Advanced PI coils

[bbsailor]

Hi Midas,

That is interesting about the Kynar having a bad DC. I just assumed that having a fancy name it must be better than PVC. Goes to show - don't just assume anything; check! All my coils now are wound with Teflon wire as I bought several large reels of different gauges from a company that closed down. I also have some stranded Litz of the same equivalent gauge as 0.25 solid. I wound two otherwise equal coils and saw no discernable difference in performance. For higher current pulses I use stranded Teflon insulated high temperature wire. This paper, and others in the references, could explain why Litz doesn't give any advantage in PI detector coils.

http://engineering.dartmouth.edu/ind...trandedopt.pdf

What is the resonant (ringing) frequency of your coil + cable.

Eric,

It is good to see you active on this forum again.

Back in 2006 when I wrote this article http://www.geotech1.com/pages/metdet...s/FastCoil.pdf , I tried making a coil with Kynar wire just to compare it to PVC and Teflon insulation. Kynar had a higher dielectric constant (DC) and thus made a coil with a lower self resonant frequency than Teflon but it also had a very thin layer of insulation compared to the others I tried. PVC has a DC that can range from about 4 to over 6. Just using a different insulation, the same coil size went from about 700KHz using PVC to 1.25MHz (indicating less coil capacitance) using Teflon insulation.

When I tried using basket weave techniques to reduce coil capacitance, I lost all benefits because the shielding of this larger wire bundle tended to increase the coil-to-shield capacitance. Stranded Teflon insulated wire seems to be the most practical way to make a coil using the thinnest wire bundle consistent with obtaining the desired TX pulse coil time constant.

Another interesting thing I discovered is that the full coil-to-shield capacitance is not imposed on the coil's self resonance. Only about 20% of the coil-to-shield capacitance causes the coil's self resonance to be lower.

Eric, do you find that Litz wire will only show its value in the very low delays of around 5 to 6 microseconds? For getting delays down to about 8 to 10 microseconds, thin strands of tin plated AWG32 seems to work good enough? The tin plating appears to provide enough resistance between strands to help break up detectable eddy currents from being generated in the coil wire itself. The Minelab coil experts also seem to recommend AWG32 as being the optimum wire strand size. However, some aftermarket Minelab coils use Litz wire. Do you have any ideas why?

I have enjoyed learning about PI metal detectors and coils mostly from your early PI classroom forum.

Thanks

bbsailor

[Midas]

Eric,

It is good to see you active on this forum again.

Back in 2006 when I wrote this article http://www.geotech1.com/pages/metdet...s/FastCoil.pdf , I tried making a coil with Kynar wire just to compare it to PVC and Teflon insulation. Kynar had a higher dielectric constant (DC) and thus made a coil with a lower self resonant frequency than Teflon but it also had a very thin layer of insulation compared to the others I tried. PVC has a DC that can range from about 4 to over 6. Just using a different insulation, the same coil size went from about 700KHz using PVC to 1.25MHz (indicating less coil capacitance) using Teflon insulation.

When I tried using basket weave techniques to reduce coil capacitance, I lost all benefits because the shielding of this larger wire bundle tended to increase the coil-to-shield capacitance. Stranded Teflon insulated wire seems to be the most practical way to make a coil using the thinnest wire bundle consistent with obtaining the desired TX pulse coil time constant.

Another interesting thing I discovered is that the full coil-to-shield capacitance is not imposed on the coil's self resonance. Only about 20% of the coil-to-shield capacitance causes the coil's self resonance to be lower.

Eric, do you find that Litz wire will only show its value in the very low delays of around 5 to 6 microseconds? For getting delays down to about 8 to 10 microseconds, thin strands of tin plated AWG32 seems to work good enough? The tin plating appears to provide enough resistance between strands to help break up detectable eddy currents from being generated in the coil wire itself. The Minelab coil experts also seem to recommend AWG32 as being the optimum wire strand size. However, some aftermarket Minelab coils use Litz wire. Do you have any ideas why?

I have enjoyed learning about PI metal detectors and coils mostly from your early PI classroom forum.

