TESORO GOLDEN SABRE

[ivconic]

"...I've tested the audio mod with transistors... used BD138 instead of BD140... not big deal... that have improved audio just sometimes, other times is not that good. The mod works... but on weak targets you could get similar problems than with just BC517... it's not a power problem but a "timing" problem.

When on for the enough time even the BC517 strategy pays: that happens on shallow targets...."

Yes this is true. It is improving audio in some percent, yet not what i wanted.

[Max]

Aah yes problems with nulling: nulling should be done if using scope in one channel mode, connected to Rx. If I try to do that with attached Tx leads - wrong, totally wrong readings. Everything under 30mV doesn't show up so i was unable to null the coil properly. I also find that Rx + cap is better than Rx with cap and dummy load, gives better readings on the scope.
The phase problem still remains, can you give me a tip there?
To measure it I am using scope connected to Tx and the other leads to pin 7 of LF353. The thing is that nulled coil doesn't means automaticly correct phase shift like I was suspecting before I started the project. And tathe evil thing here is that I can hardly affect just the shifting, which here on my board is almost the same around 30 degrees. Any ideas on that?

Good thing I like is that the signal response of the TGS is excellent, very good audio in deed, no dalays, increasingly nice when the metal object gets closer to the coil!
I love the sound of ringing coins in my pocket

Hi,
if you get 30° you can reverse leads at RX and get 180+30 = 210° that's about what you need to work good.

The nulling using just one channel you do connecting the probe (or multimeter) at output of geb sampling gate. There you'll read the voltage at ground channel in air, no targets near.

When adjusting the coil you'll try to find a zero there, like Qiaozhi stated, that means you've balanced both side of sampled wave to get that zero voltage as result.

It's not an easy stuff but easier than having two traces (tx and rx) and looking for a particular phase shift we know will null the voltage (the famous around 200°) unless you have something digital indication of actual phase... like tektronics' do, cause you can do with single channel scope or a simple millivoltmeter only, that's about accurate too cause we actually need nulling that voltage in absence of a target.

So, put the trimpot at middle, then connect the probe at sampling gate output, then move the coils relative to get minimum voltage: at that point glue them down with e.g. hot melt stuff or fast glue.

You have now ground balanced at middle position and can experience with a few degrees positive or negative using the trimpot.

I found that at total null you'll totally miss the ferrite... but miss much silver detection... better is rotating a bit the trimpot to gain silver detection, while still ignoring almost the ferrite.

Kind regards,
Max

[simonbaker]

Quiaozhi, can you help here a bit?
You mentioned monitoring of gates at TGS using scope. So i tried and there are some oscillograms at both gates, yet i dont understand how can i change something there? Turning geb, oscillograms at both gates are changing from one shape to other and that is all!?
So...i dont have reference oscillograms to compare and see if something could be better...what to change, how to influence arround zero crossing point etc.etc.

Other thing; today i made small 18cm DD coil for TGSL. I nulled it usuall way trying to achieve less possiblle residual voltage. And when nulled in that manner and tested - bad behavior, coin at 18cm with clear sound, but over 20cm up to 27cm pretty cracked and short audio. So nulling failed. Than i renulled it but this time using only a coin and ears....and achieved to null it to detect same coin at 22cm with nice audio. Again over 22cm up to 27cm cracked and short audio. Conclusion - everything is about coil!
After a while i decide to let it as it was at moment - 1e coin cleary detect up to 20cm in air and with short unreliable cracks up to 27cm.
Ok...coil is done, considering 18cm diammeter i think those air performances are optimal. But another problem appeared (deja vu) - silent audio.
Up to 15cm on coin, audio is almost good. From 15cm to 22cm it decays and become bairly audible and very, very short.
So i got mad and started to experiment wildly. Changed almost all suspectable semiconductors, changed some values...nothing, same behavior.
Arround last LM358...that 2K2 (also 10K suggested) ...when i put 2K2 even on previously well nulled coil (with almost good audio) audio starts to behave as with not well nulled coil. When 2K2 replaced with 10K than audio looks fine. Same thing with threshold components..if you change on one side - you loose on other side!?
No benefits.
All day i spent on experimenting...same story ; changing some components to achieve louder audio - loose on audio quality ; cracks appeared...
All day same story...
So i was thinking; maybe problem is somewhere else, keep in mind that i tested and experimented today with TGSL which has audio mod included (2xBC547 + BD140)....
Damn! I stucked here and cant make any move forward!
How come that Classic III has louder audio without cracks and TGS with almost same logic doesnt?
What is the catch?
I came to level where depths are not an issue any more, does not interesting me to improve it more and more. Only thing left unsolved at TGSL is damn audio. No matter what i modded and tried so far...no huge steps forward, no radical improvements...
I am considering to switch to some other design and to leave TGSL as it is.
What do you think about this problem?

