TGSL Tuning & troubleshooting

[simonbaker]

U104 was the unlucky component..

Don

Ah, U104, much the mischief maker -- well, sometimes you have to keep them in line that way...

Hmm...at first I thought: kind of makes sense, if the output pinned high. You were left with just the anemic GB channel, causing loss of depth. But then I realize that the GB channel is the weak link all the time, so pinning U104 high shouldn't cause such a loss of depth. A nice mystery to waste time on, good thing it didn't happen to me (yet)...

-SB

[dfbowers]

Good thought. I'm pretty sure one difference between your and my PCB is that my LM308s float the output at about 4 to 6 mV higher than yours and can take me into the motorboat range. Of course there may be some other factors involved too.

I have monitored the RX coil noise on both LM308s and they look quite similar so it seems both chips are working about the same -- or both equally crappy! I'm coming to the conclusion that I may be the poster boy for discount parts with questionable specs. I will no doubt be doing some parts replacements soon to see what the effects are.

Right now I am working (reinventing the wheel / fussing over nonsense) to try to troubleshoot and explain two differences in your and my PCBs:

1. Using same coil, your RX null signal is somewhere around 15% to 25% higher than mine (depends which day I test...).

2. Your noise level (in my lab) at output of LM308 seems less than mine (difficult to quantify -- 30% - 50% ?).

These two puzzles seem to actually conspire to create a puzzle "synergy", because number (1) implies you have more gain, which you would think would result in higher noise level, but vexingly, you seem to have less! A conundrum of the highest order!


I thought number (1) would be a slam-dunk to troubleshoot, but it has turned into an incredibly slippery quandry -- in fact it got so slippery that I had to take it to the quandramat to get it cleaned and start over. You wouldn't believe the ridiculous number of tests that still have not revealed the cause.

For number (2) I believe I may have deduced a plausible explanation -- it is a little surprising and not what I was expecting, but I need to confirm a few things before reporting.

Now, most people would consider this all a serious waste of time, and they are of course correct. However, it is important due-diligence for me because I want to work on my electronics troubleshooting skills. I need to keep practicing how to understand and apply the principles surrounding test measurement (such as grounding, probe capacitance, differential measurements, bandwidth limitations, etc.), and how to reconcile measurements with theory.

On the other hand you can always just start replacing parts. That actually works quite well.

-SB

P.S. Don - which LM308 did you fry?

U104 was the unlucky component..

Don

[simonbaker]

Simon,

In regards to your post that I have cited in red. I had an incident today where I accidently shorted one of my LM308s (the one next to the 1000V cap).. It's toast. As temporary fix I replaced it with a TL061, until I can get a new LM308. My symptoms were EXACTLY as you have cited, along with detection range being nearly cut in half. Motorboating below max sensitivity turning into a steady tone at max sensitivity. Changing the IC turned it back into it's old self instantly.. No motorboating whatsoever and max sensitivity just gives a little chatter.. back from 20cm detection range to 36cm. If you are still having issues with your PCB, maybe you have a bum component in there somewhere?

Don

Good thought. I'm pretty sure one difference between your and my PCB is that my LM308s float the output at about 4 to 6 mV higher than yours and can take me into the motorboat range. Of course there may be some other factors involved too.

I have monitored the RX coil noise on both LM308s and they look quite similar so it seems both chips are working about the same -- or both equally crappy! I'm coming to the conclusion that I may be the poster boy for discount parts with questionable specs. I will no doubt be doing some parts replacements soon to see what the effects are.

Right now I am working (reinventing the wheel / fussing over nonsense) to try to troubleshoot and explain two differences in your and my PCBs:

1. Using same coil, your RX null signal is somewhere around 15% to 25% higher than mine (depends which day I test...).

2. Your noise level (in my lab) at output of LM308 seems less than mine (difficult to quantify -- 30% - 50% ?).

These two puzzles seem to actually conspire to create a puzzle "synergy", because number (1) implies you have more gain, which you would think would result in higher noise level, but vexingly, you seem to have less! A conundrum of the highest order!


I thought number (1) would be a slam-dunk to troubleshoot, but it has turned into an incredibly slippery quandry -- in fact it got so slippery that I had to take it to the quandramat to get it cleaned and start over. You wouldn't believe the ridiculous number of tests that still have not revealed the cause.

For number (2) I believe I may have deduced a plausible explanation -- it is a little surprising and not what I was expecting, but I need to confirm a few things before reporting.

