TGSL Tuning & troubleshooting

[Vladimir]

1. Adjust coils to create some "null" signal. Which exact null signal we will discuss in a moment.-SB

MD sensitivity and zero signal at Tx are in inverse proportion?

[Qiaozhi]

Ok, I understand RX is the received signal. I think we have to be very specific here so I understand what you are saying.

I am assuming the following steps for setting up a coil and ground balancing a TGSL detector.

1. Adjust coils to create some "null" signal. Which exact null signal we will discuss in a moment.

2. Adjust GEB pot to ground-balance the detector using a ferrite target. The pot is turned until the ferrite target no longer makes a clear beep. I would assume the DISC pot is set for minimum discrimination during this procedure.

Now, I think you feel that the choice of the null signal (particularly the phase) in step 1 influences the position of the GEB pot setting to ground-balance the ferrite target in step 2. My feeling is that it does not.

Am I stating that correctly? If so, can you explain in detail why the null signal phase would affect the GEB pot setting? They seem independent to me, especially in terms of how the synchronous detector processes signals.

Regards,

-SB

This procedure is correct. At this point you should use the scope to see where the GEB sample pulse is relative to the RX signal (preamp output). It should be located on or close to the zero-crossing. Now, without moving the GEB control, slightly adjust the coil overlap. You will find that the sample pulse is no longer over the zero-crossing. This shows clearly that the null phase affects the GEB pot setting. If this was not the case, there would be no discussion concerning coil adjustment, as you would just go for the minimum voltage, and almost any coil from any detector would work with the TGSL, as long the residual voltage was low enough to prevent saturation of the preamp.

I have just found some notes I made in October 2007 about the Golden Sabre. Using a SPICE simulator I found that the GEB control allows for an adjustment of 45 degrees. Therefore the best position for the initial phase-shift should be 22.5 degrees, which places the GEB control in the middle.

[Vladimir]

MD sensitivity and zero signal at Tx are in inverse proportion?

Rx..........We apologize for the error

[dfbowers]

This procedure is correct. At this point you should use the scope to see where the GEB sample pulse is relative to the RX signal (preamp output). It should be located on or close to the zero-crossing. Now, without moving the GEB control, slightly adjust the coil overlap. You will find that the sample pulse is no longer over the zero-crossing. This shows clearly that the null phase affects the GEB pot setting. If this was not the case, there would be no discussion concerning coil adjustment, as you would just go for the minimum voltage, and almost any coil from any detector would work with the TGSL, as long the residual voltage was low enough to prevent saturation of the preamp.

I have just found some notes I made in October 2007 about the Golden Sabre. Using a SPICE simulator I found that the GEB control allows for an adjustment of 45 degrees. Therefore the best position for the initial phase-shift should be 22.5 degrees, which places the GEB control in the middle.

I learn something every day.. Maybe I have half a chance of making two coils where the GEB trimmer does not have to be readjusted when I swap coils!

Don

[simonbaker]

This procedure is correct. At this point you should use the scope to see where the GEB sample pulse is relative to the RX signal (preamp output). It should be located on or close to the zero-crossing.

No, it would not necessarily be near the RX signal zero crossing -- that would depend on what null phase you chose. Because the RX signal is the sum of the null signal and the target signal, and the null signal dominates.

Rather, it would be near the zero crossing of the target (ferrite) signal -- which you cannot see by itself. What you see is mainly the null signal. When you introduce the target, you will see a tiny movement of the RX signal, but the zero crossing of the RX signal is not the zero crossing of the ferrite target signal.

Now, without moving the GEB control, slightly adjust the coil overlap. You will find that the sample pulse is no longer over the zero-crossing. This shows clearly that the null phase affects the GEB pot setting.

No, the RX zero crossing has nothing to do with the GEB pot setting. The GEB pot setting is entirely dependent on the zero crossing of the ferrite target signal. You can shift that null signal all day long without changing the target signal component.

As far as I can tell, we have a linear system here, so:

Vrx = Vnull + Vtarget

where Vrx is the RX signal, Vnull is the RX "null" signal with no target present, and Vtarget is the additional signal created by a target.

