TGSL Tuning & troubleshooting

[simonbaker]

If only I had such a good depth!!!

Yesterday I played with my brand new LCR meter which showed 5.75 and 6.34 mH for my coils, so a bit low but I think usable.
I found with the function generator that the resonance frequency was good for the tx circuit, but that I had to increase a bit the Rx capacitor to get the 1.8 2 KHz off-phase frequency (so around 16.7- 16.8 I don't remember exactly...)
Therefore I think to have the correct coil values but when I try the nulling I can't get the right phase at the minimum amplitude, and the minimum amplitude is quite high.

So, to further investigate what's going wrong:

When all the coils parameters are correct, should the right phase be obtained at the minimum voltage?
What are the phisical/electrical factors that influence the rx signal phase ?
What are the phisical/electrical factors that influence the minimum rx signal amplitude at nulling point?
I noticed for example that by changing the tx frequency I can change the phase and the same changing the LC circuits tuning, but I'd need to go too far from the theoretical data and then the minimum voltage increases.
Is there any simulator or any formula that can show the behaviour of these coupled coils ?

Excellent questions; hard to answer so far.

I think the circuit is so "delicate" it is hard to simulate very well. Aziz has posted some interesting simulations using LTSpice with targets and coils, but hand-built coils are very particular.

I think best approach is to build a "laboratory" for experimenting. Some things we need:

1. Target swinger -- make a swing suspended from ceiling for swinging target repeatedly and precisely over coil. It should swing in a line, not allowed to wobble in a circle, so best if suspended by two separate cords. Ideally, it should have a small "kicker" mechanism at top to keep it swinging.

2. Phase detector. The built-in discriminator of the TGS is a phase detector, but it would be useful to build a separate one that can vary over greater range, at least 180 deg. Sort of like building a single channel of the TGSL on a separate board, with a bigger range of phase detection. You can't use an oscilloscope, because all you see is the "null" signal which is much, much bigger than the target signal.

3. Adjustable coil - I build one with plastic screw for adjusting null, but need better, more robust design.

4. Coil winder - I built one, very handy for making lots of coils. Making lots of coils is important; different sizes, wire gauges, glues, shields, inductances.

5. Outdoor oscilloscope - some way to use our equipment in outdoors or noise-free environment.

6. Variable capacitor blocks - I build one with dip switches. Handy for making adjustable TX oscillator frequency, and even adjustable RX resonant frequency (but bulky and adds noise).

7. Other ideas?

There has to be a way to get to the bottom of why a MD performs well or poorly.

-SB

[simonbaker]

Some mixed findings more.

The voltage stays below 0V at the source of TR5, whatever I do with the coils. It's not much below, only about 100mV.

TR5 is ground balance circuit, right? What happens when you turn GB pot, does voltage ever go positive? Slight negative is probably OK though.

When I sweep the coin over the coils, it gives two short beeps. It behaves the same, independent of the direction of the sweep: across the overlapping section, parallel with it or even approaching the overlap from above and then suddenly taking it back. The RX coil wires are exactly as in the gift back pictures.

Agree with other post, try reversing and always test at least 8 cm from coil. But try to make sure if you are following exactly the gift pictures, especially which RX lead is attached to inverting input of LF353. I think this is most confusing thing to figure out, have not seen perfectly clear explanation.

Does it matter, if the coil wires leave the "D" at the end of the straight line of the "D"? The pictures in the gift pack describe a coil where they leave at the middle of the straight part.

Don't think it matters.

How about the shape of the overlap? I have it rectangular as in the giftpack, but have seen also diamond shaped ones.

I think rectangular should be capable of working very well, most of Max and Ivconic's coils are like that.

I haven't been able to get any of the recommended FETs in the local shops. I have BFW61's as TR4 and TR5. I believe they should be good for the purpose.

I think Porkluvr comment is right. You just need to be sure JFet will cut off at minimum voltage. You could test your JFets. The specs are usually quite variable.

It would be interesting to hear, what is the nulling method resulting in most working detectors: zeroing down to nanovolts, zeroing the source voltage of the TR5 or making the phase shift between TX and RX as close to 20 deg as possible.

