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

it would take a ton of TVS to blow a 50 amp fuse. Maybe a crowbar circuit?

Click to expand...

Yeah... you're of course right. No realistic amount of TVS's are going to blow a 50 Amp fuse. A full crowbar would be a better plan for sure. Maybe use the dummy loads FET's as a crowbar circuit? All ten of them closed would surly blow a 50 Amp fuse. Wait.... we have had this discussion before I think... ?(0.o)

I'm just worried that, although very unlikely, the system may up and decide !OOPS! and send mains power to the dummy load. Although I *DID* just say how unlikely that is. Can't be too safe though, and it should be an easy mod.

You are right. The heat sink is at high voltage, don't ask me how I know

Click to expand...

Oh! that reminds me... the heatsink is shorted to the tab of the main power transistor from time to time. So try not to touch it while it's plugged in... OK?

()blivion said:

Oh! that reminds me... the heatsink is shorted to the tab of the main power transistor from time to time. So try not to touch it while it's plugged in... OK?

Click to expand...

Yeah, a while back I accidently brushed my hand by the heat sink. It just tingled slightly didn't hurt or anything, but I realized that it is high voltage and won't be doing that again.

()blivion said:

I'm just worried that, although very unlikely, the system may up and decide !OOPS! and send mains power to the dummy load. Although I *DID* just say how unlikely that is. Can't be too safe though, and it should be an easy mod.

Click to expand...

Like how easy? I don't really want to do it if it is going to be too involved.

Actually, I don't think I can put the thermistor on the transformer after all. There is something going on when I put it there because the circuit is behaving strangely, but only if I put the thermistor on the transformer, it's fine when I put it on the heat sink, so I think the heat sink it is. Does it even matter that the heat sink is at high voltage, I mean the thermistor is completely electrically insulated from the heat sink?

When I say the fan circuit is behaving strangely when the thermistor is connected to the transformer, sometimes the fan speeds up when I lower the current, sometimes it just turns on all the way and won't go off (this is eerily familiar to when I had the other temp control circuit going and was getting noise interference...). Now, when it is doing this, that is slowing down when I turn the current up, if I take one side of my volt meter and touch it to either the negative input of the op-amp or the ground of the op-amp it seems to negate the effects of this interference and the fan speeds back up to the normal speed it would be at if I didn't have the current turned up. Another strange thing is I have three thermistors wired up for testing purposes, two on the heat sink with different epoxies and one on the transformer, if I have one of the ones on the heat sink plugged into the circuit and then just touch ONE of the wires of the thermistor on the transformer to the negative input of the op-amp the output voltage of the op-amp go up = fan speeds up. This seems very strange to me because I am not even making a complete circuit by just touching ONE of the wires of the thermistor to the negative input. As if all that wasn't strange enough, it seems to only be doing all these things when it is colder because last night when it was running for a while I didn't notice these issues.

This is all very strange, I would be interested to know why it is doing this as some sort of science experiment if somebody knows, but it is purely educational as I am just going to attach the thermistor to the heat sink. All the above problems go away once the thermistor is not touching the transformer.

It is picking up noise from the high voltage switching through the transformer. The noise is radiated kind of like radio so the path is kind of cryptic.

I figured it was something like that...thanks ronv. So, do you foresee any real problems with putting it on the heat sink? It seems to be working fine there right now.

Maybe that high frequency switching explains why when I turn the PSU on, my FM radio blinks out.

If possible add some sort of insulation between the heatsink and the themistor. I'm not sure how good the epoxy coating is on the themistor.

I wonder if the epoxy they will be glueing it down to the heat sink with will be enough added protection, I mean, as long as they don't push it down tight against the heat sink, I can tell them to just lay it in small pool of epoxy and let it dry that way. Then there will be added epoxy between the thermistor and heat sink.

Edit:
Or do you think I should wrap some heat shrink around the thermistor (that seems like it would decrease it's temperature sensing capability some).

