TESORO GOLDEN SABRE

[xix]

The problem with trying to display the target ID on a meter - especially with the TGSL, which is an analog detector - is that the phase-shift produced by the target is dependent on the distance from the coil. There is little point in providing a ferrous/non-ferrous indication, as the TGSL always eliminates ferrous targets when DISC is switched on. The main difficulty with making your own coils is adjusting them such that the DISC control provides minimum discrimination when turned fully left, and maximum discrimination when turned fully right. This is why the initial phase adjustment is so important. Otherwise the minimum setting could end up in the middle, for example.

I see the point...
By the way, I'm suffering from problems you mentioned on adjusting the coil. I'm looking for a minimum possible level at LF353 output. Would other methods be better for nulling coils?

[Qiaozhi]

I see the point...
By the way, I'm suffering from problems you mentioned on adjusting the coil. I'm looking for a minimum possible level at LF353 output. Would other methods be better for nulling coils?

There are two possible methods:

1. Initially adjust the coil overlap until the output of the pre-amp (LF353) is close to zero. Fix the main body of the coils at this point. Then do the final adjustment of the overlap to achieve a phase-shift of 200 degrees relative to the TX signal. You will need a 2-channel scope to do this.


2. Initially adjust the coil overlap, as before, and fix the main body of the coils. Then set the GEB control to mid-position, and monitor the output of the sample gate in the ground balance channel. This is located at the source pin of TR5, which is connected to C12. I am referring to the reference designators shown in the schematic of the gift pack. Adjust the final overlap to achieve as close to zero volts as possible at this point. Once this is setup correctly, the discrimination channel will work as expected, as it samples at 90 degrees to the GEB.

Method 2 is the best approach.

[simonbaker]

The problem with trying to display the target ID on a meter - especially with the TGSL, which is an analog detector - is that the phase-shift produced by the target is dependent on the distance from the coil. There is little point in providing a ferrous/non-ferrous indication, as the TGSL always eliminates ferrous targets when DISC is switched on. The main difficulty with making your own coils is adjusting them such that the DISC control provides minimum discrimination when turned fully left, and maximum discrimination when turned fully right. This is why the initial phase adjustment is so important. Otherwise the minimum setting could end up in the middle, for example.

I'm not completely clear on this, but I see it this way so far. Because TGSL is a "motion detector", the residual nulling signal creates a constant voltage that is removed by its bandpass filters. That's because when a moving target signal is detected, it is essentially added to the nulling signal and is really independent from it. I'll try to explain.

On an oscilloscope we may see a combined signal whose phase is dependent on the nulling residual and the target signal, but the TGSL does not really respond to the phase of the combined signal (null residue + target) because the bandpass filters remove the null residue part. So it really can measure the absolute phase of the target signal by itself. (Others may disagree with me but that's what the math seems to indicate).

In fact the TGSL detects the target signal twice -- by sending the received signal (null residue plus target) down two paths, in each path multiplying it by a square wave of the same frequency but carefully chosen phase. After multiplying, you get a "DC" voltage that tells you about the received signal. Because the target signal part is momentary (as we move the metal detector coil over the target), it really produces a low frequency pulse, say 1/8 to 1/2 second. That allows us to use a very low frequency bandpass filter to remove the constant voltage from the null residue signal and just let the target signal pulse through.

But back to the received signal where we multiply and make a DC voltage. On one path, the square wave is shifted in phase about 90 degrees from the other. The result is that the DC voltage that results will depend on the phase of the target signal as well as the amplitude.

If the target signal is in phase with one square wave, it will come through loud and clear, but the other channel will have zero response, and vice versa. If the target phase is between the two square waves, each channel will have a response. So you can basically compare the two responses and determine the phase of the target signal. As it turns out, different metals produce target signals of different phase, so determining phase is how we discriminate between them.

Practically, we don't actually compute phase. What we actually do is, with our descriminator knob, shift one of our square wave signals until all signals with less than a certain phase make a negative DC voltage, and all signals greater make a positive DC voltage. The other channel we leave alone. Then we feed both DC voltages to some comparators that provide a logical "AND" function, meaning they make a high signal only if both inputs are high. So only metals whose phase shift makes a positive DC voltage in both channels will make a high signal, which is used to trigger an audio tone.

Now back to the original question, displaying the phase on a meter. We have measured the target signal on two channels. But the complication is that the amplitude of the DC signal depends on the amplitude (depth) as well as phase of the original target signal from the RX coil. We want to ignore depth and just pay attention to phase. There is actually a mathematical way to compute phase of a signal if you know the two amplitudes of the DC signals from the two channels. To do an exact computation in analog electronics is too much bother for our nice simple TGSL circuit.

