Announcement

Collapse
No announcement yet.

TEM SIGNAL PROCESSING

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

  • TEM SIGNAL PROCESSING

    We have seen the TX circuit. Now, how should the RX circuit look like?

    How are we going to process the RX signal?

    Should we sample and digitize it right at the input and do all processing digitally?

    Should we do all the processing in analog?

    Should we do some signal conditioning before digitizing?

    Below is the TX schematic again, as a reminder how it looks.

    Then we have an old analog schematic, as an example of what we could do and to give some points of reference.

    Tinkerer
    Attached Files

  • #2
    The raw signal at the input

    And here comes the screen shot of the signal, as it comes to the input, raw. NO TARGET

    Hmm, a bit of a mess. Well, as I said, it is raw. And it is a very noisy environment. And there is all sorts of clutter, like 2 mains cables and a UPS touching the coil and the coil is sitting on the floor with a re-bar mesh just 3 inches below the coil. sorry, I have not much space and the coil measures 1m x 0.5m so we just have to take it as it comes.

    The scope settings are 2V div. and 2us div. And where is the target supposed to turn up?

    OK, the second picture shows a target on the coil. Yes, something moved. We can see an offset between lead a and lead b. The RX coil is a center tapped coil, by the way.

    At least we know that the TX works and that the RX is capable of picking up a target.

    Now, what is the first step? You can look at the RX schematics above, how I did it last time. It worked quite fine.

    This time I want to do it better.

    Any suggestions?

    I think we want to do some signal conditioning before we sample. Clean it up a bit.
    Filter out what we do not need.

    Any good ideas??????????????

    Tinkerer
    Attached Files

    Comment


    • #3
      Originally posted by Tinkerer View Post
      And here comes the screen shot of the signal, as it comes to the input, raw. NO TARGET

      Hmm, a bit of a mess. Well, as I said, it is raw. And it is a very noisy environment. And there is all sorts of clutter, like 2 mains cables and a UPS touching the coil and the coil is sitting on the floor with a re-bar mesh just 3 inches below the coil. sorry, I have not much space and the coil measures 1m x 0.5m so we just have to take it as it comes.

      The scope settings are 2V div. and 2us div. And where is the target supposed to turn up?

      OK, the second picture shows a target on the coil. Yes, something moved. We can see an offset between lead a and lead b. The RX coil is a center tapped coil, by the way.

      At least we know that the TX works and that the RX is capable of picking up a target.

      Now, what is the first step? You can look at the RX schematics above, how I did it last time. It worked quite fine.

      This time I want to do it better.

      Any suggestions?

      I think we want to do some signal conditioning before we sample. Clean it up a bit.
      Filter out what we do not need.

      Any good ideas??????????????

      Tinkerer
      I got a good one for you Tinkerer. Implement a resonant RX loop. This sounds like it has a lot going for it in both improved sensitivity and noise reduction. Unfortunelty (considering the work you've put into it already) I don't think it will work with a concentric coil. You need your TX and RX to be the same size so that they can stay resonant with each other regardless of ground effect. So that probably means DD or OO.
      http://www.nexusdetectors.com/scienc...detectors.html

      Midas

      Comment


      • #4
        Originally posted by Midas View Post
        I got a good one for you Tinkerer. Implement a resonant RX loop. This sounds like it has a lot going for it in both improved sensitivity and noise reduction. Unfortunelty (considering the work you've put into it already) I don't think it will work with a concentric coil. You need your TX and RX to be the same size so that they can stay resonant with each other regardless of ground effect. So that probably means DD or OO.
        http://www.nexusdetectors.com/scienc...detectors.html

        Midas
        Thanks for the feedback.
        I will try the resonant RX-TX next, after I finish the project above. Since a long time I wanted to do a OO Tandem coil configuration.

        Tinkerer

        Comment


        • #5
          Damping the RX

          The first thing to do is to damp the RX coils. As we see in the pictures below, this makes things look a bit better.
          First picture, The signal, with the damping resistor, NO target.

          The scope settings are still the same, 2V/div. and 2us/div.

          Of course we need to make sure that we do not eliminate the target signal when tampering with the input, so the second picture shows the signal with a target. OK, the target is still there.
          But what is this other noise? Switching noise at switch ON and at switch OFF. We want to get rid of some of that noise before we amplify the signal, or it will saturate the opamp.

          So the next step is some passive filtering.

          Any suggestions? Good ideas??????????

          Tinkerer
          Attached Files

          Comment


          • #6
            Hi Tinkerer,

            That is the same sort of responce that I was getting with the Duel-Field coils and remember one of the prototype coils I have made for the project is a DD so when it is all running we can do some tests with that.

            Regards, Ian.

            Comment


            • #7
              Thanks for the feedback, Ian.
              Yes, the signal of most IB-PI look a bit like that. The Flyback has been eliminated by the Bucking coil or the overlap of the 2 coils in the case of a DD coil assembly.
              However, as precise as the induction balance may be, there is always some residual left.
              If we look at the pictures above, we see some HF noise. If we were to amplify this signal as is, by the usual 1000 times, we would grossly saturate the opamp.
              Once we have saturation, the amplification is not linear anymore. The opamp also takes some time to come out of saturation. We end up with a very distorted remnant of the original signal.

