Differential Coils ... Making

[moodz]

Have a look one page before (posting number 25):
http://www.geotech1.com/forums/showp...3&postcount=25

He is answering to someones questions.

Look before you think. Think before you post.

Aziz

Hey Aziz ... maybe this guy has discovered a new forum technique. We could ask our own questions then answer them ( in different colored text ) !!!! What a novel idea ... you never lose the argument or fail to have last word.

[Stefan]

Thanks Stefan I think your description above is partly correct however you are not taking the "centre tap" connection which is effectively at Analogue ground. With respect to this point the current in one coil is negative and the current in the other coil is positive ....or another way of visualising it is in one coil the current B flows OUT of the common point C and IN from A the other coil. So the current is not in phase as you describe. Check the dot notation of the connected coils and you will see that the current is opposite which is correct as the coils are closely coupled via mutual inductance. See diagram below.

The reason for sheilding on twisted pair is more to do with imperfect common mode rejection due to mechanical imperfection of the coil and the lack of "perfect" common mode rejection at the amplifier. If you talk to radio guys many old timers will swear by unshielded balanced lines over "shielded" coax for performance and low noise.


[ATTACH]14435[/ATTACH]

Hi Moodz,


It has taken some time but I have researched the usage of a centre tapped coil and common mode noise rejection. From what I have read there are advantages in using the centre tapped coil arrangement for common mode noise rejection. Will do some comparative tests to see what the improvement in performance is between a centre tapped coil and a single coil input into a differential amp but am basically convinced.

Regards,

Stefan

[moodz]

Hi Stefan,
I have tried the coil without a centre tap by using alternative bias however the noise did increase quite a bit. I guess if the coil had a full faraday shield it might be a fairer test ... however I am too lazy to make a shielded coil.

I used a variation of the circuit below ( which is a differential microphone preamp ) and does not require centre tap biasing for testing non centre tapped coils.

moodz

[Stefan]

Hi Moodz,

I would be using the INA163 Instrument amp in preference to the amp you have selected. The transmit arrangement is such that I can try with the centre tapped coil and compare results by removing the centre tap connection. The coil will still be damped and although the coil is floating it is referenced to ground so the instrument amp will function correctly provide CMR.

The arrangement you have will work fine although you could remove a lot of the redundant components that are not required. I suspect you have already done so. Once I get a prototype working will provide results for both however the tests will need carefull thought as the current in the centre tap arrangement will be quite high compared to the standard coil. But then again I don't need to put power into the coil to compare CMR.

The reason I have persevered with the standard arrangement is that I have a lot of coils that I have made and would like to see if I can get them to work using a differential Instrument amplifier rather than the standard single ended amp. I can make a centre tapped coil very easy with twisted litz wire and provide additional isolation for protection from the higher voltages, however will need another set of formers etc etc to do it right..

Regards,

Stefan

[moodz]

Hi Stefan,
you are fortunate to have those coils as a resource for testing and re-using is a good idea anyway. The INA163 is a fine device noise-wise however I find its high frequency response drops off at high gains ( eg 1000 ). I have found that unlike single ended designs the gain of the first amplifier can be set much higher with the differential coil because there is less noise .... the THAT1510 and 1512 exhibit bandwidths exceeding 3 Mhz at gains of upto 2000 and 1 nv noise. In a radio receiver the first amplifier determines the noise figure for the rest of the circuit ... I apply the same rule to detectors. However that said if the INA163 is used at lower gains then there is less difference .... I suppose it depends what parts you can get your hands on.

moodz.

