Dummy Load II

[jocanon]

I think it is this one that I have.

**broken link removed**

That one looks pretty nice.

 

[ronv]

Yea, I think it is like the little blurb in post 235.
The water pipe is almost an infinite heat sink but the actual chip inside the package is at 88C. Since the plastic has no heatsink it rises to the chip temperature.

 

[Mr RB]

I live by my IR themrometer, but of course ignore the laser and just put it right up to the component being tested. With many chips it will show the die temperature as the case plastic is quite transmissive to IR so the IR themometer will read much higher than a contact thermometer on the case.

As for accurcy, it meets the +/- 1 degree spec on the instruction sheet, generally reading less than 1'C error when I have tested for calibration. The only time it reads in error is with low emissive surfaces like shiny metal, you can just put a white paper sticker on the metal and read that, or some white out paint or black paint etc. Some anodised heatsinks are a bit tricky and I just stick a white sticker on them (you can even use electrical tape).

The lasers are pretty much useless as it is about an inch off axis from the sensor axis. Generally, get the IR thermometer real close <1 inch and get the highest reading, that will be the real temp of the "hot thing" in view.

 

[tvtech]

Great thread guys

I am learning here.

Regards,
tvtech

 

[jocanon]

Thanks! I am learning here too. Thanks to all the great contributors to the thread.

 

[ronv]

The more I think about it the more I think putting the temperature sensor on the plastic would be the best. It will probably respond the fastest and might also protect us from say trying to draw 4 amps at 50 volts as well as running with the water off. I'll reset it to match up with your temperature measurement at 4 amps plus a smidge.

 

[jocanon]

Sounds good to me.

 

[Dusey52]

The more I think about it the more I think putting the temperature sensor on the plastic would be the best. It will probably respond the fastest and might also protect us from say trying to draw 4 amps at 50 volts as well as running with the water off. I'll reset it to match up with your temperature measurement at 4 amps plus a smidge.

It might be wise to limit Vref based on the voltage of the DUT, so if Jeremy decides to test 4 up (50V), it would automatically cap Vref at a low enough voltage to keep the magic smoke from getting out. The same circuit could also allow higher currents at lower voltages (like a single supply).

 

[()blivion]

The more I think about it the more I think putting the temperature sensor on the plastic would be the best. It will probably respond the fastest and might also protect us from say trying to draw 4 amps at 50 volts as well as running with the water off. I'll reset it to match up with your temperature measurement at 4 amps plus a smidge.

Just to play devils advocate some more, and like I alluded to earlier, the only real problem I see with doing it this way is that the sensed temp is going to be dominated by the FET it's on. So if another FET decides to take more power than any others, unless it happens to be the one the sensor is on, the sensor will likely never see it. Then again, if it's on the pipe and the same thing happens, the change will be so small that it probably still won't see it. But keep in mind that the reverse is also true. If one decides to take less than full load, the indicated temp may com back lower than true.

So as long as we all agree that the above is not a problem with the actual implementation, then I guess I too will support putting the sensor on the FET back. It's probably more important that we see the temps the system gets to when it's properly operating. And the backs are probably closer to core temp, so it's a better indicator of how close we are to red line when heat is being shared evenly.

It might be wise to limit Vref based on the voltage of the DUT, so if Jeremy decides to test 4 up (50V), it would automatically cap Vref at a low enough voltage to keep the magic smoke from getting out. The same circuit could also allow higher currents at lower voltages (like a single supply).

Agreed. That's going to be one of the next things we are going to have to tackle to meet Jeremy's goals. I just want to make certain my core design, and ronv's implementation of it, is 100% sound before we move on. There seems to be some minor doubt, and I would like to squash it out completely before moving on. Underwear fully on before pants and so forth.

 

[ronv]

It might be wise to limit Vref based on the voltage of the DUT, so if Jeremy decides to test 4 up (50V), it would automatically cap Vref at a low enough voltage to keep the magic smoke from getting out. The same circuit could also allow higher currents at lower voltages (like a single supply).