Trying to understand your problem -- when you get crackle in audio, does output of U107a (LM358 ) look good, or does it look like it should crackle? In other words, is your mystery just after the LM358, or really the whole system?

I ask because I have been studying signals with scope and also simulating with LTSpice. And it looks to me like Tesoro has used a trick that makes less expensive manufacture, but could lead to audio weakness.

I am considering a mod that looks very interesting in LTSpice and next I will breadboard a circuit. It is a pretty easy mod. If it still looks good I will show so someone who can make PCB maybe can try. With my setup I cannot test the very deep coin region because of high noise.

On a different note, I also want to ask Qiaozhi some more about output of comparators and design of that section. (I think a very big challenge in understanding MD design is that there are two different operating regions -- large signal and ultra-small signal. There can be significant differences in how the components work and are employed in those regions. Sometimes non-ideal characteristics may be crucial.)

R40 feedback after diode maybe provides hysteresis, but also does interesting "feedforward" function that affects audio.

Let's take example of coin detection. It causes small positive pulse in both ground and disc channel. Comparators U106 will create large square pulse when both inputs go above sensitivity level. Then something interesting happens.

Resistor R40 and Diode D12 perform a "clip and add" feed-forward function. Comparators output, instead of going to V+, is clipped to level of output of U105 (which is also the input to ground channel comparator), the ground channel signal. This clip level is approximately ground. So comparators output looks like a big square pulse with a little signal on top of it. The little signal is the ground channel signal.

Now it goes to input of U107a, the final LM358. If you adjust the "threshold" pot just right, the square pulse will be level adjusted so the darlington transistor, referenced to ground, will strip off the square pulse and the driver of the Darlington will just be an amplified version of the ground signal that went into the comparator.

So what is the function in this case? The comparators simply serve to form a logical "AND" gate that passes the amplified ground signal to the audio output. The audio will be proportional to the ground signal. Actually, the diode may provide some logarithmic amplifying, which is good -- boosts small signals more than large.

However, if you adjust the threshold differently, then you will change character of the audio. In one direction, you will cause just the top of the ground signal to pass through to the audio, causing a skinny pulse. In the other direction, you include more of the square wave from the comparators, maybe saturate the LM358 and Darlington.

In addition to this, there is some low pass filtering being done with capacitors C24 and C25. The effect seems to be to round off the square pulse from the comparators. If the coin signal pulse is wide, then very little effect. If the coin signal pulse is very narrow, then the filtering can cause the square wave to never reach the peak -- and it falls beneath the Darlington cutoff, you never hear it. I think this filtering is to block noise chatter from quick noise voltage spikes triggering the comparators.

The question I'm driving at is: what was the intent of the original designers? Was it to use the ground channel signal to control the audio proportionally? If so, the threshold should be adjusted just right and never touched.

Or is the audio supposed to mainly be driven by the square pulse from the comparators? If that is true, then the feedforward circuit is just an artifact. However, in that case the audio loudness would be mainly determined by the width of the detection pulse, not the amplitude (up to a point).

So how you adjust the threshold makes completely different kinds of audio response. The proper amount of gain in the final LM358 would also be intertwined in the question of how the designers intended the audio to be controlled.

I am wondering if so many changes were made to the original circuit we lost the original idea the designers were intending.

I'm hoping these ideas can lead to adjustments to improve the audio. The difficulty is getting good audio in both cases: tiny signal, and bigger signal. There should be many things to try. It would be very useful it Ivconic could take picture of signals of crackling audio as it goes through channels, maybe we see where it happens.

Maybe these are crazy thoughts, hard to tell in middle of night...

