TGSL Tuning & troubleshooting

[dfbowers]

I have built probably 6 or 7 coils for the TGSL now..
I wanted to take a poll and see how may out there measure the inductances of their coils and how many just go by the "TGSL coil making pdf" by Ivconic. Or how many calculate inductance by other methods.

Simon had noted that the coils that I had sent him were around 6.3 and 6.9 respectively for Tx and Rx. I don't have a way to directly measure L so I have been using other methods to get close. But.. Every set of coils that I have made have been within 5 turns of each other. Even using a model for 6.0 and 6.5 mH, the difference is still 5 turns. I have two different forms.. slightly different number of turns. but still a difference of 5 turn. (105 and 110 for Tx and Rx respectively).

Now I made another set to see if I could get the Tx freq closer to 14.5kHz. So, I made a Tx coil to run at 14.5kHz. Then substitued my new Rx coil onto the oscillator circuit to shoot for 13.9kHz, (6.5mH + .020 uF = 13.9KHz) .. that should show me a difference of .5mH between Tx and Rx coils. 6.0mH for the Tx coil and 6.5mH for the Rx coil. Guess what.. Still a difference of 5 turns!! 101 turns and 106 turns.

Now with Ivconics specs, he suggests a difference of 9 turns. (98 and 107 turns). I suppose lots of variables.. different winding methods, wire sizes, bindings, etc. But if I did that, I would probably not be able to tune mine correctly. What have your experiences been?

Thanks

Don

Holy cow.. It finally happened! With the coil described above, coil #7 allows my null phase to occur close to where a ferrite slug ground balances. 31cm for a 1e.

Now.. if I can only make another one just like it.

[Qiaozhi]

So adding 5 turns gives similar change in inductance in that range. If you believe calculator.

-SB

Just make certain you are using the latest version (2.0), which I think you must be, as I get the same results.

[simonbaker]

beach

Went to beach today and took 2 MDs: dfbowers TGSL and my Tesoro Compadre with 5.75 inch coil.

Literally got my 2 cents worth . But I was with people so not focused.

One interesting test. I put a Euro on top of sand and stuck a 30 cm stick in sand. Sand was in dry area, but if dig down somewhat moist.

By eye, it seemed TGSL had range of about 15 cm to 20 cm. Compadre surprisingly was almost the same, probably a little less than TGSL.

I tried to do air test of TGSL but sorry didn't bring headphone extension cord so hard to hear audio. But maybe air test got 24 to 28 cm.

It seemed beach had somewhat more noise than in mountains. When turning sensitivity full, more chatter than in Mountains - just an impression.

I am wondering if conductivity of sand causes smaller difference between Euro and sand than with less conductive ground or air, thus smaller signal blip.

Why Compadre did quite well not sure. Maybe need to study TGSL over sand/ground some more.

-SB

[simonbaker]

I have built probably 6 or 7 coils for the TGSL now..
I wanted to take a poll and see how may out there measure the inductances of their coils and how many just go by the "TGSL coil making pdf" by Ivconic. Or how many calculate inductance by other methods.

Simon had noted that the coils that I had sent him were around 6.3 and 6.9 respectively for Tx and Rx. I don't have a way to directly measure L so I have been using other methods to get close. But.. Every set of coils that I have made have been within 5 turns of each other. Even using a model for 6.0 and 6.5 mH, the difference is still 5 turns. I have two different forms.. slightly different number of turns. but still a difference of 5 turn. (105 and 110 for Tx and Rx respectively).

Now I made another set to see if I could get the Tx freq closer to 14.5kHz. So, I made a Tx coil to run at 14.5kHz. Then substitued my new Rx coil onto the oscillator circuit to shoot for 13.9kHz, (6.5mH + .020 uF = 13.9KHz) .. that should show me a difference of .5mH between Tx and Rx coils. 6.0mH for the Tx coil and 6.5mH for the Rx coil. Guess what.. Still a difference of 5 turns!! 101 turns and 106 turns.

Now with Ivconics specs, he suggests a difference of 9 turns. (98 and 107 turns). I suppose lots of variables.. different winding methods, wire sizes, bindings, etc. But if I did that, I would probably not be able to tune mine correctly. What have your experiences been?

Thanks

Don

Time to break out... Qiaozhi's coil calculator!

I ran some cases. I'm equating a 22 cm diameter round coil with a 28 cm D shape coil. Used .25 mm wire.

I used Q's calculator to estimate inductance at different turns.

90 turns: 5.35
95 turns: 5.93 delta .58 mH
100 turns: 6.54 delta .61 mH
105 turns: 7.18 delta .64 mH
110 turns: 7.84 delta .66 mH

So adding 5 turns gives similar change in inductance in that range. If you believe calculator.