Now, most people would consider this all a serious waste of time, and they are of course correct. However, it is important due-diligence for me because I want to work on my electronics troubleshooting skills. I need to keep practicing how to understand and apply the principles surrounding test measurement (such as grounding, probe capacitance, differential measurements, bandwidth limitations, etc.), and how to reconcile measurements with theory.

On the other hand you can always just start replacing parts. That actually works quite well.

-SB

P.S. Don - which LM308 did you fry?

[dfbowers]

For these noise tests I used min-sensitivity and so the comparators were not motorboating -- however, they were being triggered by the noise. But the speaker does not sound because the pulses are too short to get through the next low pass filter.

My PCB, on my noisy workbench, does not fully motorboat with the RX coil active because the noise effectively breaks up the motorboating. If I short out the RX coil, then I can get a pretty clear motorboating at the max sensitivity.

I can't explain why, but in my field tests, I was able to get motorboating below the max sensitivity, which turned into steady tone at max sens. It's as if when I'm in the mountains, there is a bias on the comparator inputs. I previously determined it was due to the sunlight on the diodes, but this time I had it covered quite well -- just a small hole where the cable came out. Next time I'll plug that completely.

Oscillator Trimming
------------------

I tried trimming the oscillator frequency to see the effect. Nominal frequency is 13.82 kHz.

At 13.55 kHz I get a strong tendency to motorboat, more than at 13.82 kHz, but I'm not sure it's from feedback, or rather the noise signal -- or maybe a confluence of both. The noise signal looks like a lower frequency sine wave.

At 13.32 kHz, the circuit will not motorboat at any sensitivity setting. The noise chatter seems suppressed.

At 13.10 kHz, the circuit seems less noise, I can turn the sensitivity up much higher without chatter. Detection of a Euro seems cleaner and clearer also.

However, I didn't retrim the RX coil for each oscillator setting, so probably the gain was just dropping. To test that, I added 1 nF and retested and indeed noise signal was higher (probably due to higher gain). So not sure that proved much, although strong motorboating at 13.55 kHz seemed to indicate some interaction of noise spectrum with oscillator frequency.

-SB

P.S. The more I test, the more I find that slight changes in oscillator frequency make big changes in the noise characteristic, suggesting interaction with some signal source such as 60 Hz power.

But overall dfbowers seems to have better noise characteristics at LM308 output, smoother, smaller.

Also, dfbowers DC offsets are LM308 output are 1, 0 mV for DISC, GB. Mine are about 5.5, 4.5 mV, which could partly be reason I can get into motorboat region.

Simon,

In regards to your post that I have cited in red. I had an incident today where I accidently shorted one of my LM308s (the one next to the 1000V cap).. It's toast. As temporary fix I replaced it with a TL061, until I can get a new LM308. My symptoms were EXACTLY as you have cited, along with detection range being nearly cut in half. Motorboating below max sensitivity turning into a steady tone at max sensitivity. Changing the IC turned it back into it's old self instantly.. No motorboating whatsoever and max sensitivity just gives a little chatter.. back from 20cm detection range to 36cm. If you are still having issues with your PCB, maybe you have a bum component in there somewhere?

Don

[simonbaker]

Simon,

Back to our discussion on coil Q , wire size and drive efficiency. I think that we already know what the answers are from other posts here, but for educating ourselves (and maybe not contributing anything new here) we can refer to George Payne's article on ground balancing. I did a quick test of current consumed vs. wire size for a colpitts oscillator (not TGSL) and a collection of coils that I have for similar inductance. By changing the wire size from 30 to 26 gauge, I could not see much difference in current consumption despite the coil Q being halved by changing wire size. The coil with 26 gauge wire actually drew about .25 mA less than coil with 30 gauge, but that was only a measurement with 2 coils and a DVM. So, maybe a small amount of energy is being saved by having a Tx coil with smaller DC resistance? which is what you had suggested, along with the below citation. The only unknown here is what type of Tx circuit George is refering to here (Treasure Baron)? Not sure if we are dealing with Colpitts..

Don

"The Transmit wire size has very little bearing upon the coils overall sensitivity. However, it will greatly effect how much current(or power) is required to drive the Transmit coil. The designer could make the Transmit wire size very small and reduce the weight of the coil. That would be very impressive. But you would not we impressed with the battery life. The coil would draw huge currents and drain the batteries quickly."

I agree. The overall comments on this make me less enthusiastic about "hi-Q" coils as a panacea. Especially observations (mikebg) about how hi-Q circuits will be easily modulated by ground effects -- of course you might use that to advantage somehow also, but probably not easy for our TGSL circuit.

