TGSL Tuning & troubleshooting

[dfbowers]

Hi dfbowers -

Would you test how the sensitivity pot affects your air depth tests using setup similar to your video?

What is depth with min sensitivity.

What is depth with max sensitivity.

Would be interested in same tests by anyone with good working TGSL.

Regards,

-SB

Sure, I would be glad to post more videos. Just from a quick test, sensitivity set to minimum cuts off maybe 2 to 3". More than anything it just quiets it down.

Just as an experiment, I shorted the 1k ohm resistor (R37) and expected more range.. Didn't make any difference at all. The only way I could get even more range was to turn the GB pot to minimum... but it's a waste of time because it no longer ignores the ground properly.

I was out relic hunting with it the other day and it seemed a little noisy until I set the sensitivity pot to about mid-range, then it was dead quiet until I hit a target. I couldn't tell how much depth I lost in the ground. I will try to make a video of my second machine as well just to show similar, if not the same detection range, plus 3 different coils.

Don

[simonbaker]

your noise please

To everyone with a TGSL, both good and bad, please help with the following tests:

1a. With coil fully attached, please try to measure the noise at LM308 (U104) output pin 6 (using oscilloscope) for peak-to-peak measurement. I know this is difficult because it varies, make a good estimate.

1b. Please also measure the same noise with AC voltmeter and also say what frequency range your voltmeter is accurate for.

2. Please remove RX coil at PCB connection and measure DC offset voltage at LM308 pin 6, both channels.

3. Please report the air depth you measure with 1 Euro and/or US quarter. Measure for both minimum and maximum sensitivity pot setting. If too much chatter at maximum sensitivity, choose the most sensitive setting you can and describe the setting.

-----------------

Thanks. I want to do everything possible to figure out why some TGSL get poor depth, and how to fix. Your answers will help. I am thinking about the following concept:

Maybe noise level measured at LM308 (U104) pin 6 is cause of poor depth.

Noise breaks up the detected signal, and the small pieces cannot get through the low pass filter that follows.

We need to compare the background noise and DC offset of the LM308 U104 output pin 6 using the data you measure and post to this thread.

Regards,

-SB

[dfbowers]

To everyone with a TGSL, both good and bad, please help with the following tests:

1a. With coil fully attached, please try to measure the noise at LM308 (U104) output pin 6 (using oscilloscope) for peak-to-peak measurement. I know this is difficult because it varies, make a good estimate.

1b. Please also measure the same noise with AC voltmeter and also say what frequency range your voltmeter is accurate for.

2. Please remove RX coil at PCB connection and measure DC offset voltage at LM308 pin 6, both channels.

3. Please report the air depth you measure with 1 Euro and/or US quarter. Measure for both minimum and maximum sensitivity pot setting. If too much chatter at maximum sensitivity, choose the most sensitive setting you can and describe the setting.

-----------------

Thanks. I want to do everything possible to figure out why some TGSL get poor depth, and how to fix. Your answers will help. I am thinking about the following concept:

Maybe noise level measured at LM308 (U104) pin 6 is cause of poor depth.

Noise breaks up the detected signal, and the small pieces cannot get through the low pass filter that follows.

We need to compare the background noise and DC offset of the LM308 U104 output pin 6 using the data you measure and post to this thread.

Regards,

-SB

Bingo! Simon.. Check out the videos that I uploaded on YouTube..

I did some measurements around U106. I wondered why I could not get my circuit to chatter at increased sensitivity. The lowest that I could get the inputs in U106 pins 2 & 6 is 3.6mV ( 8V* (1k/10k) * (1k/ 220K) = 3.6mv. Also, I notice that the threshold for the comparaters (U106) is
-25mV, even in a noisy house!! I guess there is no offset for an LM393 .. So depending on your IC, it could be different I would think.

NOISE IS THE BIG LIMITING FACTOR..