-SB

[dfbowers]

TGSL Coil poll

I have built probably 6 or 7 coils for the TGSL now..
I wanted to take a poll and see how may out there measure the inductances of their coils and how many just go by the "TGSL coil making pdf" by Ivconic. Or how many calculate inductance by other methods.

Simon had noted that the coils that I had sent him were around 6.3 and 6.9 respectively for Tx and Rx. I don't have a way to directly measure L so I have been using other methods to get close. But.. Every set of coils that I have made have been within 5 turns of each other. Even using a model for 6.0 and 6.5 mH, the difference is still 5 turns. I have two different forms.. slightly different number of turns. but still a difference of 5 turn. (105 and 110 for Tx and Rx respectively).

Now I made another set to see if I could get the Tx freq closer to 14.5kHz. So, I made a Tx coil to run at 14.5kHz. Then substitued my new Rx coil onto the oscillator circuit to shoot for 13.9kHz, (6.5mH + .020 uF = 13.9KHz) .. that should show me a difference of .5mH between Tx and Rx coils. 6.0mH for the Tx coil and 6.5mH for the Rx coil. Guess what.. Still a difference of 5 turns!! 101 turns and 106 turns.

Now with Ivconics specs, he suggests a difference of 9 turns. (98 and 107 turns). I suppose lots of variables.. different winding methods, wire sizes, bindings, etc. But if I did that, I would probably not be able to tune mine correctly. What have your experiences been?

Thanks

Don

[dfbowers]

I did a quick test to see what kind of capacitance at C6 (parallel with RX coil) is acceptable.

By increasing C6, the gain of the received signal can be increased, but it also shifts the target signal phase, and thus can affect the discrimination, and indirectly, the ground balance.

I used:

coil from dfbowers TGSL - RX inductance 6.9 mH, TX inductance 6.3 mH.

My PCB; oscillator freq 13.81 kHz.

For discrimination, I tested with US nickel and aluminum pop-top from soda can.

My C6 soldered into circuit is 14.4 nF.

I used .9 nF, 2.0 nF, 3.5 nF, 4.0 nF trimming capacitors for test in a socket soldered in parallel with C6. This gave range of 14.4 to 18.4 nF for C6. I could verify the capacitors were making contact by seeing the null signal change amplitude.

Results:

The nickel could be discriminated with all trimming capacitors.

The aluminum pop-top could be discriminated up to 3.5 nF trim cap. With 4.0 nF cap, the pop top made a clear beep.

But note: with 4.0 nF cap, the RX null signal was quite high, meaning lots of gain. So question is whether the phase shift caused detection of pop-top at max DISC, or simply additional gain caused detection.

Because of noise in my workbench, I could not assess the depth variation of different C6 values on 1 Euro. You would think that larger values should give more depth. I will have to test outside to see.

Observation: in order to discriminate the pop-top, the disc setting seemed to cause significant loss of depth for 1 Euro. I will need to test outside for exact amount.

I found that no DISC setting would discriminate a US quarter or a 1 Euro, regardless of C6 value. A US "copper/zinc" penny was slightly discriminated at max DISC setting at lower values of C6.

-SB

I had done a similar experiment some time ago but instead of a trimmer cap, I put several taps on my RX coil and added turns . I was curious to see if I could eliminate noise by bringing the two coils closer to resonance. My results seemed similar to what you have noted above. The phases shifted too far off and I couldn't really notice any difference in detection depth..

But, in theory the closer we get to resonance, the more noise becomes rejected and the signal we are looking for is enhanced (If we have a high Q coil). But I guess as so many others have noted.. tough to do with this design.

Don

[simonbaker]

RX trimming capacitor test & discrimination

I did a quick test to see what kind of capacitance at C6 (parallel with RX coil) is acceptable.

By increasing C6, the gain of the received signal can be increased, but it also shifts the target signal phase, and thus can affect the discrimination, and indirectly, the ground balance.

I used:

coil from dfbowers TGSL - RX inductance 6.9 mH, TX inductance 6.3 mH.

My PCB; oscillator freq 13.81 kHz.

For discrimination, I tested with US nickel and aluminum pop-top from soda can.

My C6 soldered into circuit is 14.4 nF.

I used .9 nF, 2.0 nF, 3.5 nF, 4.0 nF trimming capacitors for test in a socket soldered in parallel with C6. This gave range of 14.4 to 18.4 nF for C6. I could verify the capacitors were making contact by seeing the null signal change amplitude.

Results:

The nickel could be discriminated with all trimming capacitors.

The aluminum pop-top could be discriminated up to 3.5 nF trim cap. With 4.0 nF cap, the pop top made a clear beep.

But note: with 4.0 nF cap, the RX null signal was quite high, meaning lots of gain. So question is whether the phase shift caused detection of pop-top at max DISC, or simply additional gain caused detection.