INA163
LOW NOISE: 1nV/√Hz at 1kHz
LOW THD+N: 0.002% at 1kHz, G = 100
WIDE BANDWIDTH: 800kHz at G = 100
WIDE SUPPLY RANGE: ±4.5V to ±18V
HIGH CMR: > 100dB
GAIN SET WITH EXTERNAL RESISTOR
SO-14 SURFACE-MOUNT PACKAGE

THAT 1510
• Low Noise:
1 nV/√Hz input noise (60dB gain)
• Low THD+N (full audio bandwidth):
0.001% < 40dB gain
0.005% @ 60dB gain
• Low Current: 6 mA
• Wide Bandwidth: 7MHz @ G=100
• High Slew Rate: 19 V/µs
• Wide Output Swing:
±13.3V on ±15V supplies
• Current feedback (CFB) architecture
• Gain adjustable from -6 to >+60dB with one external resistor

[Stefan]

Hi Moodz,


I would keep the first stage gain to about 50 or so. So it will be just fine on BW. The specs on the THAT 1510 look good and has a high slew rate. However I not sure where I can get the beastie so I use what I can get. Although I will see if I can hunt down a source and get a few for a try.

With regard to the coils, I make them and have been since 1996.

Regards,

Stefan

[Stefan]

Hi Moodz,


Just checked, can get the THAT 1512 (SOIC) from Mouser for about $7.00 and they have them in stock. Have an account with them anyway.

Regards,

Stefan

[AdrianM]

Is it essential (or just beneficial) that the coil is wound flat - as per the Tesla picture - or can also be wound vertically? This would make the jig fully reusable as the coil could be slipped off the pins using a similar technique to the one shown here...



photo credit Alan Yates

[moodz]

Is it essential (or just beneficial) that the coil is wound flat - as per the Tesla picture - or can also be wound vertically? This would make the jig fully reusable as the coil could be slipped off the pins using a similar technique to the one shown here...

A good idea ... I have not tried this configuration but it should still work.


moodz.

[AdrianM]

Another way to form the coil

Inspired by J_Player's picture of a spider web winding, I came up with this:

The former is 2mm rigid polystyrene sheet (sold as glazing material in B&Q for UK readers). This has a dielectric constant of 2.6 - slightly lower than acrylic at 2.8 (not far off teflon at 2.1) To tool it, I used nothing more than a length of 24mm square wooden batten with three nails hammered through it. A nail at one end is for the central pivot, another at radius (e.g 150mm) and one (e.g. 40mm) closer in to score cut the ends of the gaps between tabs.

Detailed steps for making the former:
1. Score cut sheet material down to a square of side=final diameter required (e.g. 300mm)
2. Drill central hole in sheet to exact diameter of nails used (e.g. 3mm)
3. Pilot drill 3x2mm holes for nails spaced at radius apart and 40mm in from radius into a length of 24mm square wooden batten
4. hammer nail through end hole of batten into centre hole of sheet and into backing board (or bench )
5. Deep score sheet with radius nail to at least half sheet thickness and break the four 'corner' curved pieces away - starting each break with pilers to get it going.
6. Shallow score internal ring to mark end of tab cuts
7. Mark along each side of batten on sheet with thin permanent marker to make parallel lines 24mm apart for first gap.
8. Divide circumference by odd number (e.g. 13) and measure round this far to line up batten for all subsequent gaps (I made a template out of an off-cut to mark this off from the edge of the batten)
9. Saw inwards down each marked line to inner ring.
10. With a nail pushed into the inner radius hole of the batten, and another in the pivot hole, holding the end up so you can see the end of the saw cuts, scribe out the end of each gap between tabs deep enough for breaking. Don't scribe all the way round or you'll weaken the tabs.

I used a bandsaw to cut the polystyrene but any moderately fine-toothed saw should do. When breaking, have the score facing up and try to support the edge of the break from below then push down to snap. All edges can be finished smooth with glass paper.

[AdrianM]

The bifilar winding on the above former is 21 turns of 19SWG twisted pairs. This measures 200uH to common and 800uH total. Using another PI to ping this coil O/C the natural resonant frequency is 240kHz. This would make out its intrinsic capacitance to be about 550pF. Connected to the front-end, damping is critical with 500R

As might be expected this turns in a fairly leisurely decay to first available sample of about 20uS. Really would want to better this. Even my 36-way ribbon cable experiment was letting me in at around 14us.

