Dummy Load II

[Dusey52]

ronv's files, which include the LM358 are back on post #135

 

[ronv]

Dusey, Did they work for you?

For an oldie the 358 is not to bad. The output goes to withing 20mv of ground worst case 5 mv typical. I think it does even better with the 4.7k to ground. So we might have a few tenths of an amp with the pot at 0. When we decide how many watts to make the new one maybe a little larger sense resistor would be better since we will use a higher power resistor and we can drop the maximum current per FET by a little to add some more margin all around.

 

[Dusey52]

Yes, the model files work great, thanks!

 

[()blivion]

We're using the LM358AN, ...

OK, good. Thanks.

ronv's files, which include the LM358 are back on post #135

I have them, thanks. He has more than one Op-Amp in the folder. So I didn't know which exactly to go with.

For an oldie the 358 is not to bad.

Yeah, it looks pretty good. I'm not complaining.

When we decide how many watts to make the new one maybe a little larger sense resistor would be better since we will use a higher power resistor and we can drop the maximum current per FET by a little to add some more margin all around.

As you no doubt already know, given a constant current the power/heat made in the resistor is going to be directly proportionate to the resistance, regardless of the wattage it can handle. So with the same resistance, it will get just as hot, even if we make it a higher wattage unit. So, if the objective is to make sure the resistor survives, then there is no problem. But if what we are really after is lowering heat dissipation in parts other than the FET's, then increasing the part rated wattage will have almost no effect. The only effect it could have is that the heat will be spread over a slightly larger area, so the peek temperature will be ever so slightly lower.

I assume we are not trying to shift around where the heat is dissipated? Just make sure the resistor can take what it's making? We *DID* originally decide that it would be best to lose some of the heat in the sense resistor. And I do still more or less agree with that plan. But jocanon was expressing concern with the resistors getting too hot, and we can't cool them as well. So I just want to make sure we are addressing this as well.

More "channels" should fix this too of course.

 

[Mr RB]

Mr. RB,
OK I finally see what you are saying. It is stable with respect to the 24 volts, just not with respect to the reference. The problem we have in this case is the power sequence. The 12 volts is already on when the 24 volt supply comes up. With the fast loop the current pulse is pretty short, but with the big integrator it gets pretty big. I don't know the exact power up time of the 24 volts supply. If it's real slow it may not matter so much.
I have played around a bit with the loop and can make it better, but right now it is 12 db down when it gets to 180 degrees.

That may not be a problem? Generally you do the intial power up with the dummy load turned to zero current (or pot at whatever the minimum current is). Then turn up the pot to increase load once the test PSU is running. Also on my dummy load I use a "load on/off" switch, which is very handy so likewise you power everything up first and then can turn the current on/off as needed. That's great for testing overload conditions for very short times, like one second.

Also the integrator does not need to be THAT big, it just needs to have a response time significantly slower than the control loop response and slower than any regular noise ripple etc on the feedback signal.

 

[jocanon]

()blivion, I am only concerned that the current sense resistors survive. If they can handle the heat that is all I am concerned with. When I checked the temperature they were getting very close to the max rating of 275c so it had me a little worried that we could be over doing it. So I think if the wattage is increased to 10 watts but the temperature stays the same, then I assume the rated temperature must be higher on the 10 watters (i.e. they can handle more heat so it won't be an issue anymore even if they are the same temperature).

Mr RB, you are right in saying that on power up the POT is turned to min. I think once I forgot and started it with the POT turn to max though, lucking it didn't damage anything as far as I can tell.

 

[ronv]

J,
Did you get a chance to compare the filtered meter with your regular voltmeter?

 

[jocanon]

It was too erratic to compare under load, but with no load both the little VM with no resistor or cap and the regular VM were the exact same. My watt meter was about .8 volts higher. When under load the watt meter was still about .8v higher than the regular VM (they must not be calibrated the same - these are all cheap VMs and watt meter).

 

[ronv]

How much are they bouncing around? It might be worth trying it on a battery charger to see if it is about the same. Is it worse under high load vs low load. Try measureing the 12 volt supply. You said it was stable with the cap and resistor but low. Was that against the watt meter or the other voltmeter.

 

[jocanon]

It was low against the other watt meter. It's bouncing around a volt or two sometimes.

 

[ronv]

So the little meter with the filter is steady but a bit lower than the watt meter, but probably the same as the voltmeter if it had the filter. Maybe we can try the other tests as a check.
1- low load vs higher load
2- 12 volts
3- Driving a battery charger hard (if you have a low one).

