TGSL Tuning & troubleshooting

[Mullihaka]

Epoxy & phase shift

I took snapshots of my coil nulling parameters before and after pouring the epoxy into the shell.

As so many others on this forum, I too had problems to achieve minimum null or +20 deg phase shift in the same coil positions. I made a compromise between them and ended up in a situation as in the first picture. The curves a as follows:

1 (green) TX; pin 1 of J1
2 (blue) RX; pin 7 of LF353 (U101b)
3 (red) GB, gate of TR5

I thought, that the GB signal could sample at proper part of the negative going RX signal though it is not exactly at the middle. After epoxy had cured, the GB signal was barely up at the negative going slope of the RX signal (2nd picture).

How does the real ground effect the relationship between RX and GB; does it “drop the RX out of the GB signal” or does it lower the phase shift?

Everyone likes pictures, so here’s some of my latest version of coil set.

Can someone please tell me how the sow the pictures between paragraphs; not all as one heap at the end of a message?

[johnsmith77]

[SIZE=3]

As so many others on this forum, I too had problems to achieve minimum null or +20 deg phase shift in the same coil positions. I made a compromise between them and ended up in a situation as in the first picture. The curves a as follows:

Before epoxy, you had phase shift ~ +60. I have ~ -60.
1) Do you wind coils in same direction or different?

Made some test with ferrite on Tx. It makes oscillator work on lower frequency because ferrite rises inductivity on Tx coil.

We always want achieve shift 20. Does it should be measured directly between tx vs rx or tx and lf353 pin 7 - thats the question
LF353 (Tl082) should influence shift also. Also TL082 or LF353 should change phase differently. They are not identic 100%

The most important thing is (by my opinion) phase shift between tr4, tr5, (drain and gate). It works like mixer
From where comes theory 20, none can say? I think than man which made successful project had phase shift 20 Tx, Rx. But what phase shift was tr4, tr5 (drain vs gate).
There comes Tx vs Rx signals. But they have not original Tx, Rx phase shift they are influenced by all circuits they passed. These circuit are indentical only ir schematics. In real conditions not, because components acurracy. R, C +- even 20%? So phase shift +-20%? with identicals circuits? There is the reason why each desing works differently?


I think you can compensate GEB shift by trim easily if you had shift like in the first picture, not sure about second one.. Did you measured on middle GEB trim position? This also influence TR gate GEB shift

Your phase shifts:
1pic) Tx/rx = 64 ______ +22 _________ = ____ 2pic) Tx/rx = 86 ____________________(When you applied epoxy Difference= 22)
1pic) Rx/geb = 53 ______ -22__________= ____ 2pic) Rx/geb = 31 ___________________(When you applied epoxy Difference= 22)


So if you had Tx/rx = 20 (64-44)
Rx/geb would be = 97 (53+44)

So when Tx/rx shift would be 20, rx/geb shift 97.
97 is not the middle. Middle is 90. So there is no chance achieve 20 tx/rx and rx/geb = 90 at the same time. Then question which of these shifts is more important?

[Mullihaka]

I don't care about coil winding direction. I just swap the RX wires around if it gives double beeps.

I think the 20 deg should be measured between TX and pin 7 / LF353. That's where the sampling (by the FETs) happens.

I feel, that the 20 degrees comes from the fact, that the GB adjustment circuit (U102b, C10 and R22) makes the sample to delay that 20 degrees. We would need Simonbaker to comment it, as he has (to my understanding) simulated most of the machine, piece by piece.

I tried my machine ouside, far away form anything electric. The detection distance was exactly the same as indoors.

Next I will make another coil set and measure it's resonance when attached to the detector. I have an audio analyzer with which I can feed random noise to the TX coil and monitor the output of RX coil at the same time. The max hold function of the analyzer should reveal the resonance.

I will set the resonances "by the book i.e. 14.5/16.1 kHz" and null the coils to absolute minimum. What then, I don't know yet.

The phase measurement between the RX and GB singal is there just for fun, I did not use it for anything.

[johnsmith77]

The phase measurement between the RX and GB singal is there just for fun, I did not use it for anything.

If it would be fun i would not be able correctly null my coils in 2 seconds with scope while monitoring these signals. And getting correct disc, ferr reject and so on

I will set the resonances "by the book i.e. 14.5/16.1 kHz" and null the coils to absolute minimum. What then, I don't know yet.

What RX L inductance and Rx C do you have?
What 1e coin distance achieved?

[simonbaker]

Made some test with ferrite on Tx. It makes oscillator work on lower frequency because ferrite rises inductivity on Tx coil.

That is an interesting point, depending on how much high-ferrite ground could affect the TX frequency. However, I assume ferrite will always reduce the TX frequency, which might reduce the gain/sensitivity of the TGSL by moving away from the RX resonant freq, but wouldn't cause a problem with phase shifts.