I thought about that, but then Jeremy said he could just test them as singles or be real careful. It would be fairly easy to put in a step or two at say 29 and 39 volts if we could decide on what the different voltages might be. I did some work on an analog one that allowed more current as the voltage went down, but got a little concerned about all the possibilities for what might go wrong with that. Maybe it just needs an idea.

The original intent for the temperature sensor was to protect against not having the water on. If we expand that to 15 FETs and 15 sensors.... Then we still wouldn't know which one was different so we would need some indicators..... I think we agreed to add fuses so if it goes to haywire a fuse should blow. Maybe an easy way there as well?

 

[Dusey52]

I was thinking something like the attached file that simply limits Vref based on the Vdut. The higher Vdut goes the lower the clamp on Vref goes. It's not dialed in but you get the idea.

Detecting water flow could be a simple as adding an inexpensive flow sensor and disabling the system below a given flow threshold. Something like one of these: https://www.futurlec.com/Flow_Sensor.shtml

Just throwing out ideas...

 

[Mr RB]

If you use the Vout of the DUT as Vref, it will scale the closed loop current proportional to the DUT voltage, so the entire dummy load acts like a big resistor.

That may have some value as an "open loop" mode, re previous conversation.

 

[ronv]

Great minds Ducey..

This is the one I was looking at.

The flow sensor is cute. My only concern with it is that we would now need to add a circuit to hold everything off unless the pulses were close enough together. I don't want to get into a situation where we are still debugging a year from now.
Actually, the FET is just inside its safe operating area at 4 amps at 50 volts so the temperature sensor would probably protect it if the current was turned up, but it is not that much more stuff. I could go either way. Up to you guys.

 

[jocanon]

I kind of like having the option to flip it on and leave the current below 10 amps for a second or two to see if a PSU is working without turning on the water.

 

[ronv]

OK. So I guess it boils down to whether or not you want to have the added complexity of automatically turning down the current as the voltage goes up or if you want the 30 amp setting to be 30 amps no matter what the supply (DUT) voltage is. I'm not sure how it would act if you have a bad supply that say only put out 5 volts instead of 12, I guess it would increase the current above where it would be with the normal setting. Have you had any bad ones? How did they fail?

 

[()blivion]

If you use the Vout of the DUT as Vref, it will scale the closed loop current proportional to the DUT voltage, so the entire dummy load acts like a big resistor.

Correct me if I am wrong, but I think if we did it that way it would scale a higher voltage to be a higher current... wouldn't it? High reference voltage = high current.

I was just thinking we would use an Op-Amp based inverting amplifier, then the reference/current limit would go down as the DUT voltage went up. This will keep the max watts the exact same over any range as long as the correct scaling factor was used. It should be an obvious linear relationship, as Watts = Volts * Amps.

But as Ronv just pointed out, if the DUT goes nuts, drops to a really low voltage, the system might see that and say... MORE CURRENT NOW. So this could lead to problems. First it could blow fuses for no good reason. Second, it could make the system not want to start by holding the FETs short. It's a simple problem to solve though, we just need to make the output of the scaling circuit cap off at some reasonable limit, like with a zener or something. This way, the current could never get over say... ~50 Amps no matter how low the DUT voltage was. Should be easy as pie.

Flow sensor.
I more or less like the idea of a water flow sensor, so long as the threshold is really low. Though, for use as a simple safety and not an actual analytically measurement system, it is easy enough to just make a flow sensor. All you need to do is glue a thermal sensor to a 1/4 watt resistor, putting just enough regulated power through the resistor to make it heat up a good amount, then measure the temp via the sensor. As the flow increases, the cooling will increase and the temp will decrease. If it is installed at the water inlet, it will only be affected by the water temp. If it's installed at the exhaust end, then it can be made to take into account the system output temperature as an added bonus. With water we may have to cover it in something slightly insulating like plastic, but over all this method works well.