Probably ok. Are the leads protected?

Yes, the leads are protected. I put them in electrical tape serves the purpose of making sure the leads don't short out against each other as well.

ronv said:

Here is an amplifier. I used the LM358 cause I know you have some. It will only go to within 1.5 volts of the positive supply so the maximum is 10.5 volts at about 50C. We can order one that is called rail to rail that will go all the way to 12 volts, but this will check things out.
So when you are testing it should go from about 3.8 volts at room temperature to 10.5 volts when hot.

Click to expand...

I replaced R3 with a 7.15K resistor and U1 with a rail to rail op-amp, left everything else the same , and now it is full on (12.61v) at 47amps and about 4.65v with no current (that is min speed). So I have the full range of adjustment. I am working on plotting some data today. It takes a while to do so because I have to let the PSU run for around 5-10 min in between data points to make sure and get the point at which the temperature becomes stable to see what it hovers at with each current level. I am taking readings at every 5 amps.
Only two question so far:
1. Should I put capacitors on the power input and ground of the rail-rail op-amps like we did when I built the dummy load to reduce noise? Or is that not needed on this one, overkill?
2. I don't think anything can be done about this, but I am going ask it anyway...at about 5.8v the fan starts to make a high pitch noise. It's not too loud and I bet a lot of people would not even be able to hear it, especially people that are a little bit older (since younger people tend to hear high pitches better). I can barely even hear it, but it's definitely there and it's only at 5.8v. From the little bit of knowledge I have gained in electronics over the last several months, I think it is the high frequency PWM in the fan motor that I am hearing only at that certain frequency. So, is there anything that can be done (doubt it), like somehow make the circuit skip over, or pass over quickly, the range of 5.7v to 5.9v? I mean if there is a simple way to do this, great, if not, then I don't think it is a deal breaker. The PSU fan will probably not be operating in this range that often anyway.

So, is there anything that can be done (doubt it), like somehow make the circuit skip over, or pass over quickly, the range of 5.7v to 5.9v?

Click to expand...

Not strictly hard. Theoretically it could involve a comparator circuit that would bias the fan 0.2v higher for fan voltages over 5.9v, then drop it 0.2v lower for below 5.9v. Skipping over the bad range automagically. It would need to be tuned to have only a small influence over the fan when high and low. Probably use a voltage divider configuration on the ground to change the reference.

Hm, that's very interesting, if it could be done with only a few more parts (i.e. still be a small circuit able to fit in the PSU, there is not much room in there). How would you do that? Everything I was reading about comparitor circuits say they compare two voltages and take the higher of the two...

Here is the data I got today on the fan speed controller (I graphed it in Excel in the attached file). The ambient temperature was about 70 degrees. I had 7.15K for R3 (see above post). Next I want to do the same test on another PSU. I am hoping the results are consistent within a small range, so that I can just mass produce these without having to tune each one individually...we will see.

OK, I can try and work out a circuit for the fan and the over voltage crowbar.

Which fan circuit are we working on? Is it what you will be installing in the for sale PSU's? If so, there is a spare OP-AMP/comparator not being used in it that we can use for doing the above mod. The LM358's and similar parts are dual amp circuits (as I'm sure you are aware of buy now.)

I need to know more about the exact parts you are using before I can do better, but here is a rough draft for what you will need to do, more or less...

View attachment 68198

R6 is about 100 Ohms
R7 is about 10k Ohms
R8 (unlabeled in the schematic) is any trimmer. The exact value is not critical.
The red box is the 8 pin plastic Op-Amp package, the lines from it are the power pins.

The only problem is that it is sort of a one way kind of circuit. That is to say... it will trip at the set point and raise the voltage by 0.2v like it should. But then because it has raised the voltage, and sees this raised voltage, it now has to lower down back through "the bad zone" before it will reset the right way. I'm not sure it can be done any better than this with out getting more complicated. But we will wait till ronv has a look at it. (He's better at these kind of things than I am)


Note: My idea of a crowbar circuit would be similar to this fan circuit mod, only it would take it's input from the PSU under test line in and send it's output to the control line of all the Op-Amps. Fairly clear cut and strait forward.