The synchronous detector is a linear system. It processes each component of Vrx separately. The output of the synchronous detector is:

F(Vrx) = F(Vnull + Vtarget) = F(Vnull) + F(Vtarget)

F(Vnull) is always a constant voltage, which cannot go through the bandpass filter. If you change the phase of Vnull, you just get a different constant voltage.

F(Vtarget) is clearly independent of Vnull. The GB pot is turned until F(Vtarget) is approximately zero. You do this by waving the target in front of the coil, which creates a pulse which can go through the bandpass filter. Now you have ground balanced your detector.

If this was not the case, there would be no discussion concerning coil adjustment, as you would just go for the minimum voltage, and almost any coil from any detector would work with the TGSL, as long the residual voltage was low enough to prevent saturation of the preamp.

I think there are other considerations for choosing the null.

1. The null determines the steady-state voltage on the JFet source/drain. If it goes significantly negative, you forward bias the JFets.

2. The null can also be chosen for more coil overlap to get a better magnetic footprint.

3. The null can also be chosen so that fluctuations in the null cause opposite polarity pulses in the bandpass filter and avoid chatter.

Now, if I were making a coil for an existing Tesoro machine, I would try to achieve the same null phase as their coils because I don't know what else may depend on the null phase in their design. But for the TGSL, I am inclined to use the above criteria for choosing the null. There certainly may be important reasons I haven't considered. However, the explanations you give do not make sense to me, as I described above.

I'm convinced there is some basic assumption that you and I differ on which, if we could identify it, we could focus on that rather than repeating our statements which seem not to persuade each other.

I have just found some notes I made in October 2007 about the Golden Sabre. Using a SPICE simulator I found that the GEB control allows for an adjustment of 45 degrees. Therefore the best position for the initial phase-shift should be 22.5 degrees, which places the GEB control in the middle.

That's probably a perfectly nice place to put the null phase, but again, I have to disagree with the reasons. Putting the null at 22.5 degress will achieve a zero steady-state voltage on the JFet source/drain when the pot is in the middle and no target present. However, where the pot achieves true ground balance will depend on the phase of the ferrite target signal. And that will depend on other things, such as the RX circuit resonant frequency relative to the TX frequency, the properties of ferrite, etc.

Can you describe any reason that the null signal affects the ferrite signal? Does the superposition model I made above make any sense to you? If not, why not? I can be convinced, just need to understand what theory you are using.

Regards,

-SB

[Qiaozhi]

Firstly, I think Don has understood the problem. (see quote below)
Clearly he has discovered that coils nulled slightly differently will require the GEB trimmer to be adjusted, thus confirming my statement that the initial phase-shift is important, and showing that you cannot simply null your coils wherever you like. As I said earlier, the GEB sample pulse has an adjustment range of 45 degrees. Therefore you could (if you wish) null your coils with an initial phase-shift of say 10 degrees, or even 40 degrees, but this would limit the range to either end of the GEB trimmer. To give the most flexibility, an intial phase-shift of 20 degrees (ideally 22.5 degrees) is preferred.

I learn something every day.. Maybe I have half a chance of making two coils where the GEB trimmer does not have to be readjusted when I swap coils!

Don

Now onto Simon ...

No, it would not necessarily be near the RX signal zero crossing -- that would depend on what null phase you chose. Because the RX signal is the sum of the null signal and the target signal, and the null signal dominates.

Rather, it would be near the zero crossing of the target (ferrite) signal -- which you cannot see by itself. What you see is mainly the null signal. When you introduce the target, you will see a tiny movement of the RX signal, but the zero crossing of the RX signal is not the zero crossing of the ferrite target signal.

Having read the above, I think there must be some sort of terminology problem here. When I refer to the RX signal, I am talking about the signal from the receive coil. This could be measured directly across the coil or at the preamp output. It may, or may not, contain a signal from a target. In fact, it may just be the residual signal being picked up from the TX coil. In all cases it is called the RX signal. The RX signal can change both in amplitude and/or phase depending on what type of target is close to the coil.