If the null is that important, you should be able to find it by swinging coin and finding perfect spot for maximum distance. I don't think you will find such a magic spot where suddenly depth increases. Of course, distance is not only consideration, discrimination important too. My approach is to null as close to minimum voltage as I can while still keeping the source voltage of the TR5 and TR4 positive over the range of the pots.

I wonder, if the exact frequency of the TX is important, if the RX is 9% higher.

I don't think in theory the exact freq is important, because Tesoro uses different frequencies for various MDs. Whether 9% is the best offset I don't know either -- it is a particularly interesting variable to experiment with. But practically, it may be hard to make this circuit work well at different frequencies, too many adjustments to make.

I guess I need to make a new coil, take the device out of the house and continue the nulling exercise far from everything electric.

I think that's right for now -- we don't know yet how to fully understand why a detector gets 15 cm instead of 30 cm, especially if we can't get our hands on it. As you you build more and more, you will have the power to compare and by accident see what works best. I hope you give us details of what you learn.

-SB

[simonbaker]

Mullihaka,

Double beep would suggest that you have Rx wires reversed.

Don

Agree, try it both ways just in case. But don't put coin too close to coil, should be at least 8 cm or so, otherwise multiple beeps can happen anyway.

-SB

[dfbowers]

Mullihaka,

Double beep would suggest that you have Rx wires reversed.

Don

[Stefano]

I am gettin 26 - 28 cm in air on gold rings and aluminum ring type pull tabs. Ferrite is discriminated completely.

If only I had such a good depth!!!

Yesterday I played with my brand new LCR meter which showed 5.75 and 6.34 mH for my coils, so a bit low but I think usable.
I found with the function generator that the resonance frequency was good for the tx circuit, but that I had to increase a bit the Rx capacitor to get the 1.8 2 KHz off-phase frequency (so around 16.7- 16.8 I don't remember exactly...)
Therefore I think to have the correct coil values but when I try the nulling I can't get the right phase at the minimum amplitude, and the minimum amplitude is quite high.

So, to further investigate what's going wrong:

When all the coils parameters are correct, should the right phase be obtained at the minimum voltage?
What are the phisical/electrical factors that influence the rx signal phase ?
What are the phisical/electrical factors that influence the minimum rx signal amplitude at nulling point?
I noticed for example that by changing the tx frequency I can change the phase and the same changing the LC circuits tuning, but I'd need to go too far from the theoretical data and then the minimum voltage increases.
Is there any simulator or any formula that can show the behaviour of these coupled coils ?

[porkluvr]

I haven't been able to get any of the recommended FETs in the local shops. I have BFW61's as TR4 and TR5. I believe they should be good for the purpose.

Andy

There may be other problems but I believe your BFW61 are not so good. You only have about 6.2V available for cut-off and the BFW61 might need up to 8V.

The J107 has a maximum cutoff of 5V.

[Mullihaka]

Some mixed findings more.

The voltage stays below 0V at the source of TR5, whatever I do with the coils. It's not much below, only about 100mV.

When I sweep the coin over the coils, it gives two short beeps. It behaves the same, independent of the direction of the sweep: across the overlapping section, parallel with it or even approaching the overlap from above and then suddenly taking it back. The RX coil wires are exactly as in the gift back pictures.

Does it matter, if the coil wires leave the "D" at the end of the straight line of the "D"? The pictures in the gift pack describe a coil where they leave at the middle of the straight part.

How about the shape of the overlap? I have it rectangular as in the giftpack, but have seen also diamond shaped ones.

I haven't been able to get any of the recommended FETs in the local shops. I have BFW61's as TR4 and TR5. I believe they should be good for the purpose.

It would be interesting to hear, what is the nulling method resulting in most working detectors: zeroing down to nanovolts, zeroing the source voltage of the TR5 or making the phase shift between TX and RX as close to 20 deg as possible.

I wonder, if the exact frequency of the TX is important, if the RX is 9% higher.

I guess I need to make a new coil, take the device out of the house and continue the nulling exercise far from everything electric.