 

[Mr RB]

... I am only concerned that the current sense resistors survive. If they can handle the heat that is all I am concerned with. When I checked the temperature they were getting very close to the max rating of 275c so it had me a little worried that we could be over doing it. So I think if the wattage is increased to 10 watts but the temperature stays the same, then I assume the rated temperature must be higher on the 10 watters...
...

I think running normal style power resistors over about 80'C is unacceptable from a reliability viewpoint and approaching 150'C the solder will fail and they will fall out of the connection.

It's a fairly standard problem measuring currents at 50 amps, why not just use a 50A (or better still) a 100A current shunt?

**broken link removed**

They can be found on ebay for about $10. Also small 40mm fans can be found for a few dollars and will help with resistors or shunts, they can remove a considerable amount of heat and the 12v fans consume a few mA to run.

 

[jocanon]

Hm, maybe my temperature readings were off...I mean, I haven't had any issues with solder melting off.

 

[ronv]

How about this.
Nice round number like 13 FETs.
.2 ohm 10 watt resistors like this.
https://www.mouser.com/ds/2/303/20_series-4544.pdf
The little current meter with shunt to measure total current.

J, The more I think about it the more I think your big supplies may be jumping around. Remember way back when we blew the 26 volt TVS's?

So if you get around to more experiments you might try the jumper on the big supplies to set the voltage at 24 to see if it settles down. I know you like the higher voltage -- what was it 24.6? But I'm not sure how the power supply works so it may be worth a check.
This would give you some room to run the higher voltage supplies - Not sure about 1700 watts but maybe 1500.
We are kind of getting in a box where we will have to use some fancier parts.

 

[()blivion]

Hey guys.

Apparently my internet connection went down for a little while, I *MAY* lose it again at any time, I don't really know. So don't rely on me for anything critical. And if I disappear again, that's probably why... OK?

It's a fairly standard problem measuring currents at 50 amps, why not just use a 50A (or better still) a 100A current shunt?

**broken link removed**

They can be found on ebay for about $10.

Some time during the early development for the first dummy load, I had envisioned all the FETs in parallel, using a single large shunt resistor similar to the above as a main current sense. However, someone suggested that all the FETs would not share current evenly in this configuration if the threshold voltage varied too much, and that they almost certainly would have a significant variance. I was at first reluctant to agree and wanted to at least test it my way with real hardware. Knowing FET's have a natural positive temperature coefficient I figured they would self correct to a degree. But before I went and did any experimenting, someone else said they had actually tried it and this exact problem destroyed their system within seconds of the first problem. So... without even doing any testing I did some careful research, and indeed, even a small variance in threshold would cause a significant current imbalance. An imbalance too large for the positive temperature coefficient to fix. And as critical as it is, most parts suggest a surprisingly wide spread on this parameter in the datasheet. Up to 2 volts is not entirely uncommon, which would be a HUGE difference in current. Finally, if one FET goes open circuit with a single current sense point, the circuit will see the drop in current, and automatically turn it up system wide. This makes it so all the FETs suddenly have to take the current of the missing one, stressing them even more.

Long story short, we unanimously came to the conclusion that closed loops for each FET is about the only way to virtually guaranty that all FETs are properly pulling their own weight, with the added benefit that if one goes open circuit, the rest won't draw any more current unless the user manually turns it up him or her self.

We had considered BJTs to side step the issue, but FETs basically won the coin toss.

I am only concerned that the current sense resistors survive. If they can handle the heat that is all I am concerned with. When I checked the temperature they were getting very close to the max rating of 275c so it had me a little worried that we could be over doing it. So I think if the wattage is increased to 10 watts but the temperature stays the same, then I assume the rated temperature must be higher on the 10 watters

I think running normal style power resistors over about 80'C is unacceptable from a reliability viewpoint and approaching 150'C the solder will fail and they will fall out of the connection.

How about this.
Nice round number like 13 FETs.
.2 ohm 10 watt resistors like this.
https://www.mouser.com/ds/2/303/20_series-4544.pdf
The little current meter with shunt to measure total current.

My two cents.
If one increases the resistance of a part while keeping the current the same, the percentage of watts "generated" in that part will also increase in accordance with I[SUP]2[/SUP]R. This also means that given the same heat sinking, that part will undoubtedly get hotter, regardless of it's rated wattage.

With this in mind, if the goal is to reduce the temperature, then we should do one or more of these three things.
(1) Reduce the resistance. (2) Reduce current. (3) Increase cooling.
Note that adding more channels with the same unit resistance will do all of the above, however the effect is directly related to the number of channels already existing. Meaning we will need to add lots to have any real effect.