No, the RX zero crossing has nothing to do with the GEB pot setting. The GEB pot setting is entirely dependent on the zero crossing of the ferrite target signal. You can shift that null signal all day long without changing the target signal component.

When there are no targets near the coil there is an initial phase-shift in the RX signal relative to the TX signal. Different targets produce different phase-shifts that move the zero-crossing point. To achieve ground balance you should first place the GEB sample pulse over the zero-crossing, then make final adjustments using a ferrite slug.

As far as I can tell, we have a linear system here, so:

Vrx = Vnull + Vtarget

where Vrx is the RX signal, Vnull is the RX "null" signal with no target present, and Vtarget is the additional signal created by a target.

The synchronous detector is a linear system. It processes each component of Vrx separately. The output of the synchronous detector is:

F(Vrx) = F(Vnull + Vtarget) = F(Vnull) + F(Vtarget)

F(Vnull) is always a constant voltage, which cannot go through the bandpass filter. If you change the phase of Vnull, you just get a different constant voltage.

F(Vtarget) is clearly independent of Vnull. The GB pot is turned until F(Vtarget) is approximately zero. You do this by waving the target in front of the coil, which creates a pulse which can go through the bandpass filter. Now you have ground balanced your detector.




I think there are other considerations for choosing the null.

1. The null determines the steady-state voltage on the JFet source/drain. If it goes significantly negative, you forward bias the JFets.

2. The null can also be chosen for more coil overlap to get a better magnetic footprint.

3. The null can also be chosen so that fluctuations in the null cause opposite polarity pulses in the bandpass filter and avoid chatter.

Now, if I were making a coil for an existing Tesoro machine, I would try to achieve the same null phase as their coils because I don't know what else may depend on the null phase in their design. But for the TGSL, I am inclined to use the above criteria for choosing the null. There certainly may be important reasons I haven't considered. However, the explanations you give do not make sense to me, as I described above.

I'm convinced there is some basic assumption that you and I differ on which, if we could identify it, we could focus on that rather than repeating our statements which seem not to persuade each other.

That's probably a perfectly nice place to put the null phase, but again, I have to disagree with the reasons. Putting the null at 22.5 degress will achieve a zero steady-state voltage on the JFet source/drain when the pot is in the middle and no target present. However, where the pot achieves true ground balance will depend on the phase of the ferrite target signal. And that will depend on other things, such as the RX circuit resonant frequency relative to the TX frequency, the properties of ferrite, etc.

Can you describe any reason that the null signal affects the ferrite signal? Does the superposition model I made above make any sense to you? If not, why not? I can be convinced, just need to understand what theory you are using.

Regards,

-SB

Now I'm not sure what you're not understanding about balancing the coil.
As I said before, it is simply a case of adjusting the coil overlap such that the zero-crossing of the RX signal has an initial phase-shift somewhere near the mid-range of the GEB trimmer. For the TGSL this phase-shift is approximately 20 degrees.
If you don't believe me, try setting the initial phase-shift to 60 degrees, and then attempt to ground balance. It will be impossible to achieve because the zero-crossing of the RX signal will be out of range of the GEB trimmer.

[Stefano]

As I said before, it is simply a case of adjusting the coil overlap such that the zero-crossing of the RX signal has an initial phase-shift somewhere near the mid-range of the GEB trimmer. For the TGSL this phase-shift is approximately 20 degrees.

Hi Quiaozhi,
from the tests I've done this afternoon it's not, and I've also tried with 14.8 and 10 kHz getting the same results:

If I tune the Rx-LC freq. 1.6-1-8 kHx higher than Tx I got always a signal which has a negative phase-shift. To get a positive phase-shift I would need to tune it for a frequency lower than Tx.
If I move over the null point to seek for the right phase than the signal amplitude becomes much too high and risk is to overload.
So I went for the negative phase. The GEB sampling wasn't of course at the zero crossing, nevertheless I could correctly reject the ferrite with the GEB trimmer at mid position.
Then I've tried to modify the overlap (so increasing the null signal): no changes. Ferrite was always correctly rejected at middle pos. , the discrimination worked good and the nickel coin disappeared approximately with the DISC potent. at 50%

Then I've changed the Rx tuning freq: the ferrite wasn't anymore rejected and the GEB trimmer needed a re-adjustment.
I went on changing the tuning freq. : sensibility seemed to improve but the ferrite couldn't be rejected anymore in any position of the trimmer and target response changed as well, giving problems in discrimination.
Chattering also appeared.

