TGSL Experiments

[Stefano]

My latest bench TGSL now has a case of the "beeps" -- beeps at about 10/s. Sometimes it settles down.
-SB

I had the same problem, but not always, it appeared from time to time.
I've tried shielding the circuit by covering it with a big stainless steel pot (pressure cooker ???) taken from the kitchen and connected to ground!
But it didn't fix the problem....

[dfbowers]

Using my unshielded test coil, adjustable null, I'm getting with new PCB similar to results of previous TGSL Final PCB:

Good detection of US quarter around 15 cm. Small beep to 20 cm. Sometimes get crack to 25 cm.

I tested to see how critical nulling is. I find nulling is not critical, I get similar depth results in wide range of nulling. This may be because the MD is not too sensitive to begin with. But I have always believed that nulling may not be so critical as we think, but chosen more for other reasons such as stable voltages.

I tested to see how critical nulling was to discrimination. I adjusted disc pot to just knock out a US nickel coin. When I moved null point through a range, it did not affect the discrimination noticeably.

I took MD outside to see if better environment, less noise. I did not see any improvement, which surprised me. But I will repeat that in another location some day.

I am just beginning to experiment with this PCB so have many adjustments and ideas to try. The most important goal is to get a depth of around 28 to 30 cm! I have not achieved that and so far it seems a difficult goal.

-SB

Wow.. pretty cool. I thought of trying that before but got caught up in playing around with etching. Simon, I'm just curious. What version did you build? Do you have a list of parts that you used and how do they compare to the list that I had posted?
Don

[dfbowers]

Using my unshielded test coil, adjustable null, I'm getting with new PCB similar to results of previous TGSL Final PCB:

Good detection of US quarter around 15 cm. Small beep to 20 cm. Sometimes get crack to 25 cm.

I tested to see how critical nulling is. I find nulling is not critical, I get similar depth results in wide range of nulling. This may be because the MD is not too sensitive to begin with. But I have always believed that nulling may not be so critical as we think, but chosen more for other reasons such as stable voltages.

I tested to see how critical nulling was to discrimination. I adjusted disc pot to just knock out a US nickel coin. When I moved null point through a range, it did not affect the discrimination noticeably.

I took MD outside to see if better environment, less noise. I did not see any improvement, which surprised me. But I will repeat that in another location some day.

I am just beginning to experiment with this PCB so have many adjustments and ideas to try. The most important goal is to get a depth of around 28 to 30 cm! I have not achieved that and so far it seems a difficult goal.

-SB

Simon,

One thing I noticed. At least with both of my TGSLs.. There is a significant difference between a U.S. Quarter and a 1e as far as detection distance in air goes. I just double checked this. 25cm is not that far off of my observations with a U.S. Quarter BUT.. I get 29cm with a 1e coin. It seems like this circuit is hot on low conductors, not high. Mis-adjusting the GB pot will get you a little more on the high end but there is nothing to gain outside in the ground. A gold wedding band gets me over 30cm.
If you can get your hands on a 1e that would be the thing to compare. If nothing else, I will send you a pair of coils to compare results.

Don

[pebe]

My latest bench TGSL now has a case of the "beeps" -- beeps at about 10/s. Sometimes it settles down.

I'm trying to pin down if it is mechanical or electrical. It seems suspiciously mechanical, since if I move suddenly the oscillations, visible on the scope, get larger and then settle down. But once the oscillations seemed to slow in frequency, which may argue against mechanical, although this building may have many modes.

I tried disconnecting speaker, but oscillations still visible on the scope (I put sweep at about .1 sec / division and monitor pin 6 of LM308's).

So I don't think it is magnetic feedback from speaker.

When I disconnect Rx coil, output is very quiet -- so definitely something picked up or involving the coil.

Using the discrim pot, I can discriminate out the beeps -- they are still visible on the scope, but now the two channels are out of phase and do not trigger the speaker.

Will be interesting to try to hunt this down.

My current feeling is that if you can get rid of these noise signals (maybe go outside), that you can play with increasing the gain to get more depth.

I'm thinking that maybe there is room to boost the gain; for some reason, those of us with only 20 cm depth just aren't amplifying the signal enough. I'll be testing, but the ambient noise and these "mechanical" oscillations are interfering for the time being.

