I have a 0-1kg load cell with 4 wires: Excitation + and -, and Signal + and -. I haven't had any problems wiring them up correctly without amplification, and my voltmeter shows me a 1 mV difference between no load and full load (1 kg).
The signal has to feed into an ATmega microcontroller, so ideally I would like to be able to use the full 10-bit analogue scale when reading the values (voltages from 0 to 5 V). In any case, as it stands now, I only get 1 / 1024 as a reading under full load.
I have a prototype deadline for a product soon, and silly as I am I don't have any instrumentation amps left. I tried wiring the signal wires to a general-purpose NPN transistor, but I'm not getting any amplification. This is the transistor configuration that I tried: ignore the double cell symbol for voltage, I couldn't find anything better.
I also tried adding resistors between drain and ground, but that didn't help.
Is there any way I could obtain a usable reading using components I have right now? Ordering a proper In-amp would take too long, and I don't really know what kind of op-amp would be suited for this kind of measurement (although that might be an option).
I don't have a big selection of stuff available, mainly different kinds of transistors. And of course diodes, capacitors, resistors, and the like.
Best Answer
You will never be able to get accurate amplification of a 1mV full scale DC signal with a simple circuit using discrete components (I do know a way, but it's not simple). Even a crappy LM324 will do better (because the transistors are matched and on one chip).
Tell us what you have and we can, perhaps, help. An instrumentation amplifier can be made with three op-amps, and as I said, even a crappy 10-cent LM324 will be better than trying to use discretes.
In general, an LM324 will probably drift about 5-10uV/K so you can expect (at best) errors commensurate with that- about 5g ~ 10g/°C. A better amplifier would reduce those errors.
By comparison an B-E junction (as in your voltage follower) changes -2mV/K so you'd see 2kg/°C change- worse than useless.