TGSL Experiments

[simonbaker]

Right now i am reconsidering option to put back on pcb that fet and the rest of components... Simonbaker, if i do so, than it would be your "guilt"!

That might be too much for me...

I am greedy for depth, and removing JFet makes maximum TX signal -- that's good!

So I am trying other small mods so "wild" TX signal does not disturb the discriminator so much -- like a voltage divider, but not so easy I have found. Any little phase shift is important it seems. Sometimes I make a small mod that I think should be excellent and... terrible. TGSL scolds me, it wants to be just how Ivconic makes it.

If you do put back JFet, I think we need to add an adjustment pot somewhere so we can get max TX signal without distortion. Not recommended for commercial detectors because adjustment might slip, but for hobbyists who can tune, an advantage I think.

For now I'm trying many things and trying to understand what makes a "sweet" TGSL and why others no good when everything seems fine on the oscilloscope. Some subtle thing in there. I hope to learn before I'm crazy(er).

Quem deus vult perdere, dementat prius.

-SB

[Fred]

I remember that the zener value played an important role in frequency, also somewhat in amplitude.

[ivconic]

I remember that the zener value played an important role in frequency, also somewhat in amplitude.

"..Zener diode provides the voltage reference for the peak sine wave amplitude; this is rectified and fed to the gate of the fet, thus varying its channel resistance and hence, loop gain."

No really, i remember, in the past when experimented much, i tried there various zeners; 3V9,4V3,4V7 and 5V1. I dont remember that i saw any differences than. But at the time i was without scope so...maybe there were some differences and i couldn't notice those.
In time i decided to exclude that part from pcb, initially to save some space on it, or to make smaller pcb..
Now, understanding better its role there, i am tempted to return it to pcb and to do a bit more experimenting. But...time, i dont have time! I have 6-7 various projects waiting on my table to be finished! Ooh! If i am strong enough not to sleep!
I think winter is the best time for such thing.
Regards!

[simonbaker]

Hello,
Its look like I used wrong wire.I bought 0,25 bare wire.And because its from specialized shop I tough that I can take this as a true.
Now same guy that sold me this wire,take a measure and tell me its 0.21))))
And I bought new wire which have to be 0,25.
Strange now I have higher resistance with same former,but still have to make some extra winds so I can get right inductance.
Hum.....Still some doubts on wire I use.
With new wire I made 2 coils.Data's are:
TX 100w
5,58mH 25,2ohm
RX 105w
5,94mH 26,1ohm
Frequency 14,6 28,9 at G TR4
And
TX 106w
5,95mH 26,7ohm
RX 111w
6,67mH 27,8ohm
Frequency 14,1
Maybe its my former.I use last former posted in TGS thread which is in PDF.And maybe when I print,dimensions are changed.
Damn.This took too many time and efforts.
But I am curies what is that I miss so far.
Just for test I have add the missing TX part with JFet.No big diferance.But coil is balanced much more easy.Dont now if its because of the different wire or JFet.
Same wrong frequency at G TR4.
Sometimes I gain same as TX but sens is still too low.20cm.

I agree with Ivconic -- difficult to fix over internet -- but I am interested in learning as much as possible to compare with same problems I have.

First: I am surprised that you still get double frequency at G of TR4 with JFet installed. Maybe it is not regulating correctly? Is it still installed?

Here are some questions I'm interested in:

1) What is the peak-to-peak voltage of your TX signal at 2907 collector with JFet installed?

2) If you turn the discrimination pot from one end to the other, does the voltage on capacitor C15 (on source of JFet TR4) stay postive always? Same test with ground balance and C12?

2.1) What are dimensions of coils? (probably not that important).

2.2) Is your sensitivity pot turned for max sensitivity? Are there any chatters?

2.3) Does your ground balance control affect the depth?

2.4) How did you shield your coil?


If you have patience, it is time to get the oscilloscope involved, see photos.

3) Need to see TX signal.