Andy

[Jerry]

simonbaker:

Thank you for your detailed explanation and suggestions.
I will be leaving for a 10 day trip to warmer climate (I hope) so cannot post any scope pictures right now. Will pick this up again when I get back.

Thanks

Jerry

[simonbaker]

I have a TGSL built and working but have not completely eliminated the problems with GEB and Disc. It chatters in all metal mode but stable in Disc mode. I am gettin 26 - 28 cm in air on gold rings and aluminum ring type pull tabs. Ferrite is discriminated completely. Unfortunately, so are US dimes and quarters. It seems I have made a MD that really likes Gold

I know it will do much better than what I have now because I see good signal response at the outputs of the lm308's well above 32 cm. Of course the only hits that get through to the speaker are those in Phase at those outputs. (Gold & Pulltabs)

So I think I have a GEB, Disc or coil null/phase problem. I Have taken the time to read and try to digest everything that has been written about the TGSL on Geotech to try and avoid asking questions that have been covered before.

It would be very helpful for me to understand exactly how the TGSL works if someone could explain the sampling process at the Disc FETs. I have printouts of all the scope pictures taken and sometimes I think I get it and then read another post on the subject and get confused again.

It would help to show what part of the sample pulse actually does the sampling (start to finish of sample) and what area it takes of the received signal.

This may seem very basic to some of you but I think it will be a key in helping me understand this detector.

Thanks for reading.

Jerry

The JFet is just a switch that turns on when the sample pulse at the gate goes the most positive -- in this case, near zero volts. The switch is on for half of a cycle duration of the TX/RX signal. The .22u capacitor after the JFET (source side) will integrate (accumulate) the RX signal "pieces" that comes through the JFet switch while it is turned on. Therefore, the voltage on the capacitor depends on whether the sample pulse phase coincides with the positive half of the RX signal (positive result), the negative half (negative result), or somewhere in between (near zero).

The discrimination knob shifts the phase of the sample pulse so that you can choose to ignore (near zero accumulation at capacitor) RX signals that are 90 degrees different from the sample pulse, while detecting signals that are more in phase with the sample pulse.

The capacitor that accumulates the RX signal actually discharges slowly back through the JFet throught the 20k and 1k resistors so it acts as a low pass filter, and creates a low frequency voltage pulse as you sweep your coil over a target (which creates a tiny RX signal of the correct phase that your JFet detector will accumulate on the capacitor).

Ok, now what do you do. First, congratulations, you have a nice metal detector with excellent sensitivity. Many of us have not achieved that. You must have made a good coil I would guess.

However, you seem to discriminate out metals like silver and copper. I believe this means that your sample pulse needs to be able to shift more "to the right" as viewed on an oscilloscope (triggered by the TX signal). (It is a little hard to tell without actually testing your MD to make sure everything is hooked up as I'm familiar with.)

There isn't much you can do to adjust the discrimination range without significant changes to the circuit. But it would be useful to see on an oscilloscope what the output of your LF353 pin 1 looks like -- is it a clean sine wave or does it have lots of spikes in it? Can you post a picture?

Then I would check the 100 pf capacitor C9 (pin 2 of LF353) to make sure it is the correct value; even try replacing it. Make sure all the resistors in the discrimination circuit are correct also. I would do the same for the ground balance circuit also (U102 LM393 pins 5,6,7).

Douple-checking values of caps and resistors in the RX circuit (LF353 pins 5,6,7) is a good idea also; wrong values can shift phase.

Definitely stick with this, it seems you are very close.

You have a heck of a gold detector in any case!

Cheers,

-SB

[Jerry]

I have a TGSL built and working but have not completely eliminated the problems with GEB and Disc. It chatters in all metal mode but stable in Disc mode. I am gettin 26 - 28 cm in air on gold rings and aluminum ring type pull tabs. Ferrite is discriminated completely. Unfortunately, so are US dimes and quarters. It seems I have made a MD that really likes Gold

I know it will do much better than what I have now because I see good signal response at the outputs of the lm308's well above 32 cm. Of course the only hits that get through to the speaker are those in Phase at those outputs. (Gold & Pulltabs)

So I think I have a GEB, Disc or coil null/phase problem. I Have taken the time to read and try to digest everything that has been written about the TGSL on Geotech to try and avoid asking questions that have been covered before.