I was thinking maybe we could try just taking the ratio of the two DC voltages (V1 / V2) somehow. That's still pretty tricky in analog electronics probably. But the idea is that because the depth of the target would equally affect V1 and V2, but the phase would affect V1 and V2 differently, dividing them would remove the effect of depth and leave the effect of phase.

Well, that was a lot of wind, but that's the idea. Maybe there is a clever way to do it without using tons of circuitry. It's not important if the meter measures phase accurately, just so long as it shows different values for different phases, and is independent of depth. With the right circuitry, theoretically you could make it measure depth and ignore phase when you flip a switch. But don't expect much reliability -- targets in the real world are full of nasty treachery that can fool circuits a lot of the time.

Cheers,

-SB

[Aziz]

Hi SB,

..
I was thinking maybe we could try just taking the ratio of the two DC voltages (V1 / V2) somehow. That's still pretty tricky in analog electronics probably. But the idea is that because the depth of the target would equally affect V1 and V2, but the phase would affect V1 and V2 differently, dividing them would remove the effect of depth and leave the effect of phase.
..
-SB

Indeed... analog division and multiplier is really complex one:
lV1 = ln(V1) (logarithmic amplifier)
lV2 = ln(V2) (logarithmic amplifier)
for division: tmp = lV1 - lV2 (subtracter)
for multiplying: tmp = lV1 + lV2 (adder)
Result: exp(tmp) (exponential back transformation, anti-logarithmic amplifier)

All these stages would add much noise to the signal. It is a cost effective feature. Of course, you can take multiplier and divider chips. They need often at least 12..18 V bipolar supply voltage. And they are very expensive. Doing these steps in digital manner is much cheaper!

Aziz

[Qiaozhi]

I'm not completely clear on this, .....

Simply monitor the output signal from the pre-amp relative to the TX signal. You will need a 2-channel scope to do this. Trigger on the scope channel that is connected to the TX, and watch the RX signal shift as you bring a target close to the coil. It is then easy to see that the phase-shift of the target is dependent on depth.

[Geo]

Yes it was last year. I made 12 devices using "Goldy" pcb. All works until today very good.
I rearranged new pcb than to suit enclosure i got than - CZ5 enclosure. Than i got 1265/66 enclosure and draw another pcb, than i got Gold Bug 2 enclosure and draw new one....than i got MD3009 and 3009a and draw another pcb...huuuh!
All this vould be voided if i had same type of enclosure from a very start.
But i didnt, so ...
But initially, "Goldy" was the very FIRST one we (Nakky,ApBerg and me) did using TGSL schematic - TGS without notch part.

No chatters and hums (at least not significant) on Goldy IF coil is made and nulled well. So i can only recommend it also as others, to be used.
Regards!

Hello Ivica Do you know if Roberts will be at belligrad at 7....8/11. I want to meet him
Please contact with him and tell me.
Regards

[ivconic]

Simply monitor the output signal from the pre-amp relative to the TX signal. You will need a 2-channel scope to do this. Trigger on the scope channel that is connected to the TX, and watch the RX signal shift as you bring a target close to the coil. It is then easy to see that the phase-shift of the target is dependent on depth.

Indeed, and real goal would be to achieve less phase shifts, influenced by depth rises. Same old problem; some medium sized iron at greater depths will be detected as "coloured" by many detectors not only TGSL in the soil especially.
TGSL has weak TX. Is it possiblle to overcome phase issue with stronger output? At least to expand accuracy at a bit greater depths... I think it is possiblle. Stronger output will expand accuracy and phase would be slightly shifted at greather depths than it is now. But stronger TX will cause many other problems, stabillity issue at first place...
Would be good to analyze disc circuitry at White's Eagle, XLT,MXT,M6 and DFX serie. Those are accurate up to some point and VDI shows real numbers, but after some point those do not showing any number, just undefined audio on deep detections.... So problem is not solved at those models but also those do not "claim" deep items as "coloured" or iron. Just producing audio at detection of those deep items. Smart done.

[ivconic]

Hello Ivica Do you know if Roberts will be at belligrad at 7....8/11. I want to meet him
Please contact with him and tell me.
Regards

So as i heard from him, he is in Germany and will stay there till christmass.

[Qiaozhi]

Indeed, and real goal would be to achieve less phase shifts, influenced by depth rises. Same old problem; some medium sized iron at greater depths will be detected as "coloured" by many detectors not only TGSL in the soil especially.
TGSL has weak TX. Is it possiblle to overcome phase issue with stronger output? At least to expand accuracy at a bit greater depths... I think it is possiblle. Stronger output will expand accuracy and phase would be slightly shifted at greather depths than it is now. But stronger TX will cause many other problems, stabillity issue at first place...
Would be good to analyze disc circuitry at White's Eagle, XLT,MXT,M6 and DFX serie. Those are accurate up to some point and VDI shows real numbers, but after some point those do not showing any number, just undefined audio on deep detections.... So problem is not solved at those models but also those do not "claim" deep items as "coloured" or iron. Just producing audio at detection of those deep items. Smart done.

