Chance PI coil...

[baum7154]

I was messing around with this coil design in a program called Maya. This python code might be useful for some of you.

import math

sin = lambda angle: math.sin(math.radians(angle))
cos = lambda angle: math.cos(math.radians(angle))
blend = lambda start, end, weight: (start * (1-weight)) + (end*weight)

def makeSpline():
    radiusData = [9.5, 11.5]
    thickness = 0.1
    totalTurns = 6
    notches = 37 * 2
    profileData = [1, 7, 8, 9, 8, 7, 1]
    notchSkip = 5

    offsetScale = thickness
    notchDegree = 360.0 / notches
    numProfilePoints = len(profileData)
    pointRot = notchDegree / (numProfilePoints - 1)

    allPoints = []
    accRot = flipFlop = count = 0

    for i in range(totalTurns * notches):
        flipFlop = i % 2

        if count>notches:
            count=0
            offsetScale+=thickness

            if radiusData[0]<radiusData[1]:
                radiusData[0]-=thickness
                radiusData[1]+=thickness
            else:
                radiusData[0]+=thickness
                radiusData[1]-=thickness

            accRot+=notchDegree*2
            allPoints.append([sin(accRot)*radius, 0, cos(accRot)*radius])

            radiusData = [radiusData[1], radiusData[0]]
                    
        else:
            for j in range(numProfilePoints):
                accRot+=pointRot*notchSkip
                if j==0: accRot-=pointRot*notchSkip
                weight = j/float(numProfilePoints-1)
                offset = profileData[j]*offsetScale/9

                if flipFlop:
                    radius = blend(radiusData[0], radiusData[1], weight)
                else:
                    radius = blend(radiusData[1], radiusData[0], weight)
                    offset*=-1

                allPoints.append([sin(accRot)*radius, offset, cos(accRot)*radius])
            count+=1

    #import pymel.all as pm
    #circle = pm.circle(radius=thickness/2.0)[0]
    #curve = pm.curve(d=2, p=allPoints)
    #pm.extrude(circle, curve, extrudeType=2, fixedPath=True, useProfileNormal=True, useComponentPivot=True)
    return allPoints

allPoints = makeSpline()
print allPoints

Does the reference to 'Notchskip=5' refer to the winding pattern where the wire is placed in every third notch?

If you can run the 8 X 6" ellipse can you run it for 30 turns?

THanks,

Dan

[baum7154]

More pictures, this time of an experimental Teflon insulated 23 AWG solid wire wound spider coil form, and a different graphite shielded coil wound with 26AWG solid enameled wire. The teflon insulated 23AWG solid wire came from a plenum rated Cat6 cable. The graphite shielding was a mixture of varnish and powdered graphite and was applied at a rate of 1300 to 1500 ohms/inch. All coils are 8" diameter. The shielded coil was fiberglassed with two layers of woven fiberglass mat and a lexan cross support and mounting ears were 'glassed in' just like the coil in the prior posting.

Theses polycarbonate coil forms were machined on a milling machine using a rotary index table and a 1/8" cutter. Using a cutter with a larger diameter than the thickness of the material causes less bending stress on the wire/insulation because the wire is bent less than 45 degrees as it is passed through the radial slots and also minimizes parallelism of alternate wires within a slot resulting in less capacitance. This process is preferable to just slotting the form material with a hacksaw. That said,hacksaw slots can be hand filed to depth using a 1/8" fine round file to achieve similar results.

[ATTACH]27961[/ATTACH][ATTACH]27962[/ATTACH][ATTACH]27963[/ATTACH]

I did find another image of the 1st layer of the fiberglass applied over the shielded coil described above. It is below. The 'serpentine skin' appearance of the coil is due to the hospital tape overlap of the shield layer and the inability of the resin to penetrate one layer but not both layers of the overlapped tape. After curing of this layer of fiberglass a Dremel tool was used to remove whiskers and to smooth the resin on the surface. A second fiberglass layer was then applied with subsequent smoothing and a thin finish layer of resin only was applied. Also attached are two more views of this finished coil.

Dan

[baum7154]

Reposted from the tech forum DAMPING RESISTOR thread as it belongs here too:

Just ran SRF on 4 coils

#1. 8" 335uh spider coil with 33" feed of twisted pair continuation of coil wire. Wire is 600 Volt PTFE insulation 24 AWG 19 strand twisted silver plated wound on .093" thick polycarbonate toroidal form. No shield
SRF =1.056 mhz.
NOTE: I was mistaken that this coil was wound with 26 AWG until I put a micrometer on it tonight...It is really 24 AWG wire. I'm sure 26 AWG would make it even faster.