If we go to a whopping 20 channels, and 20 of the 10 Watt 200 milliohm resistors, AND good cooling fans, we will be beyond safe on all fronts. And the total resistance will be a nice clean 10 milliohms. (not that this really changes anything, I don't think.) Also, 18 channels + 180 milliohms, or 16+160mΩ, or 14+140mΩ are all acceptable arrangement's. Could even do odd numbers too, but I'm very partial to symmetry myself.

 

[Dusey52]

Something like this for the sense resistor solves the cooling issue (assuming it can be soldered to the water pipe along with the Mosfets).



They're a little more expensive and some have a ceramic back but a suitable one should be available for < $3. The money saved on fans may even make up the cost difference.

Here's one example: https://www.digikey.com/product-detail/en/PF2203-0R2F1/PF2203-0.200-ND/2447810

 

[ronv]

Hey guys.

My two cents.
If one increases the resistance of a part while keeping the current the same, the percentage of watts "generated" in that part will also increase in accordance with I[SUP]2[/SUP]R. This also means that given the same heat sinking, that part will undoubtedly get hotter, regardless of it's rated wattage.

With this in mind, if the goal is to reduce the temperature, then we should do one or more of these three things.
(1) Reduce the resistance. (2) Reduce current. (3) Increase cooling.
Note that adding more channels with the same unit resistance will do all of the above, however the effect is directly related to the number of channels already existing. Meaning we will need to add lots to have any real effect.

If we go to a whopping 20 channels, and 20 of the 10 Watt 200 milliohm resistors, AND good cooling fans, we will be beyond safe on all fronts. And the total resistance will be a nice clean 10 milliohms. (not that this really changes anything, I don't think.) Also, 18 channels + 180 milliohms, or 16+160mΩ, or 14+140mΩ are all acceptable arrangement's. Could even do odd numbers too, but I'm very partial to symmetry myself.

I like symmetry to, don't know why I just do.

I agree with everything except when you use a bigger resistor it has more surface area so it doesn't get as hot. I think the 10 watt ones would only be about 65C if I remember right. Still toasty, but way cooler then the 5 watt ones.

The problem with going to many channels and low resistance is that the sense voltage gets pretty small so the offset in the op amp starts to be a problem especially at low currents. Right now the op amp is only 4 mv so with .1 ohms we would have an potential error of 40 ma in each channel so our minimum current might be .4 amps. With 20 twice as much, plus the sense voltage would be down in the mud. Thats why I made it a little bigger. We could probably find a better op amp I suppose.

The other danger I see is with J's desire to have a higher voltage we can't protect against the pot being left at a high current setting from testing 24 volt supplies then switching to 50 volt supplies. Right now it can't be set to more than about 6 amps, but 6 amps at 50 volts is a lot more power than at 24 volts. So if we need to do this we may need some type of over current protection.

Dusey, I like those to, but since we are still using them as fuses I'm not sure about unsoldering them in case of an oops. Lets see what J thinks.

 

[()blivion]

Something like this for the sense resistor solves the cooling issue (assuming it can be soldered to the water pipe along with the Mosfets).

According to the datasheet, the electronics are insulated from the tab, so we could in fact solder them to the pipe with great effect. I would rather use these, but there is the problem that ronv pointed out that we were planing on using the sense resistors as fuses. Also, there is the problem of having to solder more things to the pipe in the first place, taking up more area on it. I can't help but think that things would start to get crowded, but maybe not. As far as performance and precision though, those units are the way to go. Resistors will lose precision as they change temperature, and the normal ones are harder to keep stable.

If there is plenty of area to solder on 30~40 FET like things to the pipe, then I too would suggest we use these and just install proper fuses. Making the sense resistors dual function is great and all, but if we can get better performance with a different config, I think that would be better. Auto fuses are dirt cheep and can be socketed. And it would be best if all our heat was going to the one heat sink, where we can keep an eye on it.

when you use a bigger resistor it has more surface area so it doesn't get as hot. I think the 10 watt ones would only be about 65C if I remember right. Still toasty, but way cooler then the 5 watt ones.

It's true enough, bigger is better and higher wattage resistors are generally bigger. The old 5 watt resistors have a surface area of ~1201mm, and the new 10 watt units have a surface area of ~3593mm. So, going on surface area alone (i.e. assuming the same thermal transfer per unit area) they should have about 3X the cooling. However, as far as I know, the wattage is the max rating that the package can safely handle when properly cooled, not an indicator of how well it cools. And note that even a higher rated resistor will over heat if the watts being made are not being removed just as fast. Even with a 100 Watt resistor at only one watt power, if it were perfectly insulated it would still easily overheat.