So it seems to me that the target phase response changes with tx/rx tuning freq. difference and this circuit had been projected for target phase behaviour as per 1.6-2 kHz of difference. I feel like there's room for improvement in this direction (I mean having closer frequencies) but something in the circuit should be changed.

Regards

[simonbaker]

Hi Quiaozhi,
from the tests I've done this afternoon it's not, and I've also tried with 14.8 and 10 kHz getting the same results:

If I tune the Rx-LC freq. 1.6-1-8 kHx higher than Tx I got always a signal which has a negative phase-shift. To get a positive phase-shift I would need to tune it for a frequency lower than Tx.
If I move over the null point to seek for the right phase than the signal amplitude becomes much too high and risk is to overload.
So I went for the negative phase. The GEB sampling wasn't of course at the zero crossing, nevertheless I could correctly reject the ferrite with the GEB trimmer at mid position.
Then I've tried to modify the overlap (so increasing the null signal): no changes. Ferrite was always correctly rejected at middle pos. , the discrimination worked good and the nickel coin disappeared approximately with the DISC potent. at 50%

Then I've changed the Rx tuning freq: the ferrite wasn't anymore rejected and the GEB trimmer needed a re-adjustment.
I went on changing the tuning freq. : sensibility seemed to improve but the ferrite couldn't be rejected anymore in any position of the trimmer and target response changed as well, giving problems in discrimination.
Chattering also appeared.

So it seems to me that the target phase response changes with tx/rx tuning freq. difference and this circuit had been projected for target phase behaviour as per 1.6-2 kHz of difference. I feel like there's room for improvement in this direction (I mean having closer frequencies) but something in the circuit should be changed.

Regards

Hi Stefano:

I'm not sure what your exact question is. Is it about choosing a null phase, or trying some improvements like moving Tx freq and RX resonant freq closer?

As you have noticed, my opinion is that null phase does not significantly affect where you set the GB pot or the Disc pot for targets. So choose null phase for good overlap of coils and try to keep voltage on C12, C15 above -.5 volts.

-----------
Checklist for coil:

Are your coils oriented correctly? Do you get a single or double beep with a target about 8 cm distance?

What diameter wire are you using? I have experimented with thicker wire and it can make trouble with the unmodified TGSL.

When you play with nulling your coils, can you get positive voltages on C15 and C12 with DISC and GB pots set to minimum? Can you change the null and get negative voltages?

Shields - I think the way we ground and wire our shields can have an impact on target phases. Have you doublechecked that your shield wiring and grounding are per standard advice?

----------

Moving the RX resonant frequency around can definitely affect target phase and therefore the GB pot setting also, especially if you move it closer to TX frequency than the standard. mikebg and others' graphs show the idea.

In fact, it's not a good idea to move RX res freq much closer to TX freq - there is a point where it suddenly will throw off your detector.

However, I think there may be some interesting designs where you can move them closer, but it would probably be much harder to stabilize the design. I plan to try some anyway.

Cheers,

-SB

[Stefano]

Hi Stefano:
I'm not sure what your exact question is.

It's not a question, it 's just a report of what I did and what I noticed seems to conifirm your opinion about nulling.

For your information: I used yesterday an 8" original Tesoro DD coil where I change the null moving a small loop of few turns in series with the Rx coil.

Moving the RX resonant frequency around can definitely affect target phase and therefore the GB pot setting also, especially if you move it closer to TX frequency than the standard. mikebg and others' graphs show the idea.

In fact, it's not a good idea to move RX res freq much closer to TX freq - there is a point where it suddenly will throw off your detector.