Unfortunately you are up against some basic physics here. The way the Whites (or any other digital detector) displays "accurate" VDI numbers, is by having a lookup table in ROM of various phase angles measured for specific targets at various depths. By monitoring the amplitude and cross-referencing this with the lookup table, it is possible to "guess" the identity of the target. This is only really accurate when the target is relatively close to the coil, or when performed in an air test. The ground causes lots of distortion in the RX signal, which makes the target id inaccurate.
There is no easy way around this problem with an analog detector.

[simonbaker]

Simply monitor the output signal from the pre-amp relative to the TX signal. You will need a 2-channel scope to do this. Trigger on the scope channel that is connected to the TX, and watch the RX signal shift as you bring a target close to the coil. It is then easy to see that the phase-shift of the target is dependent on depth.

Not really IMHO. The phase shift of the combined signal (null residue + target) is dependent on depth. That's because the null residue dominates for weak (deep) targets, and the target signal dominates for strong (shallow) targets.

But the absolute phase of the target part of the signal is not dependent on depth. The TGSL responds only to the target signal, so the design should discriminate independent of depth (theoretically).

The reason is the bandpass filters. They remove the effect of the null signal. Therefore, theoretically, the exact null value is not important for the TGSL design.

At least that is what the math indicates. I don't see any other explanation. The received signal Vrx = Vnull + Vtarget. The synchronous detector converts each part of the signal into a DC (or very low frequency) voltage. Because Vnull is a steady signal, it becomes a constant voltage which is blocked by the bandpass filter. The Vtarget becomes a very low frequency pulse, which passes through the bandpass filters.

The math indicates that Vnull does not affect Vtarget, they are independent. It doesn't matter what the combined signal Vrx looks like. Even if you can't see Vtarget, it is there. Vnull will not go through the bandpass filter, only Vtarget will. Even if Vtarget is very small, the discrimination should work the same.

If in reality the discrimination does change with depth, it is because of some reason we are not talking about. I would guess noise is blurring the phase of the target signal or some other effect. I don't believe it is directly due to the relative size of the null residue and the target signal. That would be true if zero crossings were use to detect phase, but the synchronous detector works differently.

There probably is a good reason to set the null voltage precisely, but the basic design and equations don't show it. The equations say don't worry too much about the null. However, because the LM358 U103 is directly coupled to the output of the synchronous detector, too large a null residue would saturate it or overload it -- that is a real reason to keep the null residue quite small. The amount of coupling with the TX coil also has complex effects because the TX coil is also a receiver for noise and target, and also can couple a load to the RX preamp input.

That's how I see it so far, I enjoy the discussion and ideas of everyone, just trying to understand it better and better.

Regards,

-SB

[Qiaozhi]

Not really IMHO. The phase shift of the combined signal (null residue + target) is dependent on depth. That's because the null residue dominates for weak (deep) targets, and the target signal dominates for strong (shallow) targets.

You need to do the actual experiment, and not rely just on the simulation.

[simonbaker]

You need to do the actual experiment, and not rely just on the simulation.

You're right!

-SB

[simonbaker]

Indeed, and real goal would be to achieve less phase shifts, influenced by depth rises. Same old problem; some medium sized iron at greater depths will be detected as "coloured" by many detectors not only TGSL in the soil especially.
TGSL has weak TX. Is it possiblle to overcome phase issue with stronger output? At least to expand accuracy at a bit greater depths... I think it is possiblle. Stronger output will expand accuracy and phase would be slightly shifted at greather depths than it is now. But stronger TX will cause many other problems, stabillity issue at first place...
Would be good to analyze disc circuitry at White's Eagle, XLT,MXT,M6 and DFX serie. Those are accurate up to some point and VDI shows real numbers, but after some point those do not showing any number, just undefined audio on deep detections.... So problem is not solved at those models but also those do not "claim" deep items as "coloured" or iron. Just producing audio at detection of those deep items. Smart done.

Did you try running oscillator straight from battery, any difference in depth and discrimination depth?

-SB

[ivconic]

Yes i was thinking to try like that. More power there would do some good i presume. I will try in next TGSL.

[Eugene52]

TESORO GOLDEN SABRE

I think that is a Great Idea , and also run audio on a separate supply voltage . Speaking of voltage , can anyone tell me why engineers / Technicians use the ICL7660 on circuit designs ?? I would like to see an "Equivelent Circuit" of voltage conversion....................Regards.......Eugene