TGSL Tuning & troubleshooting

[Dennis the Mennis]

Hi Dennis,

Ok , made some tests , I explane you ... when I finish the coil (before Yarn) I measure 1st inductance (3400 uH), so I yarn and then I measure more then inductance (4600 uH) , last glue , dry and measure (6000 uH) ....

is difficult estimate a exact inductance .

regards Alex.

Hi Alex ,

Now I understand , I haven't glued my coils (yet) so I did not know that the glue could effect the inductance that much ...
It must be very though for you to get the right values for the coils .

succes

Regards

Dennis the Mennis

[Mullihaka]

dfbover's winding machine

Hi dfbovers,

What is the thickness of the spacers between the top and bottom plates of your winding machine? Does the wire make neat&even layers easily in the slot or do you need to guide it back and forth in the slot with fingers?

I wound another coil set and got 8cm extra distance. I feel, that the missing 10cm (from the expected 30cm) is somewhere in the coils.

Andy

[simonbaker]

Hi dfbovers,

What is the thickness of the spacers between the top and bottom plates of your winding machine? Does the wire make neat&even layers easily in the slot or do you need to guide it back and forth in the slot with fingers?

I wound another coil set and got 8cm extra distance. I feel, that the missing 10cm (from the expected 30cm) is somewhere in the coils.

Andy

That's very significant. Can you compare the two coils to see what the differences are? Inductance, resistance, self-resonance, null voltage and phase, shielding, etc? Is the cable the same in both coils?

-SB

[Mullihaka]

I only have 0.28mm wire (0.32 with enamel insulation). The former is the famous 255*137mm. The cable between the coils and the electronics is taken from a thin USB cable, connectors snipped out.

First coil set
TX 100 turns, 5.4mH taped tight, 19 ohms.
RX 105 turns, 5.8mH tapef tight, 20 ohms.
The frequency was 15.0kHz.
I tried with shielding and without, not finding much difference.
1 eur coin detected at 12cm.
All experiments were random, so I don't have any other inforamtion written down.

Second coil set
TX 109 turns, 6.2mH taped tight, 20 ohms.
RX 115 turns, 6.8mH tapef tight, 21 ohms.
The frequency is now 14.1kHz.
This set has copper tape shielding.
1 eur coin detected at 20cm.

Until yesterday, I kept the coils wired as described in the gift back. With the first coil set I tried to set the coils for:
- min voltage at the output of the preamp
- 0V at the source of the GB FET
- RX signal 200 deg lag
- RX signal 20 deg lag

I hadn't figured out which of those to select. Yesterday I started reading the thread backwards and picked up the latest one; that was 20 deg lag.
I also swapped the RX coil wires around, so I don't know which of the changes is in charge of the extra 8cm.

The best detection I got when the lag was 60 degrees and the level at the preamp output 450mV (~10mV at the preamp input).

When I furter adjusted the coils for lag closer to 20 degrees, the preamp output goes to 4V and the detection drops back to 12cm.

I feel (without any concrete reason), that I need to get the coils tighter.

Could it be that if not tight enough, the field pattern is not so concentrated thus decreasing the detection?

Could it be that the 20 degree lag _together_ with the 10-15mV residual at the RX coil wires is the key to success?

I have now made the same arrangements for coil winding as dfbovers. I'd just like to know if the cross section of his coils is square or flat.

Andy

[simonbaker]

I only have 0.28mm wire (0.32 with enamel insulation). The former is the famous 255*137mm. The cable between the coils and the electronics is taken from a thin USB cable, connectors snipped out.

First coil set
TX 100 turns, 5.4mH taped tight, 19 ohms.
RX 105 turns, 5.8mH tapef tight, 20 ohms.
The frequency was 15.0kHz.
I tried with shielding and without, not finding much difference.
1 eur coin detected at 12cm.
All experiments were random, so I don't have any other inforamtion written down.

Second coil set
TX 109 turns, 6.2mH taped tight, 20 ohms.
RX 115 turns, 6.8mH tapef tight, 21 ohms.
The frequency is now 14.1kHz.
This set has copper tape shielding.
1 eur coin detected at 20cm.

Until yesterday, I kept the coils wired as described in the gift back. With the first coil set I tried to set the coils for:
- min voltage at the output of the preamp
- 0V at the source of the GB FET
- RX signal 200 deg lag
- RX signal 20 deg lag

I hadn't figured out which of those to select. Yesterday I started reading the thread backwards and picked up the latest one; that was 20 deg lag.
I also swapped the RX coil wires around, so I don't know which of the changes is in charge of the extra 8cm.

The best detection I got when the lag was 60 degrees and the level at the preamp output 450mV (~10mV at the preamp input).

When I furter adjusted the coils for lag closer to 20 degrees, the preamp output goes to 4V and the detection drops back to 12cm.

I feel (without any concrete reason), that I need to get the coils tighter.

Could it be that if not tight enough, the field pattern is not so concentrated thus decreasing the detection?

Could it be that the 20 degree lag _together_ with the 10-15mV residual at the RX coil wires is the key to success?