However, I think it's worth putting to the test. I'm pretty sure manufacturers covered this ground in spades and found the sweet spot for performance, weight, etc. But let's not forget, they have different objectives than we do. They want to sell metal detectors. We want to customize MDs for our own use. Maybe there is a specialty use for a high-Q coil -- someone has to try it (and believe in it) to find a possible niche for it.

Your results seem to confirm what I think should happen -- a hi-Q coil should oscillate with less battery draw and hopefully more current in the coil.

Here's an experiment I'd like to try. It is very difficult to measure the actual current in the TX coil without disrupting the circuit. The closest I can think of is to use a little "sniffer" loop to sample the magnetic field. This won't give you amperes unless you do some math, but it should allow comparisons.

I'd like to tape a little sniffer loop to the underside of the search head, centered in the TX coil.

Then lower the coil head over the ground and note how the signal in the sniffer loop changes in the presence of the ground.

Repeat this with different Q coils. Repeat with different ground types (ferrite-like, conductive-like, etc.)

My point is: how resonant are we anyway, and how much does the ground change it, and can we make use of higher Q coils.

Oops -- is our null signal a sniffer loop already? Just monitor that???

BTW -- thanks for reminding me about C9, C10 -- definitely should try replacing at some point. I notice sometimes parts get fearsome hot during soldering - anyone know what that does to capacitors?

-SB

P.S. Mystery of the Contrary PCBs continues soon in TGSL Experiments thread... new results.

[dfbowers]

Very useful test, it's something I've wanted to know about.

Perhaps the law of diminishing returns kicks in. However I still wonder if maybe there is some other adjustment necessary to take advantage of such a coil because of some negative factors that come in.

There are a few things to speculate about. One is the "sixth power" law that someone claims -- I think it states that the received signal strength drops off as the sixth power of the distance beyond a certain point, so that it takes enormous additional power to get meager improvement.

Intuitively, it seems to me that the S/N should improve, if the main noise is EMI. One way to test that would be to measure the noise power at the output of the LM308 chips. It should be the same for both TX coils. If the hi-Q coil has higher noise, then we can suspect some kind of jitter or oscillator noise.

It would be interesting to compare the magnetic field strength of the two coils also. I would make a small "sniffer" loop, maybe a centimeter or two in diameter with twisted pair leads, mounted on a stick. Then sample the magnetic field by putting the loop in the field and observing the signal on an o-scope.

In some ways, if the thinner wire works just as well as thicker, we're happy coil builders. But it is one of those puzzles I don't understand.

I encourage you to make the finished coil and see if you can explore why it does or does not work better.

-SB

Simon,

Back to our discussion on coil Q , wire size and drive efficiency. I think that we already know what the answers are from other posts here, but for educating ourselves (and maybe not contributing anything new here) we can refer to George Payne's article on ground balancing. I did a quick test of current consumed vs. wire size for a colpitts oscillator (not TGSL) and a collection of coils that I have for similar inductance. By changing the wire size from 30 to 26 gauge, I could not see much difference in current consumption despite the coil Q being halved by changing wire size. The coil with 26 gauge wire actually drew about .25 mA less than coil with 30 gauge, but that was only a measurement with 2 coils and a DVM. So, maybe a small amount of energy is being saved by having a Tx coil with smaller DC resistance? which is what you had suggested, along with the below citation. The only unknown here is what type of Tx circuit George is refering to here (Treasure Baron)? Not sure if we are dealing with Colpitts..

Don

"The Transmit wire size has very little bearing upon the coils overall sensitivity. However, it will greatly effect how much current(or power) is required to drive the Transmit coil. The designer could make the Transmit wire size very small and reduce the weight of the coil. That would be very impressive. But you would not we impressed with the battery life. The coil would draw huge currents and drain the batteries quickly."

[Qiaozhi]

Sheerluck Ohms

That's funny.

[dfbowers]

Because I haven't been posting enough lately...

Did a couple of quick tests. I tested the gain of the main filter section (U103, U104); and swapped some chips.

1. Injected sine wave into non-inverting input pin 3 of U103 (LM35 to calculate gain at output of U104 (LM308 DISC channel) for my PCB and dfbowers PCB.

I used a small signal so output looked undistorted by feedback diode D10 on the LM308.

Both gains were virtually identical (approx 700) comparing my and dfbowers PCBs.

(input signal: .287 mV pp output signal: .2 V pp Frequency: 10 Hz)

2. Swapped LF353 (U101) and LM393 (U102) between my and dfbowers PCB.

I did not see any noticeable change in the results. Our noise signals have similar look, but my noise seems higher, maybe 50% to 100%.