-SB

Hi Simon,
I think you have the problem known as ‘motor-boating’. It is electrical and is usually caused by positive feedback from one stage to an earlier stage. Because of the RC time constants involved the coupling effectively makes a low frequency oscillator. In many cases this is due to poor decoupling of the supply rails, and in others it can be common coupling due to the wiring layout.

In your case it would seem to be the layout. The +12V and ground wires are single thin wires and go all the way round the board with no decoupling capacitors anywhere. I think you should increase the diameter of both wires, especially the ground wire from the regulator to the early stages, U101 etc. Double or even triple the number of wires would help. Also a couple of 25µF caps across the supply rails at various points would probably help.

The original TGS circuit used an LM2930 regulator that had a low ‘drop out’ voltage of 0.1V. That meant the battery could discharge down to 8.1V before regulation was lost. The replacement 7808 regulator is not a good choice because it has a dropout of 2V, so as soon as the battery falls to 10V it won’t regulate any more.

The other thing is that all manufacturers of the 78xx or 79xx series regulators recommend a 220nF capacitor on the input and a 100nF on the output to prevent RF instability in the regulator. If it oscillates the regulated voltage can go up. You may get away with fitting electrolytics at the moment, but some time in the future when they have dried out a bit and their ESR has gone up you will almost certainly have problems.

Hope that helps.

[Stefano]

One thing I was thinking about these days is : why we make all our rx coil of 6.5 mH? I've made a 10" trying to achieve this value, but according to this table
http://www.geotech1.com/forums/showp...64&postcount=2 it appears that for a 11" DD, Tesoro's Rx coil value is 5.38 and it's indeed confirmed by the lower than tx coil resistance.
Well, 5.38//15 = 17.72 kHz.
Would the target behave the same in terms of phase and sensibility if I look for that frequency with a 6.5mH coil? It woud be as to have a 12.4nF capacitor...
Or is that table wrong not only in the Lobo data?

[simonbaker]

Wow.. pretty cool. I thought of trying that before but got caught up in playing around with etching. Simon, I'm just curious. What version did you build? Do you have a list of parts that you used and how do they compare to the list that I had posted?
Don

Hi Don:

I basically took TGSL Final layout and re-drew it using PCBExpress to fit the .1" perfboard. I will definitely post the files, I'm waiting a little in case I made a mistake which would come out in testing. But so far seems to be a working circuit.

1. I did add the little voltage divider circuit I mentioned.
2. removed a couple of resistors off the unused U107 op amp (stabilizing instead with single wire neg feedback according to recommendation I saw somewhere).
3. used single MPSA13 darlington instead of dual NPN transistors for speaker.
4. Used 22k instead of 24k resistors.

Attached schematic shows circuit. Red circled areas shows changes I made to TGSL Final schematic. Green circled areas show parts substitution I made during construction - board accepts either parts.

If I can get this board working well, I will try a really screwy idea, which is to use a CircuitWriter conductive pen to draw most of the traces and see what happens. The idea is to make a design which is really easy for anyone to make with limited equipment, even though not the best or slickest model in the long run. Just something to have fun with.

I'm trying to work on easy techniques for coils also.

-SB

[simonbaker]

Simon,

One thing I noticed. At least with both of my TGSLs.. There is a significant difference between a U.S. Quarter and a 1e as far as detection distance in air goes. I just double checked this. 25cm is not that far off of my observations with a U.S. Quarter BUT.. I get 29cm with a 1e coin. It seems like this circuit is hot on low conductors, not high. Mis-adjusting the GB pot will get you a little more on the high end but there is nothing to gain outside in the ground. A gold wedding band gets me over 30cm.
If you can get your hands on a 1e that would be the thing to compare. If nothing else, I will send you a pair of coils to compare results.

Don

If I could get a solid 25 cm with quarter I would be pleased, but I think I'm more around 20 cm at best.

Comparing with your coils is an exciting idea, let's keep that in mind, thanks for the offer. It would really help identify how important the coils are in the performance vs. possible circuit limitations. It will take me a while to feel like I have a stable platform to test with here. Right now, things seem to change somewhat day-to-day and I don't trust my observations yet. It's hard to be rigorous, so many things to play with.

Of course, do we know if coils would survive the journey unchanged? I guess after returning them you can test again to verify.

I will work toward getting ready to try your coils some day; very kind of you.

I'm working on getting a 1 Euro - my wife has friends in England.

Cheers,

-SB

[simonbaker]

Try putting the PCB in a metal enclosure with the case connected to 0V.

Not perfect, but I made a crude shield with cookie tin. I did not see enough change to feel it was the main problem.

