TGSL Tuning & troubleshooting

[simonbaker]

my dfbowers mod

Whatever the signficance, that's quite the observation that you have made! Who would have ever thought? I would almost be tempted to cover them with a small piece of electrical tape and repeat the test just to make sure that it is light and not heat. Now when you say the DC bias on the LM308 are you speaking of DC offset? If so, are the offsets different?


Simon, I'm still with you on the idea that the offset voltage on the LM308s can have some impact on performance. If the outputs of u104 and u105 are not in completely in sync we must be losing some part of the signal to u106. Maybe that's why some circuits seem to have a more robust audio or at least contribute to it?

With my next experiment, I am going to kludge in an offset adjustment to TL071s in place of the LM308 (u104 and u105). Just looks like I need two small pots with the wipers tied to the -5v rail.. I can probably just solder the pot to the IC legs of the TL071s and swap the LM308s out. Waste of time or not.. I need to know!! Need to order a set, so stay tuned!!

Don

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...

[dfbowers]

If you think the problem is caused by having different DC offsets on U104 and U105, then you could try disconnecting pin 3 (non-inverting input) of U104, and connecting a trimmer preset to allow adjustment of the opamp reference voltage. First measure the DC offset voltage at pin 6 U105, and then adjust the trimmer to get the same voltage at pin 6 U104. Some detectors have this adjustment, which is referred to as DBIAS.

You might also want to try removing the limiting diodes D8, D9, D10 and D11. They're not really needed anyway.

I can see where adding a trimmer to U104 pin 3 would do the same thing as adding a DC offset adjustment to U104 and/or U105. I guess the other consideration would be to determine exactly where both U106a and U106B will trigger.. and adjust accordingly to allow them trigger more in unison. More things to investigate.

[Qiaozhi]

Whatever the signficance, that's quite the observation that you have made! Who would have ever thought? I would almost be tempted to cover them with a small piece of electrical tape and repeat the test just to make sure that it is light and not heat. Now when you say the DC bias on the LM308 are you speaking of DC offset? If so, are the offsets different?


Simon, I'm still with you on the idea that the offset voltage on the LM308s can have some impact on performance. If the outputs of u104 and u105 are not in completely in sync we must be losing some part of the signal to u106. Maybe that's why some circuits seem to have a more robust audio or at least contribute to it?

With my next experiment, I am going to kludge in an offset adjustment to TL071s in place of the LM308 (u104 and u105). Just looks like I need two small pots with the wipers tied to the -5v rail.. I can probably just solder the pot to the IC legs of the TL071s and swap the LM308s out. Waste of time or not.. I need to know!! Need to order a set, so stay tuned!!

Don

If you think the problem is caused by having different DC offsets on U104 and U105, then you could try disconnecting pin 3 (non-inverting input) of U104, and connecting a trimmer preset to allow adjustment of the opamp reference voltage. First measure the DC offset voltage at pin 6 U105, and then adjust the trimmer to get the same voltage at pin 6 U104. Some detectors have this adjustment, which is referred to as DBIAS.

You might also want to try removing the limiting diodes D8, D9, D10 and D11. They're not really needed anyway.

[Vladimir]

With my next experiment, I am going to kludge in an offset adjustment to TL071s in place of the LM308 (u104 and u105). Just looks like I need two small pots with the wipers tied to the -5v rail.. I can probably just solder the pot to the IC legs of the TL071s and swap the LM308s out. Waste of time or not.. I need to know!! Need to order a set, so stay tuned!!

Don

For the selection of the OP amp 103a / b, LM358 is not the best choice .... and there may be a problem? 442 is a better solution for me.

[Vladimir]

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

[dfbowers]

I think I have conclusive results now.

I tried hair dryer -- got much hotter than I realized (ouch) -- did not affect circuit that I could notice.

I took it outside again, and using small shadow maker, isolated the effect. It is caused by sun on the LM308 components, and I presume the diodes.

I brought it back indoors and verified that my energy-saver light does not affect it.

I used an old incandescent 60 watt bulb and it seemed I could hear a little more chatter.

So I put a DC voltmeter on the LM308 output and shorted the RX coil inputs.

With the 60 watt bulb off, the DC bias on the LM308 output was about 16 mV. When I turned the bulb on, it jumped to about 35 to 40 mV.

I put tape over the feedback diodes on the LM308s and confirmed that the voltage did not jump when I turned the light on.

So I feel that particular kinds of light on the D10, D11 diodes (LM308 feedback) affect the output offset bias and thus the sensitivity threshold.

Of course this shouldn't affect a final TGSL circuit in its box. But it could be a factor for anyone doing bench testing or testing outdoors.

Probably only applies to diodes with clear packages.


Forever in your service,

-SB

Whatever the signficance, that's quite the observation that you have made! Who would have ever thought? I would almost be tempted to cover them with a small piece of electrical tape and repeat the test just to make sure that it is light and not heat. Now when you say the DC bias on the LM308 are you speaking of DC offset? If so, are the offsets different?


