Announcement

Collapse
No announcement yet.

DEEPER PI DETECTION DEPTH

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

  • #76
    Originally posted by Midas View Post
    Hi All,

    Well this has been somewhat enlightening for me. It would seem that a faster switch off is in fact allround beneficial. Intuitively equating inductance with momentum it seemed to make sense that superfast pulses might not be the best way to transfer a given quantity of energy. Kind of like trying to get a car moving by shooting bullets at it. At some very high frequency I'm sure that would prove to be true. ie. if you tried to pulse it faster than the resonant frequency of the target nugget. But since thats probably somewhere in the MHz (anyone ever tried to work it out?) it can probably be completely ignored. And since Aziz has quite rightly pointed out that increased skin effect is less detrimental than the benefits of higher frequency, bring on the near vertical current ramps...

    Carl, I'm not sure alumimium foil is much good as a 'standard'. 0.022 millimetres is apparently typical thickness but I highly doubt its universal across all brands. Coins are an obvious choice for consistancy though I guess there not small enough for low TC nugget simulation.

    Midas
    I think we should keep looking at the TX pulses for a bit longer, from all different angles.

    For a traditional PI TX pulse, we can consider each pulse as an individual transient event.

    For example:

    The eddy currents generated by a linear pulse ramp penetrate about 2mm deep into a gold nugget, within about 250us.
    When we switch off, the current inverses, we could say it comes out of the gold nugget. The rate of change of the magnetic field is now much higher that when it went in, therefore the eddy currents it generates, are of much higher amplitude.

    Does this make sense?

    I am sure we can find a few more angles to look at it. Let's see which one we decide to be the most correct one.

    It also would be very interesting to build a real accurate LTSpice model for simulating PI detectors. The spice model I use above, gives results that are roughly similar to the results obtained with a real part circuit, but it could be improved a lot.

    Is somebody willing to help me with building a more perfect PI LTSpice model?

    Tinkerer

    Comment


    • #77
      Originally posted by Tinkerer View Post
      The eddy currents generated by a linear pulse ramp penetrate about 2mm deep into a gold nugget, within about 250us.
      When we switch off, the current inverses, we could say it comes out of the gold nugget. The rate of change of the magnetic field is now much higher that when it went in, therefore the eddy currents it generates, are of much higher amplitude.

      Does this make sense?
      No, I don't think it does. The initial current ramp is creating eddy currents in the target that at switch off must effectively be cancelled before the current can begin to flow the other way. So your final eddy current amperage would actually be less than if let the current flat top allowing eddy currents in the target to settle before you commence switch off. I doubt it has that significant a negative effect, on account of the switch of di/dt being so much larger but I can't see it being a positive thing.

      I doubt I'll be much help on creating a better LTspice target since I know very little about the program, but if I can help I will. I am curious as to the variables included in your current simulated targets. What do you have inductance, capacitance, resistance, surface area?

      Comment


      • #78
        Originally posted by Midas View Post
        No, I don't think it does. The initial current ramp is creating eddy currents in the target that at switch off must effectively be cancelled before the current can begin to flow the other way. So your final eddy current amperage would actually be less than if let the current flat top allowing eddy currents in the target to settle before you commence switch off. I doubt it has that significant a negative effect, on account of the switch of di/dt being so much larger but I can't see it being a positive thing.

        I doubt I'll be much help on creating a better LTspice target since I know very little about the program, but if I can help I will. I am curious as to the variables included in your current simulated targets. What do you have inductance, capacitance, resistance, surface area?
        Yes, of course, the switch off eddy currents will cancel the switch on eddy currents. But how can we quantify the effort needed to do that?
        What would be a good test method for quantifying the proportion of response loss due to the cancelling of switch on eddy currents?

        I am only at the beginning of the learning curve for LTSpice. So far I am using resistance and inductance to simulate the target metal and target size.
        The experiment with adding target capacitance has not yielded results. I also tried some magnetic response. This seems to give some kind of reasonably similar results to a real component response.

        Volume? surface area? How to do that?

        Tinkerer

        Comment


        • #79
          It is true that the eddies from an unsettled turn-on will partially cancel the turn-off eddies, but I also doubt it is very significant. It would be interesting to test for this, by using a series resistor to flat-top the current on a wide TX pulse width, then removing the series R and reducing TX pulse width until the peak coil current at turn-off is the same as the flat-top current.

          Comment


          • #80
            Originally posted by Carl-NC View Post
            It is true that the eddies from an unsettled turn-on will partially cancel the turn-off eddies, but I also doubt it is very significant. It would be interesting to test for this, by using a series resistor to flat-top the current on a wide TX pulse width, then removing the series R and reducing TX pulse width until the peak coil current at turn-off is the same as the flat-top current.
            You could also argue that, although the unsettled current opposes the turn-off eddies, it is really the di/dt that matters, and reversing the unsettled current is part of the total di/dt and has the same sign (no reversal) -- potentially could even help, although my guess is it gives about the same di/dt function, just starting from a different initial condition.

