Constant current load drifts

[Alex1]

Hi, I built the constant current load that dave made but there seems to be a problem where the current will slowly drift up at around 10mA a minute.

Any idea what could be causing this? I checked with a scope and nothing is oscillating (I have the 4.7nf cap from op amps output to the inverting input). It happens with all sorts of load currents.

Thanks.

[robrenz]

TCR temperature coeficient of resistance of the current sense resistor might be the culprit.

[Alex1]

TCR temperature coeficient of resistance of the current sense resistor might be the culprit.

Sorry but is that something to do with the temperature of the current sense resistor? If that is the case shouldn't the op amp adjust the gate voltage of the mosfet to compensate?

My current sense resistor is very oversized and doesn't get hot.

[SeanB]

Probably an offset from thermal EMF causing it. Post the circuit and a photo or two of the layout and as built. The cheap resistors often have a pretty terrible TC with even a moderate rise above ambient.

[Alex1]

Probably an offset from thermal EMF causing it. Post the circuit and a photo or two of the layout and as built. The cheap resistors often have a pretty terrible TC with even a moderate rise above ambient.

I have a fan for cooling. I just don't understand as many people have built this simple circuit and have had no issues like this. My sense resistor is a 0.1 ohm and it doesn't get very warm.

Will post some pics when I get a chance.

[codeboy2k]

The current sense resistor doesn't have to feel warm to be changing its resistance value.. any small current will heat it up, and this will change its value.

Your feedback loop assumes the resistor has a specific value.  It will be wrong if this value changes, and any little current through a current sense resistor certainly change its value, but how much depends on the tempco value.

If the tempco was 300ppm and the resistor rises just 15C (which can be hard to feel), it will be 300ppm/C x 5C = 4500ppm or 0.45% .. then your 0.1 ohm resistor is now at .10045 and if this is in a feedback loop with a gain of 10, so that the set voltage is 0 to 1V = 0 to 1A, then your setpoint at 0.5V for 500mA will be trying to maintain 0.05V across the 0.10045 ohm resistor, and this is I = V / R = 497 mA or 3mA off from your setpoint.

If your feedback loop uses 0-5V = 0-5A and you are setting it for 5A, so your set voltage is 5V, then the feedback loop will try to maintain a voltage of 0.5V across the sense resistor, and this will be causing 0.5 / 0.10045  = 4.977A which is 22mA off your setpoint.

So you see it's easy for the tempco of the sense resistor to cause such a drift... that said, it *should* stabilize, just be off by an error.

If it's not stabilizing after some time has passed, then something else is wrong. Check your set point voltage, is that drifting at all?

[Paul Price]

Hi Alex,
Where do you download a schematic of the Constant Current Load?
Maybe I can help if I can see the roadmap.
You say the current drifts...but from what current level setting?
Are you possibly exceeding the constant current compliance voltage of the  device being loaded?

[Harvs]

How are you measuring the current and what current is the happening at?

Reason I ask, handheld DMM's typically have relatively small current shunts due to their constrained size, so will drift with a decent input current.  On all the meters I've used, they have a positive tempco, and hence the reading increases when in fact it's not.

[Paul Price]

My Fluke meter does not vary because of heating while measuring current.

[Alex1]

Thanks for all of the help!

I have replaced the 0.1ohm current sense resistor and it now seems to be working, still drifts down a little bit at heavier (2 amps+) currents but I can't really complain as the two green sense resistors get very hot at these currents.

Never thought that the original resistor dissipating such little power could drift in value so much, but it was from maplins so go figure lol.

There was also a small grounding problem when the fan turned on. Its 90mA return current caused a small voltage shift to appear at the potentiometers, raising the set current by about 8mA.

I also managed to destroy a MOSFET, which went 130ohm between drain and gate (seems odd to me as drain to source is open circuit). I think it might have been rapid heating that caused it as I unintentionally turned the course potentiometer up very fast without knowing what current it would draw. 

This was before I had thermally calibrated the fan activation temperature too, so it wasn't running at the time.

Picture of the monstrosity (please don't flame me for the warped strip-board  )




The waveforms all look ok, no oscillations on the gate or feedback loop and I had it running for a few hours yesterday drawing 2 amps from a 12v source and it was ok.

The relay you can see next to the load input was originally a range selector between a 1 ohm and 0.1 ohm current sense resistor, but I am going to turn it into some sort of thermal cut out since the 0.1 ohm turned out to be unsuitable. The 1 ohm "stack" does the job fine besides getting hot, so I am going to solder on a small fin to help out a bit with the heat dissipation.

TL;DR Replacing the current sense resistor fixed it, and for its intended purpose of being able to quickly load test the DC-DC converters I use in many of my projects I think it will be adequate.