Thanks ()blivion, unfortunately I got done checking the second PSU, and it did not make the noise in the same spot, so I think each one is going to be different as far as that goes. Fortunately the range of temperature adjustment was very close, it was off by about .2v which is close enough, so I think I will be able to make one circuit that will fit all.

OK, so I just measured something that I found interesting. The fan inside the PSU does speed up as it heats up when I just ground pin 4 without installing the temperature control circuit we made. But it only speeds up to a max increase of about 2.41 volts. At min when cold the voltage across the fan was 4.67, then I turned the dummy load up to 50amps and let it heat up until it reached the max temperature and measured the voltage across the fan, it was 7.08v, an increase of 2.41. I let it run for a long time and it did not go above 7.08, so that is where it maxed out. I still think the temperature control circuit is needed because the fan needs 12v at max, not 7.08v. If no mod is done to short pin 4 to ground the fan runs on full speed. But, does this mean that there is some sort of temperature sensor already in there, or do you still think the increase in speed could be related to the high frequency switching and causing it to increase as temp rises?

I can't imagine they would put one in there and then not use it. What did you use for ground? Was it - sense tied to the other pins or real ground? Not sure what would make it do that with the input grounded.

I used the regular DC ground on the main 12v power rail. I didn't connect the sense pins to anything. It was behaving strangely, kind of like the temp control circuit did when I had it glued to the transformer and was getting the noise...the strange behavior was when I touched my leads to test the voltage across the fan the voltage would slowly rise on the fan and the fan would speed up, then when I take the leads off the voltage would drop back down to 7.08 and the fan would slow back down a little. When I first turn the load up the fan doesn't speed up immediately, but waites until it gets hot, but then if I turn the load bank down to 0 amps after it's hot, the fan speed immediately drops back down to min speed, then when I turn it back up the fan immediately speeds back up with current when it's already hot, so it's not just temperature controlling it, but it does have something to do with temperature because like I said, when it's cold it's always on min speed, even when I turn up the current but when it's hot the fan speeds up and down with current.

Without the temperature control circuit, just shorting pin 4 to pin 8, the fan speeds up to about 7.4 volts max. I ran it for about 20 min this way until the internal temperature of the PSU stopped rising. The transformer inside the PSU got up to a whopping 245 degrees Fahrenheit! Then I did the same test but with my temperature controlled circuit installed so the fan was going at max, the same transformer only got up to 165 degrees Fahrenheit, a difference of 80 degrees, I think that is a significant improvement. As a side note, the case was almost too hot to touch without the circuit installed, it was 120 degrees, with the circuit it was only 80 degrees.

The following is speculation based on observation so take it for what its worth.

- The power supplies have built in thermal protection which includes controlling the fan speed to suit the needs of the power supply. This has been observed and if you think about it, the server manufacturer would be crazy not to do it this way. These are field replaceable units, if the thermal control was in the server and it failed, then both (redundant) power supplies would be toast and so would their replacements.

- Pin 4 is there so the host server can call for additional cooling (for the server, not the PS).

- jocanon's fan test peaked at ~ 7v because the PS internal temp wasn't hot enough force it any higher (see below).

- The servers usually live in a comfortable 65-70 degree environment however they can withstand much higher temperatures. We've had computer room cooling failures where I work and the servers start shutting themselves down at about 95 degrees ambient.

If you're going to offer a temp 'assist' circuit then why not pick a temperature where the speedup starts and another temperature where the fan reaches full speed? Load by itself doesn't mean nearly as much as load AND ambient temperature. Just measure the temperature of the air comming out the back of the PS. I've measured several servers in operation and the temp comming out the back is 110-120 degrees and their fans are all running at their slowest speed.