It would be very helpful for me to understand exactly how the TGSL works if someone could explain the sampling process at the Disc FETs. I have printouts of all the scope pictures taken and sometimes I think I get it and then read another post on the subject and get confused again.

It would help to show what part of the sample pulse actually does the sampling (start to finish of sample) and what area it takes of the received signal.

This may seem very basic to some of you but I think it will be a key in helping me understand this detector.

Thanks for reading.

Jerry

[Qiaozhi]

If I'm right this is the way the sample puls aprox. should look like ?

[ATTACH]11517[/ATTACH]


regards

Dennis the Mennis

You need to shift the GEB sample pulse slightly to the left, but otherwise the signal relationship looks ok.

[Dennis the Mennis]

Hi Dennis TM,


It would be good to see the nulling in practise (Youtube?) to know, if my procedures are anything like those resulting a good detector.

Andy

http://www.geotech1.com/forums/showpost.php?p=105383&postcount=4026

For your comfort , I did manage to get about 12-15 cm in air on a 2 euro coin with coils complete out range , and I thought I was close ....not.
So don't you have any doubts on your pcb or its components ?
Your coils look better as the ones I was testing with at that moment ...

At this moment the most strange things are happening when I shield my coils , so I'm workin on why and what...

regards

Dennis the Mennis

[Dennis the Mennis]

I think I have spotted your mistake.
The sampling pulse should not start at the zero-crossing of the RX signal. It is the middle of the sample pulse that must be at the crossing point. Also, for a DD coil, the sampling must take place at the negative-going transition, not the positive.

And one last point ... all targets should cause a phase-shift to the right.

If I'm right this is the way the sample puls aprox. should look like ?




regards

Dennis the Mennis

[simonbaker]

I think I have spotted your mistake.
The sampling pulse should not start at the zero-crossing of the RX signal. It is the middle of the sample pulse that must be at the crossing point. Also, for a DD coil, the sampling must take place at the negative-going transition, not the positive.

And one last point ... all targets should cause a phase-shift to the right.

This is a tricky issue which Qiaozhi and I occasionally have slightly different interpretations of. And I don't have concrete proof for my ideas. But...

My opinion is that the null signal is very deceiving to use for phasing decisions because it does not really indicate the phase of the actual received signals. The actual received signals are very tiny and not visible and have their own phase which is not directly related to the phase of the null signal. I believe you can vary the phase of the null signal by shifting the coils here and there without much changing the phase of the received signals. But don't believe me yet -- it needs to be demonstrated.

However, based on my opinion, it means the "sample pulse" needs to be properly phased relative to target signals, in other words, don't worry so much about the null signal for that issue. Of course the sample pulse phase moves around when you rotate the discrim pot, so we always need to mention the location of the discrim pot when we show the "sample pulse", otherwise we are comparing different circuits.

As it turns out, there is no adjustment in the TGS circuit to "bias" or shift the basic phase range of the discrim "sample pulse" -- just the discrim knob itself. So for now, we just build the circuit and hope for the best. I'm actually trying to experiment with some ways to adjust the range over which the discrim control works to help troubleshoot if that may be the cause of poor performance.

Anyway, what this all means is that I think the phase of the null is mainly important for biasing that voltage at the Source side of the JFet. However, I realize that must be an oversimplification because of all the importance placed on nulling. But from my tests so far, that's how I see it.

I know I need to learn more and understand better, so if I completely change my opinion later, don't be surprised. But the main reason I make this point is that I think that with some coils you can null for the rest of your life and they'll never work any better. It's something else about the coil that is deficient and it is a mystery right now. So I would follow one of Ivconic's recommendations which is: null your coil by swinging a coin above it for maximum depth. If you don't get good depth, build another coil!

I hope to know more some day. I'm in that group that has not made a TGS with more than 15 to 20 cm depth.

-SB

[Stefano]

Time at the moment time is my greatest problem , but I'm still reading these posts hoping someone has the solution ...

That's exactly my same problem!!! .....

Thanks Qiaozhi for your hints: always very appreciated