Thanks

bbsailor

Hi bbsailor,

Are you sure you can even get tin plated stranded Teflon insulated wire?
This somewhat contradictory website seems to think that it would either stick together during the higher temperature extrusion process or redissolve back into the copper during its high temperature service. Not sure how they can know the latter if they never made any though..
http://www.logwell.com/tech/shdwe/teflon_wire.html
So they seem to think the choices of coating are either silver or nickle. Silver probably negates any benefit (apart from being more durable) of being stranded since its so conductive and nickle sounds like it would be bad because its magnetic, although perhaps its so thin it doesn't matter. Anyway silver plated is all I've managed to find so far.

Midas

[Ferric Toes]

Magnetic Viscosity Plots of Different Materials 1.

Here is a report that shows how frequency dependent susceptibility relates to the viscosity response from a PI metal detector. The plots show the log amplitude/log time relationship of this type of magnetic decay, and that it falls in a very narrow band around t^-1 for the different materials tested. I hope to get even greater acccuracy shortly by using a uP timed viscosity meter. It appears to be these small variations in slope that affect the accuracy of the GB setting of a detector.

The Chico soil from California is reported to give VLF machines the fits, but you can see from the susceptibility data that the Wedderburn ironstone has 12 times the reading.

Eric.

[Ferric Toes]

More information on soils and rocks.

This information gives coil and electronics designers an insight into magnetic ground that can cause problems. PI IB coils are particularly vulnerable in ground with high susceptibility levels and the ON time signal is processed.

Eric.

MVM Report 2002.pdf

[Ferric Toes]

Eric,

Eric, do you find that Litz wire will only show its value in the very low delays of around 5 to 6 microseconds? For getting delays down to about 8 to 10 microseconds, thin strands of tin plated AWG32 seems to work good enough? The tin plating appears to provide enough resistance between strands to help break up detectable eddy currents from being generated in the coil wire itself. The Minelab coil experts also seem to recommend AWG32 as being the optimum wire strand size. However, some aftermarket Minelab coils use Litz wire. Do you have any ideas why?

I have enjoyed learning about PI metal detectors and coils mostly from your early PI classroom forum.

Thanks

bbsailor

Hi bbsailor,

It's good to talk again, and I have been quite enthused by things that are happening here in PI. I haven't found Litz wire to be useful at all in PI so far. Tin plated stranded wire works fine down to very short delays. I use 10/0.1 stranded in small coils down to 1uS delay, although for a production unit I used a multilayer printed coil with a fine track. This was back in the early '90s when I worked for Pulse Technology on industrial PI. Silver plated stranded is no good as the inter-strand resistance is too low.

My coil of the moment is Moodz bifilar wound differential mono. This I think has great potential and I have been testing one today on a Goldquest V3 circuit board. I modified the NE5532 dual stage preamp to have a differential input with a total gain of 400. TX was one end to centre tap ground and RX was total coil across the two inputs with CT ground. I had damping across each coil half which included a cermet preset for fine adjustment. Peak V across the damping was about 60V so no avalanche, shock, or sparking over to worry about. Damping wasn't critical and I was able to sample at 7.5uS and be well clear of the "knee". This setup is good because you can have a very fast TX as there is only 80uH inductance, but the total inductance as seen by the preamp is 320uH. I have a 27R series resistor in the TX circuit so the current flat tops in 13uS. Pulse rate is 13Kps (kilopulses per sec) so S/N is good after integration.

Although from one aspect screening the coil is not as necessary because of the common mode rejection, I have screened this coil to reduce higher frequency differential mode radio pickup. It also reduces higher harmonic emissions to keep the EMC boys happy. It certainly looks clean and quiet on the scope, but I have yet to compare it with a similar standard mono to see how significant the improvements are. I'm sure it can be better with an instrumentation type differential preamp and faster switching off of the TX. I seem to lose a couple of uS here.

Long live the mono, in its different(ial) clothing. Top Hats not allowed .

Eric.

[6666]

Eric
did you happen to catch this thread when it was active ?
some interesting results were being discussed.
At least in the first few pages.
http://www.geotech1.com/forums/showt...ighlight=byv28

[Davor]

I'm sure it can be better with an instrumentation type differential preamp

A simple two transistor differential amplifier can do that beautifully, and with acceptable CMMR. There are several configurations that can compare to THAT amplifiers, yet for a fraction money and time to come by.

[Ferric Toes]

Eric
did you happen to catch this thread when it was active ?
some interesting results were being discussed.
At least in the first few pages.
http://www.geotech1.com/forums/showt...ighlight=byv28

Hi 6666,
No, I had not seen that thread. Back about that time I did not inhabit forums much, and therefore missed out on some good info.

