Resistance drift vs. Temperature

[M84AB1]

Hi guys,

so I am building a constant current dummy load, similar to the one dave did a video about some time back. I am using 10 x 10ohm 5W Metal Oxide Film Resistors to make a 1ohm 50W resistor.

Everything works fine but the issue that I am having is that as the resistors start to heat up (current drawn 2+ Amps), I notice that the current starts to go up (by about 1mA every few seconds), although the voltage across the resistors does not change. This implies that the resistance is dropping. To further prove my point, If i take a 80mm fan and start blowing on the resistors as they start to drift, the current drift starts to reverse but it is still not good enough to completely stop the drift, only slow it down significantly.

Am I doing something wrong or is this to be expected?

This is the original shcematics




Here I am trying to stabilize the current by feeding the amplified current voltage back to the inverting input of the op-amp. The aim is to have the opamp reduce the output voltage so to match the input (non inverting input) with the inverting input. This causes MASSIVE oscillations. And I have tried placing a RC low pass filter between the differential amp output and inverting input of the op-amp.

[jahonen]

Try limiting the bandwidth of your error amplifier. You can do this by putting a series resistor between your current measurement amplifier and error amplifier and then a capacitor from error amplifier output to its inverting input. Fiddle with the values until it is stable enough.

Regards,
Janne

[Dr. Frank]

Hello,
both effects are to be expected.

In the first case, the resistors are loaded with 2-25W Watts  (I*I*1Ohm), 200mW - 2.5W for every single one.

You do not specify the type , case size and max. power of the resistors, neither the OpAmp type.

If you write Metal Oxide / 5W, then I assume, you are using cheap thick film power resistors, which have a very high T.C. >= 100ppm
Look for wirewound precision power resistors instead.

Better, you reduce the power dissipation by reducing the resistance and the voltage over them by a factor of 10 at least.

There are either power resistors with cooling case, or precision resistors, as thin film, or wire wound with low T.C. of < 10ppm/K or <3ppm/K, respectively.
For first try, ordinary 50ppm/K thin film resistors may be good enough.

 

The 2nd schematic does not work, will oscillate, because the 100x OpAmp is lacking feedback.
Please, compare your circuit with standard inverting/noninverting circuitry of OpAmps.

In your configuration you need a special "instrumentation amplifier", instead of that ordinary OpAmp.

But you can at first discard the first buffer OpAmp, because the 2nd OpAmp also has low input current on the '+' input.

Then, I would simply use one low offest type or Chopper OpAmp (e.g. OP01, 7650) only.

Divide the input voltage by 10, i.e. 0 .. 0.5V, then directly on the chopper '+', and then use 10 * 1 Ohm metal film (TC 50) or better resistors.

Also, you have to design a 4 point Kelvin arrangement on your PCB, for those 10 resistors, i.e. 2 lines for the current, from the FET to GND, and two lines from the resistors to the '-'  of the Chopper, and the GND from the potentiometer/reference.

Frank

[Marco]

There's 2 opamps and 10 resistors too many in the bottom schematic, the 0.01 resistor is the shunt ... the 10 1 Ohm resistors are just a non functional heating element.

Why do you say A=100 for the bottom opamp? Did you just leave off the feedback network? Or is it a fixed gain amplifier?

Any way, if you want to use a smaller shunt make sure you are using an opamp with a sufficiently low offset voltage ... and as Dave explained in the video you only really need one, he used two because he had a dual on hand. So just go back to the original schematic, drop the buffer opamp and put a resistor above the pot to increase the sensitivity (with the smaller shunt the max voltage across the shunt at usable currents will be smaller).

[M84AB1]

Hi guys,

let me start by thanking you all for the pretty awesome replies!!! Great community we have here.

In the first case, the resistors are loaded with 2-25W Watts  (I*I*1Ohm), 200mW - 2.5W for every single one.