Stefano

[Qiaozhi]

Hi Stefano,

When you set the coil overlap to give an initial phase-shift of around 20 degrees (using the coil inductances and cap values shown in the schematic) what is the residual voltage on the RX coil? Also, if it has not overloaded the preamp, what voltage is at the output?

[Stefano]

Hi Stefano,

When you set the coil overlap to give an initial phase-shift of around 20 degrees (using the coil inductances and cap values shown in the schematic) what is the residual voltage on the RX coil? Also, if it has not overloaded the preamp, what voltage is at the output?

With the standard 15nF cap @ 14.5 kHz the +20° is almost impossibile to achieve for me. I'll make a video of what happen at the waveform on the oscilloscope in at least 8-10 days. Right now I'm about to leave for 1 week of holiday.

These days I've seen a wonderful brainstorming and prolific activity on the subject here. I hope to find more as soon as I'll get back.

Nice summer to all of you!

Stefano

[simonbaker]

For your information: I used yesterday an 8" original Tesoro DD coil where I change the null moving a small loop of few turns in series with the Rx coil.
Stefano

Interesting technique. Hopefully not such a big loop to add much to the overall inductance.

Keep us informed of your tuning efforts so we can learn and/or help.

Regards,

-SB

P.S. have a great holiday!

[Qiaozhi]

More fuel for the fire!

I couldn't find the SPICE simulation mentioned earlier, so I hacked together a simulation for LTSPICE:

* GEB Sample Pulse Simulation
C10 TX_DELAY TX 100p
R22 TX_DELAY 0 100k
V1 P8V 0 8
V2 N5V 0 -5
V3 TX 0 SINE(0 6 14.5k)
XU1 0 TX_DELAY P8V N5V SAMPLE1 LM393
RPULLUP SAMPLE1 P8V 3k
.tran 0 200u 0 200n
.lib LTC1.lib
.backanno

* LM393 VOLTAGE COMPARATOR "MACROMODEL" SUBCIRCUIT
* CONNECTIONS:
* NON-INVERTING INPUT
* INVERTING INPUT
* POSITIVE POWER SUPPLY
* NEGATIVE POWER SUPPLY
* OPEN COLLECTOR OUTPUT
.SUBCKT LM393 1 2 3 4 5
*
F1 9 3 V1 1
IEE 3 7 DC 100.0E-6
VI1 21 1 DC .75
VI2 22 2 DC .75
Q1 9 21 7 QIN
Q2 8 22 7 QIN
Q3 9 8 4 QMO
Q4 8 8 4 QMI
.MODEL QIN PNP(IS=800.0E-18 BF=2.000E3)
.MODEL QMI NPN(IS=800.0E-18 BF=1002)
.MODEL QMO NPN(IS=800.0E-18 BF=1000 CJC=1E-15 TR=807.4E-9)
E1 10 4 9 4 1
V1 10 11 DC 0
Q5 5 11 4 QOC
.MODEL QOC NPN(IS=800.0E-18 BF=20.29E3 CJC=1E-15 TF=942.6E-12 TR=543.8E-9)
DP 4 3 DX
RP 3 4 46.3E3
.MODEL DX D IS=800.0E-18
*
.ENDS
.end

This represents the GEB sample pulse circuit (C10, R22 and U102b). Run the simulation and view TX, TX_DELAY and SAMPLE1. Measure the time difference between the negative-going zero-crossing of TX and the middle of the next GEB sample pulse on SAMPLE1. This will be found to be approximately 8.6us. Since the TX frequency is 14.5kHz, then 8.6us is equivalent to 45 degrees. Reducing R22 to 100 ohms results in a value of about 5us. Next, setting R22 to 50k (equivalent to putting the GEB trimmer in its mid-position) gives a delay of 24 degrees. Not quite the 22.5 degrees I said before, but close. As you can readily see, in order to allow the TGSL to ground balance with the control in mid-position, you need an initial phase-shift of around 20 degrees. This is of course for a DD coil. For a concentric coil the RX signal is inverted.

