My first thought was like you, that copper would have a horrid tempco, but thinking about it a little more, it gets interesting. The deflection of the meter is proportional to the current in the windings, which will be proportional to the resistance of the shunt and inversely proportional to the resistance of the coil. So, you have a point, the tempco should drop out, neglecting any self-heating differences.
A nichrome coil makes no sense. Resistance in the coil would make it less efficient. These movements are all about efficiency (turning the smallest possible current into the desired deflection). Resistance just reduces accuracy (draws more power from the DUT).
Try it. There are two things to measure: the meter's input resistance tempco, and the meters's current-deflection tempco. Or just its deflection vs temp with a constant-voltage input.
The shunt may self-heat from the load current, which would complicate matters if you used a copper shunt. Current shunts are usually manganin. If you want a copper shunt, you'd probably have to make your own.
How accurate do you need this to be? Copper is about 0.4% per degree C.
John Larkin Highland Technology, Inc
jlarkin at highlandtechnology dot com
Copper and nichrome is (very similar to) a copper/chromel thermocouple. So, you can expect a bit of a Seebeck coefficient, about 20 uV/K. It's not just resistance that gives you a thermal effect, but thermal gradients (you wouldn't want a nichrome shunt to be mounted vertically, self-heating would warm the top more than the bottom).
The meter is a bit hard to judge; the magnetic field, the coil dimensions, the hairspring return force, and the wiring resistance ALL have some thermal response. It'd be easy to make the meter with a series thermistor if copper-versus-nichrome-shunt TC of resistivity needed nulling.
The meter drive of 100uA +/- copper variation is a minute fraction of your 10A full scale measured value. Go figure the error due to meter being in parallel to the shunt. Think of the shunt as a very low source impedance voltage source, thus the meter resistance is close to irrelevant.
No, if the meter resistance changes its part of the current and therefor its deflection will change. Not withstanding the above the real question remains. Just how accurate do you think you can measure with an uncertified meter. :-?
OK - from the feedback so far is looks like I am right i.e. a shunt with a "zero" TC with a copper-wound moving coil meter is going to create a 0.4% degC error just due to the ambient temp variation.
Actually I think 4 % variation over a 10C range is a lot! One can read even a cheap meter to 1% and this meter has a mirror scale.
It's a little project somebody is building. Funnily enough, in the
1970s when at college, I built a similar but much more sophisticated version, which had the meter in the feedback loop of an op-amp based constant current source so the meter's TC was irrelevant. Remember an AD504 op-amp? ;)
But the chap's point about shunt self heating is also valid and would probably make a bigger error, because a 0.01 ohm shunt at 10A will dissipate 1W which is a lot for a shunt of any realistic size. As a rough order, 1 square foot of 16swg ally is 1degC/W.
So the best way to do this is to use a zero TC shunt (I have a 0.01 ohm precision resistor handy) and then separately temperature compensate the meter itself with a thermistor os something...
I worked for many years at a company (Edgerly Instrument Labs, or EIL*),
where they modified and calibrated meters of all types, and I soon learned that meters with a full scale current larger than the meter movement itself contained a shunt made of manganin or constantan wire or bus bar, and the meter movement had a "swamping" resistor in series that made it a voltmeter with FS of 50 mV or 100 mV.
Analog meters designed for measuring DC current were usually about 4 mA and
10 mV (2.5 ohms), and a 10 ohm series resistor made it 50 mV. Thus the tempco was "swamped" by a factor of 5 and good enough for standard 2% and 1% meters. Also, the coils of these meters were often wound with a copper alloy that had a lower tempco than pure copper. 100 uA meters usually are about
200-500 ohms and rarely as much as 1000, so they could also be used with a swamping resistor for 100 mV shunts and sometimes also for 50 mV shunts. For self-contained meters less than 1A FS, shunts could be more than 100 mV.
I also learned that blank scale meters from the factory had magnets that
were fully charged so that the sensitivity was about 10% higher than their nominal rating. So you only needed to use a standard resistor value (based on ohms per volt for voltmeters), and then the meters were demagnetized with a pulsating magnetic field until they read correctly. A fully charged meter movement would gradually lose some magnetism and sensitivity, and even after demagnetizing they would drift as much as 1%, so for highest accuracy they would be allowed to "age" for a few hours and then finally calibrated exactly.
We made some analog meters that were specified as 1/4%, and they had to be calibrated very carefully. They even needed hand drawn scales because the movements were not exactly linear - probably slight variations in the magnetic gap. I also found that the linearity could be tweaked slightly by varying the angle of the movement in the demagnetizer coil.
IIRC, these meters also had a switch for several current ranges, and they used a precision copper resistor as part of a temperature compensating network.
These were special meters used for a military application, and sold for several hundred dollars (1974 dollars). They probably only needed to be
1%, but each middleman in the chain of command probably demanded twice the accuracy and so we had to take extra time and effort to get them to an accuracy much better than they were originally designed for. But we still made a nice profit, since the blank meters were only maybe $10 or so.
On this project, I can make the shunt voltage 300mV rather than 100mV and that that will improve the ambient tempco by 3x.
We also have a linear resistance range (done with a floating current source done with an op-amp, with several resistors switched to give different currents i.e. resistance ranges) and that feeds into the rest of the multimeter on a 1V range, so the tempco of the movement will be swamped by 10x.
I think times have moved on in that nobody makes the 10uA movements which were used in the best analog multimeters in the 1970s (100k ohms per volt) and the smallest meters one can easily get are 100uA (maybe
50uA if you are not fussy about the other stuff) and the ones I have seen are 1k.
Good point. There are materials that have low tempcos and a low thermocouple coefficient against copper. I think the best is manganin, which is about -600 nV/K against copper. Temperature-compensating the meter itself is a better idea.
Interesting stuff, thanks. I'm a huge fan of old Avometers, especially the Model 8 Mk IV. (The Mk V was the beginning of the downward slope.) I have a Mk IV from the collection of a guy who also collected Rolls Royces--not a trace of stickiness, and accuracy well within 0.5% on volts and amps.
Well, let us consider what the delta temperature of the meter coil is for full scale deflection and the change in resistance in delta percent. For valid readings we would want both the short term and the long term delta percent resistance to be the same, ignoring the differences in delta temperature itself. Does that lead to using the same materials for both devices?
Well, let us consider what the delta temperature of the meter coil is for full scale deflection and the change in resistance in delta percent.
The dissipation in a 1000 ohm coil passing 100uA is only **10uW** at FSD.
Ambient temp error is the only issue.
For valid readings we would want both the short term and the long term delta percent resistance to be the same, ignoring the differences in delta temperature itself. Does that lead to using the same materials for both devices?
** Cooper is impractical as a shunt cos the resistance is too low.
Eg: for 10mohms, 44cm of of 1mm dia wire is needed.
If made from Nichrome, the wire is then only 0.67cm long.