What is the standard way to sample DC offset of fast analogue signal

Discussion in 'Analog & Mixed-Signal Design' started by Domophone, Sep 14, 2016.

  1. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Hi Everyone,

    I'm having a bit of trouble identifying some anomalies in an analog signal. I am hoping somebody could point me in the right direction.

    I currently have a high speed (50-100MHz) analog signal arriving from another board (via coax). On my board, this signal travels about 2" and then to a BNC connector where it leaves the board again. The purpose of my board is to measure the DC offset of the signal.

    The Signal is composed of a transient non-sinusoidal signal (of roughly 50mV, ~50MHz) riding on a DC offset of roughly 200mV. The signal has an output impedance of 50 Ohms at it's origin, and is terminated as 50 Ohms off the board (on the BNC side).

    I need a circuit in which i can sample the DC offset of this signal without interfering or distorting the signal. I've attempted to make a 50Ohm Impedance trace between co-ax and BNC connector on my board. This trace has a small stub coming off it going to a buffer voltage follower amp (LT1077) via a ~50kOhm resistor. The output of the buffer amp is then fed into other circuitry that requires lowish impedance, hence the buffer.

    If i introduce this buffer op-amp into the circuit, i get an odd signal distortion, roughly 500ns after the signal first appears . If I remove it, it seems to attenuate significantly (though not entirely). If I bypass the board entirely, it disappears.

    I've attached two images: One is the transient with the buffer amplifier in place via a 50k resistor, the other with the amplifier removed (via removal of the resistor). Both images are pictured averaged over 25 transients. Like previously mentioned, the problem disappears when the board is entirely bypassed.

    Does the distortion pattern look familiar to anyone?


    Buffer Removed.jpg Buffer through 50kOhm.jpg

    Could this be due to external noise being picked up by the board and made worse when the buffer amplifier is connected? The acquisition is triggered via laser, which can create a decent amount of electrical noise, this might also explain why it does not average out over 25 shots? The electronics are all within a grounded aluminum box, with a line filter on the power input.

    Perhaps the answer is as simple as swapping the Op-Amp for something with a higher impedance input?

    Any and all suggestions/criticisms welcome.

    I'm sure I'm missing some info relevant to this question, so please ask if I've omitted anything.

    Thanks
     
  2. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    I suspect the "stub" feature may be to blame.

    Try using an inductive pick up with solid copper magnet wire and no physical connection at all. Basically an antenna on the OP amp input all inside your shielded box.
     
  3. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    Scratch that. You want to know dc offset. I just described how to sample the signal sans any DC component at all.

    Does the DC component serve some purpose? Why are you interested in it?
     
  4. OBW0549

    Well-Known Member

    Mar 2, 2015
    1,308
    884
    Why did you choose such a slow op amp for your voltage follower? If you're trying to buffer a signal with 50-100 MHz bandwidth, the LT1077 is too slow by a factor of over 1000.
     
  5. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0

    Sorry, to clarify, the LT1077 is buffering only the DC portion of the signal. The transient part of the signal should be, under ideal conditions, unaffected by this board. I just want to extract the DC offset.
     
  6. OBW0549

    Well-Known Member

    Mar 2, 2015
    1,308
    884
    Ah. OK.
     
  7. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    ?
    DC and secrets. Seems to be a lot of that these days.
     
  8. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Hey,

    Ya, the DC component of the signal is important. It is later processed and used to decide whether the the transient signal is acceptable. Also, in the example I attached earlier, the DC level does not shift during the event, but under other circumstances it can (on the scale of miliseconds). The initial DC offset has to be measured, along with the DC offset at several millisecond intervals
     
  9. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Sorry, I was just answering questions in order of ease, not trying to skip yours
     
  10. crutschow

    Expert

    Mar 14, 2008
    13,003
    3,232
    Do you need the instantaneous DC offset at some particular point in time or is the average offset including a signal okay?

    500ns is a very long time for any type of signal reflection. It corresponds to over a 150 feet of transmission line so that is a puzzle. :confused:
     
  11. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Practically, it can be averaged to include the signal. Ideally, though, I'd like it to be sampled about 100ns prior to the signal. It can change, but only over ~10^(-4) second or longer timescales.

    This was my thought regarding reflections, or anything to do with the quality of my pcb transmission line, thinking it may be just picking up switching noise from the laser.

    My question then would be, is there a reason adding an op-amp off a stub would make it that much more susceptible to noise?
     
    Last edited: Sep 14, 2016
  12. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    Guessing that you have a relatively long time between signal events, I would suggest using the signal to start a delayed sampling.
    Only wanting the offset between events makes this much more complex if using only analog devices.
    A micro with A/D would be a good choice here. You could detect the event and declare a delay in the code to put your sampling points between pulses.

    The decaying sinusoidal look of your problematic glitch looks to be parasitic inductance and capacitance in the form of a tank circuit.
    Something is resonating/ringing it.
     
  13. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Any suggestions on how i could further investigate this? If it was a tank circuit, would i see another further attenuated glitch ~800-1000nanoseconds?
     
  14. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    Not likely. The ringing seems to be totally decayed within 3 full cycles. Since it occurs well after the steep peak of your signal your guess that it is from the laser power supply could be correct.

    Changing some features might eliminate it. Ringing usually occurs when the start event freq. is resonant with the tank component's values. Slightly different spacing between pcb tracks or slightly longer or shorter coax may either increase or decrease the ringing.

    What is it about the laser Power supply that makes you suggest that as a possible source?
     
  15. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    Have you tried adding some capacitance after your sampling resistor, i.e. make a low pass filter at the input of the buffer?
    Not that it should mitigate the effect, but might give you some clues.
     
  16. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    Another thought.
    The glitch seems to occur before the signal has fully decayed to a point where only the DC offset is present.
    If you decide to go with my sampling idea, you could just ignore the glitch and begin the offset readings after it has decayed fully.

    Is the glitch interference a source of other problems in your load circuit?
     
  17. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Thanks, I will try changing some coax lengths and report any changes

    I just suspect the laser because it is full of switched noise that is basically synced with the laser firing. This glitch, if it is noise, is in the always in the same location relative to the signal. The laser pulse triggers the acquisition.

    I'm also going to try measuring the noise in the vicinity of the laser, and see if there is any signal showing up at identical time.
     
  18. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    So are you suggesting, only connecting the dc sampling circuit after it has decayed fully? I've got to sample it directly before the shot.

    Just to clarify, I can acquire the DC offset without a problem. The glitch interfering with the DC signal isn't a problem, rather the DC aquisition messes with the Transient signal that is supposed to be just passing through my board. The glitch appears to stem from my passing the signal my board, and is made worse when the addition of the branch with the voltage follower for measuring DC offset. Perhaps due to the layout/components chosen, perhaps because my board is inadvertently acting as an antenna and picking up external noise (like the laser)
     
  19. Kermit2

    AAC Fanatic!

    Feb 5, 2010
    3,786
    945
    If the glitch has no effect on the target load and is occurring before a steady state DC is achieved, then yes, ignore it and sample the DC offset at some point in time afterwards.
     
  20. Domophone

    Thread Starter Member

    Sep 10, 2013
    42
    0
    Ah, I see. The problem with that is the "target load" is actually just analysis of the signal trace, so having a glitch in the signal trace makes analysis of the signal and curve fitting less accurate later in the process. The signal is measuring a characteristic of a chemical reaction.
     
Loading...