We always want achieve shift 20. Does it should be measured directly between tx vs rx or tx and lf353 pin 7 - thats the question

That may be desired for certain TGSL constructed a particular way. I personally suspect however it is not necessary in general, and some MDs will naturally need a different phase shift. I'm working on finding a way to explain why and how to predict the correct phase shift for each TGLS.

LF353 (Tl082) should influence shift also. Also TL082 or LF353 should change phase differently. They are not identic 100%

I would be interested in seeing an actual comparison test with real chips and oscilloscope pics.

The most important thing is (by my opinion) phase shift between tr4, tr5, (drain and gate). It works like mixer
From where comes theory 20, none can say? I think than man which made successful project had phase shift 20 Tx, Rx. But what phase shift was tr4, tr5 (drain vs gate).
There comes Tx vs Rx signals. But they have not original Tx, Rx phase shift they are influenced by all circuits they passed. These circuit are indentical only ir schematics. In real conditions not, because components acurracy. R, C +- even 20%? So phase shift +-20%? with identicals circuits? There is the reason why each desing works differently?

Don't forget -- the null signal phase (20 deg or whatever) has nothing to do with the target signal phase. The phase between the drain and gate of TR4, TR5 that you observe is only for the null signal, not the target signal. The null signal only makes a DC voltage at the source of TR4, TR5 -- which is ignored by the next op amp filter!

So the null signal phase is not really important in the way you think (I think).

What matters is the relationship between the sync pulse phase (gate of TR4, TR5) and the target signal, which you cannot see normally, it is so small. You can change the null signal phase all you want and it does not change the phase of target signals. So don't worry too much about the null signal phase as a way to null your coils -- use other criteria. Except this warning -- make sure the DC voltage on the source of TR4, TR5 does not go too far less than ground, or you will forward bias the JFET and make a mess. In other words, shift your coils to pick a different null signal phase.

I have not completely made my mind up on the effect of the null signal phase on ground balancing. Qiaozhi and I had a long confused discussion on this and I accepted the idea that ground/ferrite would amplitude-modulate the null signal without changing its phase. Since then I have been considering that this is only partially true -- that the null signal may be composed of a "magnetic" component and also a "non-magnetic" component, in which case ground/ferrite would only modulate the "magnetic" component. I also have been thinking that ferrite truly may be a "zero phase" target, which really is different from simply modulating any null phase.

Sorry, getting too complicated. I need to do much more thinking about it. But my final advice is don't get too concerned about your null signal phase for the purpose of nulling your coils. And don't worry that the wrong null signal phase will hurt your DISC control -- it should not. Do make sure your JFet gate-source voltage is not forward biasing the JFet junction.

-SB

[simonbaker]

I feel, that the 20 degrees comes from the fact, that the GB adjustment circuit (U102b, C10 and R22) makes the sample to delay that 20 degrees. We would need Simonbaker to comment it, as he has (to my understanding) simulated most of the machine, piece by piece.

I think that 20 degrees cannot be easily calculated from my simulations. I don't remember exactly, but my basic simulations model TX, RX coils as a transformer and actually predict about 5 degrees maybe, and the actual 20 degrees (or what ever you get for your MD) is mostly determined by some effect of the shield and some effect of capacitance perhaps from the cable or other. There is also the question of how you ground your RX coil (if at all) and the effect on the null signal phase, and I need to continue that LTSpice analysis.

It will be interesting to try to include the shield and capacitance and maybe other effects to get a better prediction. I'll keep trying.

-SB

[johnsmith77]

Don't forget -- the null signal phase (20 deg or whatever) has nothing to do with the target signal phase. The phase between the drain and gate of TR4, TR5 that you observe is only for the null signal, not the target signal. The null signal only makes a DC voltage at the source of TR4, TR5 -- which is ignored by the next op amp filter!

So the null signal phase is not really important in the way you think (I think).

You can change the null signal phase all you want and it does not change the phase of target signals. So don't worry too much about the null signal phase as a way to null your coils -- use other criteria. Except this warning -- make sure the DC voltage on the source of TR4, TR5 does not go too far less than ground, or you will forward bias the JFET and make a mess. In other words, shift your coils to pick a different null signal phase.

-SB

I think a little different. I try give you an example and explain how i imagine this when no target applied.

On Tr gates comes signal 14.5khz from Tx coil.
On Tr drain comes signal 14.5khz received by Rx coil, shifted and modulated by Tx signal from Tr gate. Because of it we don't see nice sinus on Tr drains as wee see on LF353 pin 7 - it is modulated here. Depending on phase shift between Rx/Tx on Tr drain signal will be positive or negative, also it influence Tr source voltages.

When disc pot are turned, shift also changes, this changes voltage on Tr 4 source.
If we have wrong phase shift at min disc pot position, then we will no be able achieve correct disc at all pot positions.
As you know disc pot changes cut off point for target metal.

Example again (if we have wrong phase shift tx/rx) then imagine:

disc pot min (phase shift between Rx/Tx disc sampling pulse -40) then voltage on Tr s = -V.
disc pot middle (phase shift between Rx/tx disc sampling pulse 0) then voltage on Tr s = --V (more negative).
disc pot middle (phase shift between Rx/tx disc sampling pulse 40) then voltage on Tr s = -V.