Because of noise in my workbench, I could not assess the depth variation of different C6 values on 1 Euro. You would think that larger values should give more depth. I will have to test outside to see.

Observation: in order to discriminate the pop-top, the disc setting seemed to cause significant loss of depth for 1 Euro. I will need to test outside for exact amount.

I found that no DISC setting would discriminate a US quarter or a 1 Euro, regardless of C6 value. A US "copper/zinc" penny was slightly discriminated at max DISC setting at lower values of C6.

-SB

[Vladimir]

Vladimir,

Did you add the filter to pin 3 and 5 or just pin 5 of U106?

Don

Only the pin 5.

[simonbaker]

Well, I concluded my experiment with the DC offset on U104 and U105. I did not have a TL071 on hand but had a TL061.. Not my first choice, but good enough to experiment with. So, I just piggy backed a pot on top of the IC and connected the pot per DC offset diagram in the datasheet.

Results: I have to conclude that my TGSL circuit is pretty optimal to begin with. I first substituted the TL061 with the adjustable offset into the LM308 for the AM channel. In tweeking the offset, I noticed no improvement in audio, but changed where the sensitivity adjustment needed to be... and at the expense of where the circuit would ground balance... if I could balance at all.

Likewise, substituting the TL061 with adjustable offset for the LM308 for the disc channel would alter the sensitivity setting but at the expense of where the TGSL would disc.

So, I don't believe that there is much point in chasing this idea any further. It might benefit a circuit that is not optimal to begin with but mine seems to be working as good as advertised by others.

If we want to split hairs, I can continue playing around with it but in actual use outside the benefits would be negligable if any benefit at all.

Oh well.. I just had to know!

Don

Good experiment. I wouldn't mind seeing a photo of your piggyback if easy. Can you tell us what the baseline output offset voltages were and how you changed them?

At first I was totally perplexed by your seeing a shift in where DISC and ground balancing occurs. But I guess anything that affects the trigger point of the beep would give an apparent shift in those settings. But I hope there wasn't much shift, unless the TGSL was very suboptimal to begin with.

As a practical summary to this, there are a couple of measurements I would take to make sure the sensitivity circuit is tuned up. First, I'd want to establish that the two LM393 comparators have the same threshold. If not, then we probaby want to compensate by having different output offsets for U104, U105 (LM308 ). But, if the thresholds are the same, I'd want to hand-pick some LM308 chips to get the output offsets really close to the same value. The newer TL0.. chips may be better in that regard.

Once those two measurements are correct, I would put in the "dfbowers mod" to the sensitivity control so it has more range to operate.

For now, on my PCB, I have put in the dfbowers mod and hand-picked the LM308 chips to have an output offset within 1 to 2 mV of each other, absolute value of about +4, +5 mV. It's too difficult to test the LM393 thresholds at the moment, so I'll hope they are quite similar, being on the same chip.

For next trip to low-noise area, I'll take along some trimming capacitors for the RX coil, and keep the sun off my circuit! Wish I had a nice dank basement.

-SB

[dfbowers]

I noticed that the real conditions on the ground, sometimes noise is small, sometimes no noise. I suppose that different conditions affect such change. That's why I installed an external control of fileter (as shown). With additional small loss in sensitivity, improves the stability.
With basic filter 1EUR coins in the air ... 36-38cm (sometimes rare peaks occur on the ground)
With 47pF + ... 32-34cm (MD is completely stable on the soil, but in the air make small peaks)
With 68pF + ... 30-32cm (MD is completely stable in the ground and air)

Regards

Vladimir,

Did you add the filter to pin 3 and 5 or just pin 5 of U106?

Don

[greybeard]

i haven't looked to hard at the cct but it might be the change in offset rather than the magnitude that is causing an issue.
in which case a few 10's of mVolts either way isn't going to make a difference.

[dfbowers]

Looks good. Theoretically, you only need to adjust one of them for balance, then use the normal sensitivity control (with your mod) for overall level. Is that your plan?

I'm still not sure I like pin 3 for noise amplification reasons, but probaby negligible. It's practical.

One thing we want to do is "gear down" the adjustment as much as possible for stability. Otherwise mechanical vibration of the pot will make a big effect.

-SB

Well, I concluded my experiment with the DC offset on U104 and U105. I did not have a TL071 on hand but had a TL061.. Not my first choice, but good enough to experiment with. So, I just piggy backed a pot on top of the IC and connected the pot per DC offset diagram in the datasheet.

Results: I have to conclude that my TGSL circuit is pretty optimal to begin with. I first substituted the TL061 with the adjustable offset into the LM308 for the AM channel. In tweeking the offset, I noticed no improvement in audio, but changed where the sensitivity adjustment needed to be... and at the expense of where the circuit would ground balance... if I could balance at all.