I see the argument about the relevance of the back emf spike had raised its head again. It is the rate of cutoff of the TX field that is important, and the spike is just the effect of that relative to the inductance of the coil. However, faster and faster cutoff becomes irrelevant when it is 10X less that of the object TC. I am losing a bit of speed at the moment because the gate drive to the mosfet is not switching the gate off fast enough - gate capacitance and all that. I will attend to that today and see if I can sample at 5uS after the initiation of cutoff. Sooner than that is OK for some industrial applications, but too sensitive for some hobby uses. e.g. saltwater beaches or many fields in UK, where you would forever be finding fragments of metal you can hardly see. You could of course have a variable delay (say 5 - 50uS), but many users just crank it to max sensitivity and not use it intelligently.

Eric.

[Aziz]

Some Insight into 3D Vector Analysis

Hi all,

I have been very busy with the coil software upgrade.

I did find (hopefully) a good approximation of aligning the target coil into an optimum response orientation (good inductive coupling to both TX and RX coils - note: complex & arbitrary coils). But don't know, whether it is correct or not (see below - sorry for German comments and mixed language style). The code is giving reasonable orientation proposals but didn't check and test everything. I will need more time, that's for sure.

I just want to show you, what kind of work is involved (B-fields, vectors, matrix, dotproduct, crossproduct, sin, cos..).
(If you find a bug, just let me know - Ok? )

..
                  // Magnetfelderregung der Spulen am Punkt P ermitteln:
                  //   Spulen-Strom gleich, daher egal welcher Wert -> pElectric
                  // TX B-Feld am Punkt P ermitteln
                  BFeldSpule(_BTX, P, pTXSpule, pElectric);

                  // RX B-Feld (ja RX) am Punkt P ermitteln
                  BFeldSpule(_BRX, P, pRX, pElectric);

                  // optimale Achse (maximale Kopplung) berechnen : B(opt) = B(TX) + B(RX)
                  VekAdd(_Bopt, _BTX, _BRX);
                  //EinheitsVektor(_Bopt);

                  // Spulenachse nun an optimale Achse ausrichten (rotieren) und 
                  //   TG auf Position setzen
                  RotateToAxisAndOffsetCoil(P, _Bopt, pTG);


..

//-----------------------------------------------------------------------------
// Spulenachse an eine neue Achse ausrichten (rotieren) und an Position vP setzen
// vP:    Position
// vA:    Achsenvektor (muss nicht Einheitsvektor sein)
// pCoil: Zeiger auf Spule
BOOL RotateToAxisAndOffsetCoil(LPVEK vP, LPVEK vA, LPSPULE pCoil)
{
  VEKTOR  vR;
  LEITER *pLeiter;
  int     i, nLeiter;
  double  cosa, w;
  MATRIX  MR;

  if (pCoil == NULL)
    return FALSE;
    
  pLeiter = WireList_Base(&pCoil->WireList);
  nLeiter = WireList_Count(&pCoil->WireList);
  
  // für alle Leiterelemente der Spule
  for (i=0; i < nLeiter; i++, pLeiter++)
  {
    // Leiterelementkoordinaten (bezogen auf Spulenmitte M) in den 
    //   Ursprung (0/0/0) verschieben
    //
    #ifdef _HFELD_KORREKTUR
    VekSub(pLeiter->LK, pLeiter->LK, pLeiter->M);
    #endif
    VekSub(pLeiter->L1, pLeiter->L1, pLeiter->M);
    VekSub(pLeiter->L2, pLeiter->L2, pLeiter->M);
    //VekSub(pLeiter->M,  pLeiter->M,  pLeiter->M); // wird zu 0-Vektor
    VekZero(pLeiter->M); // besser so

    // Rotationswinkel ermitteln
    //  Winkel zwischen Vektoren Spulenelementnormale dF0 und Achsenvektor vA
    //    cos a = Skalarprodukt(dF0, vA) / (|dF0|*|vA|)
    cosa = CosAlphaVek(pLeiter->dF0, vA);
    w    = NACH_GRAD * acos(cosa); // Winkel in Gradeinheit

    // Rotationsachse ermitteln (genialer Einfall)
    VektorProdukt(vR, vA, pLeiter->dF0);