Either my front end gain in the LF353 is 50 to 100% higher (contradicted by the null signal voltage), or maybe extra noise is introduced during demodulation of the RX signal by the synchronous detector.

Sync noise could be responsible. It seems it must come from oscillator, because both channels have virtually same noise signal.

Or something else...

-Sheerluck Ohms

Hey Simon,

I would just "shot gun" the thing and change all IC's between PCBS to see if the symptom moves, again trying to divide problems. Also, I can't remember if I asked, what type of caps did you use and where? Maybe compare specs of your caps if you can identify what they are.

I seem to remember in a thread somewhere in a discussion about C9 and C10 and how the quality of them can noticably affect things.

If nothing else, build a few more circuits. ~($40.00)?

Don

[simonbaker]

Gain test; chip swap

Because I haven't been posting enough lately...

Did a couple of quick tests. I tested the gain of the main filter section (U103, U104); and swapped some chips.

1. Injected sine wave into non-inverting input pin 3 of U103 (LM35 to calculate gain at output of U104 (LM308 DISC channel) for my PCB and dfbowers PCB.

I used a small signal so output looked undistorted by feedback diode D10 on the LM308.

Both gains were virtually identical (approx 700) comparing my and dfbowers PCBs.

(input signal: .287 mV pp output signal: .2 V pp Frequency: 10 Hz)

2. Swapped LF353 (U101) and LM393 (U102) between my and dfbowers PCB.

I did not see any noticeable change in the results. Our noise signals have similar look, but my noise seems higher, maybe 50% to 100%.

Either my front end gain in the LF353 is 50 to 100% higher (contradicted by the null signal voltage), or maybe extra noise is introduced during demodulation of the RX signal by the synchronous detector.

Sync noise could be responsible. It seems it must come from oscillator, because both channels have virtually same noise signal.

Or something else...

-Sheerluck Ohms

[simonbaker]

GB vs DISC channel noise

Very interesting -- I'm looking at the noise at the LM308 outputs for the GB and DISC channels.

The DISC and GB pots are both at minimum.

The two "noise" signals are almost identical amplitude and phase. Slight phase offset, maybe 20 to 40 deg? Fairly sinusoidal looking.

If I move the DISC or GB pot, I can change the phase relationship.

My first conclusion is that it is not due to noise in the long DISC pot leads. Whatever it is is common to both channels.

Video: http://www.youtube.com/watch?v=8ON0TXzIiHY

-SB

[simonbaker]

Have you tried swapping ICs between the PCBs? I believe you said somewhere that they were both socketed.

I haven't gotten to that yet, but good idea.

I could see it making a difference in threshold voltages, like the LM308 output biases, but can't imagine a difference in gain anywhere due to feedbacks.

I could try injecting a signal at the input of the first LM358 just to make sure the main filter section has the same gain.

It's hard to work on dfbowers PCB because he is so tidy there ain't much to hook onto!

Tidiness is always a suspect in my PCBs, but so far I'm believing that the dominent noise is RX coil-sourced -- or has to do with demodulating the RX signal -- that is one heck of an antenna and dwarfs any little leads here and there. And short-jumpering the RX coil effectively eliminates the noise I'm measuring.

But I just said the magic words -- demodulating the RX signal. Noise that could jitter the sync pulse could be the culprit, and tidiness could be a factor there. I mentioned long leads on my DISC pot. I could try very short leads and see what happens. My board is getting pretty beat up though...

Thought... I can compare the noise at the DB channel to the DISC channel. The GB pot is on board and has no dangling leads. I'll try to make an assessment of that. In the past I concluded the noise was about the same -- I'll re-check.


-SB

[Qiaozhi]

Hold the phone.

I just checked null signal amplitude at output of LF353.

Mine: 1.08 V pp

dfbowers: 1.42 V pp

Aside from fairly large null signal, it looks like dfbowers has higher gain by 40%. I thought I adjusted my C6 cap same as his using in-circuit capacitance measurements.

Well -- I can retrim my C6 to get approx same gain. I put in a crude trim and got my null signal up to 1.52 V pp.

Now I'm pretty sure my noise signal at LM308 output is higher than dfbowers, maybe 1.5X to 2X as high.

Why? is the question.

Is some of the noise oscillator noise? That could explain the difference. Of course dangling trim capacitors on my oscillator don't help that probably. But maybe it has more to do with the transistor or other components? Is my USB connector shabby?

Or is there some other reason for null voltage besides gain? Of course he does have a little more gain due to 4.7k vs. 5.1k resistors, which is a little different than adjusting resonance like I did.

Have to sleep on it, design more tests.