Is it time to ditch the magnet wire (as some have argued in this thread)? I got almost half the capacitance with a similar coil made from standard PVC sheathed equipment wire (stranded TCW). This was just a bundle of 21 turns of bifilar wound twisted pair 300mm in diameter. It measures 690uH and resonates at 380kHz indicating intrinsic capacitance of 254pF. Damped with 1K, this also lets me in at 14us.

[AdrianM]

Made a few more coil armatures so here's a shot of the "magic tool" that does it all I like the speed with which this allows me to make an easily reproducible coil. Simply hammering in nails at different positions will give a range of diameters for accommodating the windings. The little slice of pie with the crosshatch is the off-cut that lines up the next position for the batten. I chose the size of the gap empirically, based on what felt comfortable when crossing the twisted pair from one side to the other (and the nearest lump of wood to hand). Using a permanent marker is great as it can be easily removed (!) using IPA on a cloth afterwards.

For a finished detector head I'm thinking of sandwiching the armature between a pair of plain discs and using polystyrene cement to glue a strip around the outside to enclose it. The windings could do with some spray varnish or something to keep them still though.

[mickstv]

1.
Two strands of 19SWG ( 1mm ) insulated solid copper wire are twisted together at approx 1 turn per inch to form a single twisted pair strand. ( You can do this with an electric drill at one end of the pair and a nail on the fence at the other ... stretching it out to avoid kinks)

2.
Wind the twisted pair onto a suitable coil former. Length required will depend on diameter and number of turns.

3.
If one of the conductor pairs is A and the other B. Cross connect the A conductor from one end of the twisted pair coil to the B conductor of the other end . This junction is the "common". The remaining two ends are the differential ends of the coil.

4.
Connect the coil to the electonics using a 3 conductor lead ( 3 conductor mains flex works for me )

For reference ....
A 32 turn 22 cm dia coil has a decay of 24 us and requires a 1K damping resistor. In receive the coil is equivalent to a 64 turn mono coil.
A 16 turn 18 inch flat wound coil has a decay of 20 us and requires a 750 ohm damping resistor.


Simplified diagram of a flat wound coil with no twists below ... with thanks to Tesla.

Hi Moodz, I was always of the understanding that you need as fast a coil as possible 10 us delay or less especially with regards to gold prospecting to be able to pickup small nuggets 1g or less. Is this the same for your differential coil setup. Or is something going on that I don't understand as yet that enables the coil to pickup these small nugget sizes.

Have you tried your detector out in the field and how does it compare to manufactured models ie :- Minelab.



Mick

[mickstv]

Is this as good as it gets here. Thanks LOL



Mick

[moodz]

Is this as good as it gets here. Thanks LOL



Mick

Good to see it keeps you amused at any rate ....

No I have not compared the diff coil to a minelab ( dont have one ) ... though these do appear to be nice machines. To be honest there has not been enough experimentation with the diff coil yet to determine an answer to your question. AdrianM has been doing some great work with alternate coil configs to lower the capacitance and it may transpire that this is what is required to detect the very small targets.
As for me I find that the smaller target still produces a response ...much smaller but still there ... well beyond the Tc you would expect for a traditional PI. ( my typical coil starts sampling at 22 us )
I do know that we can run much higher amp gains ( prior to sampling) on the first amplifier for a diff coil than you can for a traditional PI because the noise is much lower and we are taking advantage of common mode rejection.
The synchronous digital filtering I use ( nothing to do with diff coils ) lets me lift small signals from below the noise floor. The combination of the diff coil and synchronous filtering does seem to provide definite improvement over your standard signal PI signal chain.
My main problem is that I have limited time and I work for a company where my employment contract has a clause that says anything I invent (company time or my own time ) belongs to them ...even though this has nothing to do with the job I normally do ( telecommunications ).

regardz from moodz.