I have now made the same arrangements for coil winding as dfbovers. I'd just like to know if the cross section of his coils is square or flat.

Andy

It would be interesting to measure null voltages and phases before shielding put on and compare with after shielding put on. I think shielding can move phase around in various amounts depending on type of shielding - but does it matter?

The day someone builds a 30 cm depth coil, they should try to build another identical one to see if it is repeatable.

Of course the remaining question is: is it possible to take almost any coil and adjust the main circuit components to make it work great. So far answer is no, but why not?

-SB

[dfbowers]

Andy,

I have been getting consistent results with the 3 coils that I have made so far. The space between formers I use is 2.6mm and the coils come out with a square cross section. My current problem is getting the varnish on my wire to dry inside the winder.

How do you keep track of how much wire you use in each coil?

Don

[Mullihaka]

Phase shift vs residual voltage?

I just count the turns in my head. I made a third one, wound it around the former made of nails. The nails were bent at 90 degreeangle, so the thickness of the coil was much like yours: 3mm. I guess it's of no use to make more of them before knowing what to look for in the nulling.

I wrote the following in the "main" TGSL thread, but I'm not sure if anybody reads it anymore.

I have a request for all members who have a TGSL which has detection distance of 30cm or more for a (1 eur) coin in the air.

Would you be so kind and measure:
- the phase shift between TX signal and RX signal (at pin 7 of the LF353)
- residual voltage at the preamp output (pin 7 of LF353)

I'm trying to figure out what kind of combination of the abovementioned I should try on my TGSL to get it beyond 20cm.

I can get the phase shift of 20 degrees or residual voltage of 450mV, but only one at time. Having either of them "as it should", the other jumps far away.

All results appreciated,

Andy

[dfbowers]

Andy,

I did not think about it much as performance was acceptable from the first coil that I built. I measured phase difference of 3 coils that I made. In the one below, the Rx lags Tx by maybe 40 degrees, and the residual voltage is about 1v p -p measured at pin 7 of the LF353, Another one, the phase difference is something less than 20 degrees and distance with 1e is even better, but it tends to be a little unstable. Maybe because it is closer to resonance? The only odd thing that I have noticed with my TGSL is the all metal mode does not work quite right. Iron has to be oriented just right to pick it up, unless it is large , but the disc mode works perfectly. I'm thinking of building a second circuit just to compare.

Don

[Stefano]

In the one below, the Rx lags Tx by maybe 40 degrees

I'd say this to be around 90 degrees, but doesn't matter. The important is to be able to understand what is the more desirable phase and/or whether it is important or not. In this sense I think Mullihaka's question was quite right.

[dfbowers]

Now that you commented, yes, it's probably closer to 90 degrees. Now as I remember, I had everything all set in the enclosure before I filled it with epoxy.. When it hardens, it changes things a bit. The GEB trimmer is still set at nearly midrange and rejects a ferrous rod just fine and can still pick up a thin silver coin.

[Mullihaka]

Getting more depth

Finally I got wire which is 0.24mm without insulation. The 100/105 turn coil set with 25cm diameter gives 5.7/6.2mH taped tight. The frequency is 14.7kHz.

I made a capacitor bank like Simonbaker and replaced the 15nf C6 with it. It has four dip switches and selected capacitors with which the bank covers 13...17nF with 250p increments.

I started from 15nF upwards. After each increment I adjusted the overlap for 20 degree lag. I can peak the distance (23cm) with 15.5nF. When further increasing the capacitance, the distance gets shorter and the device unstable.

The sound is not a nice beep anymore at the maximum distance, it's very short crack.

The LM393s I have are "GL393", the manufacturer is unknown to me. There's no logo at all. I wonder if they are some second class (high noise) components thrown away by some recognized manufacturer an sold bargain under "GL".

Has anyone tried to measure the noise level of their device with no metal present? The noise could mask the weakest signals thus reducing the sensitivity.

Any thoughts?

Andy

[simonbaker]

Finally I got wire which is 0.24mm without insulation. The 100/105 turn coil set with 25cm diameter gives 5.7/6.2mH taped tight. The frequency is 14.7kHz.

I made a capacitor bank like Simonbaker and replaced the 15nf C6 with it. It has four dip switches and selected capacitors with which the bank covers 13...17nF with 250p increments.

I started from 15nF upwards. After each increment I adjusted the overlap for 20 degree lag. I can peak the distance (23cm) with 15.5nF. When further increasing the capacitance, the distance gets shorter and the device unstable.

The sound is not a nice beep anymore at the maximum distance, it's very short crack.

The LM393s I have are "GL393", the manufacturer is unknown to me. There's no logo at all. I wonder if they are some second class (high noise) components thrown away by some recognized manufacturer an sold bargain under "GL".

Has anyone tried to measure the noise level of their device with no metal present? The noise could mask the weakest signals thus reducing the sensitivity.

Any thoughts?

Andy

I have monitored the noise by scoping the output of the LM308 amps. I put the trace really slow, like .2 div/second.

I found that in my "lab", the noise is fairly high, probably around the threshold of the sensitivity when the pot is most sensitive. That would make it about 5 millivolts, although I think.

