Cut Through The Noise, See Tiny Signals

An oscilloscope is a handy tool for measuring signals of all kinds, but it’s especially useful if you want to measure something with a periodic component. Modern oscilloscopes have all kinds of features built-in that allow you sample a wide range of signals in the hundreds of megahertz, and make finding and measuring your signal pretty easy, provided you know which buttons to push. There are some advanced oscilloscope methods that go beyond the built-in features of even the best oscilloscopes, and [AM] has a tutorial on one of them.

The method used here is called phase-senstitive detection, and allows tiny signals to be found within noise, even if the magnitude of the noise is hundreds of times greater than the signal itself. Normally this wouldn’t be possible, but by shifting the signal out of the DC range and giving it some frequency content, and then using a second channel on the oscilloscope to measure the frequency content of the source and triggering the oscilloscope on the second channel, the phase of the measured signal can be sifted out of the noise and shown clearly on the screen.

In [AM]’s example, he is measuring the intensity of a laser using a photodiode with a crude amplifier, but even with the amplifier it’s hard to see the signal in the noise. By adding a PWM-like signal to the power source of the laser and then syncing it up with the incoming signal from the photodiode, he can tease out the information he needs. It’s eally a fascinating concept, and if you fancy yourself a whiz with an oscilloscope this is really a tool you should have in your back pocket.  If you’re new to this equipment, we do have a primer on some oscilloscope basics, too.

16 thoughts on “Cut Through The Noise, See Tiny Signals

    1. That’s exactly a lock-in amplifier. The author is using a wrong name for it. It is not called a “phase sensitive detection”, and the key element for why that technique is working is not the phase, but the frequency lock. Lock-in amplifiers can achieve some incredible measurements, indeed.

      Shahriar did a great presentation of a lock-in amplifier, how and why does it work, and some nice experiments:

      Not to be confused with a chopped amplifier.
      :o)

      1. I think the video meant that phase sensitive detection is used for measuring distances with lasers.
        (It’s only ever mentioned at the end when that’s the topic being discussed, right?)
        The author just misunderstood the video and thought the (lock-in-amplifier-)technique used in the video was called ‘phase sensitive detection’

  1. This is the basic reason why the trigger input exists.

    Since his “noise” was the mains hum, he could just trigger every cycle of that at a known phase angle and achieve the same result without the 5 kHz signal injection, which often isn’t available.

    1. Or more appropriately, try to lock on to the zero crossing of the AC waveform, so the amount of hum doesn’t change the absolute value.

      And since he had two probes in the scope, he could have taken a loop of wire to the other probe and set it close to the circuit, then subtract A from B in the scope to get rid of the hum. You simply change the loop area to get the error signals to match in amplitude. This works as long as the noise isn’t random.

  2. There are technologies that can help pick signals out of noise but you also have to be careful. 30 or so years ago I was involved in some research and we were moving a sample fractions of a millimeter with an MTS servo hydraulic setup. The dither on the servo hydraulics was borderline significant, that was how small our stimulus was. The saving grace was our signals were subsonic and the dither was around 4Khz if I recall. Our data was collected through a vast array of transducers, but most of it was buried in the noise. One of the guys with a guy from the applied math department developed an amazing filter. You took our noisy data in and it spit out nice clean results. We were all really happy. Until one day I started the data acquisition system moments before I started the servo hydraulic system. When we got to looking at the data later on down the road we noticed we had good looking results with no input. OOps. In some situations this may not be a show stopper. Computer data for example, it would become readily apparent that the received data was garbage so tring to use it would not hurt anything. In that case having wrong data is not worse than having no data. In our case however, having wrong data was much worse than having no data as there was no way of really telling it was wrong. The point I am trying to make is you have to be careful if you go down this path.

    1. Yup. and the problem exists all over. If you carefully filter out all noise by whatever means, all that’s left is something you expect to see, and most filters will turn white (or other semi random) noise into just that. So at some point it becomes hard to learn anything new.
      In the case of a lock-in amp or a boxcar averager if you have two signals that aren’t harmonically related and can only lock to one, the other becomes noise – even if in the overall scheme of things it was important. Not much problem in some experiment designs like stimuli-evoked response measurement.

      But at say CERN, where the data can utterly overwhelm the ability to write it down, someone has to figure out a relatively simple trigger mechanism (my friend there uses FPGA stuff in the detector front end) to decide what to save.

      To be fast enough, it has to be simple criteria (relatively speaking). So you wind up tossing some baby with the bathwater and missing some of the kinds of things that are *almost* the stuff you don’t think you need to see anymore while looking for other specific rare stuff…but which might represent new science. It’s something they know about and agonize over to some extent. They simply can’t follow the old time religion of “write it all down and we can analyze till the end of time to see what we might have missed” that scientists used to preach and practice when possible. Even in my dinky low-end lab, as these things go, it’s a problem…

      So when one can, one strives for solving some by experiment design, and by doing whatever one can to get that signal out of the noise without removing info – because what you thought was noise might just be another interesting signal…

    2. Interesting comment and thought provoking thread too, thanks :-)
      FWIW. Late 1970’s worked with a fellow engineer now in USA with strong theoretical base who developed a brilliant Kalman based sample data systems filter I implemented in a psuedo Forth code base on the NS pace-16 micro (paper tape initially then eeprom!) for nucleonic iron ore flow measurement. It fed into a Foxboro controller for adjusting angle feeder rates on the grading screens at the beneficiation plant Mt Newman Western Australia. Few tinkering cycles with coefficients to deal with odd skewed std deviation of the scintillation counter rates being fed from Cs137 or Co60 beta sources was interesting whilst watching 100+Kg chunks of ore fall through both radiation beams across the flow curtain – part of the ore grading design by Dave Gibson of Pretron Electronics circa 1975.
      Nowadays that filter type can be more easily applied to these sorts of issues as per this thread given the processing speeds so much higher and many ai like patterning advances too for Kalman variations in hundreds of journal papers since that time, lots of opportunities.
      ie Trundling along at a GHz or so for only a few $ (and even fitting in a probe compensation enclosure) can lock onto a wide range of diverse patterns and adaptive as well once certain types of boundaries set so getting all sorts of signals not mere count shifts can show up all sorts of oddities.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.