Apologies for not making myself clear, those are 10 x 10 Ohm @ 5W 5% each. That would be the equivalent of 1 ohm @ 50W. At about 2.5A, that is 2.5 / 10 = .25A, therefore power dissipated by each resistor should be .25 x .25 x 10 = .625W. That is 1/8th of their rated power and yet they heat up significantly if left without cooling.

If you write Metal Oxide / 5W, then I assume, you are using cheap thick film power resistors, which have a very high T.C. >= 100ppm
Look for wirewound precision power resistors instead.

There are either power resistors with cooling case, or precision resistors, as thin film, or wire wound with low T.C. of < 10ppm/K or <3ppm/K, respectively.
For first try, ordinary 50ppm/K thin film resistors may be good enough.

Yes, I am using the cheap ones from ebay. I paid about $3 for 10 of them a few months back.
http://www.ebay.com.au/itm/10-x-5W-700V-1-2-ohm-Metal-Oxide-Film-Resistors-5-Watt-/160937209997?pt=AU_B_I_Electrical_Test_Equipment&hash=item25789af08d

Can you provide a link to wirewound precision resistors? I bout a 1ohm 100W tone aluminum case resistor for $3.5 just to see if that would make a difference.
http://www.ebay.com.au/itm/Gold-Tone-Aluminum-Case-Resistors-100W-Power-1-Ohm-5-/150827235981?pt=AU_B_I_Electrical_Test_Equipment&hash=item231e00fa8d

The 2nd schematic does not work, will oscillate, because the 100x OpAmp is lacking feedback.
Please, compare your circuit with standard inverting/noninverting circuitry of OpAmps.

In your configuration you need a special "instrumentation amplifier", instead of that ordinary OpAmp.

Sorry, I did not provide the full detail schematics first time around. I assumed it would be understood that if there is no feedback circuitry that a differential/instrumentation amplifier is ued. I am using a INA122 instrumentation amplifier. Gain is set to 100 via a prevision trimpot.

Divide the input voltage by 10, i.e. 0 .. 0.5V, then directly on the chopper '+', and then use 10 * 1 Ohm metal film (TC 50) or better resistors.

Yeah I see what you mean. It makes sense. I will simply use the a by 10 divider between the two buffer amps. I have 10 x 1ohm 3W Metal Film Resistors 1% tolerance.
10 x 1ohm 3w = 0.1ohm 30W

So to put things in perspective, Say i want to draw/load a total of 7 amps
- 7A / 10 resistors = 0.7A per resistor
- 0.7 * 0.7 * 1 = 0.49W dissipation., That is 1/6 of its max W rating. I hope that should be enough to keep it cool.

Try limiting the bandwidth of your error amplifier. You can do this by putting a series resistor between your current measurement amplifier and error amplifier and then a capacitor from error amplifier output to its inverting input. Fiddle with the values until it is stable enough.

Regards,
Janne

That did the trick mate. It is FULLY stable now, oscillations wise that is. The cheap resistors that I am using are still drifting, not by much but still enough to annoy me.

There's 2 opamps and 10 resistors too many in the bottom schematic, the 0.01 resistor is the shunt ... the 10 1 Ohm resistors are just a non functional heating element.

Why do you say A=100 for the bottom opamp? Did you just leave off the feedback network? Or is it a fixed gain amplifier?

Any way, if you want to use a smaller shunt make sure you are using an opamp with a sufficiently low offset voltage ... and as Dave explained in the video you only really need one, he used two because he had a dual on hand. So just go back to the original schematic, drop the buffer opamp and put a resistor above the pot to increase the sensitivity (with the smaller shunt the max voltage across the shunt at usable currents will be smaller).

INA122 which I am using has 0.15mV offset, at 100gain i get about 15mV offset. Not zero but I will eliminate the offset in the microcontroller software.




This is what I will try tomorrow. I have a good feeling it will work ok.

[Marco]

It can work ... but :

- Those 10 resistors are there to get hot and do nothing useful.
- You use an instrumentation amplifier to amplify the voltage over a ground referenced shunt.
- You needlessly use a separate amplifier in the feedback path.

You're creating a Frankenstein monstrosity :/