The interesting thing is that the noise mostly comes from the coil/cable. If I disconnect the RX coil/cable, the noise is about 1/10 of what it was, nice and low. Because of the low frequency filtering of the circuit, the observed noise is very low frequency, probably in the 0 - 20 hz range.

So in my tests, it seems the circuit past the JFets does not make important noise contribution, rather it seems to be the coil. I need to take the coil and my scope outside somehow, away from EMI interference.

If I still get coil noise when outside, then probably there is some instability in the coil/cable, maybe mechanical.

Another test would be to feed a small signal from the TX oscillator to the RX input of the LF353 and see what the noise at the LM308 outputs is. If the noise is high, then I would suspect some kind of phase noise caused by LF353 or LM393 circuit, or maybe even the JFets.

As for the oscillator, theoretically any oscillator instability should be canceled out by the synchronous detector, but maybe it can get through. It is very likely that the oscillator has all kinds of little variations, but supposedly shouldn't matter much. Maybe check by hooking in a crystal oscillator.

It is also possible that the discrimination and ground-balance pots can make noise -- if they vary a tiny percent in resistance, it could shift the phase accordingly. We probably should use very high quality pots (which I don't). Even sound or motion vibrations could affect those pots.

Definitely need to understand what is causing the noise I see... Maybe will know in a few weeks if my next PCB gets finished...

Regards,

-SB

[Cyclonite]

An interesting read about noise coming from the coil and cable. It would also be interesting to compare a coil wound from regular Copper Wire to that of OFC (Oxygen Free Copper) Wire, which is also used in High-End Audio applications. And maybe this will also prove to be an easily, and readily available source to obtain OFC. Even cheap Sennheiser headphones and Ear Plugs use OFC nowadays, and also Polypropylene Microphone Cable has an OFC core, which is then specified by an OFC print on the cable mantel.

Also, using potentiometers of half the value specified will reduce the noise figures. This noise exists in nearly all pots that have been unused for a period of time, including those fresh from the manufacturer. When a potentiometer has passed its useful life, this noise can be a result of a worn and dried out resistance element. But, long before this point has been reached, the potentiometer may appear to be beyond recall because of noise caused by high contact resistance. Incidentally, no lubricant or other material should be applied to any part of a potentiometer. To do so, will cause deterioration of internal parts.

And how about Conductive Polymer (Conductive Plastic) Potentiometers ?

They have excellent noise figures, please check out the conductive polymer potentiometers from Bourns® and Omeg® (as examples).

Bourns® Precision Potentiometers and Turns-Counting Dials:
http://www.bourns.com/pdfs/bourns_pr...short_form.pdf

Omeg® Potentiometers and Switches:
http://www.omeg.com/

And a little off-topic, the Digital Chip solution from Dallas Semiconductor Maxim.

Dual, NV Audio Potentiometers Provide Low-Noise, Low Distortion Digital Volume Control:
http://pdfserv.maxim-ic.com/en/da/DS1881-AD.pdf

Just throwing a few stones in the pond, and waiting to hear about the ripple

Regards,

Robert

[dfbowers]

Andy,

* Which version of the TGSL are you working with?
* Is there a chance that you have the shield in your coils grounded to the wrong pin on U101a?
* What kind of cable are you using for your coil?
* If you scope out both channels, does one signal appear to be much weaker than the other when measuring target response on Pins 3 and 5 on U106?


If you want, I can send you a pair of coils, just so we can do a comparison.

I have only built one circuit so far and a bunch of coils.. Every once in a while I get a coil that's a dud for some reason.. I'm going to build a second circuit board just for a comparison. Too bad we can't all get into the same room somewhere for a day just to compare!!

[simonbaker]

Andy,

* Which version of the TGSL are you working with?
* Is there a chance that you have the shield in your coils grounded to the wrong pin on U101a?
* What kind of cable are you using for your coil?
* If you scope out both channels, does one signal appear to be much weaker than the other when measuring target response on Pins 3 and 5 on U106?


If you want, I can send you a pair of coils, just so we can do a comparison.

I have only built one circuit so far and a bunch of coils.. Every once in a while I get a coil that's a dud for some reason.. I'm going to build a second circuit board just for a comparison. Too bad we can't all get into the same room somewhere for a day just to compare!!

I'll chime in with my experience so far because those are good questions... on my workbench, I have not found that grounding one way or the other makes a brilliant difference... in fact I seem to get a little better result by not grounding the RX lead at the coil shield... but I know my whole rig is suboptimal, so I'm trying to build a better PCB and get things more solid.

I definitely see one channel much weaker than the other. The Ground Balance channel is weaker.

I plan to do some tests ignoring the GB channel completely (tying the output high, or feeding the disc signal into both channels) to check several things. Using only the larger signal should show whether you have a chance at 30 cm air depth or not, or whether something else is limiting it.

I have tried two different cables. I use simple RCA cables, and also a Belden 8723. I terminate both into simple RCA connectors, don't know if I lose something there. The choice of cable hasn't made a significant difference for me, because I think something else is limiting the operation.

-SB