    // Rotationsmatrix erstellen (Drehung an Rotationsachse um Winkel w)
    Rotiere_an_Achse(MR, vR, -w);

    // alle Koordinaten und Vektoren um die Rotationsachse rotieren
    //
    
    // absolute Koordinaten
    #ifdef _HFELD_KORREKTUR
    Matr_Mul_Vek(pLeiter->LK, MR, pLeiter->LK);
    #endif
    Matr_Mul_Vek(pLeiter->L1, MR, pLeiter->L1);
    Matr_Mul_Vek(pLeiter->L2, MR, pLeiter->L2);
    Matr_Mul_Vek(pLeiter->M,  MR, pLeiter->M);
    
    // Vektoren
    Matr_Mul_Vek(pLeiter->r,  MR, pLeiter->r);
    Matr_Mul_Vek(pLeiter->ds, MR, pLeiter->ds);
    Matr_Mul_Vek(pLeiter->dF0,MR, pLeiter->dF0);

    // Leiterelement (Koordinaten) auf Position vP verschieben
    //
    #ifdef _HFELD_KORREKTUR
    VekAdd(pLeiter->LK, pLeiter->LK, vP);
    #endif
    VekAdd(pLeiter->L1, pLeiter->L1, vP);
    VekAdd(pLeiter->L2, pLeiter->L2, vP);
    VekAdd(pLeiter->M,  pLeiter->M,  vP);
  
  } // for i
  return (nLeiter != 0);
}

//-----------------------------------------------------------------------------
// Führt eine Rotation an Rotationsvektor v um Winkel phi (Grad)
// -> Rotation um beliebige Achse (=Vektor v)
//  return: Rotationsmatrix M
void Rotiere_an_Achse(MATRIX M, LPVEK v, double phi)
{
  double x,y,z, xy, xz, yz, cosphi, sinphi, one_mi_cosphi;
  VEKTOR _v;
 
  // Vektor v muss Einheitsvektor sein
  VekCopy(_v, v);
  EinheitsVektor(_v);
  x  = _v[VX];
  y  = _v[VY];
  z  = _v[VZ];
  xy = x * y;
  xz = x * z;
  yz = y * z;
  phi          *= NACH_RAD;
  cosphi        = cos(phi);
  sinphi        = sin(phi);
  one_mi_cosphi = 1.0 - cosphi;
  // Rotationsmatrix aufbauen
  //
  Einheitsmatrix(M);
  // Zeile 1
  M[VX][VX] = x*x * one_mi_cosphi +   cosphi;
  M[VX][VY] = xy  * one_mi_cosphi - z*sinphi;
  M[VX][VZ] = xz  * one_mi_cosphi + y*sinphi;
  // Zeile 2
  M[VY][VX] = xy  * one_mi_cosphi + z*sinphi;
  M[VY][VY] = y*y * one_mi_cosphi +   cosphi;
  M[VY][VZ] = yz  * one_mi_cosphi - x*sinphi;
  // Zeile 3
  M[VZ][VX] = xz  * one_mi_cosphi - y*sinphi;
  M[VZ][VY] = yz  * one_mi_cosphi + x*sinphi;
  M[VZ][VZ] = z*z * one_mi_cosphi +   cosphi;
}

Bugger me!, I hate math! Really!

Cheers,
Aziz

[bbsailor]

Hi bbsailor,

It's good to talk again, and I have been quite enthused by things that are happening here in PI. I haven't found Litz wire to be useful at all in PI so far. Tin plated stranded wire works fine down to very short delays. I use 10/0.1 stranded in small coils down to 1uS delay, although for a production unit I used a multilayer printed coil with a fine track. This was back in the early '90s when I worked for Pulse Technology on industrial PI. Silver plated stranded is no good as the inter-strand resistance is too low.

My coil of the moment is Moodz bifilar wound differential mono. This I think has great potential and I have been testing one today on a Goldquest V3 circuit board. I modified the NE5532 dual stage preamp to have a differential input with a total gain of 400. TX was one end to centre tap ground and RX was total coil across the two inputs with CT ground. I had damping across each coil half which included a cermet preset for fine adjustment. Peak V across the damping was about 60V so no avalanche, shock, or sparking over to worry about. Damping wasn't critical and I was able to sample at 7.5uS and be well clear of the "knee". This setup is good because you can have a very fast TX as there is only 80uH inductance, but the total inductance as seen by the preamp is 320uH. I have a 27R series resistor in the TX circuit so the current flat tops in 13uS. Pulse rate is 13Kps (kilopulses per sec) so S/N is good after integration.