#2. 8" 371uh Original Chance coil design with 48" RG-62B/U coax (13.5 pf/ft). Wire is 23 AWG solid copper with PTFE insulation on .093" polycarbonate form. No shield
SRF =1.065 mhz

#3. 8" 378uh Original Chance coil design with 35" continuation of twisted coil wire as feed. Wire is 26 AWG 7 strand twisted silver plated PTFE insulation on a .055" thick polyethylene form with No shield
SRF =1.251 mhz

#4. 8" 410uh spider coil with 34" of RG-62B/U. Wire is 26AWG enameled copper wire wound on a .093" thick polycarbonate toroidal coil form. Shielding is varnish/graphite applied at 1300 to 1500 ohms per inch.
SRF =825 khz

Dan

[baum7154]

Reposted from Tech Forum Damping Resistor thread as it belongs here too

the coil/feed capacitance calculated from the SRF above are:

#1 68pf,
#2 60pf
#3 43pf
#4 91pf

Removing the calculated feedline capacitance based on 13.5pf/foot of RG-62B/U and 1 pf/foot for PTFE insulated twisted pair gives the following coil capacitance:

#1 35pf
#2 6pf
#3 8pf
#4 52.75pf

#2 & #3 are the Original Chance coil design. It is hard to believe the capacitance numbers for the coils calculated without the feedlines. I'll run SRF on the #3 coil tonight with the feedline removed.

Dan

[baum7154]

Reposted from Tech Forum Damping resistor thread as it belongs here too

Removed feedline from #3 coil last night. Twisted feedline (26 AWG PTFE insulation)measures 31pf with my LC meter. Coil w/o feed measures 368.8 uh with LC meter. Did SRF on coil only - 2.435 mhz. Coil capacitance calculates from this to be 12 pf. Adding the feedline capacitance back into the calc gives SRF @1.263, within reasonable margin of error. Conclusion is that this form of coil with this wire is about 12pf.

[green]

Tested detector with 1110 ohm coil damping on 1/2' target tonight over wet grass and saw no degradation in signal strength from last night's test over wet grass. Got the same 8 signal strength units at maximum proximity to target. Coil is 8" diameter.

Other targets were tested in air in the basement as follows:
TARGET DISTANCE
1 sq in x .001" thick aluminum foil 9.5"

1/2 sq in Aluminum foil 8"

1/4 sq in Alum foil 7"

1 cm sq Alum foil 4"

US Nickel 12"

US copper Penny 9"

US silver dime 7"

10.5 grain dental gold 6"

8.4 grain dental gold 4"

First, your recent test shows it's a good way to make a low capacitance, fast coil. Was looking at this thread to see how to build one when I saw this reply. From some of the plots I've made I was surprised to see the distance you were getting with the 1/4 in square. I thought I was starting to understand how a PI worked. probably not. Including two plots I had done awhile back. Targets cut from aluminum can instead of aluminum foil. Five inch coil instead of eight. First plot shows more than a 2 inch lose between each size for the same amplitude. Sample was averaged between 5 and 10 usec. The second plot shows a difference in time constant. The amplitudes are about the same, but were taken at different distances. Depending when the sample is taken, I would think the distance difference might be greater. Maybe some one could explain where my thinking is wrong. Maybe it's my data or the hardware used to take the data. Any help appreciated.

[baum7154]

First, your recent test shows it's a good way to make a low capacitance, fast coil. Was looking at this thread to see how to build one when I saw this reply. From some of the plots I've made I was surprised to see the distance you were getting with the 1/4 in square. I thought I was starting to understand how a PI worked. probably not. Including two plots I had done awhile back. Targets cut from aluminum can instead of aluminum foil. Five inch coil instead of eight. First plot shows more than a 2 inch lose between each size for the same amplitude. Sample was averaged between 5 and 10 usec. The second plot shows a difference in time constant. The amplitudes are about the same, but were taken at different distances. Depending when the sample is taken, I would think the distance difference might be greater. Maybe some one could explain where my thinking is wrong. Maybe it's my data or the hardware used to take the data. Any help appreciated.

First of all I have not tried to detect a 1/4" X 1/4" target of either foil or aluminum can. 1/4" X1/4" is 1/16 of a square inch. I have detected a 1/4 square inch target of aluminum can and it is a great target for testing, measuring 1/2" X 1/2". Perhaps the language I used caused some confusion. Sorry if it did. If you are detecting targets 1/4" X 1/4" of aluminum can that is impressive. My most difficult standardized target so far is the 1/2" X 1/2' .001" thick aluminum foil target.

Regards

Dan

[green]

[ Perhaps the language I used caused some confusion.]

Thanks for the reply. Your language is correct. My error

[baum7154]

First, your recent test shows it's a good way to make a low capacitance, fast coil. Was looking at this thread to see how to build one when I saw this reply. From some of the plots I've made I was surprised to see the distance you were getting with the 1/4 in square. I thought I was starting to understand how a PI worked. probably not. Including two plots I had done awhile back. Targets cut from aluminum can instead of aluminum foil. Five inch coil instead of eight. First plot shows more than a 2 inch lose between each size for the same amplitude. Sample was averaged between 5 and 10 usec. The second plot shows a difference in time constant. The amplitudes are about the same, but were taken at different distances. Depending when the sample is taken, I would think the distance difference might be greater. Maybe some one could explain where my thinking is wrong. Maybe it's my data or the hardware used to take the data. Any help appreciated.

