Extract a micro V signal from a 120V transient

Discussion in 'The Projects Forum' started by Monolith, Oct 20, 2009.

  1. Monolith

    Thread Starter Member

    Jun 21, 2008
    13
    0
    I want to extract a signal of micro Volt amplitude that is superposed on a 120V transient.
    The transient duration is of about 15uS and I want to take a 2uS sample of the signal that is superposed.

    The repetition rate of the transient with the signal is 1Khz.

    How do I go about that? High voltage differential amplifier?

    Any ideas?

    Monolith
     
  2. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    When you say the 120 volt signal is a transient, the usual meaning is that it is recurrent, but not predictable. So -
    - tends to say the 120 volt signal is recurrent and predictable.

    If that is the case, something like a synchronous detector might be able to grab your microvolt signal.

    Tell us everything about the 120 volt transient.
     
    Last edited: Oct 21, 2009
  3. Monolith

    Thread Starter Member

    Jun 21, 2008
    13
    0
    The apparatus is a pulse induction metal detector.

    The 120 V transient happens every 1000us, when the current in the sensor coil is switched off.

    The desired signal is of microvolt amplitude and appears on the 120V, with the presence of metal near the sensor coil, it goes more positive for gold and more negative for iron, than the 120V.
    I can detect the signal if it is near 0V, but how do I get it on top of the 120V?
     
  4. someonesdad

    Senior Member

    Jul 7, 2009
    1,585
    141
    As beenthere asked, you'd best describe in detail the signal setup. I'd suggest showing a drawing/sketch of exactly what's going on. It sounds like an interesting problem, but, on the surface, all we readers have to go on so far is that you're trying to detect some signal at μV levels on top of a signal at 100 V levels. Thus, you're trying to detect something 8 orders of magnitude less than an existing signal. Not an easy task, especially if it needs to be done 1000 times per second.

    It may be possible to do, but you'll have to leverage knowledge of what the waveforms look like (i.e., their time behavior). I'd expect that, if it can be done, it will involve looking at "phase" differences between the signals or some kind of lock-in (synchronous) technique, as mentioned by beenthere.
     
  5. t06afre

    AAC Fanatic!

    May 11, 2009
    5,939
    1,222
    A lock-in (synchronous) technique, is only effective then the signal is not distorted. If the signal is say clipped the lock-in technique will perform poorly.
     
  6. Monolith

    Thread Starter Member

    Jun 21, 2008
    13
    0
    Thanks for the replies.
    Attached is a scope picture that shows the transient. This transient repeats 1000 times per second.
    There is a sampling time window of about 15us when the transient voltage is about 120V, kind of a flat top.
    When there is a metal target near the sensor coil, there is a signal present during that time, this signal can give information about the size and the conductivity of the target. The signal also indicates if the metal target is of the magnetic or non magnetic kind.

    The problem that I have, is the high voltage of the sampling window.

    I can extract a faint echo of the signal after the transient drops to the lower level, but by that time most of the information has already faded away.

    I know it can be done, because I know somebody who can do it, but he does not share the information.
    So I want to find out myself how to do it.
     
  7. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    I would start by using analog switching techniques and level shifting circuits. Use the initial rising edge as a trigger, and for a certain time window after that, subtract 120 V from the signal. You will end up with a more manageable signal that is easier to work with. The signal will be zero volts, except during your capture window, for which the signal is now in the +/- 10 V range.

    After that, you can try some high pass filtering to remove the DC offset and 1 KHz fundamental component of the signal. I'm assuming that your useful information is at a much higher frequency than 1KHz.

    Perhaps, this new signal will be more amenable to standard techniques.
     
  8. Monolith

    Thread Starter Member

    Jun 21, 2008
    13
    0
    thanks for the help.

    It is the subtracting of the 120V that I don't know how to do. Once I can get the signal into an opamp, I can extract it.
     
  9. steveb

    Senior Member

    Jul 3, 2008
    2,433
    469
    It would be helpful to know more about the signal. What is the frequency range for this? If this is at a reasonable frequency < 10 MHz, and preferably less than 1MHz, there are some tricks that can be done.

    It's also helpful to know your expertise and knowledge about circuits in general.

    Remember caution with high voltages that we are talking about here.

    You don't really need to subtract the signal, but can simply reference that signal to a 120 VDC very low noise reference point. Isolation of circuits is needed. There are too many details to try and lay out, and I'm sure I would need quite a bit of work to do a system design for this. But the basic principle may guide you.
     
    Last edited: Oct 21, 2009
  10. Monolith

    Thread Starter Member

    Jun 21, 2008
    13
    0
    More about the signal...
    the desired signal is basically a DC signal, if we look only at the sampling window of about 15us, where I want to extract the information.
    Looking at the scope screen, it is a line that is about straight. After amplification it becomes more curvy.
    The curves give the information about the target, by taking several 2us samples at different points.
    I can isolate the RX by using a center tapped, transformer coupled inductor.
    This worked well as long as I was using very low power, so that the Flyback voltage stayed within the range of the differential input opamp and the Sample and Hold.
     
Loading...