So I did this. Replace 36 with a 100k pot. One end to +8V, the wiper goes to R39 and kluge in the other end to -5V. At the very least, I gain several inches of detection range. At the most, I gain is 10cm for 1e on top of my 30!!! Outside where it's less noisy the mod is acceptable. The potential flaw in the mod is that 5v may not be regulated enough, but I have not played with it much.

Also.. keep in mind that my observations have not been field tested. This may be highly impractical.. but could be a source of other peoples problems.

Don

[Stefano]

Wow!!!! I can't wait to be at home to test your mod and make the measurement suggested by SB.
Before leaving, following SB's clues I checked quickly the output of U104 and I found to be very noisy, but since it was syncronous, I was thinking of it being related with the -5V conversion.
Anyway this is a very good news !!!!
We look forward to confirming a field test improvement...

[Cyclonite]

This is indeed the best way to go, by comparing standardized replicable measurements and procedures by those who have built the TGSL.

It would also be interesting to see measurements performed on the OP07, as done on the LM308. This modern LM308 follow-up chip has much better specs, and is sold at an even lower price too. The availability of the OP07 should not be a problem, for it is not classified as an obsolete component, thus forging this component should not be a problem yet, and least of all being a very lucrative business based on it's $1 price tag. The 100pf C22 capacitor may be omitted when an OP07 is used.

Regretfully, I can't do these experiments myself, for I have not built the TGSL yet.

Here you can download the OP07 Data Sheets:

OP07 Datasheets (2 x RARed Part-Files):

OP07 - Ultralow Offset Voltage Operational Amplifiers (Analog Devices).pdf
OP07 - Ultralow Offset Voltage Operational Amplifiers (Intersil).pdf
OP07 - Ultralow Offset Voltage Operational Amplifiers (Maxim).pdf
OP07 - Ultralow Offset Voltage Operational Amplifiers (National Semiconductor).pdf
OP07 - Ultralow Offset Voltage Operational Amplifiers (Raytheon Semiconductor).pdf
OP07 - Ultralow Offset Voltage Operational Amplifiers (STMicroelectronics).pdf

Regards,

Robert

[lunamay]

hi

i have a question..is there any method to check a phase shift without a scope?? maybe i could check it with a unimeter somehow??

best regards

[dfbowers]

To everyone with a TGSL, both good and bad, please help with the following tests:

1a. With coil fully attached, please try to measure the noise at LM308 (U104) output pin 6 (using oscilloscope) for peak-to-peak measurement. I know this is difficult because it varies, make a good estimate.

1b. Please also measure the same noise with AC voltmeter and also say what frequency range your voltmeter is accurate for.

2. Please remove RX coil at PCB connection and measure DC offset voltage at LM308 pin 6, both channels.

3. Please report the air depth you measure with 1 Euro and/or US quarter. Measure for both minimum and maximum sensitivity pot setting. If too much chatter at maximum sensitivity, choose the most sensitive setting you can and describe the setting.

-----------------

Thanks. I want to do everything possible to figure out why some TGSL get poor depth, and how to fix. Your answers will help. I am thinking about the following concept:

Maybe noise level measured at LM308 (U104) pin 6 is cause of poor depth.

Noise breaks up the detected signal, and the small pieces cannot get through the low pass filter that follows.

We need to compare the background noise and DC offset of the LM308 U104 output pin 6 using the data you measure and post to this thread.

Regards,

-SB

Simon,

I took a few measurements around lm308 pin 6 (both ICs)

1a. Noise measurement is less than 40mV P-P with a scope.

1b. I cannot get an accurate reading with a DVM.

2. With BOTH coils disconnected, output of either lm308s (pin 6) goes to + 1.6v. I cannot see any d.c. offset whatsoever with coils connected.. The noise level is right around 40mV p-p.


Hope this helps..

I will try to check TGSL #2 (TL071 version) as a comparison.. but performance is the same..

Don

[simonbaker]

Bingo! Simon.. Check out the videos that I uploaded on YouTube..

