as much as possibly a necessity. Op-Amps tend to self oscillate (fish tail out of control) Such things could spell disaster.

#4 Get two RNF14FTD23K2 and any random switch I guess. That will give you a 24v mode and a 12v mode. Don't run it off 24v in 12v mode.


*Bonus round*
Though I haven't tested it yet, I think we can make the device run off the power provided by the DUT (Device Under Test) by simply putting 1 or 2 1k resistors from the MOSFET gate pins to ground. Which as we discuses, would be preferred but not necessarily required.

I'll be testing this later today if you don't mind waiting to buy parts YET AGAIN...?

OK, Resistors of any kind holding the gates down does not appear to work enough to let the thing start. Still looking into other options, hopefully you will not have to bite the bullet and just end up using an independent power source. If we do find a solution, it will be only a few extra parts added and a few extra bucks/cents.

If you want, you can get the parts we have picked out now, as you are still going to need at least these. Then later buy the rest when we figure out what your going to need to fix this problem. You will of course need to pay shipping a second time. And you loose the opportunity to get discounts from buying in bulk, should we be buying more of the same parts you already have picked out. I prefer having everything hammered out, THEN buying parts. But then again, I have infinite free time too.

The problem restated

In the current version of the circuit the op-amps power pins are attached directly to the DUT input power. With this configuration, when our circuit is underpowered, the device sits in a state where the MOSFET's conduct a great deal more current than they should. Presumably this is because of the op-amps running improperly when not powered, (as should be expected). Or possibly because the MOSFET's are somehow self powering when the gates are floating. Whatever the cause, the op-amps never get any power because of it, and thus the op-amps do not activate enough to preform their intended function. And the load continues to be seen as a short and the DUT shuts off from over current protection.

Can anyone think of a solution to this paradox that does not involve an independent power source?

You should check the input common mode voltage range of the tl082. It doesn't go down to the negative rail. Try an LM358 instead, which has more appropriate inputs