[SeanB]

The 5w resistors are not really stable, they are often made with steel wire for the lower values, and have quite a high TC.

[Alex1]

The 5w resistors are not really stable, they are often made with steel wire for the lower values, and have quite a high TC.

So would I be right in thinking that I need to look for resistors with a low ppm/°C rating?

[BravoV]

The 5w resistors are not really stable, they are often made with steel wire for the lower values, and have quite a high TC.

So would I be right in thinking that I need to look for resistors with a low ppm/°C rating?

Lower ppm will yield better result, suggesting to look for good quality low ohm resistor.

For example like the one below with TO-220 body 16 watter has the rating of 200 ppm/°C, although not the best, but definitely better than just common low ohm resistor like yours.

Also with TO-220 body for example, if attached to a heatsink, it will dissipate heat better, hence lower temperature = higher precision, rather than rely air convection alone to cool the resistors like yours which is difficult and slow to dissipate the accumulated heat.

[codeboy2k]

Get this one.  just 84 cents, compared to over $4 - $8 for the Caddock brand, depending on which one you actually pick.

.100 ohms, 1%, 3W, 20ppm/C

http://www.digikey.ca/product-detail/en/BR3FBR100/BR3FBR100-ND/2075010



If you put 2A through it, it will be about 400mW.  This bare metal element will likely rise about max 45C - 50C / Watt in open air.  So expect a maximum temperature rise of 0.400W * 50C = 20C above ambient, so it will be at 45C.  At 20ppm/C that's just 40 uOhms, or 0.04% of the initial 0.1 ohms...

The maximum you can put through this one (in open air) is about 5.4 Amps (at which point the metal could be about 150C or so, and at 20ppm/C that means a change of (150-25_Ta)*20ppm = 2.5mOhms (or 2.5% of the initial 0.1 ohms)

It will be even better if you have some air flowing over the resistive element.

So if you want to consistently put 5A through it, you need a fan or find a higher wattage version. But this element is very inexpensive and at 20ppm it's pretty good, especially if you actively cool it and keep it well under its max rating.

[thm_w]

Get this one.  just 84 cents, compared to over $4 - $8 for the Caddock brand, depending on which one you actually pick.

.100 ohms, 1%, 3W, 20ppm/C

Nice part!

[codeboy2k]

Get this one.  just 84 cents, compared to over $4 - $8 for the Caddock brand, depending on which one you actually pick.

.100 ohms, 1%, 3W, 20ppm/C

Nice part!

I hope it turns out to be. I am thinking about jumping on the active load bandwagon here on Dave's blog.. Or the lab PSU wagon... and I will use a part like this one or similar from it's series.

Like The Boy Scouts, Dave needs to give out badges to users  that have completed a specific project lol

[Alex1]

Get this one.  just 84 cents, compared to over $4 - $8 for the Caddock brand, depending on which one you actually pick.

.100 ohms, 1%, 3W, 20ppm/C

http://www.digikey.ca/product-detail/en/BR3FBR100/BR3FBR100-ND/2075010

Wow they look good, especially when you consider the price! I will definitely be ordering some.

[TorqueRanger]

Get this one.  just 84 cents, compared to over $4 - $8 for the Caddock brand, depending on which one you actually pick.

.100 ohms, 1%, 3W, 20ppm/C

http://www.digikey.ca/product-detail/en/BR3FBR100/BR3FBR100-ND/2075010

Wow they look good, especially when you consider the price! I will defiantly be ordering some.

Can you post up a schematic of your design ?

[Alex1]

Can you post up a schematic of your design ?

Sorry I had not seen your post until now. At the core it is just the same constant current load circuit as Daves except I used one of the spare op-amps as a temperate controller for the fan, the small trimmers were used so that I could set separate fan on and off temperatures (hysteresis).

The relay does a few things, firstly it acts as an over temperature cut off using a second spare op-amp. Secondly there is a small switch on the board that allows me to manually de-energize the relay and disconnected the power supply being loaded.

And it also implements a small turn on delay when the control section is first powered up, that way the control is always "ready" first.

The blue and yellow capacitors and resistors around the MOSFET just prevent unwanted oscillation caused by the gate capacitance of the MOSFET and circuit parasitics, without them I had a nice 750khz oscillator 

And of course some indicator LED's are needed to make it look nice in the dark

[David Hess]

Given the temperature coefficient of the sense resistor and the apparent effort put into your design, it may be worth lowing the sense voltage and using a more precise operational amplifier to balance the errors contributed by TCR, thermocouple effects, and offset.

I wonder how feasible first order compensation of the sense resistor TCR would be.