I did some measurements around U106. I wondered why I could not get my circuit to chatter at increased sensitivity. The lowest that I could get the inputs in U106 pins 2 & 6 is 3.6mV ( 8V* (1k/10k) * (1k/ 220K) = 3.6mv. Also, I notice that the threshold for the comparaters (U106) is
-25mV, even in a noisy house!! I guess there is no offset for an LM393 .. So depending on your IC, it could be different I would think.

NOISE IS THE BIG LIMITING FACTOR..

So I did this. Replace 36 with a 100k pot. One end to +8V, the wiper goes to R39 and kluge in the other end to -5V. At the very least, I gain several inches of detection range. At the most, I gain is 10cm for 1e on top of my 30!!! Outside where it's less noisy the mod is acceptable. The potential flaw in the mod is that 5v may not be regulated enough, but I have not played with it much.

Also.. keep in mind that my observations have not been field tested. This may be highly impractical.. but could be a source of other peoples problems.

Don

Excellent test. This is exactly what I think we need to dig into. I had forgotten to think about the actual comparator input threshold as being part of the equation, but of course it is.

In any case, you read my mind with your senstivity threshold mod. I have been thinking of going a step further and experimenting with a separate threshold for each channel, because of the "lower" gain to targets on the GB channel, and to give better ability to adjust for chip variations.

Probably a working design would have an additional trim pot to "balance" the two channels so that the normal sensitivity pot would be most effective. Kind of like the stereo balance and volume knobs. The trim pot could be adjusted once and left alone.

As you pointed out though, noise is ultimately the toughest problem, you can only lower the threshold so far. Maybe the problem is everybody's workshop has different noise. This is why I asked for data, to see if there is a pattern. Maybe we just have to get our MDs away from noise, then look at the threshold problem and either replace IC chips or maybe adjust some thresholds with trimming resistors.

You have certainly confirmed once again that Ivconics designs can give excellent depth and performance, so that's encouraging for us all.

There are actually some tradeoffs that could possibly help reduce noise and improve depth, but those are for the next generation design experiments (which I'd like to get to if I could only get my MD to work like dfbowers' does!).

Maybe if we all move next door to dfbowers our MDs will work.

Anyway, stunning results.

-SB

[simonbaker]

Test points

lunamay asked about where to measure voltages.

Attached are images showing the voltage test points. The are simply the outputs (pin 6) of each LM308 chip. PCB is TGSL Final I think.

lunamay -- use voltmeter to measure both DC and AC voltages, with RX coil attached and also unattached, as described previously. Put on most sensitive settings.

Find the specs on your voltmeter and see what freq range the AC volts are accurate for.

-SB

[pebe]

.....Probably a working design would have an additional trim pot to "balance" the two channels so that the normal sensitivity pot would be most effective. Kind of like the stereo balance and volume knobs. The trim pot could be adjusted once and left alone.....

....There are actually some tradeoffs that could possibly help reduce noise and improve depth, but those are for the next generation design experiments (which I'd like to get to if I could only get my MD to work like dfbowers' does!)......

-SB

Hi Simon,
If you are thinking of doing a second generation circuit, here are a couple of things that you might like to consider (I have done a search but cannot find the original circuit for TGSL so I am going by the circuit you provided that had the voltage divider to C9 and C10).

U101 as shown is not balanced. The input impedance from RX1 to pin6 is 5K1, but the impedance from RX2 into pin5 is 5K1 + 220K. Also there is no DC isolation between the inputs. It would be better if R16 was 5K1 and R14 were replaced by a 100nF capacitor.

The two signal paths to the right of R32 and R34 are identical and are finally mixed in U106a&b. You could take the output from R32 to pin2 of U105 and mix them there (this is a normal mixing arrangement in audio amplifiers). U104 and U106A would then be redundant. The LM393 has an open collector so even as the circuit stands at the moment, R54, C24, and D12 are not required. You could use the spare LM358 in place of U106b but then you would need to leave in D12.

R43 causes U107 pin6 to go negative. That biases the output pin7 high, so MPSA13 is switched on and SP1 is passing current when there is no signal. In the original TSG circuit SP1 was driven by an NPN transistor so there was no current with no signal and so less drain on the battery.