4) Then signal at LF353 pin 1; is it clean signal, or clipped, or raggedy? Does it cross the zero line cleanly, or with bumps?

5) Then check square wave at gate of TR4 (in discrimination mode).

A) Is the square wave symmetric, or is one half longer than the other?

B) Is there an extra pulse in it (probably is, which makes double frequency)?

Those are some basic questions.

Now, the problem is: how to fix and make deeper? I think I know how to fix problems like double frequency, but making deeper -- that is trick I don't know yet.

Your circuit is probably correct as Ivconic designed it; the magic, I think, is making a coil just like Max and Ivconic does. I have not done it either. However, I want to see if it is possible to adjust the circuit so it works better with any coil I make.


I agree this takes much time and patience. I am spending tons of time to try to understand it better.

I will keep trying if I have time available and show any conclusions here.

Cheers,

-SB

[simonbaker]

I remember that the zener value played an important role in frequency, also somewhat in amplitude.

It shouldn't have that effect hopefully. I just tried in LTSpice and there the zener affects amplitude but not frequency significantly. As Ivconic showed purpose is to control amplitude and linearity of oscillator.

By the way... without JFet, oscillator is perfectly stable in frequency, that is not a problem... but it is very nonlinear and makes some distortion and also high amplitude.

Is distortion important? Depends on amount and where. The "bump" can potentially be a little problem in discrim circuit because LF353 and C9 is a differentiator which exaggerates any bumps in the waveform. It can affect zero crossing point and make final square wave assymetric or even extra pulse -- bad. Or make discrim pot a little erratic.

But bigger problem for me is the amplitude -- it exceeds the rails of LF353 and really creates crazy output there. Different coils can make different amplitude so everyones' MD a little different. So amplitude is double-edged sword... it improves depth, but maybe drives the LF353 nutty.

And so the saga continues...

-SB

[Ultimate_Haze]

Yes,TX JFet is still installed.Ugly but I just solder missing parts.
Around 15V peak to peak
2.1) What are dimensions of coils? (probably not that important).
27DD used last former PDF posted in TGS thread
2.2) Is your sensitivity pot turned for max sensitivity? Are there any chatters
Sens at max all the time.No chatters.
2.3) Does your ground balance control affect the depth?
little.
2.4) How did you shield your coil?
Like Ivconic show on the pictures in TGS thread.With Al tape.
3) Need to see TX signal.
Ok you have it
4) Then signal at LF353 pin 1; is it clean signal, or clipped, or raggedy? Does it cross the zero line cleanly, or with bumps?
Looks rely ugly.But why?
5) Then check square wave at gate of TR4 (in discrimination mode).
Ok
A) Is the square wave symmetric, or is one half longer than the other?
Lower are longer then picks!
B) Is there an extra pulse in it (probably is, which makes double frequency)?
Nope just lower parts are little longer.
Will post this too.

Hope we find it))
Left is LF353 PIN1/Right is TX vs RX when adjusting

[Fred]

No really, i remember, in the past when experimented much, i tried there various zeners; 3V9,4V3,4V7 and 5V1. I dont remember that i saw any differences than.

It shouldn't have that effect hopefully.-SB

It was a long time ago but as i didn´t have the zener i tried different diodes configurations, then when i got the right zener i (think to) remember frequency changed.Between 3v9 and 4v7 too, but i may be wrong.

[simonbaker]

It was a long time ago but as i didn´t have the zener i tried different diodes configurations, then when i got the right zener i (think to) remember frequency changed.Between 3v9 and 4v7 too, but i may be wrong.

I'm sure very possible depending on other parts, like the JFet, caps, etc. Some oscillators maybe very sensitive.