Likewise, substituting the TL061 with adjustable offset for the LM308 for the disc channel would alter the sensitivity setting but at the expense of where the TGSL would disc.

So, I don't believe that there is much point in chasing this idea any further. It might benefit a circuit that is not optimal to begin with but mine seems to be working as good as advertised by others.

If we want to split hairs, I can continue playing around with it but in actual use outside the benefits would be negligable if any benefit at all.

Oh well.. I just had to know!

Don

[simonbaker]

dfbowers mod alt

Just for the record, the "dfbowers mod" shouldn't need the resistor R37, which may in fact have been his original mod. Either way is OK.

-SB

[simonbaker]

It looks like we have at least 2 choices for playing with offset and either may let us arrive at the same result. I like Qiaozhi's suggestion on biasing pin 3 of u104 but I may try playing with the offset pins. If I cannot get the range required I will try his suggestion.

I was just going to piggy back a small trimmer on a pair of TL071s and solder directly to the legs, then run a small wire to the -v5 rail.

Don

Looks good. Theoretically, you only need to adjust one of them for balance, then use the normal sensitivity control (with your mod) for overall level. Is that your plan?

I'm still not sure I like pin 3 for noise amplification reasons, but probaby negligible. It's practical.

One thing we want to do is "gear down" the adjustment as much as possible for stability. Otherwise mechanical vibration of the pot will make a big effect.

-SB

[dfbowers]

Hi Don:

Yes, I am talking about the LM308 output "offset voltages", meaning the offset from the ideal zero average output (best seen with shorted RX coil input).

And yes, they are different by about 20 mV! ...validating your desire to test separate offset adjustments, which I feel we fanatics would be pleased to have as a useful feature because of off-spec chips. In fact, with such adjustments, the need for precision op amps may disappear, since we are happy to tweak and are not mass-producing these MDs.

*** Note -- when adjusting separate thresholds, just looking at the U104 (LM308 ) outputs is not enough -- we need to consider out-of-spec input thresholds of the LM393 as dfbowers once noted. So the real way to calibrate is to separately test the thresholds by watching when the LM393 output flips as we make adjustments.

Back to the "light effect" -- in fact I did the tape test and it seemed to confirm light on the diodes to be the culprit -- although what part of the spectrum I'm not sure. I first used some cheap blue electrical tape and did not see the effect go away to my surprise, until I checked and saw that light went throught the tape! More opaque black tape confirmed it. Also the hair dryer test convinced me it wasn't due to heat -- it was hot enough believe me, there was a toasty smell in the air!

The "light" incident reinforced my feeling that the DC offsets at U104, U105 (LM308 ) along with the threshold of the U106 LM393 inputs is critical for detecting small signals (and explaining the differences in peoples TGSLs) because I believe the light boosted the LM308 output offset and increased my sensitivity.

**** I have now implemented the "dfbowers" mod on my PCB and look forward to testing it. My first test was to crank up the sensitivity in my lab, expecting to hear the audio go to a solid beep. To my surprise, it did not, it just got more chatter. This makes me feel that my LM393 chip may have a bigger threshold than I assumed (those sneaky discount parts...).

So, I think we may want to extend the mod even further just to make sure we have absolutely full sensitivity control. I'm thinking maybe replacing the 220K R39 with 100K. This should change the min threshold from -17 mV to -37 mV, hopefully enough to overcome any out-of-spec LM393 and LM308. Of course, going to 50K gives you even more range, but makes the sensitivity control more "touchy".

Looking forward to your (dfbowers) tests with TL071 chips. I assume your point is to test using separate adjustments of the offset/thresholds on each input to the comparator to optimize those thresholds. Because the DISC and GB channels have different gains, this might be useful. We might also want to consider literally boosting the gain of the GB channel a little to see how that plays out.

Where exactly will you (dfbowers) put the extra adjustments? I guess there are a number of places as Qiaozhi noted. The only caveat to using the non-inverting input to U104 (LM308 ) would be that the stage gain for non-DC signals maybe could magnify noise from the pot. A bypass capacitor around it might be a good idea. In fact, a bypass cap across resistor R38 in the sensitivity control might be a good idea in principle.

If you can identify an optimal ratio of the DISC sensitivity threshold to the GB threshold, we might be able to just add another resistor to the sensitivity threshold circuit and pick off two voltages that move together nicely.

Regards,

-SB

PS - I know my screeds are too windy for easy consumption; sometimes lots of issues to sift through to get near the reality...

It looks like we have at least 2 choices for playing with offset and either may let us arrive at the same result. I like Qiaozhi's suggestion on biasing pin 3 of u104 but I may try playing with the offset pins. If I cannot get the range required I will try his suggestion.

I was just going to piggy back a small trimmer on a pair of TL071s and solder directly to the legs, then run a small wire to the -v5 rail.

Don