Now a couple of points that you can probably help me with. I built an IB detector about 30 years ago but nothing since, so I am a relative newcomer to these circuits. I am trying to get to grips with the nulling. When I built mine the coils were nulled to give zero induced signal from the TX coil, and phase didn’t enter the equation. But I notice here that coils are nulled for a certain phase shift. Could you explain for me? Also, I cannot understand the arrangement of C10/R22. It would seem that the loading of C10 on the tuned TX coil varies with R22.

[simonbaker]

recommendations

Thanks to dfbowers tests including modification to sensitivity pot wiring, I'm encouraged maybe we get closer to understanding why some of us don't get 28 cm or more depth. Maybe not just because of bad coils (although that's still very important).

Right now this is my recommendation for people with TGSL that gets less than 28 cm:

1. Test the "offset" DC voltage of both LM308 chips at output pin 6 with RX coil disconnected. If they are very different, replace one or both chips until the outputs have about the same offset (within 5 mV of each other).

2. Do the modification that dfbowers showed, connecting the 10K pot directly to -5V buss instead of R37. You can leave the wire to the R37 resistor temporarily and waste 5 mA. (please wait until we confirm how dfbowers wired his mod)

3. Take TGSL out of the workshop to area where it is very low electrical noise. Take test equipment, at least a good voltmeter, and some ferrite and metal targets. Then test air depth, playing with sensitivity pot to find best depth without much chatter. You should start with minimum discrimination and ground balance pot settings.

4. If you don't get good depth, use voltmeter to measure the AC and DC voltages at both LM308 output pin 6. Use millivolt settings. It may help to disconnect RX coil to measure DC voltage. But do connect the RX coil to measure the AC voltage. If you have a portable oscilloscope, make same measurements with scope.

5. If you get good depth, 28 cm or more, hooray! Time for champagne (or substance of choice). Now try setting ground balance with ferrite target. Also check discrimination with piece of aluminum foil or US Nickel or whatever you use.

6. Please report your findings here to help others.

Regards,

-SB

[simonbaker]

Hi Simon,
If you are thinking of doing a second generation circuit, here are a couple of things that you might like to consider (I have done a search but cannot find the original circuit for TGSL so I am going by the circuit you provided that had the voltage divider to C9 and C10).

Yes, it's pretty close to Ivconic's TGSL Final.

U101 as shown is not balanced. The input impedance from RX1 to pin6 is 5K1, but the impedance from RX2 into pin5 is 5K1 + 220K. Also there is no DC isolation between the inputs. It would be better if R16 was 5K1 and R14 were replaced by a 100nF capacitor.

I have always wondered about that tuned preamp circuit, and I don't have a good understanding of it right now. There is something puzzling about the choice of resistors there. Also remember that the RX coil to J2-2 is grounded in the coil head. Hard for me to understand all the effects.

Maybe we can use LTSpice to try your ideas and see what it means. What is your idea with replacing R14 with 100 nF capacitor? Would that change RX tuned circuit though?

The two signal paths to the right of R32 and R34 are identical and are finally mixed in U106a&b. You could take the output from R32 to pin2 of U105 and mix them there (this is a normal mixing arrangement in audio amplifiers). U104 and U106A would then be redundant.

This is an interesting idea I have wondered about too. There are some complications though. By mixing at LM393 as normal, you get a "logical" mixing instead of analog mixing. With analog mixing, the problem is that the two channels do not have the same gain to the targets, partly determined by the disc and GB pot settings. So one channel response is much bigger than other, and mixing them analog does not achieve the logic we want.

However, it may be possible to increase the gain of the GB channel to get a more logical mixing at the LM308 input as you suggest. The gain ratio is not fixed though, because the pots change it. But maybe it would work OK, expecially for a simplified, cheaper model of the TGSL.