              The usual method is to clamp the signal with 2 fast switching diodes. This truncated signal has lost most of the information it contained.

              So we want to do it differently.

              In my sample circuit, above there are diodes too. But the only function of these diodes is to protect the input from overload during the balancing procedure when building the coil, or if something goes wrong. The signal amplitude is kept below the conducting level of the diodes at all times.

              With the 1m x 0.5m coil, we will not be searching for very small targets. Therefore, we can low-pass filter the signal with a Fo of about 150k. This will greatly reduce the HF noise we see.

              Now, what is the best filter design for this purpose? This is where I hope to get some help.
              In my trials, I have found that the filter can drastically reduce my target signal, but I have not come to understand exactly why.

              In the above circuit, I use single opamps and limit the bandwidth with a feedback capacitor. the result is quite OK.
              With the new design I want to use an Instrumentation Amplifier. I can not add a feedback capacitor, therefore I want to filter the signal before the InA.

              Then there is also the choice of InA. I have some AD620's, but the the slew rate and the bandwidth is very low at 120kHz with the gain at 100. The gain of 100 is OK for this purpose, so we are not too far off.
              A bit better is the SSM2019, with 200kHz at G 100. This will do. But I am sure there are still better choices available.

              Any ideas????

              Suggestions????

              Below is the first part of the new circuit. The filter is far from good. Any help would be appreciated.

              I added the TX-BU-RX coil configuration to give an idea how the whole front-end looks like.

              Tinkerer
              Attached Files

              Comment


              • #8
                Hi Tinkerer

                Instead of trying to filter out the noise you could use analogue switches to isolate the RX coil during the switching like Moodz' or Jon-ecm's design. More complex but I think the minimal attenuation when the switch is on and you want to sample is always going to be better than what you get with a passive filter.

                Midas

                Comment


                • #9
                  Attached is the filtered and amplified RX signal.
                  The gain is 5.5
                  There is a passive differential filter and a passive common mode filter.

                  The SSM2019 opamp is doing a good job.

                  The target response to large targets is good, but the response to small targets we will only see once the signal is further amplified.

                  Above the signal, we see the Sample pulse. This shows the part of the RX signal that we take for sample #1.

                  Next we need to decide if we want a "Motion type detector" or a "Non Motion type detector.

                  For most uses, the motion detector works better. However, for some specific purpose, we would like to be able to detect with minimum movement of the coil.

                  Therefore we are going to look a both modes and try to obtain good results either way.

                  Does anybody have some suggestions? Clever ideas?

                  Tinkerer
                  Attached Files

                  Comment


                  • #10
                    Tinkerer is the 300uH across A and B ?
                    Attached Files

                    Comment


                    • #11
                      Originally posted by 6666 View Post
                      Tinkerer is the 300uH across A and B ?
                      The TX coil and the Bucking coil are wound in series, opposing.
                      The RX coil is wound on the Bucking coil and is center tapped.

                      The TX can be anywhere from 100uH to 1000uH, the RX coil should be anywhere from 300uH to 1000uH.

                      Cable capacitance has little influence.

                      The sample #1, is taken on the point of highest target signal.

                      You could try Aziz's "Top Hat" coil configuration. I will give it a try myself.

                      Tinkerer

                      Comment


                      • #12
                        TEM signal processing step by step

                        In post # 9 we see the signal wave shape and the #1 signal pulse. We take the #1 sample at the point in time where the desired target response has the highest amplitude. This spot can be changed somewhat, to attenuate the response of certain undesired targets. The UNIPI timing control allows for precise sampling time adjustments.

                        #2 signal pulse is taken at a later time, at the point in time where the signal has the least target response. It is also the sample that we use to adjust the GB. Moving this point in time a little bit earlier or later allows for precise Ground Balance adjustments.
                        This sample also compensates for the EF, the signal generated by the coil moving through the Earth’s magnetic field.

                        At the output of the analog switch, the sample has an offset of about 3.5V. For now, we simply use an AC coupling to remove the offset.

                        Then we amplify the 2 sampled signals. Let’s try a gain of 1000.

                        At his point, with the AC coupling, we only need the delta of the motion by the coil over the target. Say 10Hz maximum for a coil of this size. Therefore we restrict the bandwidth of the amplifier to that frequency.

                        Now, let’s look at the output. Do we have a target signal?

                        Picture #1, NO TARGET_TOTAL GAIN 5500
                        Picture #2, Silver dollar at 28cm
                        Picture #3, Steel bolt at 28cm.


                        The oscilloscope is set to 100mV/div., so we see about 100mV of signal for the silver dollar at 28cm.
                        The Steel bolt gives very little signal, so we do not want to use this setup when searching for steel targets.

                        Tinkerer
                        Attached Files

                        Comment

                        Working...
                        X