Simon, I'm still with you on the idea that the offset voltage on the LM308s can have some impact on performance. If the outputs of u104 and u105 are not in completely in sync we must be losing some part of the signal to u106. Maybe that's why some circuits seem to have a more robust audio or at least contribute to it?

With my next experiment, I am going to kludge in an offset adjustment to TL071s in place of the LM308 (u104 and u105). Just looks like I need two small pots with the wipers tied to the -5v rail.. I can probably just solder the pot to the IC legs of the TL071s and swap the LM308s out. Waste of time or not.. I need to know!! Need to order a set, so stay tuned!!

Don

[mikebg]

hmmm, got to be wary of those solar cells

However there is a patent for photovoltaic powered metal detector.

[greybeard]

Probably only applies to diodes with clear packages.

hmmm, got to be wary of those solar cells

[simonbaker]

heat/light test

I think I have conclusive results now.

I tried hair dryer -- got much hotter than I realized (ouch) -- did not affect circuit that I could notice.

I took it outside again, and using small shadow maker, isolated the effect. It is caused by sun on the LM308 components, and I presume the diodes.

I brought it back indoors and verified that my energy-saver light does not affect it.

I used an old incandescent 60 watt bulb and it seemed I could hear a little more chatter.

So I put a DC voltmeter on the LM308 output and shorted the RX coil inputs.

With the 60 watt bulb off, the DC bias on the LM308 output was about 16 mV. When I turned the bulb on, it jumped to about 35 to 40 mV.

I put tape over the feedback diodes on the LM308s and confirmed that the voltage did not jump when I turned the light on.

So I feel that particular kinds of light on the D10, D11 diodes (LM308 feedback) affect the output offset bias and thus the sensitivity threshold.

Of course this shouldn't affect a final TGSL circuit in its box. But it could be a factor for anyone doing bench testing or testing outdoors.

Probably only applies to diodes with clear packages.


Forever in your service,

-SB

[simonbaker]

Maybe it's a heat effect rather than light. How about trying a hairdryer?

Good idea. It was a very quick effect using the shadow of the sun, but I'll give the hairdryer a try.

-SB

[Qiaozhi]

(I wish the edit control would tell you how much time until expires... starting again...)

I just tried the light experiment in my workshop with fairly bright energy-saver bulb (which they say puts out significant UV).

I could not see any effect of the light at all on my PCB. The DC offset of the LM308 did not budge as light was turned on and off. I could not make the sensitivity control cause a continuous beep like when in the sun. I'm a little surprised. I even tried shorting the RX coil to make it as quiet as possible, but no effect noticeable.

I'll try to replicate the sun effect again outside.

A mystery... wouldn't have it any other way.

-SB

Maybe it's a heat effect rather than light. How about trying a hairdryer?

[simonbaker]

light effect at bench

(I wish the edit control would tell you how much time until expires... starting again...)

I just tried the light experiment in my workshop with fairly bright energy-saver bulb (which they say puts out significant UV).

I could not see any effect of the light at all on my PCB. The DC offset of the LM308 did not budge as light was turned on and off. I could not make the sensitivity control cause a continuous beep like when in the sun. I'm a little surprised. I even tried shorting the RX coil to make it as quiet as possible, but no effect noticeable.

I'll try to replicate the sun effect again outside.

A mystery... wouldn't have it any other way.

-SB

[simonbaker]

Have you tried the same experiment at home using a light bulb instead of sunlight?

I will do that post-haste. I recall now having similar problem before with some circuit (maybe earlier TGSL prototype) - I learn slowly...

I tried to use LTSpice to simulate a photodiode in place of D11, but I don't think I have realistic model yet. It shows huge effect, like .7 volt offset in output.

Regards,

-SB

[Qiaozhi]

The fascinating thing was the affect of sunlight on the MD -- dramatically different behavior vis a vis the sensitivity control. With sun on the PCB, I was able to tune it to get more air depth, but the control could also go beyond chatter into constant tone.

Have you tried the same experiment at home using a light bulb instead of sunlight?

[simonbaker]

Hi Simon,

Now you can understand why Tesoro don't bother to do this. Can you imagine the problems in manufacturing? For stability and ease of production, it is much easier to stay with the 20cm to 25cm detection distance. Most of the issues you are experiencing can be put down to component tolerances, and maybe PCB layout and general wiring.

Good point. I'm actually pleased in many ways that my PCB did as well as it did -- I feel I could make a serviceable MD with it. Of course in this forum obsession is a virtue, and I confess I have been bitten by the bug...

The fascinating thing was the affect of sunlight on the MD -- dramatically different behavior vis a vis the sensitivity control. With sun on the PCB, I was able to tune it to get more air depth, but the control could also go beyond chatter into constant tone.

My first guess is my 4148 diodes -- they have clear packages and work quite well as photo-diodes I think. I suspect the culprits could be the feedback diodes on the LM358 and LM308 amps, particularly the LM308 -- the diodes could be influencing the output offset. If so, then dfbowers sensitivity mod may provide a similar "boost" as the sunlight did. We'll see.

-SB