I have a back-burner project to take an old bench power supply and retro-fit it with two 4 digit LED displays for current and voltage. The power supply is 3A/30V, so four digits means 3000 count and 12-bit conversions (4096 = 2^12).
I've been trying to design an analogue current measuring block that just has a pair of current shunt resistors with a relay and some range detection circuitry etc. But I've bumped into a weird question about which resistors to use.
Ignoring the range selection, let's assume that I choose resistor and current sense amplifier so that 3A => 3V. Then, in order to get 1mA accuracy, I have a maximum error of 333ppm. Round that down to 200ppm to give the later ADC some headroom. Now, it seems that a reasonable temperature range is maybe 10°C - 30°C, which basically corresponds to "rooms I can function in". The 15°C range from the midpoint then says that, assuming I null out errors correctly, I can have a maximum TC of 13.3ppm/°C.
Even ignoring any errors from the current sense amplifier, I don't think that one can buy such resistors! (Or not easily). Looking hard at Farnell/Digikey/Mouser, I can find a few 20ppm/°C resistors but that's it.
So the question: Have I messed up my calculation horribly? If not, how do people build multimeters...?