-SB

Have you tried swapping ICs between the PCBs? I believe you said somewhere that they were both socketed.

[simonbaker]

gain gain gain

Hold the phone.

I just checked null signal amplitude at output of LF353.

Mine: 1.08 V pp

dfbowers: 1.42 V pp

Aside from fairly large null signal, it looks like dfbowers has higher gain by 40%. I thought I adjusted my C6 cap same as his using in-circuit capacitance measurements.

Well -- I can retrim my C6 to get approx same gain. I put in a crude trim and got my null signal up to 1.52 V pp.

Now I'm pretty sure my noise signal at LM308 output is higher than dfbowers, maybe 1.5X to 2X as high.

Why? is the question.

Is some of the noise oscillator noise? That could explain the difference. Of course dangling trim capacitors on my oscillator don't help that probably. But maybe it has more to do with the transistor or other components? Is my USB connector shabby?

Or is there some other reason for null voltage besides gain? Of course he does have a little more gain due to 4.7k vs. 5.1k resistors, which is a little different than adjusting resonance like I did.

Have to sleep on it, design more tests.

-SB

[simonbaker]

similar noise personality

Adjusting my oscillator as close as possible to dfbowers made our two noise signals look much more similar.

So at least in my environment, you have to be very careful in trying to compare apples to apples. Oscillator frequency very important.

I still have the impression that dfbowers noise may be slightly less amplitude, but it is very difficult to tell -- in the time it takes to move the probes over and reconnect the coil, the noise character easily changes quite a bit. My oscillator has drifted down a little, maybe between 13.71 and 13.72 kHz.

However, the two PCBs now are more similar in measurements, including air depth test -- both fairly lousy due to noisy environment, so not useful really to compare here.

But let's assume dfbowers has lower noise signal at output of LM308. The noise clearly comes from the RX coil, because shorting the RX coil makes the noise go away. Why would his PCB detect a lower noise signal from the same coil? I don't know. A gain difference would explain it. I tried to match the RX resonant tank, but a difference there would make a different gain. We do have different resistors in the LF353 preamp (4.7k dfbowers, 5.1k SB), but his gain should be higher, not lower. His slightly lower capacitor at C14, C15 could boost gain a tad. Otherwise, I don't know -- and it may just be an illusion.

The best I can think of to truly measure this noise would be to hook up to a digital sampler and take averages, switching the probes exactly every 30 seconds or so for several hours. Not about to happen soon.

-SB

[porkluvr]

For a comparator used in the non-inverting configuration, the hysteresis voltage can be calculated with the following equation.
HYST = R1 * (VOH - VOL) / R2
where R2 is the positive feedback resistor.
In that case the hysteresis works out as zero volts, because R1 equals zero ohms.

I'll go along with that. I overlooked that there is a formula for hysteresis and that an input resistor is needed. I should glue the formula to my forehead so I won't forget it.

I have always believed that the 100k resistor is a "feedforward" path whose purpose was simply to make the TGSL audio louder for stronger targets. It effectively sets the pull-up voltage of the comparators to the output of the GB LM308, which is then passed through to the audio signal by the next stage. Note also that in the original TGS, the GB LM308 does not have the limiting.

-SB

Feed-forward makes sense. I've read that before. When it's late at night, my wishful thinking can sometimes lead to confabulation having no basis in reality.
That also holds true in the daytime, come to think of it.

Porkluvr, excuse me for joke
<snip>
Experiment showed, that TX circuit diagram of Eagle Spectrum has worst properties than

all other circuits we have tested with salty water. When operates at low power mode as

marked in posting #529, by lowering search head even to not enough conductive earth, the amplitude of oscillation so diminishes that change can be measured by a conventional AC voltmeter. This is because LC tank is pumped by constant current portions as in your circuit with 555 (posting #585). My circuit in posting #595 simply increases or decreases the current portion stabilysing amplitude across TX coil.

I wasn't mad. I was laughing too hard to be mad.

What you're suggesting is that the Lobo style system will lose power at the beach. You might be right but my "Plan A" PCB is already pretty far along in the design stage. I could hunt gold near my neighborhood without going to the beach.

I do have a plan B. I chose the Lobo drive circuit over that found in the Minelab Relic Hawk because it is simpler and uses less power.
But the RH circuit looks like a PID as you describe, and my experiments show that it is readily adaptable to the 14kHz range. It actually seems to work better above the "stock" 5kHz frequency because it NEEDS a high Q coil which is easy to devise at 14kHz. I do not know the math behind the PID controller but trial and error is a viable tool if I let LTspice do the heavy lifting.