Edit: Also note, that the above is talking about something that is almost no different from simply slapping a thermal sensor on the pipe. If it gets over a certain temp, it's almost certain that the water is not flowing.

In an automobile mass air flow sensor, they have a small length of thin resistance wire that they put a regulated current through, then periodically measure the resistance of the wire to infer air flow. Or instead, as stated in the wiki article, they will measure the current going through it when it reaches thermal equilibrium. Fairly straightforward system.

 

[ronv]

Totally forgot about the fuses (). Good catch! You right though we could make it for 4 amps max at 12 volts. There might still be a little problem with a bad supply, but we could clamp it somehow.

 

[jocanon]

I have yet to have a DUT fail while connected to the dummy load. I did have one go POP! but that happened the instant I plugged in AC mains before I even powered it up and connected it to anything.

 

[Dusey52]

Correct me if I am wrong, but I think if we did it that way it would scale a higher voltage to be a higher current... wouldn't it? High reference voltage = high current.

I was just thinking we would use an Op-Amp based inverting amplifier, then the reference/current limit would go down as the DUT voltage went up. This will keep the max watts the exact same over any range as long as the correct scaling factor was used. It should be an obvious linear relationship, as Watts = Volts * Amps.

But as Ronv just pointed out, if the DUT goes nuts, drops to a really low voltage, the system might see that and say... MORE CURRENT NOW. So this could lead to problems. First it could blow fuses for no good reason. Second, it could make the system not want to start by holding the FETs short. It's a simple problem to solve though, we just need to make the output of the scaling circuit cap off at some reasonable limit, like with a zener or something. This way, the current could never get over say... ~50 Amps no matter how low the DUT voltage was. Should be easy as pie.

Flow sensor.
I more or less like the idea of a water flow sensor, so long as the threshold is really low. Though, for use as a simple safety and not an actual analytically measurement system, it is easy enough to just make a flow sensor. All you need to do is glue a thermal sensor to a 1/4 watt resistor, putting just enough regulated power through the resistor to make it heat up a good amount, then measure the temp via the sensor. As the flow increases, the cooling will increase and the temp will decrease. If it is installed at the water inlet, it will only be affected by the water temp. If it's installed at the exhaust end, then it can be made to take into account the system output temperature as an added bonus. With water we may have to cover it in something slightly insulating like plastic, but over all this method works well.

Edit: Also note, that the above is talking about something that is almost no different from simply slapping a thermal sensor on the pipe. If it gets over a certain temp, it's almost certain that the water is not flowing.

In an automobile mass air flow sensor, they have a small length of thin resistance wire that they put a regulated current through, then periodically measure the resistance of the wire to infer air flow. Or instead, as stated in the wiki article, they will measure the current going through it when it reaches thermal equilibrium. Fairly straightforward system.

The circuit in my post does invert Vdut and ramps the Vref limit down as Vdut goes up. Note, it just limits Vref if it's too high for a given Vdut, it never increases Vref so the scenario where Vdut suddenly drops would not result in an increase in current. The other circuit scales Vref in response to changes in Vdut. One way lowers the 'ceiling', only changing the setting if it's above limits, the other way always scales. Minor difference I guess but the first method preserves the amp setting unless it's in the 'danger zone', the second method changes the setting anytime Vdut changes which seems less desirable to me.

I like the home made flow sensor idea and I think it would be more effective at saving the system because it will respond much quicker than a sensor attached to a large piece of metal.

 

[()blivion]

In case you missed it, I was responding to Mr RB in post #272. I hadn't even examined you and ronv's circuit yet. But at a glance I suspected that is about what they did. And yes, your way of "dropping the ceiling" is better, even if similar to mine. We could call it "semantics", then again, we could call it "details" too. Either way, as I figure it, so long as the control voltage can never get over a certain presumed safe limit, then it should work with both methods.

As for the flow sensor, I believe I first saw it in a Microchip application note actually. Something about fuzzy logic fan control or whatever... don't exactly recall the details.