Note there is one extra resistor in the netlist called RPULLUP. This is because the LM393 has an open-collector output. In the TGSL circuit this resistor is missing, as it is not really necessary for the circuit to work, but in the simulation we cannot see the sample pulse without it. The TGSL uses n-channel JFETs for the sync demod, and these need a negative voltage on the gate relative to the source to pinch off the channel. In other words, for the device to be turned on (and sample the RX signal) it just needs to go non-negative.

[simonbaker]

I couldn't find the SPICE simulation mentioned earlier, so I hacked together a simulation for LTSPICE:

* GEB Sample Pulse Simulation
C10 TX_DELAY TX 100p
R22 TX_DELAY 0 100k
V1 P8V 0 8
V2 N5V 0 -5
V3 TX 0 SINE(0 6 14.5k)
XU1 0 TX_DELAY P8V N5V SAMPLE1 LM393
RPULLUP SAMPLE1 P8V 3k
.tran 0 200u 0 200n
.lib LTC1.lib
.backanno

* LM393 VOLTAGE COMPARATOR "MACROMODEL" SUBCIRCUIT
* CONNECTIONS:
* NON-INVERTING INPUT
* INVERTING INPUT
* POSITIVE POWER SUPPLY
* NEGATIVE POWER SUPPLY
* OPEN COLLECTOR OUTPUT
.SUBCKT LM393 1 2 3 4 5
*
F1 9 3 V1 1
IEE 3 7 DC 100.0E-6
VI1 21 1 DC .75
VI2 22 2 DC .75
Q1 9 21 7 QIN
Q2 8 22 7 QIN
Q3 9 8 4 QMO
Q4 8 8 4 QMI
.MODEL QIN PNP(IS=800.0E-18 BF=2.000E3)
.MODEL QMI NPN(IS=800.0E-18 BF=1002)
.MODEL QMO NPN(IS=800.0E-18 BF=1000 CJC=1E-15 TR=807.4E-9)
E1 10 4 9 4 1
V1 10 11 DC 0
Q5 5 11 4 QOC
.MODEL QOC NPN(IS=800.0E-18 BF=20.29E3 CJC=1E-15 TF=942.6E-12 TR=543.8E-9)
DP 4 3 DX
RP 3 4 46.3E3
.MODEL DX D IS=800.0E-18
*
.ENDS
.end

This represents the GEB sample pulse circuit (C10, R22 and U102b). Run the simulation and view TX, TX_DELAY and SAMPLE1. Measure the time difference between the negative-going zero-crossing of TX and the middle of the next GEB sample pulse on SAMPLE1. This will be found to be approximately 8.6us. Since the TX frequency is 14.5kHz, then 8.6us is equivalent to 45 degrees. Reducing R22 to 100 ohms results in a value of about 5us. Next, setting R22 to 50k (equivalent to putting the GEB trimmer in its mid-position) gives a delay of 24 degrees. Not quite the 22.5 degrees I said before, but close. As you can readily see, in order to allow the TGSL to ground balance with the control in mid-position, you need an initial phase-shift of around 20 degrees. This is of course for a DD coil. For a concentric coil the RX signal is inverted.

Note there is one extra resistor in the netlist called RPULLUP. This is because the LM393 has an open-collector output. In the TGSL circuit this resistor is missing, as it is not really necessary for the circuit to work, but in the simulation we cannot see the sample pulse without it. The TGSL uses n-channel JFETs for the sync demod, and these need a negative voltage on the gate relative to the source to pinch off the channel. In other words, for the device to be turned on (and sample the RX signal) it just needs to go non-negative.

Hi Qiaozhi:

Can you attach the actual LTSpice file we can run?

Regards,

-SB

[Qiaozhi]

Hi Qiaozhi:

Can you attach the actual LTSpice file we can run?

Regards,

-SB

File attached. There is no schematic, as it was quicker to just write the netlist by hand. Open the GEB.cir file in LTSPICE and run the simulation as normal. Right mouse click in netlist and select "Visible Traces" to see TX, TX_DELAY and SAMPLE1 in the plot pane.