Cheers,

-SB

[eclipse]

Aah yes problems with nulling: nulling should be done if using scope in one channel mode, connected to Rx. If I try to do that with attached Tx leads - wrong, totally wrong readings. Everything under 30mV doesn't show up so i was unable to null the coil properly. I also find that Rx + cap is better than Rx with cap and dummy load, gives better readings on the scope.
The phase problem still remains, can you give me a tip there?
To measure it I am using scope connected to Tx and the other leads to pin 7 of LF353. The thing is that nulled coil doesn't means automaticly correct phase shift like I was suspecting before I started the project. And tathe evil thing here is that I can hardly affect just the shifting, which here on my board is almost the same around 30 degrees. Any ideas on that?

Good thing I like is that the signal response of the TGS is excellent, very good audio in deed, no dalays, increasingly nice when the metal object gets closer to the coil!
I love the sound of ringing coins in my pocket

[Max]

Quiaozhi, can you help here a bit?
You mentioned monitoring of gates at TGS using scope. So i tried and there are some oscillograms at both gates, yet i dont understand how can i change something there? Turning geb, oscillograms at both gates are changing from one shape to other and that is all!?
So...i dont have reference oscillograms to compare and see if something could be better...what to change, how to influence arround zero crossing point etc.etc.

Other thing; today i made small 18cm DD coil for TGSL. I nulled it usuall way trying to achieve less possiblle residual voltage. And when nulled in that manner and tested - bad behavior, coin at 18cm with clear sound, but over 20cm up to 27cm pretty cracked and short audio. So nulling failed. Than i renulled it but this time using only a coin and ears....and achieved to null it to detect same coin at 22cm with nice audio. Again over 22cm up to 27cm cracked and short audio. Conclusion - everything is about coil!
After a while i decide to let it as it was at moment - 1e coin cleary detect up to 20cm in air and with short unreliable cracks up to 27cm.
Ok...coil is done, considering 18cm diammeter i think those air performances are optimal. But another problem appeared (deja vu) - silent audio.
Up to 15cm on coin, audio is almost good. From 15cm to 22cm it decays and become bairly audible and very, very short.
So i got mad and started to experiment wildly. Changed almost all suspectable semiconductors, changed some values...nothing, same behavior.
Arround last LM358...that 2K2 (also 10K suggested) ...when i put 2K2 even on previously well nulled coil (with almost good audio) audio starts to behave as with not well nulled coil. When 2K2 replaced with 10K than audio looks fine. Same thing with threshold components..if you change on one side - you loose on other side!?
No benefits.
All day i spent on experimenting...same story ; changing some components to achieve louder audio - loose on audio quality ; cracks appeared...
All day same story...
So i was thinking; maybe problem is somewhere else, keep in mind that i tested and experimented today with TGSL which has audio mod included (2xBC547 + BD140)....
Damn! I stucked here and cant make any move forward!
How come that Classic III has louder audio without cracks and TGS with almost same logic doesnt?
What is the catch?
I came to level where depths are not an issue any more, does not interesting me to improve it more and more. Only thing left unsolved at TGSL is damn audio. No matter what i modded and tried so far...no huge steps forward, no radical improvements...
I am considering to switch to some other design and to leave TGSL as it is.
What do you think about this problem?

Hi Ivconic,
I've tested the audio mod with transistors... used BD138 instead of BD140... not big deal... that have improved audio just sometimes, other times is not that good. The mod works... but on weak targets you could get similar problems than with just BC517... it's not a power problem but a "timing" problem.

When on for the enough time even the BC517 strategy pays: that happens on shallow targets.

I'm not totally sure of that but think...that the problem is that in TGS the timing of some blocks is made in a way you get weak targets as very fast transients on signal path. Too fast to get useful audio.

The device allow too fast recovery of amplifiers, so you'll get very weak audio cause of that when signal amplitude at coil is low and fast swinging like in the case of very small targets or far.

If you change the timing you can change this behaviour I think.

I will check at integrators... I suppose that in MDs like ClassicIII you have longer time constants so the integrators take their time before turning back to threshold, that's mainly due to cap values at there.

So, machines like ClassicIII you can sweep also more slowly, but TGS you're supposed to sweep at higher speed to detect stuff... you have lower weight also and machine is made for shallow nugget hunting... small stuff at few depth with large , fast swings!

The initial TGS intented use is what maybe turned design to that behaviour, but I have to say that situation at bandidoII is similar also, so it's probably a Tesoro's way of doing stay fast in the integrators chain and point to fast sweeps.