            -SB

            Comment


            • #81
              It may be possible to grow the rising edge slowly - to very short flat top, then abrupt shut off as normal. Easy with an DtoA from a micro


              1)Rising edge introduces little or no target eddys (di/dt rising-edge = small)
              2)Saves battery life (current/area under the curve)

              Steve

              Comment


              • #82
                OK, I've had my laugh now..

                Will SOMEONE PLEASE tell me WHY people have an insistence of putting a micro in a PI?

                You are looking for nV level signals and to be honest none of the PCB's I've seen here so far (with ALL due respect to the hard, excellent work done by those who post them) are what might be termed "low noise designs".

                Try this, run a PIC in a SIMPLE timing loop for a PI and put it next to an AM radio receiver. Listen to how much noise there is! Well, ALL of that is getting coupled into your PI RX side too (as it's a broadband system by nature). Sooooooo, keep micros of ANY sort WELL away from PI designs, go for a low noise PCB layout (WATCH OUT for ground loops especially) and you wont go far wrong.

                Oh and remember, DECOUPLE EVERYTHING!

                Comment


                • #83
                  You do have a point, micros emit a load of noise.

                  You have to be clever to get rid of it/reduce it to a tolerable level.

                  It can be weeks of work too. Months on complex radios with heavy processor power / data flying round.

                  I can do this stuff but Ive had years of fiddling with filters, screens, copper tape, conductive polymers, layout care and have ulcers and grey hair from the experience of doing it for customers like Tosh, HP, Siemens, etc.

                  S

                  Comment


                  • #84
                    Originally posted by golfnut View Post
                    You do have a point, micros emit a load of noise.

                    You have to be clever to get rid of it/reduce it to a tolerable level.

                    It can be weeks of work too. Months on complex radios with heavy processor power / data flying round.

                    I can do this stuff but Ive had years of fiddling with filters, screens, copper tape, conductive polymers, layout care and have ulcers and grey hair from the experience of doing it for customers like Tosh, HP, Siemens, etc.

                    S
                    Would you like to apply all this experience into a fun project? To design an extreme deep detecting machine?
                    They tell me that having a lot of fun heals the ulcers and prolongs life.

                    Tinkerer

                    Comment


                    • #85
                      Originally posted by Sean_Goddard
                      ***** Deleted at request of poster - sensitive information. *****
                      You should sell the secret, not the MD! That's the best pitch I've heard!

                      -SB
                      Last edited by Qiaozhi; 01-27-2012, 09:47 PM.

                      Comment


                      • #86
                        Originally posted by Sean_Goddard View Post
                        *** Deleted at request of poster - sensitive information. ***
                        LOL. ... I would have to disagree ..... Of course digital circuits generate hash. ... The problem is that the analogue ccts pick this up ...oh I see what the problem is ... It's the analogue ccts . So really it's the analogue ccts not the CPU that is the problem.

                        Moodz
                        Last edited by Qiaozhi; 01-27-2012, 10:21 PM.

                        Comment


                        • #87
                          *LOL*

                          No, no!

                          It's the coil, that picks up the noise.
                          *LOL*

                          Aziz

                          PS: µC noise isn't much an issue. Can be handled well.

                          Comment


                          • #88
                            Actually it's BOTH!!

                            Going back to what I said, if the PCB's are designed properly then there should be much less of a problem. If this weren't the case, why do most of the units I've built from on here sound off when you tough parts of exposed metalwork and WHY even when using a genuine search head, do most of these machines have poor performance?

                            No criticism of the hard work people do, but SOMETHING MUST BE MISSING in the final analysis. That's a fact!

                            Comment


                            • #89
                              You could try an LTSpice simulation to see if you could demonstrate a significant advantage to locating the cap near the coil.

                              Again I wonder if one advantage of putting the cap on the PCB is to filter out some cable noise.

                              There could be some advantage to putting a high-Q TX coil-cap tank in the search head to minimize the driving current in the cable -- that is, if you ever wanted to play with high-Q TX tanks, which may have some disadvantages such as ground sensitivity.
                              Sean your clearly a real expert.
                              Steve

                              Comment


                              • #90
                                No, No expert mate, just pragmatic about what I THOUGHT I knew, which turned out to be JACK S@*T.

                                I thought "It should be REALLY easy to design a good deep PI" so after AMY different builds, ALL with **** poor performance, I decided to run some sims of existing machines to see what got. Pretty similar results as all the machine had pretty similar performance.

                                I looked at what I had and made a few changes, a few did sod all, BUT most did. the opposite to what I expected. SO much for theory!

                                SO I want back and did the OPPOSITE of what I thought would work and.....I started to get results.

                                Moving from simulation (Matlab, Simulink) to breadboard proved I was on the right track, so I do more of what I "wasn't doing" and even better results appeared. Then I over did it and it stopped working altogether.

                                Backing off a little got me to an "edge" where the machine was extreme in depth, but sometimes broke into oscillation, backing off 25% made it stable but lost a little depth. Refining what I was left with will be the machine I go with.

                                Comment

                                Working...
                                X