Still digesting your tests and questions. Chance uses a multipulse tx and from what I can gather about this it pumps up the field with many pulses before the coil is allowed to flyback. This has been reported to have a beneficial effect on the smaller targets. I will try to detect a 1/4" X 1/4" target tonight and let you know the result, if any. Right now I am struggling to determine the optimum damping resistor value for my Chance and this 8" coil. I have had values from 980 ohms to 1260 ohms and it looks like the answer lies in the range of 1020 ohms to 1150 ohms. I do know that the value does have an impact on discrimination and I want to keep that as accurate as possible while not sacrificing too much sensitivity.

Are you using a single pulse rate in your testing?

Dan

[green]

[Are you using a single pulse rate in your testing?]
The mono coil tests are done with an inverting amplifier with R in being R damping. Fixed coil on time usually between 50 to 100 usec. The IB coil tests use a 2 amp instrument amplifier connected to the Rx coil. For time constant tests I record a no target signal, a target signal and plot the difference. Distance plots are the change in volts out of the integrator/(amplifier*integrator gain). Sample average (5 to 10 usec)after fet gate off. The setup is for testing and learning on the bench. Been playing with circuits for GEB and discrimination. One of these days I need to try and build a PI detector. Have you tried to measure the CHANCE sample times yet?

[baum7154]

No I haven't but that is high on my list as I really need to get some definite correlation between Guard Interval values 10 through 99 and delay in microseconds. The designer has never given these values that are pretty critical. I first need to get my damping optimized and then move on to the sample times. One note: as I was trying to measure resonance of the ringing on my coil I set the Guard Interval to 99 and the fast spider coil ran without damping but I could see the ringing as it ran. Did not try to see how low GI would go before it quit working. All of my testing is done with damping resistor and set to a GI =10...wide open and the lowest this software will allow.

Dan

[6666]

The IB coil tests use a 2 amp instrument amplifier connected to the Rx coil.

Green is there a schematic on the web for what you are useing ?

[Tinkerer]

Green is there a schematic on the web for what you are useing ?

Many schematics there. The problem is matching the resistors and capacitors in the feedback.

https://www.google.ch/search?q=2+opa...w=1707&bih=838

[Tinkerer]

[Are you using a single pulse rate in your testing?]
The mono coil tests are done with an inverting amplifier with R in being R damping. Fixed coil on time usually between 50 to 100 usec. The IB coil tests use a 2 amp instrument amplifier connected to the Rx coil. For time constant tests I record a no target signal, a target signal and plot the difference. Distance plots are the change in volts out of the integrator/(amplifier*integrator gain). Sample average (5 to 10 usec)after fet gate off. The setup is for testing and learning on the bench. Been playing with circuits for GEB and discrimination. One of these days I need to try and build a PI detector. Have you tried to measure the CHANCE sample times yet?

I notice that your graphs show target response in the low micro volts. Obviously you have deduced these results by calculation. <volts out of the integrator/(amplifier*integrator gain)>

It might be a good time to take a closer look at the actual total gains of our detectors, as there are wide discrepancies between the numbers circulating on the web.
First the pre-amp stage: When an opamp saturates, the gain is not linear anymore. The recovery and settling time varies widely between different opamps.

Next we add LP filters which often reduce the gain for short TC targets considerably. We actually need to keep the cutoff frequency at higher than 200kHz for short TC targets.
Then we sample and integrate: If we take a 10us sample, we need 10,000 samples to get the full voltage in 100ms. Integrating 100 samples, we get 1% of the signal voltage. So if we add a gain of 100, we end up with no gain at all.

And so it goes on to the end of the processing.

By the way, using an IB coil and no saturation, very much more information about the targets becomes available.

[baum7154]

This a repost of a post in the Tech Forum - Damping Resistor

Ran coil #3 (an original CHANCE pattern coil) this AM on my CHANCE PI and with an adjustable network set at 1360 ohms this coil would run at a Guard Interval setting of 11, but never at the minimum of 10. Had to unwind a turn off the coil (now it is 368 uh) to get the solder joints out of the coil in order to get it to run at 11. Discrimination and sensitivity were not good. Large Gold Ring was represented at about 75% up the scale. This coil definitely needs shielding at the low GI settings.

Decided to put the prescribed 390 ohm damping resistor in and see how it behaved. The lowest Guard Interval it would run at was 18. At this setting it would not see 8.8 grain dental gold. It does see a 10.8 grain piece of dental gold at 1.5". Discrimination on a 6mm wide men's 10K wedding band is odd. Face on to the coil it displays 75% up the scale, but on edge it is solidly on the left end as Gold. Detection distance on the ring face on is about 9" and on edge to the coil is about 7"

This coil does not serve my purpose to find small gold. While it is quite fast it badly needs shielding but when that is applied it will lose much of it's speed. With it's broad flat profile even with 8 mm dielectric shield spacing it will be much diminished. A spider coil like #1 in post 48 of this thread is much better for my purposes.

Dan