Understanding 1/f Noise in DC Voltmeters: Causes and Solutions

[fluidistic]

I've been told that the 1/f noise for DC voltmeters is a problem when they measure a very low voltage (i.e. around 1 nV). Some readings can take weeks or even months. However, a DC voltmeter would have no problem to return a stable reading of a simple resistive circuit, of the order of the volt.
Physically I do not understand what's going on and why there's a difference between the 1 nV and the 1 V case regarding the pink noise. I'd like to see equations that would tell me whether the 1/f noise would be a problem, for a DC reading.

Note that DC corresponds to f = 0 Hz, so infinite noise in principle.

 

[berkeman]

Could you say more about Pink noise, 1/f noise and thermal noise in the context of your measurement? What is the source impedance of the voltage that you are wanting to measure so accurately?

EDIT/ADD -- BTW, I don't have anything in my lab that can measure in the 1nV range. I do have a pA meter that we use for specialized measurements...

 

[dlgoff]

BTW, I don't have anything in my lab that can measure in the 1nV range.

I just did a google search and found this really good voltmeter but if I understand the specs correctly, the best it can do is 6nV. i.e. the peak to peak dc noise. Here's the pdf for the Keithley 2182A.

 

[sophiecentaur]

What is the source impedance of the voltage that you are wanting to measure so accurately?

Also, what bandwidth is involved? Do you not need to know that in order to asses the noise performance of a device?

 

[fluidistic]

I am sorry, I am not measuring anything in particular, I ask only to know more.
Let's forget about thermal noise by considering a material near 0 K. I did not want to involve that noise at all.
Let's say the material has a resistance of about 1 ohm and we perform a 4-points probe method on it, to read out its voltage when we use a DC of 1 A. By reversing the current (but still a DC), we should in principle be able to measure a different voltage. The difference in the voltages when the current passes in 1 direction compared to the other might be in the nV range due to some physical effects I do not want to go into here. Apparently using a DC voltmeter (the Keithley is a good example I think) is problematic due to the 1/f noise.
On the other hand, using a "high" (not too high though) frequency AC is better suited to get that small nV voltage difference, which can be done in seconds.
I have no idea about bandwidth. The bandwidth of what exactly?

 

[sophiecentaur]

Like the Poor and Taxes, Noise is always with us and the noise power is dependent on the system bandwidth. To discuss operating near 0K, every item in the circuit (before the first stage of amplification) has to be that cold.Thermal noise power is kTB where B is the measurement bandwidth. You have to add to that the Noise Figure of your amplifying stage. That's always a consideration in communications (receivers) but it applies everywhere. This link can give you some information.