Although from one aspect screening the coil is not as necessary because of the common mode rejection, I have screened this coil to reduce higher frequency differential mode radio pickup. It also reduces higher harmonic emissions to keep the EMC boys happy. It certainly looks clean and quiet on the scope, but I have yet to compare it with a similar standard mono to see how significant the improvements are. I'm sure it can be better with an instrumentation type differential preamp and faster switching off of the TX. I seem to lose a couple of uS here.

Long live the mono, in its different(ial) clothing. Top Hats not allowed .

Eric.

Eric,

it seems that you derive additional sensitivity by operating at 13Kpps and then integrating those RX samples to improve the signal to noise ratio. The PI MD design is then extracting a weak target signal burried in noise by acting like a "Lock-in Amplifier". http://electron9.phys.utk.edu/optics...m10/tn1000.pdf

With a 13Kpps pulse you would have a square wave of 1/13000 or 76.9 uS divided by 2 or about 38.5uS for a square wave. So if the waveform is flat topping then it must be at or near 5 Time Constants on the coil charge side. 80uH/30 ohms (27 ohm resistor plus coil resistance and MOSFET on- resistance) or about 2.6 us so the a pulse width of 7.8 uS provides about 85 percent max current and 13 uS provided 100 percent of max current. As you mention the S/N is good after integration because your High Frequency design makes up for the "Brute Force" method of putting a lot of current in the coil by now emphasizing the receive side of detecting small target signals by integrating many signals.

Is there an optimum number of RX signals to integrate?

At 13Kpps scanning using a 10 inch actual size coil a sweep width of 40 inches or approximately 1 meter per second puts a target under the coil for about one fourth of the time or 13000/4 or 3250 pulses being integrated.

Are all size targets equally sensitive to sweep speed or does changing sweep speed effectively retune the lock-in amplifier filtering effect to favor different target sizes and TCs or affect some target sizes and TCs the least?

One of last areas of PI coil design has been ignored so far. I believe that putting the first amplifier/buffer stage in the near the coil will allow faster pulses to be integrated by lowering the capacitance seen by the coil by active RX buffering at, in, or near the coil. When the damping resistor is a higher value it means that there is less capacitance to damp and thus the coil discharge curve is faster and thus can better stimulate smaller targets. A 320 uH coil that is center taped will have each coil be 1/4 the total inductance or 80uH. What value damping resistors do you have across each coil? I suspect the sum of the two is a higher value than a single damping resistor would be on a typical mono PI coil.


Thanks

bbsailor

[Ferric Toes]

I hate maths too but programming like the above is a step too far. Still it's good to see Aziz having a go, and coming up with some interesting results. I was personally pleased to see that the Dual Field Coil came out with some benefits exactly the way Whites and I had hoped. Didn't lose much in depth, but improved on the pinpointing.

I have had a less brain numbing day doing a few more tests with the bifilar differential arrangement. I think now that I am losing some time in the slower than optimum recovery of the 5532. I'll try a couple of LME49990 as I have already seen an improvement in one of my other detectors by using these. Anyway, here are some pictures of scope traces with the cursor at 7uS. Which is where I am sampling at the moment. See the benefit of early sampling for the 0.5gm nugget where there is nothing left after 20uS. The 0.9gm nugget goes on for about twice as long. Both nuggets were close to the coil winding to get sufficient signal.
The tray of gravel, which covered most of the coil area shows a typical viscosity response. An initial fast decay which slows up as time goes on. Even at the far right of the screen, it is still not quite back to the zero line.
The last picture is the no signal response of the coil and preamp. The spikes along the trace are breakthrough from the clock generator. Some of the wires going out to the controls (frequency control especially) should be shielded to prevent this. Anyway, this is just a lashup on the bench to see what works. The spikes don't interfere as they are synchronous.

Eric.

