How to measure uV voltages buried in noise?

Question

I'm working on a project in which I am analogly processing ultra low signals. The phase, harmonics and signal shape are critical to circuit functions. The signal on the input comes from an antenna type device, so as you can guess, it picks up a ton of noise.

My minimum input voltage is 2uV AC ranging anywhere from 10-10kHz.

The circuit applies a gain of 1000 and does some other jazz to the signal. The problem is being able to measure the input signal and compare it with the output.

I'm even having trouble measuring the output properly as even my Tektronics scope does not go anywhere near that low. I'm currently using a 6 1/2 digital bench DMM but I can only measure amplitude.

In noisy, low voltage situations like this, how do I go about measuring this signal?

If I amplify everything more I fear the noise to signal ratio will be so large I will only see noise anyway.

Would I need to use a band pass?

Answer 1

anywhere from 10-10kHz.

Would I need to use a band pass?

Sounds like you at the very least would want to use a low-pass filter; that'd remove the noise power from above 10 kHz.

In noisy, low voltage situations like this, how do I go about measuring this signal?

You'd apply as much gain as necessary for your measurement device or ADC to measure the signal. There's not much you can do if you know nothing else about the signal! If your signal is unknown, and so is your noise, you simply can't tell them apart.

I do, however, get the feeling that you might know a lot more about the signal in question than you describe in your question. In that case, signal processing can help you tell signal from noise, and estimate what you need to estimate. As an example:

The millivolts you're getting are pretty much signal, when you consider that GPS receivers work with signals down to -160 dBm – that's 10⁻¹⁹ watts, or ca. 30 nV amplitude if we assume that the GPS receiver has an input resistance of 100 Ω. Most importantly, that's far below thermal noise floor over the same bandwidth! Still, GPS works reliably by employing processing gain based on knowledge of the signal structure.

Answer 2

Place the system inside a screen-room. Drop the Efields. Get all motors, black-brick power supplies at least 1,000 meters away. Only use twisted-pairs for AC power cables.

And use a input shunt-to-ground termination resistor of 62 ohms, which is 1 nanovolt / rtHz random noise, thus in 10,000Hz bandwidth produces only 100nanoVolts RMS noise against your 1,000 nanoVolt signal.