TESORO GOLDEN SABRE

[simonbaker]

Hi Simon,

you have an excellent nulling machine! Really intelligent solution.

I am using two component epoxy glue. If the nulling wasn't good, then ripping the coils off and trying again and again...

Regards,
Aziz

I'm like woman, need to change mind all the time, so do anything to avoid solder or glue!!!! (just kidding, old joke about women, shame on me)

Tektronix 465 - bought it new full price with first paychecks of first job -- better than food! Don't look up when it was made -- but I feel like I'm 22 too...

Cheers,

-SB

[Aziz]

Hi Simon,

you have an excellent nulling machine! Really intelligent solution.

I am using two component epoxy glue. If the nulling wasn't good, then ripping the coils off and trying again and again...

Regards,
Aziz

[ivconic]

Eclipse i envy you on LT1008 !!!

Ha,ha...if only i could find those in local shops! I am sure LT1008 would be much better there at TGSL than LM308!

What a nice pcb!!! I envy also on that! You are the artist! Bravo!
Most beautifull pcb i saw so far. Excell job! Bravo!

And Simon... i envy you on "nulling machine" and scope !


Ok...today and tomorrow i am stuck repairing TS800 and TS1000 coils for Musketeer. Day after tomorrow i will return to TGSL .
The hell with Minelab cables there! Both cables are "munched" in several places...like rats did a job!
So low profile cables? I dont beleive my eyes!?
I have Relic Hawk coil with extra high quality cable. What happened with Minelab latelly? Are those latest produced in China?
Although this is a way off topic i would give you a short and valueable hint;
TS800 and TS1000 are having exactly the same inductances and pretty close resistances as my 39cm coil for Relic Hawk. So....
Conclusion; those could be succesfully used with Relic Hawk and vice-versa, 39cm RH coil can be used with Musketeer's.
One more thing; connections in connector must be corrected.Pay attention on that. Those jokers from Minelab intentionally used different connections in conectors for RH and Musketeer, so laic user can not dig out this hot and sweet secret!!!

You own me for this hint!

[simonbaker]

More Pics

Continued...

Pic 1 - one side of null.
Pic 2 - other side of null.

Cheers!

-SB

[simonbaker]

nulling and RX connections

I saw Max advice on RX coil phasing and nulling, 200 deg, etc, but now I have some more confusion.

I use my "nulling machine" for experiment (what, you never saw it yet? Here's a good laugh, see pictures).

Using screw, I adjust my coil to get a good null (less than 2 mv signal, just noise left, see first oscilloscope picture. Top trace is TX signal at collector, 5v/div, bottom trace is RX coil by itself, 5mv/div. Coils are grounded together with shield wire.)

Now I connect coil to circuit -- not sure which way, just pick one.

Then I look like Max says at pin7 of LF353 and compare to TX signal. See next picture (top trace TX, bottom pin7 20 mv/div). Phase is less than 90 deg looks like.

Now I adjust null of coil teeny, tiny, bit, just breath on it, go slightly to other side of null. See picture, phase completely shift, like turn upside down.

You already know this -- when right at null center, very slight shift either side makes huge phase difference. That must be why someone says don't null exactly, go more to one side.

So now I'm not sure which way to connect RX coil leads to circuit. I think we need total explanation starting at coils.

First, which leads are shielded and connected together, based on direction of windings. Then, which pin on ciruit gets shielded lead. Then, using Max's advice for 200 deg phase shift, we can decide which side of null to adjust coil on.

I think it's been said but in pieces; if someone could go start to finish it would be good reference.

Thanks!

-SB

P.S. More pics next post...

[simonbaker]

Don't want to spoil the discussion, though the page is already going to end..
here are some images from today
1. the sweet ICs im going to use 2. bottom layer with gnd plane 3.time to solder

Those are most beautiful!

I'm very interested in effect of ground plane, what you find.