Tr s voltage when disc pot min and max will be equall so disc at min and max will act the same. Only from Min to middle and from middle to max different.
Disc can't act the same on min and max. Also this applies to geb.

There is how i think

Also depending on this, some TGSL made projects has iron rejection at min pot, some not.
Mine has iron rejection when disc pot turned at 1.
Depending on phase shift i can make that iron will be rejected when disc pot turned at 0 or at 3.

[johnsmith77]

The phase between the drain and gate of TR4, TR5 that you observe is only for the null signal, not the target signal. The null signal only makes a DC voltage at the source of TR4, TR5 -- which is ignored by the next op amp filter!


-SB

Yes, it can be. U voltage on Tr sources ignored.
But signal from target - delta U, not ignored.
And delta U- signal depends on phase shift between Rx/ disc sampling pulse.
It can go to +V or -V. It is my theory

[Mullihaka]

7cm more depth

I measured the voltage at pins 2/6 of 106. Riding on the DC voltage there was 40mV ripple at the frequency of the TX. I soldered 1uF, 470nF and 100nF capacitors just across the R38. That reduced the ripple to 20mV. Additionally, I soldered 1uF capacitor across the C5 and the ripple was reduced to 10mV.

With these changes, the detection distance of 2 euro coin increased from 15cm to 22cm in the air.

I'm now trying to find the reel of 220nF SMD capacitors I bought a year ago. I'm going to bypass some other stages as well. Maybe the collector of TR7 to the ground and pins 4/7 of U102 to the ground too.

Every stage that has big swing of the voltage is likely to draw significant current from the plus and minus rails. That, in turn, generates the ripple. Bypassing every oprational voltage line and voltage reference points would not do any harm in here. The TX frequency is not in sync with the detected target, therefore ripple can mask weak signals or, as in this case, raise the detection threshold.

All improvement ideas more than welcome!

[johnsmith77]

I soldered 1uF, 470nF and 100nF capacitors just across the R38. That reduced the ripple to 20mV. Additionally, I soldered 1uF capacitor across the C5 and the ripple was reduced to 10mV.

With these changes, the detection distance of 2 euro coin increased from 15cm to 22cm in the air.

All improvement ideas more than welcome!

Can try dfbowers mod for sensitivity, r37, maybye you will achieve 35cm then

[Mullihaka]

Can try dfbowers mod for sensitivity, r37, maybye you will achieve 35cm then

I have that already but it didn't help. I think that this mod does not help as long as the spikes all around mask the weak detected signals.

I soldered some 330nF SMD caps in several places around the PCB. Now I get occasionally 28cm, but the tone is short and faint.

I still haven't got rid of the spikes all around the voltage rails. I think it's even not possible with a design like this with chopping negative voltage generator.

I guess I will next disconnect TR7 and TR8 and supply the negative voltage from an additional battery to see, if this makes the detection better.

I might make another PCB, where the upper side is solid copper (apart from openings for component leads). That way I could solder the SMD capacitors on the upper side of the PCB from any component lead to solid ground. That is not possible with the currect design, as there's not enough ground points available on solder side.

Does anyone have a one side PCB design with ground "everywhere"?

[Jerry]

I have that already but it didn't help. I think that this mod does not help as long as the spikes all around mask the weak detected signals.

I soldered some 330nF SMD caps in several places around the PCB. Now I get occasionally 28cm, but the tone is short and faint.

I still haven't got rid of the spikes all around the voltage rails. I think it's even not possible with a design like this with chopping negative voltage generator.

I guess I will next disconnect TR7 and TR8 and supply the negative voltage from an additional battery to see, if this makes the detection better.

Another thing you could try is shorting out the input terminals from the rx coil to see how much of that noise goes away. In other words, is it internal or external to the board?

Let us know if the battery makes any difference.

Jerry

[Mullihaka]

The interference is definitely internal; it's the TR7/TR8, which "modulate" both positive and negative voltage rails with the TX frequency. The rise time of the pulse is short, approximately 150ns. I'm wondering, if TR7/TR8 could be changed to ones with lower frequency range, just to lower the ripple.

[Jerry]

The interference is definitely internal; it's the TR7/TR8, which "modulate" both positive and negative voltage rails with the TX frequency. The rise time of the pulse is short, approximately 150ns. I'm wondering, if TR7/TR8 could be changed to ones with lower frequency range, just to lower the ripple.

I remember some conversation about this a long time ago. As I recall the scheme is set up so the transmit frequency is divided down to perform other functions so that any transients that are generated would be in sync with the transmit freq and not cause problems. Do not remember much in the way of details but that the gist of it.

I am sure others can fill in the blanks.

Jerry

[johnsmith77]

I guess I will next disconnect TR7 and TR8 and supply the negative voltage from an additional battery to see, if this makes the detection better.

Do we have test results?