[Ferric Toes]

Eric,

it seems that you derive additional sensitivity by operating at 13Kpps and then integrating those RX samples to improve the signal to noise ratio. The PI MD design is then extracting a weak target signal burried in noise by acting like a "Lock-in Amplifier". http://electron9.phys.utk.edu/optics...m10/tn1000.pdf

With a 13Kpps pulse you would have a square wave of 1/13000 or 76.9 uS divided by 2 or about 38.5uS for a square wave. So if the waveform is flat topping then it must be at or near 5 Time Constants on the coil charge side. 80uH/30 ohms (27 ohm resistor plus coil resistance and MOSFET on- resistance) or about 2.6 us so the a pulse width of 7.8 uS provides about 85 percent max current and 13 uS provided 100 percent of max current. As you mention the S/N is good after integration because your High Frequency design makes up for the "Brute Force" method of putting a lot of current in the coil by now emphasizing the receive side of detecting small target signals by integrating many signals.

Is there an optimum number of RX signals to integrate?

At 13Kpps scanning using a 10 inch actual size coil a sweep width of 40 inches or approximately 1 meter per second puts a target under the coil for about one fourth of the time or 13000/4 or 3250 pulses being integrated.

Are all size targets equally sensitive to sweep speed or does changing sweep speed effectively retune the lock-in amplifier filtering effect to favor different target sizes and TCs or affect some target sizes and TCs the least?

One of last areas of PI coil design has been ignored so far. I believe that putting the first amplifier/buffer stage in the near the coil will allow faster pulses to be integrated by lowering the capacitance seen by the coil by active RX buffering at, in, or near the coil. When the damping resistor is a higher value it means that there is less capacitance to damp and thus the coil discharge curve is faster and thus can better stimulate smaller targets. A 320 uH coil that is center taped will have each coil be 1/4 the total inductance or 80uH. What value damping resistors do you have across each coil? I suspect the sum of the two is a higher value than a single damping resistor would be on a typical mono PI coil.

Thanks

bbsailor

Hi bbs,

It is a little different in that the TX pulse, when sampling at 7uS, is 21uS wide and looking at the voltage, or current, waveform, it has just about reached its max at shut off. It I drop the clock frequency so that the sampling delay is 10us, then the TX width is 30uS. Always 3X. So for a 20uS delay, TX would be 60uS. As I have it set up, with a delay of 7uS, then the period is 70uS i.e. 14.28 kps. A little up on yesterday due to driving the mosfet better and winning a uS.

I have only 1ft of lead between coil and electronics so if I had a standard cable length, then things would be degraded. I just have three wires plaided as it is a differential system. For fast industrial systems the coil was wired to the electronics by no more than two or three inches of cable. OK as long as there is no relative movement and components with plated steel wires kept at a distance. 1N4148s for instance. Also electrolytic caps with ali cans.

Damping is a 1K fixed resistor in series with 1K cermet preset. One across each half of the winding. One opamp input is inverting so that presents a parallel resistor. Each preset has to be adjusted separately with the one on the non-inverting input seeming to be more critical. As I say, the preamp as it is, is not ideal and a better one will be built for this coil.

Response is very fast and I will probably increase the integrator TC a bit to smooth out the noise still further. I also have a switchable 1st order/2nd order LP filter after the integrator which sharpens up the overall response nicely for accurate pinpointing. With everything running and audio on phones, the current consumption is 125mA at 12V, whatever the TX pulsewidth and pulse rate.

Eric.

[mickstv]

Hi Eric, this is the diff amp circuit I've been using in my latest experiments. The resistor values are for reference only.

[Aziz]

Hi all,

I have bad news:
The simple assumption for the approximation of the target orientation doesn't work. It's obviously quite non-trivial and requires either much much more processing power by trying many many different random orientations (the Monte-Carlo Method) or another approximation idea.
Exact solution would be:
Solving very very complex integral equations.
(Who is going to dare it? Nooooo!, don't look at me! I'm not that smart. )

But the code for aligning the target orientation (aligning the target coil axis into a given vector-axis) works perfect.
(That wasn't trivial too. But a brilliant idea made it easy.)

The Monte-Carlo Method would blow up the processing time (although much easier to code). We should continue to use the fixed target orientation analysis until I get another brilliant idea or implement the Monte-Carlo Method.

Ok, what's left over? Any forgotton coil types? Wishes?
Let's invent together the "World's Best Coil Technology" (WBCT) now.

Cheers,
Aziz

[deemon]

Hi all,
Ok, what's left over? Any forgotton coil types? Wishes?
Let's invent together the "World's Best Coil Technology" (WBCT) now.

Believe me or not , Aziz , but the "world best coil technology" is just the simple monocoil with additional balance circuit ...