Announcement

Collapse
No announcement yet.

PI holes

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • PI holes

    There has been some discussion of GEB in at least two threads lately. [TDI ground balancing scheme, twice over] and[ Replies to Minipulse Plus]. I charted some data to see if I could understand the problem. The lower right chart is a log-linear plot of ground (1/t) with an amplitude of 20 at 1usec and target time constants with an amplitude of 100 at zero time vs time in usec. The other charts plot signal amplitude vs target time constants sampling at different times. The lines with two samples reject ground, ((sample 1 amplitude - (sample 2 amplitude x sample 2 time/sample 1 time)). The single sample plots the percent amplitude based on 100 at zero time for the different time constant targets.
    Attached Files

  • #2
    Excellent!
    The zero crossing is a hole. It can also serve as a discrimination criterion.
    You may also try adding several samples with durations in a power of two order, and see what happens

    Comment


    • #3
      Originally posted by Davor View Post
      Excellent!
      The zero crossing is a hole. It can also serve as a discrimination criterion.
      You may also try adding several samples with durations in a power of two order, and see what happens
      Thanks for the reply. A couple questions I had, how far apart the sample 1 and sample 2 should be and how far apart the second set of samples should be. S2 time/S1 time=3 is better than S2 time/S1 time=2. Looks like doubling the time for the second set of samples is a good place to start. I'm guessing there is a limit with the time difference with the PI detector where it gets worse. Excel isn't limited by noise.
      Attached Files

      Comment


      • #4
        Now try adding those values and see what you get

        Comment


        • #5
          Originally posted by Davor View Post
          Now try adding those values and see what you get
          If I add them I'll end up with a hole. I've bee thinking two channels, one each ear or different frequency for each channel. Maybe there is a better option.

          Comment


          • #6
            Of course you get a hole, but observe where it is, and the shape of a positive side before the hole
            Remember that the tau past some 300us or so is meaningless anyway.

            Comment


            • #7
              Originally posted by Davor View Post
              Of course you get a hole, but observe where it is, and the shape of a positive side before the hole
              Remember that the tau past some 300us or so is meaningless anyway.
              I'm not understanding what to add. It looks like the hole is going to be at less than 30usec.

              Comment


              • #8
                Yes, you'll always have a hole with paired samples.
                But, if you take samples in sequence one after the other, and their durations are following some order, say power of 2, and their polarities alternate of course, and you add them all up, you'll end up with something quite interesting.
                Besides, you may always wiggle between two different sampling schemes, and use span between the holes as a sort of discrimination.

                Comment


                • #9
                  Originally posted by Davor View Post
                  Yes, you'll always have a hole with paired samples.
                  But, if you take samples in sequence one after the other, and their durations are following some order, say power of 2, and their polarities alternate of course, and you add them all up, you'll end up with something quite interesting.
                  Besides, you may always wiggle between two different sampling schemes, and use span between the holes as a sort of discrimination.
                  Am I thinking right? If I sample ground each sample should be the same amplitude. Odd numbered samples would equal ground amplitude and even numbered samples would be zero.

                  Comment


                  • #10
                    The scheme I propose for Minipulse (and further) is a sequence: delay=1T, +sample=1T, -sample=2T, +sample=4T, -sample=8T, +sample=16T, -sample=32T, ... and appropriate EF at the end.
                    For simplicity I'm using only delay=1T, +sample=1T, -sample=2T, +EF=1T and surely I get a hole, but I can wiggle T and fix it that way.

                    Comment


                    • #11
                      Originally posted by Davor View Post
                      The scheme I propose for Minipulse (and further) is a sequence: delay=1T, +sample=1T, -sample=2T, +sample=4T, -sample=8T, +sample=16T, -sample=32T, ... and appropriate EF at the end.
                      For simplicity I'm using only delay=1T, +sample=1T, -sample=2T, +EF=1T and surely I get a hole, but I can wiggle T and fix it that way.
                      An attempt at charting it. Used same data that was used for #1 and reply#3. Forgot to label X axis (target time constants) y axis (amplitude)
                      Attached Files

                      Comment


                      • #12
                        OK, now just add those 3 traces together and see what you get.

                        Comment


                        • #13
                          Originally posted by Davor View Post
                          OK, now just add those 3 traces together and see what you get.
                          Chart with 3 traces added.
                          Attached Files

                          Comment


                          • #14
                            Well, that's as far as this concept goes. With 10us as a time base, you'd get a hole beyond 100us, a flat response for nearly all targets, and meaningful ground balance as well. What you don't get easily is EF, which is 1+4+16= 21T ... unless you go for alternating pulsing and get rid of EF altogether.
                            Each pair is ground balanced so there are several other possible useful scenarios:
                            1. You may have VLF IB kind of discrimination by using a hole-less response as "all metal" channel, and use a response with a hole as discrimination channel for discrimination gating purposes.
                            2. You may shift timebase T in time (say 10Hz) to achieve shifting of a hole, and use polarity of the response for modulating audio properties, thus having simple and effective discrimination with only one pair of samples, and virtually no hole
                            3. You may use a hole to notch out the offending target, say a pulltab.


                            You may not achieve discrimination of iron this way. It may be possible to pass the sampled flyback peak voltage envelope through a motion filter and use that as an indication of changing permeability. In essence the VLF IB devices do precisely that.

                            Comment


                            • #15
                              One GB newbie question: the 1/t ground response is opposite or same polarity as a conductive target? Thanks!

                              Comment

                              Working...
                              X