Kind regards,
Max

TESORO GOLDEN SABRE

Ivconic , Run audio section with its own supply voltage , a separate battery !!! Just for testing , see what happens ?? ................Eugene51

[ivconic]

Quiaozhi, can you help here a bit?
You mentioned monitoring of gates at TGS using scope. So i tried and there are some oscillograms at both gates, yet i dont understand how can i change something there? Turning geb, oscillograms at both gates are changing from one shape to other and that is all!?
So...i dont have reference oscillograms to compare and see if something could be better...what to change, how to influence arround zero crossing point etc.etc.

Other thing; today i made small 18cm DD coil for TGSL. I nulled it usuall way trying to achieve less possiblle residual voltage. And when nulled in that manner and tested - bad behavior, coin at 18cm with clear sound, but over 20cm up to 27cm pretty cracked and short audio. So nulling failed. Than i renulled it but this time using only a coin and ears....and achieved to null it to detect same coin at 22cm with nice audio. Again over 22cm up to 27cm cracked and short audio. Conclusion - everything is about coil!
After a while i decide to let it as it was at moment - 1e coin cleary detect up to 20cm in air and with short unreliable cracks up to 27cm.
Ok...coil is done, considering 18cm diammeter i think those air performances are optimal. But another problem appeared (deja vu) - silent audio.
Up to 15cm on coin, audio is almost good. From 15cm to 22cm it decays and become bairly audible and very, very short.
So i got mad and started to experiment wildly. Changed almost all suspectable semiconductors, changed some values...nothing, same behavior.
Arround last LM358...that 2K2 (also 10K suggested) ...when i put 2K2 even on previously well nulled coil (with almost good audio) audio starts to behave as with not well nulled coil. When 2K2 replaced with 10K than audio looks fine. Same thing with threshold components..if you change on one side - you loose on other side!?
No benefits.
All day i spent on experimenting...same story ; changing some components to achieve louder audio - loose on audio quality ; cracks appeared...
All day same story...
So i was thinking; maybe problem is somewhere else, keep in mind that i tested and experimented today with TGSL which has audio mod included (2xBC547 + BD140)....
Damn! I stucked here and cant make any move forward!
How come that Classic III has louder audio without cracks and TGS with almost same logic doesnt?
What is the catch?
I came to level where depths are not an issue any more, does not interesting me to improve it more and more. Only thing left unsolved at TGSL is damn audio. No matter what i modded and tried so far...no huge steps forward, no radical improvements...
I am considering to switch to some other design and to leave TGSL as it is.
What do you think about this problem?

[simonbaker]

Yes, the machine is upgraded now: it has more power over the user

Ha ha ha ha ha ha ha... best one yet!

-SB

[simonbaker]

Are you sure that was an upgrade?

Nope! New laptop, they had me over a barrel...

Maybe it will run DOS...

-SB

[Fred]

Are you sure that was an upgrade?

Yes, the machine is upgraded now: it has more power over the user

[simonbaker]

Hi,
indeed it is. If you look at Jack Gifford's patent about tesoro's discriminator circuit you'll see it actually treat steel bottle caps as regular target... that could be avoided using disc knob (understanding it's steel before digging) but actually really far in phase from pure, soft iron.

That's the truth about iron disc: soft iron is easy rejected by disc circuit... while steel isn't and is a task for the user do that or dig it anyway in case of dubts.

Today it's a problem... many steel washers and screws around... also steel nails are a big waste of time... not to talk of horseshoes etc etc steel is everywhere....today

Sure, in more advanced machines VID informations solve that issue. Problem is that many times, expecially at depth, VID stuff is unreliable... and you'll end up digging every target to be sure don't miss the good stuff.

Kind regards,
Max

I just did additional test with Compadre (5.75 inch coil). I'm reporting so I don't leave bad information.

If turn discrimination all way max on Compadre, silver clad quarter still beeps, but steel washer mostly disappears. Very old silver-half dollar (not clad, mostly silver) has very weak signal.

This is not quite in agreement with my tests I did with my TGSL circuit, because I measure steel washer phase as greater from TX signal than silver clad quarter.

(I measure phase to right of TX signal; not think of it as left shift from null signal).

But I measured these phases very close to coil.

This is very useful to know about steel and iron. Now with Compadre if I thumb the disc knob and if it shows quarter, I can go further and if it disappears, maybe not quarter, maybe steel.

