Noise Cancellation (Math Function)?

Discussion in 'Wireless & RF Design' started by mossman, Nov 21, 2013.

  1. mossman

    Thread Starter Member

    Aug 26, 2010
    131
    3
    I'm attempting to measure some very low amplitude signals on a circuit board and ambient noise is a huge problem. The probe I am using is very directional, but still picks up quite a bit of noise. I am curious if there is some way to cancel the noise signal only leaving the signal of interest. For example, by using a second omnidirectional probe in close proximity to the directional probe in order to pick up a common noise signal, then inverting the waveform on the scope and adding the two waveforms to cancel out the noise, essentially leaving only the differential signal (signal of interest). Will this work? FYI, the signal of interest has a data rate of about 40 Mbps.
     
    Last edited: Nov 21, 2013
  2. Papabravo

    Expert

    Feb 24, 2006
    10,157
    1,795
    You'll be far better off eliminating the source of the noise. Second best is to shield the board from the source of the noise. In my experience, noise cancellation techniques along the lines you suggest produce very little positive benefit.
     
  3. PRFGADGET

    Active Member

    Aug 8, 2011
    51
    7
    Years ago, I had a similar situation.
    Used two identical sensors wired back to back but with crossed polarity.
    it worked, but you will need to match the sensors and also match them to the frequency range of interest.
    Also helps if you can get the device in question into as near a sound proof environment as possible (foam lined box or sound proof room).
    This sort of thing has also been used for years in the audio industry with stage productions and even in the communications industry for noise cancellation in high noise environments.
     
  4. MrChips

    Moderator

    Oct 2, 2009
    12,449
    3,365
    That was the idea behind the "hum bucker" employed in guitar pickups in order to reduce AC line pickup.

    We would be able to better assist you if you told us everything.
    What type of signal are you measuring?
    Is the signal to be measured periodic, such as a sine wave?
    Can you show us a circuit diagram?

    There are ways of reducing noise but we have to know the situation.
    Full disclosure is called for.
     
  5. mossman

    Thread Starter Member

    Aug 26, 2010
    131
    3
    Unfortunately I can't provide a lot of detail, but it is a 40 Mbps digital signal flowing through a circuit trace and I am attempting to measure the emissions with a "sniffer probe". I was wondering if I could connect a second sniffer probe (same model) in close proximity, and use the Channel 1 - Channel 2 feature on my scope to cancel out the common mode noise, thereby leaving only the signal of interest.
     
  6. MrChips

    Moderator

    Oct 2, 2009
    12,449
    3,365
    What are the "emissions"?
    What is the "sniffer probe"?

    Too bad. Without more information I'm outta here.
     
  7. mossman

    Thread Starter Member

    Aug 26, 2010
    131
    3
    RF emissions from the current (40 Mbps digital signal) traveling on the PCB trace, and the sniffer probe is a B. Carsten E101 magnetic field probe.
     
  8. Papabravo

    Expert

    Feb 24, 2006
    10,157
    1,795
    So you have a magnetic field probe and you are trying to separate one of many signals on the board from all the other signals on the board. I think you will have a very difficult time with this one.
     
  9. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    Every loop with a time-varying current is generating a time-varying magnetic field, with amplitude proportional to the loop's enclosed geometric area. And who knows how they're all oriented. If there's a ground plane, most of the loops won't have much enclosed area, so they'll be low level as well as noisy and all mixed together.

    If there's only one such signal trace on your board, you might have a chance. Otherwise, maybe there's an "averaging" mode on your instrument. Try a decent spectrum analyzer. They usually have averaging. Or maybe you could record the data and then do the averaging off-line. It would need to be a stable periodic signal. If you record a short stream of it and then overlay shorter lengths of that, and average all of them, it should pretty-much remove most of the un-correlated stuff, if you average enough of them.

    I'm not sure how you plan to use the information but to measure the RF emissions, in what to me might be a somewhat useful way, and this is just off the top of my head, you would need to use a set of (probably directional) broadband calibrated measurement antennas, one at a time to cover the entire spectrum of interest (assuming one antenna would not be able to do it all), in a proper anechoic chamber, and record the received signal level at closely-spaced points all over an imaginary sphere with the board at its center. It helps (a lot) to use an automatic positioner and a network analyzer, with a computer that has software to automate the whole thing.

    If you just want some rougher information about the maximum total effective radiated power (in some direction), you could probably just use a broadband RF power meter, or a spectrum analyzer with a "channel power" measurement mode, and manually wave an antenna around the board, looking for peaks.

    There are also some articles on the web about making tiny RF Sniffer probes, using a very small loop of wire and an amplifier for the sensed current. But again, I don't know how you plan to use the data you get and I don't see a lot of use for the data from one small area, very near the board, unless you're just looking for RF hotspots.

    There are some little log-periodic antennas that are available, on small triangular-shaped PCBs, a few inches in size, that might work for you (again, depending on what your actual goal is). I forget who makes them.

    By the way, there are also electric field probes available.
     
  10. mossman

    Thread Starter Member

    Aug 26, 2010
    131
    3
    Everything will be automated and we are using a stage with 1 nm precision. I believe the probes we are using are the best that there are for this type of work (we have H-field and E-field). I'm assuming an anechoic chamber is crucial then?
     
  11. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    Maybe an anechoic Faraday cage. They make RF isolation boxes, with built-in arm holes with metallized gloves so you can reach inside. We have them at work but I forget the name of the manufacturer. I could find out on monday.

    Is this for EMI/EMC purposes? You might want to just throw a CAD model of the board into one of the electromagnetic simulator software packages.
     
  12. gootee

    Senior Member

    Apr 24, 2007
    447
    50
    If you know what the digital signal is, in a pcb trace of interest, and you just want to measure the em component due to that trace alone, then knowing what signal to look for might be very helpful.
     
Loading...