So I switched my "adjustable" coil for a glued coil that I made a year or two ago.

Here's where I start going nuts. It seemed there was still a low freq oscillation of some kind (at output of LM308s), but a little higher freq than before, and much less sensitive to vibration (like thumping the floor).

So I went to bed to continue next day. Today, the signal looks more random, I'm not seeing the clear oscillation, and not getting regular beeps, just more random ones.

Could there be different electrical interference (EMI) today? Maybe different temperature?

I will re-hookup other coil (adjustable) and see if the beeping comes back.

Then I'll continue testing with coils.

-SB

[simonbaker]

Hi Simon,
I think you have the problem known as ‘motor-boating’. It is electrical and is usually caused by positive feedback from one stage to an earlier stage. Because of the RC time constants involved the coupling effectively makes a low frequency oscillator. In many cases this is due to poor decoupling of the supply rails, and in others it can be common coupling due to the wiring layout.

In your case it would seem to be the layout. The +12V and ground wires are single thin wires and go all the way round the board with no decoupling capacitors anywhere. I think you should increase the diameter of both wires, especially the ground wire from the regulator to the early stages, U101 etc. Double or even triple the number of wires would help. Also a couple of 25µF caps across the supply rails at various points would probably help.

The original TGS circuit used an LM2930 regulator that had a low ‘drop out’ voltage of 0.1V. That meant the battery could discharge down to 8.1V before regulation was lost. The replacement 7808 regulator is not a good choice because it has a dropout of 2V, so as soon as the battery falls to 10V it won’t regulate any more.

The other thing is that all manufacturers of the 78xx or 79xx series regulators recommend a 220nF capacitor on the input and a 100nF on the output to prevent RF instability in the regulator. If it oscillates the regulated voltage can go up. You may get away with fitting electrolytics at the moment, but some time in the future when they have dried out a bit and their ESR has gone up you will almost certainly have problems.

Hope that helps.

I think those are excellent points - it does sound like "motorboating", and I agree those rails are long and winding roads and make me worry. I have thought that a "star arrangement" would be better, but hard to lay out.

However, one idea would be to change where the speaker ground connects -- instead of the end of the long ground rail, why not directly to the battery ground point? The speaker is the main load that comes and goes and the main culprit for motorboating.

But --- when I disconnected the speaker, I still saw the oscillations, which goes against motorboating -- unless perhaps two things where happening and combining to give the motorboating. The circuit may be borderline stable, and the other oscillations triggered it? But my impression was that the oscillations were making the beeps and kept going when the speaker was detached.

I totally agree about the LM7808 - I have plans to use a low dropout reg in a future design - and maybe run off a lower voltage battery pack, such as 9 volts or so. In fact, rechargables have a very level discharge curve until they poop out, I'm wondering if a budget design with no regulator at all would work with rechargables.

Your point about the caps on the regulator is interesting too. I have wondered what is the best arrangement. At first I didn't understand why the big cap at the input of the regulator, but then I noticed that the speaker is connected directly to the battery. I thought maybe the big cap at the input was a holdover from designs where a rectifier circuit with ripple feeds into the regulator. So I have a big cap at input and output, but I agree you may need that smaller cap for the high freq feedback.

A while ago I experimented with bypass caps along the rails on a breadboard circuit of TGSL. I was surprised that they didn't seem to affect much, including noise spikes on the rails. But maybe they are important for motorboating.

I could use thicker buss wires for sure. I wondered about that, but thought the 30 milliamps wouldn't stress them. But that speaker pulse maybe does. As mentioned before, moving the speaker ground connection to the battery connection should be even better to remove any ground current along the buss. But would spreading the speaker wires apart make more emi and require other shielding? I hope not.

Ivconic did add a whole bunch of rail bypass caps, which were not on the original schematic, to his TGSL Final PCB, and maybe those are valuable. I wanted to see first how it would work without them.

The strange thing about the motorboating is that it sometimes comes and goes, so it seems to be a borderline condition. But these are all good ideas to work on it.

-SB

[simonbaker]

One thing I was thinking about these days is : why we make all our rx coil of 6.5 mH? I've made a 10" trying to achieve this value, but according to this table
http://www.geotech1.com/forums/showp...64&postcount=2 it appears that for a 11" DD, Tesoro's Rx coil value is 5.38 and it's indeed confirmed by the lower than tx coil resistance.
Well, 5.38//15 = 17.72 kHz.
Would the target behave the same in terms of phase and sensibility if I look for that frequency with a 6.5mH coil? It woud be as to have a 12.4nF capacitor...
Or is that table wrong not only in the Lobo data?