Regards,

-SB

[simonbaker]

Yes,TX JFet is still installed.Ugly but I just solder missing parts.
Around 15V peak to peak
2.1) What are dimensions of coils? (probably not that important).
27DD used last former PDF posted in TGS thread
2.2) Is your sensitivity pot turned for max sensitivity? Are there any chatters
Sens at max all the time.No chatters.
2.3) Does your ground balance control affect the depth?
little.
2.4) How did you shield your coil?
Like Ivconic show on the pictures in TGS thread.With Al tape.
3) Need to see TX signal.
Ok you have it
4) Then signal at LF353 pin 1; is it clean signal, or clipped, or raggedy? Does it cross the zero line cleanly, or with bumps?
Looks rely ugly.But why?
5) Then check square wave at gate of TR4 (in discrimination mode).
Ok
A) Is the square wave symmetric, or is one half longer than the other?
Lower are longer then picks!
B) Is there an extra pulse in it (probably is, which makes double frequency)?
Nope just lower parts are little longer.
Will post this too.

Hope we find it))
Left is LF353 PIN1/Right is TX vs RX when adjusting

Thanks for the excellent data!

1) From looking at your TX waveform, I can see the JFet is not regulating the oscillator much... are you sure you have the correct JFet and diodes installed in the right direction? I can tell because your TX waveform still has the "bump" in it.

However, I think it has reduced your TX voltage, is that correct? If so, you may just need to increase the zener diode voltage some more -- I sometimes just add regular diodes to get the voltage I want. LEDs can be added also, they're more fun. Just make sure to orient them same as the 1N4148 (I think).

By the way, the "nulling" signals you show in the photo look like they would make negative voltage at C15 (bad). Normally you null on the "other side", more like 180 deg apart. Just make sure voltage on C15 and C12 are positive.


4) Ugly signal is very typical -- I see it too on mine. I'm quite sure it is because the TX signal (without JFet) is larger than on original TGS and exceeds the input limits of the LF353. The jagged part is because the LF353 with feedback has a resonance at very high frequency (this appears in LTSpice sim also) and any sudden change (like exceeding input limits) causes small oscillations.

The JFet, if operating correctly, should be able to make this signal cleaner, by lowering the TX signal. However, it will cost you some depth that way.

I am trying to solve the problem by using a voltage divider to take a smaller fraction of the TX voltage, without using the JFet. But if not done right, it can add a little phase shift and I want to avoid that. A small capacitor in parallel with the feedback resistor also can get rid of jagged part, but that also makes a phase shift, so not recommended by itself.

This ugly signal definitely messes up the discrimination signal at TR4 gate when the disc pot is on that side.

I think we can fix the bad disc signal, but will it help depth??? Maybe not much, but still important to get fixed.

It may help a little, because if square wave at TR4 gate is not symmetic (equal length up and down), your detection is losing a little gain. But probably not much. For example, when you turn disc knob to reject certain metals, you lose some gain but for some reason not too noticeable, although the math says it should be.

B) Is there an extra pulse in it (probably is, which makes double frequency)?
Nope just lower parts are little longer.
Will post this too.

Put scope on gate of TR4 and move disc pot back and forth slowly. Does the square break up sometimes?


2.1) You coils seem good. I still don't know how Max and Ivconic make a perfect coil that works so well. I am just testing many things for now, see if it is possible to adjust the circuit to work well with whatever coil you make.

I am trying to make a "capacitor substitution" tool so I can check effect of varying TX frequency, RX resonant frequency, etc.

More to come...

-SB

[Ultimate_Haze]

B) Is there an extra pulse in it (probably is, which makes double frequency)?
Now I saw this.
When I put scope on G TR4 again.There is extra pulse.And shape is strange.
When I turn disc signal brake sometimes.
I am shore that added Jfet and the rest of missing parts.
I respected original schematic.Disconect 47K R and add other parts.I am using 4V3 now.Not much change.Bumb in TX signal is because TRs are not perfect devices.And its normal.