Case in point, the Mk1 has ten, 5 watt, 0.13Ω, resistors. At 50 Amps total current each resistor will be dissipating 3.25 watts. Even so, they were still getting quite hot despite not being at the max wattage. This clearly indicates that the package max wattage alone can not be relied upon to indicate the final temperature.

We really won't know how well they cool when compared to the others until we try it though.

The problem with going to many channels and low resistance is that the sense voltage gets pretty small so the offset in the op amp starts to be a problem especially at low currents. Right now the op amp is only 4 mv so with .1 ohms we would have an potential error of 40 ma in each channel so our minimum current might be .4 amps. With 20 twice as much, plus the sense voltage would be down in the mud. Thats why I made it a little bigger. We could probably find a better op amp I suppose.

IMHO, the solution to all this is to use better Op-Amps, and use a single current sense resistor for the indicated current, like what Mr RB has shown us. This way the readout will show us the true current, but the dial will control the current loops still. Also, the total current measurement will be more accurate.

That, or as Dusey52 suggested, just use the better resistors. This should give us more accurate current-to-voltage sensing with out a single main sensor. Then it's just a matter of making sure we are measuring it with enough precision, and with enough immunity to noise. Again, better Op-Amps will help with this.

The other danger I see is with J's desire to have a higher voltage we can't protect against the pot being left at a high current setting from testing 24 volt supplies then switching to 50 volt supplies. Right now it can't be set to more than about 6 amps, but 6 amps at 50 volts is a lot more power than at 24 volts. So if we need to do this we may need some type of over current protection.

I think we are going to need higher voltage FET's to do higher voltage system also, and the system should have overcurrent protection just on principle IMO. It's quite a powerful circuit after all. And no matter what, if it is intended to be a wide dynamic range dummy load, there is going to need to be some protection to prevent a miss configuration. The only way to make it intrinsically safe is to make it survive max voltage X max current, which is getting less and less feasible.

 

[Mr RB]

...
The problem with going to many channels and low resistance is that the sense voltage gets pretty small so the offset in the op amp starts to be a problem especially at low currents. Right now the op amp is only 4 mv so with .1 ohms we would have an potential error of 40 ma in each channel so our minimum current might be .4 amps. With 20 twice as much, plus the sense voltage would be down in the mud. Thats why I made it a little bigger. We could probably find a better op amp I suppose.
...

I've done closed loop power control with cheap opamps and only 200mV or so for max volts on the current sense resistor. They were perfectly stable especially with plenty of integration. The offset voltage is morte an issue in instrumentation (ie open loop precision measuring).

In a closed loop system you set the current with the metering, then for whatever the opamp offset is, it was already nulled out and won't change much because the entire closed loop will remain at the same current and sense voltages.

The big issue is resistance drift of your current sense resistors, which might give a couple percent error as the resistors get hot. You can counter that somewhat by using larger resistors run under half their power rating, and by a "tweak" adjustment once it is hot. That will always have to be done regardless how "perfect" the opamp is.

If you used 0.1 ohm resistors times 8 on 50A max that is 6.25A through each resistor, max voltage of 625mV, that is enough voltage in a current control loop to be extremely stable. Max dissipation is 3.9W each resistor, using eight 10W resistors and a small fan that would be reliable.

 

[ronv]

Mr. RB, The only problem I see with the offset is when you have ask for zero with the pot. At that point you can't dial it out. I'm not sure its a problem, but () was a little concerned in one of his simulations.

The plan with this one is to measure total current with a current shunt like you have shown. Which brings us to another problem:

J, I think the ones you bought need their own dedicated power supply. In other words it's ground can't be tied to the system ground. Remind me what are we running off the 12 volt supply. If I remember right it is just fans and our logic. Maybe we can free it from the logic and go back to powering from the DUT.

The first load had 10 fets and resistors in parallel. The number went down when Jeremy accidentally shorted the input to the FET to +12. This of course turned it full on leading to the destruction of the resistor and the FET. I'm not sure what happened to the second one, but probably lost trying to repair the first one. Unfortunatly the over current detect in the supply didn't shut it down fast enough.

So they were really only running 5 amps each at .13 ohms or 3.25 watts. But once in a while Jeremy wanted to trip the over current in his big supplies so it ran as high as 6 amps. Now they really are way to hot.