I'm very interested in super-simple, cheap versions that work pretty well, even if not as good. Another idea is to leave out the GB channel completely. Would have to try it to see how it behaved.

The LM393 has an open collector so even as the circuit stands at the moment, R54, C24, and D12 are not required. You could use the spare LM358 in place of U106b but then you would need to leave in D12.

Great ideas to try for simplifying. Again, I don't know all the subtleties of that part of the circuit, although I played with it some. (One of my ideas is to provide some positive feedback around the LM393 including a capacitor to make a short-term latching of the target signal, to hold it a little longer. But that is another story...)

Capacitor C24 and R54 are part of a low-pass filter that filters the pulses that come out of the LM393 chips. But you probably could combine them into C25 and the LM358 U107 filter parts.

I have thought a lot about diode D12 and resistor R40. R40 seems to be a "feed-forward" resistor with the idea of providing a small amount of audio-depth sensitivity; in other words, stronger targets sound louder. I don't think it is that noticeable or really that useful, but that's what I think it was designed for. You could probably just connect R40 to ground for basically good results.

Diode D12 may actually provide an "assymetric" filtering, so that the pulse is filtered faster in one direction that the other. This may help with the audio and target response. Maybe better ways to do that (probably so since Tesoro moved on to other designs).

Overall, I find it a very peculiar part of the circuit, with hard-to-understand levels and filtering. Good area to experiment.

R43 causes U107 pin6 to go negative. That biases the output pin7 high, so MPSA13 is switched on and SP1 is passing current when there is no signal. In the original TSG circuit SP1 was driven by an NPN transistor so there was no current with no signal and so less drain on the battery.

I don't think so. Although U107 pin 6 is slightly negative, the non-inverting pin 5 is normally more negative because the LM393 pulls it down to -5V when no target signal is present, and so the output is not normally high.

Now a couple of points that you can probably help me with. I built an IB detector about 30 years ago but nothing since, so I am a relative newcomer to these circuits. I am trying to get to grips with the nulling. When I built mine the coils were nulled to give zero induced signal from the TX coil, and phase didn’t enter the equation. But I notice here that coils are nulled for a certain phase shift. Could you explain for me?

There are differences of opinion on the nulling, at least I am different. For this circuit, I have not been convinced that the tiny difference in nulling to one side or other is important. But I need to test once I get a really sensitive detector. dfbowers could probably test this right now as he has excellent models.

One argument (mine) is that you prefer to null (220 deg) so as to keep the voltage on capacitors C15 and C12 positive so you don't forward bias the JFets TR4 and TR5 when the gate is pulled up to ground level. However, with normal coils, I don't think you get a very negative voltage if you null on the other side, so this argument may be not too important. If you fool with very high gain coils (like I do), the null signal can be very large and then maybe it matters.

On the other side, if you null with the other phase (20 deg), it usually corresponds with moving the coils to a slightly more overlapped position. This is a little more ideal for the RX coil seeing the magnetic field from the TX coil. But it is a tiny difference, not sure how important.

If you null right in the middle (minimum), I'm not sure but there may be stability problems where the coils are more sensitive to vibrations, since small deformations create a greater phase shift (I speculate). Again, probably not that big a difference.

People may have other reasons, but I have not heard something which really grabs me. If you study the math for a "Synchronous Detector", you will see that it is a linear system, and that combined signals are detected separately. A constant phase offset (such as the null signal), creates a DC voltage at capacitors C15 and C12, which is essentially ignored by the bandpass filter that follows. So to a first order, the particulars of the null is ignored.

If you go too far from null with your coils, they become coupled and that introduces some other stuff.

Bottom line: I'm not sure null phase is really important here, but it has a history of obsession, and maybe there is some truth to it, just not convinced what that is.

Also, I cannot understand the arrangement of C10/R22. It would seem that the loading of C10 on the tuned TX coil varies with R22.