[simonbaker]

B) Is there an extra pulse in it (probably is, which makes double frequency)?
Now I saw this.
When I put scope on G TR4 again.There is extra pulse.And shape is strange.
When I turn disc signal brake sometimes.
I am shore that added Jfet and the rest of missing parts.
I respected original schematic.Disconect 47K R and add other parts.I am using 4V3 now.Not much change.Bumb in TX signal is because TRs are not perfect devices.And its normal.

Ok. Wanted to doublecheck.

I agree, bump is usually not a problem itself, but indicates the JFet is still not reducing the voltage much. I have one TGSL circuit with the JFet and extra diode and there is no bump -- but very low TX voltage (10 v p-p) and hardly any depth.

So I think problem with disc is TX voltage so high. I am trying to find best solution but not sure yet.

As I said before, getting more depth is another question. Don't understand yet.

I think you probably have cable shields, grounds, and coil wires correctly assigned, but maybe check those again too. There are several combinations.

Regards and thanks for quick replies.

-SB

[simonbaker]

I want to add...

I'm probably too focused on discrim circuit right now... so what if a little "grungy"? I don't think that is main problem with depth. Depth problem is something else I think.

So I will concentrate more on why sometimes depth no good. Coil, of course!!!!! But there is a mystery there, because some coils seem very well made, what is wrong?

Anyway, I will try to investigate. I'm building capacitor substitution circuit so I can try different TX frequencies and compare.

More to come...

-SB

[simonbaker]

Attached is the circuit mod I use to make clean(er) discrim signal at LF353 pin 1 when no JFet in oscillator.

It is simply voltage divider with an additional capacitor that compensates for small phase shift. This stops overloading inputs and gets rid of small oscillations. There is still big "notch" due to bump shape in TX signal, but not a problem I think.

Resistors chosen for compromise. If make larger, introduce more phase shift. If make smaller, they suck voltage from oscillator.

This does not help depth for me -- just clean up discriminator.

-SB

[simonbaker]

Experimenting trying to learn what makes "deep" TGSL.

1. I think it is useful to vary the TX frequency to find best point. Change oscillator cap C2 (.022uF) up and down by small amounts to change freq -- maybe between 12 kHz and 16 kHz, see if there is a best point for you.

I have not found magic point that makes suddenly "deep detector", but it does make some difference for sure.

2. I think to understand what makes good MD, I need to start with good MD by Ivconic or Max, then change something and see what makes it worse, where sensitivities are. Does cable matter? How important type of coil shielding? How important exact inductances? How important coil resistances? How important resistor and capacitor values? How important ICs? How important box shielding? etc.

3. I keep coming back to the following logic (which may be wrong):

Depth requires gain. And noise limits useful gain. If your detector does not have much noise but has poor depth, you need to increase gain. TGSL circuit does not have gain control, so gain mainly due to TX oscillator voltage/current, RX coil resonant circuit. Perhaps perfect coils will make correct gain????????

Circuit gain due to preamp gain (LF353) and bandpass filters (LM358, LM308 ) gain. Preamp gain affected by 4.7k/5.1k resistors -- could 4.7k give slightly better gain than 5.1k, enough to make difference???? I wouldn't think so, but....

What about back-to-back 10uF caps instead of 4.7uF non-polarized? I wonder if back-to-back 10uF are same as 5uF maybe can be like 10uF? First, 5uF slightly bigger than 4.7uF, so makes slightly slower filter, maybe more gain and less noise (longer integration). But not clear how back-to-back caps work. For small signals (under 1 volt), maybe like a 5 uF cap. But for bigger voltages, maybe one cap is "forward biased" and does not contribute same capacitance... this could change filter properties.

So people who use 4.7uF non-polar cap maybe have different gain than those who use back-to-back 10uF????

Just questions to keep checking...

-SB

[nakky]

So people who use 4.7uF non-polar cap maybe have different gain than those who use back-to-back 10uF????
Simon I don't no nothing about GAIN and ELEKTRONIKA but TGS works perfect,solder two 10 uf together +--+ and measure ,I am only COPY things and sometimes it works.Grt Nakky.