Use a DVM to directly digitize low-frequency noise (Part 1)
Use What You Have - Overlooked Capabilities Might Really Simplify Things
I have been doing some low-frequency, low-noise work lately and I was looking at implementing a peak-to-peak detector circuit like Jim Williams presented before .
Jim was right, the only way to really be sure of your peak-to-peak noise specification is to make many readings and run a statistical analysis to see if the measurements are really random or something else. It is just far too easy to kid ourselves that a few readings are significant when they might not be.
The project I was working on involved measuring 0.1 to 10 Hz noise not only for RMS and peak-to-peak values in the time domain, but using an FFT Analyzer in the frequency domain also.
I was going to use my trusty Agilent 34401 DMM to read the analog detected peak-to-peak voltage when I remembered that these meters are really quite fast in making readings and can really zip readings into the internal memory. Looking at the data sheet reminded me that indeed the reading rate can be greater than 1000 readings per second.
The other thing that caught my attention was that the integration period is selectable from 100, 10, 1, 0.2 and 0.02 power line cycles. The integration period is like a traditional sample-and-hold circuit's Aperture Time (At) and this is really what sets the achievable speed, resolution and noise bandwidth.
Integrating for a longer time (or longer aperture time) decreases the noise bandwidth and the resolution can increase because of the lower noise.
The second parameter that is adjustable in setting the sampling rate is the trigger delay. Thus by setting the integration time and the trigger delay between samples it is possible to set a sampling repetition rate.
Table 1 details the performance I managed to get out of the 34401 with my simple USB to GPIB converter on a laptop .
Table 1 – These are the maximum sampling rates that I could get from my USB-to-GPIB-converter-connected Agilent 34401 DMM. It is important to remember to turn Autozero and the Display Off to achieve these fast reading rates.
Since I wanted to sample a 10 Hz low-pass filtered signal, sampling at a 40 Hz rate would be appropriate and saving to 401 points in internal memory would give me a 10-second total measurement window. That is just what is required for a 0.1 to 10 Hz noise measurement as measuring for 10 seconds is like adding an additional high-pass zero at 0.1 Hz to the resulting data.
After some experimenting I found that I could set the integration period to 1 NPLC (Number of Power Line Cycles) which is 16.7 milliseconds and still achieve the desired 40 samples per second rate by adding extra trigger delay.
Setting the integration period like this has another digital effect on the readings. Sampling a signal with a sample and hold adds a Sinc-like function to the frequency response of the measured data. The first null of this sinc function is that 1/At, or in this case 60 Hz, which is really nice for rejecting that pesky 60 Hz noise which is everywhere. The 3 dB bandwidth of this sync function is at 0.45/At or in the case here: 27 Hz .
That inherent low-pass filtering is very useful as this adds more out-of-band noise rejection for free, with no added code or analog parts required. The passband "Droop" of this filter at the 10 Hz band edge is less than 0.5 dB, so this is not a problem for the measurements that I was making.
Next month we will take a look at the verification of using this DVM as a low-frequency digitizer. Not only in the time domain, but in the frequency domain as well.
Figure 1 – When we think of DMM's we normally think of measuring very slowly moving DC signals. Next month we will take a look at what it takes to use a DVM as a Waveform Digitizer.
 Jim Williams "775 Nanovolt Noise Measurement for A Low Noise Voltage Reference", AN 124, Linear Technology, 2009
 A LAN to GPIB converter or a Desktop with a PCI Card will give faster results than the USB / GPIB converter setup I used in this article.
 HP3458A Product Announcement, HP Journal, April 1989