C10 / R22 is a traditional phase shifter, ideally shifting the phase from 0 to 90 deg magnitude. I agree that R22 does couple in varying degrees of capacitance to the TX oscillator. It is a small capacitor of course compared to the oscillator compacitors. TX frequency is not that critical and I don't think you'll see much affect. Of course if you turn the pot to zero ohms you may not get anything useful out of the comparator, so normally you just don't go to that extreme. Usually GB pot is set somewhere near the middle.

Note that with my voltage divider mod, the effect of C10 / R22 is a little different. It probably isolates the oscillator somewhat from C10, but causes a little coupling of the disc and GB channels. But usually the GB pot is not moved much once set, so coupling not too important I think.

Long answer but lots of interesting ideas there .

Regards,

-SB

[simonbaker]

Wow!!!! I can't wait to be at home to test your mod and make the measurement suggested by SB.
Before leaving, following SB's clues I checked quickly the output of U104 and I found to be very noisy, but since it was syncronous, I was thinking of it being related with the -5V conversion.
Anyway this is a very good news !!!!
We look forward to confirming a field test improvement...

Hi Stefano:

If you see synchronous noise, like voltage spikes, that is another issue -- it may or may not be important.

I am more interested in the low frequency random noise, best seen if you put your scope very slow, like .1 seconds/div. If your scope has a low freq filter, than might help to mask out the spikes.

For some reason, the fast synchronous spikes just may not matter. But it's also possible they are part of the problem, so we'll wait and see.

Don't forget to check the offset voltages with the RX coil disconnected also.

Regards,

-SB

[simonbaker]

imagining it?

After a pitiful attempt at a field test (described in TGSL Experiments thread), I came home, soldered the broken battery pack lead, and hooked it all up again.

Nulled the coil and moved the scope probes away from the coil cable for less shaking and any other interaction.

After a while, it seemed the noise was lower, less chattering than usual.

I used alligator clip to jumper the sensitivity pot directly to -5V, similar to dfbowers mod.

Carefully playing with sensitivity control, I achieved a point where (I have to be careful here) I think perhaps (could it be?) that, even though some chatter was present (still, I think I was not imagining it), by waving my quarter rather quickly in the vicinity (all the while suppressing the temptation to cheat) of (dare I say) 30 cm, it seemed to me (no other witnesses were present) that in fact (I don't believe I was dreaming) there was actual detection of the quarter (ok I said it). I would say incontestable around 25 cm.

I was using dfbowers coils with a makeshift foil shield around the RX coil only.

While I admit there was room for subjectivity in that test, it really left me in a state of believing that such a depth is actually achievable and might at some point be achievable by my very own humble circuit.

However, I'm a little at a loss as to why I am (I can only assume temporarily) having a little hint of success here. Some thoughts:

1. Re-situating the coils after transporting them to the field and back managed to put them in a more mechanically stable position.

2. The surrounding EMI temporarily subsided for some reason, refuting Murphy's Law.

3. Shoving probes and wires around got rid of some kind of noise or vibration source. (I find shoving wires around often a successful way to solve problems, which does not please me to be frank).

4. The dfbowers mod allowed the extra depth while the noise happened to be low, confirming that the LM308 outputs are a hopeful place to work on tuning our TGSLs.

The mission has not changed - to get every cranky TGSL working to its full potential. I am less dubious that this is realistic - in fact, we're going to do it (ok I said it).

[simonbaker]

surprise

In an unbelievable gesture, it looks like dfbowers may actually loan me one of his completely assembled and working TGSL models.

This will allow really comparing the circuits, coils, and noise environment, and hopefully understanding what is necessary for a good working TGSL (besides being dfbowers, ivconic, max, etc).

If this happens, pressure will be on me to report findings, which I will do to the best of my ability. As you may imagine from the number of posts I have put here, I need to put a little time in on my job also because I have been having too much fun with noise and offsets and such and have not checked my bank account lately and am a little afraid to.

Anyway, thanks again dfbowers!

Let's all continue looking at the outputs of LM308, find less noisy environment to test, select good IC chips, and play with dfbowers sensitivity mod. Of course, build excellent coils too.

Cheers,

-SB