I play today with TGSL circuit, and steel washer now act a little differently -- maybe depends on whether neighbors turn on vacuum cleaner -- anyway, better discriminated. Maybe I had GB at one end.

As usual, lot's of playing needed to get good understanding of TGSL pots and performance.

Cheers,

-SB

[Qiaozhi]

(just upgraded to Windows Vista -- smileys all dancing around now, whoa!)

Are you sure that was an upgrade?

[Max]

I tried normal iron nail, and it discriminated correctly. (By the way, coil much more sensitive to nail when perpendicular to straight edge of D coil).

I also tried same steel washer, etc. with my Tesoro Compadre, and it also sounds like a target (not discriminated), so I think TGSL working same as Compadre.

Now I think maybe a notch filter would be really useful to have, even if just to get rid of the steel. My measurements show steel quite a bit past silver in phase.

Cheers,

-SB

Hi,
indeed it is. If you look at Jack Gifford's patent about tesoro's discriminator circuit you'll see it actually treat steel bottle caps as regular target... that could be avoided using disc knob (understanding it's steel before digging) but actually really far in phase from pure, soft iron.

That's the truth about iron disc: soft iron is easy rejected by disc circuit... while steel isn't and is a task for the user do that or dig it anyway in case of dubts.

Today it's a problem... many steel washers and screws around... also steel nails are a big waste of time... not to talk of horseshoes etc etc steel is everywhere....today

Sure, in more advanced machines VID informations solve that issue. Problem is that many times, expecially at depth, VID stuff is unreliable... and you'll end up digging every target to be sure don't miss the good stuff.

Kind regards,
Max

[simonbaker]

I think you may have solved it! I'm using steel to test, steel washer, bottle caps. I think maybe we cannot discriminate it because it is even past silver. That would explain it, iron vs. steel. I'll test.

Thanks, regards,

-SB

I tried normal iron nail, and it discriminated correctly. (By the way, coil much more sensitive to nail when perpendicular to straight edge of D coil).

I also tried same steel washer, etc. with my Tesoro Compadre, and it also sounds like a target (not discriminated), so I think TGSL working same as Compadre.

Now I think maybe a notch filter would be really useful to have, even if just to get rid of the steel. My measurements show steel quite a bit past silver in phase.

Cheers,

-SB

[simonbaker]

Sorry for the delay, im attaching now the eagle file. Its made with Eagle 5.
Don't remember free or professional version, if I did it with prof, there is a big chance that you won't be able to open it. Check for errors, use it as is
I found the problem causing that I was unable to null the coil properly.
And next thing in the menu: changing the phase shift between Rx and Tx signals seems almost impossible just using the method of moving the two coils. The thing is I am now able to adjust the amplitude having 5mV/Div on the scope and Its hard to see even the peaks in the voltage, less than 1mVpp for sure! How much exactly I don't know perhaps around 0.1-0.2Vpp.
BUT: I can't adjust phase in this case. I can do that measuring on the output of LF353, but this will automaticly disable the good ballance achieved!
So the dilemma now is ways to achive phase shifting, not using Rx and Tx:

1. Using metal pieces, What type: Al, Cu, Fe... Drawbacks of this method? I know it's used by DIY-ers not by famous manifacturers (mineleb, tesoro) this must have a reason.
2. Using different values of resistors in the Rx preamp part. I've seen circuits which you can change resistor's values in Rx part to achieve phase shift
3. Others??

Regards

Thanks for posting Eagle file.

Glad your coil nulling now. Can you tell us how you fixed problem nulling? Would be helpful information.

I am trying to understand importance of this phase of nulling residue signal. I would like to see comparison of real TGSL with one coil with wrong phase and one with "correct" phase, both nulled to same voltage.

My thinking is it only should matter for non-motion pinpoint mode.

The reason is: for motion mode, the nulling residue should just make a constant voltage after the JFet detector (filtered by capacitor).

Each target has it's own signal which is added or subtracted. Motion detector responds to difference, ignores constant nulling residue.

Why should it matter whether nulling residue voltage is tiny negative or tiny positive because only target signal passes through bandpass filter?

But proof is in the pudding -- would like to see actual tests. So please save those bad coils!!!

Regards,
-SB

(just upgraded to Windows Vista -- smileys all dancing around now, whoa!)