# What current values are prone to noise?

I'm using voltage dividers to drop down from high voltages like 327V peak to 2.5V peak and read it with an ADC. (Also reading lower voltages like 5V from various sources, so the above was just an example)

In that particular case I'm using 1.3Mohms with 10Kohms resistors, giving a current of 0.25mA, with 80mW power.

Should I use 130Kohms and 1Kohms? Power will get a bit high. Also, if I'm drawing 2.5mA with my measuring it might affect the circuit I'm reading.

The question is: At what current values does noise become a thing to look out for?

• It's not just down to small currents being easier to disturb; resistor noise is also a consideration, and that scales with resistance: en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise – Nick Johnson Apr 18 '15 at 9:22
• what noise are you referring to? the higher the current, the better in this case. I think you can find insulated inamps that can step down your voltage as required, that would be a much better choice. – Vladimir Cravero Apr 18 '15 at 9:48
• Any kind of noise, but I'm assuming induced noise (inductive coupling?) would be highest. From various electronics like monitors nearby or florescent lighting. – Dragos Puri Apr 18 '15 at 9:56

Your 1.3M + 10K divider has a source resistance (looking back from the ADC input) of 1.3M||10K ~= 10K. The Johnson noise from ideal resistors is likely negligible for your purposes-- 13nV/$\sqrt{Hz}$- so if your bandwidth is 1000Hz your RMS noise is 411nV or perhaps 2-3uVp-p. So if your full scale is 2.5V it would be good for about 20 bits, which is about the maximum that is probably achievable anyway.

There are additional source of noise- noise current and noise voltage of the ADC, and leakage current. That depends a lot on the ADC- if unbuffered 10K is probably a good maximum number, but if you buffer it you could likely (depending on required accuracy) go considerably higher.

So, for 'noise' I would look at the ADC requirements in relation to your accuracy spec, and also look carefully at the stability of (especially) the 1.3M resistor. Permanent drift of the resistance value with time under application of high DC voltage/moisture is not uncommon. You can get an idea of the stability with the resistor "load life" specification, but check the test conditions.

Also, when you have high voltages on a PCB try to maximize creepage distances (for example by milled slots) and/or use guard traces to conduct leakage current away from sensitive nodes. A leakage resistance of 1G$\Omega$ in parallel with the 1.3M would cause 0.1% error.