I don't think you will gain an advantage by raising the RX resonant freq from 16.1 kHz to 17.7 kHz. You will lose some gain. The phase probably won't change much because both 16.1 kHz and 17.7 kHz are fairly far from the oscillator freq of 14.5 kHz, the difference in phase shift should only be a few degrees.

My feeling is it is more interesting to try to lower the Rx resonant freq to increase the gain -- but then you do start to shift the phase of the target signal, according to my LTSpice simulations, and that can eventually mess up the disc, GB, and detector.

You can certainly try these ideas out. I spent quite a bit of time varying the oscillator frequency from below to above the Rx resonant frequency in small steps. Generally it seemed best near the designed specs, but you may make a small improvement varying the frequency, especially if your coils are not at spec.

I believe it may be possible to make a detector with Tx oscillator and Rx resonant freq closer together for better gain and signal-to-noise, but would require redesign of disc and GB circuits to provide appropriate phases. The design would be less stable, maybe require extra manual or automatic adjustments.

-SB

[simonbaker]

I had the same problem, but not always, it appeared from time to time.
I've tried shielding the circuit by covering it with a big stainless steel pot (pressure cooker ???) taken from the kitchen and connected to ground!
But it didn't fix the problem....

I know, it's not always there... once with previous pcb I just moved wires, shook things around, it went away. Some mystery there.

-SB

[Stefano]

I don't think you will gain an advantage by raising the RX resonant freq from 16.1 kHz to 17.7 kHz. You will lose some gain. The phase probably won't change much because both 16.1 kHz and 17.7 kHz are fairly far from the oscillator freq of 14.5 kHz, the difference in phase shift should only be a few degrees.
-SB

Ok, but what I meant was just to point out the strange low Rx coil value on that table.
By the way: if you need a 1 euro coin you can PM me your address: I'll be in Savannah GA next week and I can send you the coil from there if you want.

[simonbaker]

schematic, pcb files

Someone was interested in the ExpressPCB schematic and PCB files I used, so I will post them now. Hopefully there are not errors.

Screenshots for reference only, not clear enough for making boards.

Free schematic and PCB software can be obtained here: http://www.expresspcb.com/


-SB

[simonbaker]

Ok, but what I meant was just to point out the strange low Rx coil value on that table.
By the way: if you need a 1 euro coin you can PM me your address: I'll be in Savannah GA next week and I can send you the coil from there if you want.

Yes I agree, good question. I wonder how inductances, frequencies chosen. Also whether table of values is correct. I think some new Tesoros have really strange values which I don't understand. Which is best, and why? I would think you want similar values for coils, but some new designs have very different inductances. Maybe someone can explain the concept. It may depend on Q of coils somewhat.

I would like to try making Lobo one day since it is maybe the best Tesoro model.

Ok, let's try with Euro! I'll PM you.

-SB

[simonbaker]

vibrations

I reconnected my adjustable coil to check again the beeping problem.

I am fairly convinced that it is mechanical vibration, because I can make very large oscillations which almost die out if I wait. It is possible the building has small constant vibrations that never completely go away.

So I have nice seismometer which will tell me a few seconds before building falls down.

I guess my adjustable coil is not useful unless I find a solid platform for testing. I had better luck before I changed my workshop location to different room.

It is time to make another coil I think. My last glued coil is very unusual. I used thick wire so each coil is only about 6 ohms. Probably OK for TX coil (maybe very good), but it makes RX coil high Q and maybe picks up more noise at resonant frequency, and alters phase a little at the oscillator freq.

I also used Elmers glue, a common water soluable all-purpose white glue. When I first glued the coil, I noticed the coil really "sucked" after gluing -- it was dramatically worse in depth and response. I was very surprised. I never understood why. Was it moisture in the glue? I just don't feel that would be so important. You can put a chunk of metal on you coil and it will work pretty well, try it. Why would glue matter so much? Did the null change? My measurements didn't indicate significant change. Or was it coincidence, and I messed up the PCB with other tests?

However, it has had a year or two to dry, and now it may have gotten better. It works OK with my latest PCB, although still not great.

Anyway, more coils you have, the better for testing. Next RX coil I'll make with thinner wire to be more standard.

-SB