-SB

[eclipse]

Don't want to spoil the discussion, though the page is already going to end..
here are some images from today
1. the sweet ICs im going to use 2. bottom layer with gnd plane 3.time to solder

[simonbaker]

1. Schematic from "Gift Pack" shows R47, R55 on output of last LM393 comparators, but other schematics do not have. They must affect audio I think, but has anyone tested?

I tested those already. I havent spotted any difference than.

I'm playing with those resistor, at first thought: make bigger, stretch pulse, make trailing edge slow... but now I think, no, that will increase false signals from chatter by allowing many pulses to combine. However, still may be good idea because deep coins may combine with noise and make "stutter" signal, so good to combine in that case! It depends on how much noise is around.

No. I reffered plug thing in situation ONLY when plug exist,connected in setup but speaker is used. Cracks appeared on speaker - not on ears?? Funny and strange? Once plug removed - speaker sounds perfectly!? Funny!?

That is spooky! I would think it is that little spring connector that makes contact when you pull out earphone plug -- maybe metal-on-metal you made a diode! I wonder if speaker noise (or magnet) is vibrating the contact.

What bias?

Just resistor adds to emitter-follower, increases impedance at base, less drag on input signal - kooky thought.

Ha! I have idea for you; try to locate and remove 1M resistor parallel to diode at last stage! (if there is any 1M resistor there in Compadre). Than you will get same audio as on TGSL.

Hey, nice idea -- when I get the b--ls to cut into it!

This is what i thought also. I experimented with R45....no significant differences....sheeeeeshh! That's why i suggested extra amplification, in my last post...

I think careful design with pulse shape going into LM358 will be the answer, whatever it is.

When I look at design I think most detections will make same volume, only very, very deep coins at edge of detection will be quieter, so I think maybe really is not proportional audio design in general.

Also, threshold for LM358 U1007A is very awkward design for me, I am struggling with it, won't obey me or make mods easy. Everything is relative near negative rail, very awkward, but I think I need to fix something too.

You will have it in no time. Just do a proper pcb.

I will be excited to do it, your encouragements very appreciated, you got it all rolling with excellent PCBs! I study and appreciate layout even more, you kept "Sync Detector" drive signals nicely away from sensitive areas. I'm not good with PCBs or chemicals (or spending money), so my first try will probably be transfer to copper board from printer, then dremel to cut away copper between wires -- more craziness. But I amuse myself...

Regards,

-SB

[simonbaker]

Hi,
"When you look at circuit, you wonder: why is audio proportional? Comparators U106 make a binary signal, original signal amplitude is forgotten -- right?."

No, it isn't. In theory any comparator has a binary output (+Vcc or -Vee) but it really isn't on practice: simply doesn't exist a comparator that give output at 2 discrete levels of voltage... not considering also that you have also output stage voltage drop... in real devices.

What happens really: you have an op amp that works in open loop at highest possible gain... that is similar in behaviour to an ideal comparator ...that doesn't mean you have really, always 2 discrete levels at output and no intermediate voltage.

When you consider gain you have the answer: consider +8 and -8V as supply voltages and open loop gain of 200000 (as typ of lm393)... how much differential input voltage is required to saturate device ?

deltaV+ = +8/200000 = +40uV
deltaV- = -8/200000 = -40uV

So... you have full saturation voltages out the interval +40 to -40uV at differential inputs.

But what you'll get inside the interval, what if you have e.g. +10uV there ?

V=10uV --> Vout = +10uV*200000= +2V

+2V , not +8 or -8V... so it's an "intermediate" voltage out of discrete limits.

Why I'm doing all this ? Simple... you actually HAVE 10uV and similar voltages there... expecially when you'll measure submicrovolt influence of a small coin at 30cm from coil bottom! So you'll get not full saturation at comparators... but intermediate values and then (sadly) intermediate sound... cracks.

So, when you consider solid signals... like coins at 15cm... you'll get full sound... and real saturation of device, but on weak signals you have linear amplification and intermediate values, thus resulting in poor sound.

This explains why the audio on TGS is proportional... after a particular depth/size combination... so input level threshold... the device will give you a proportional tone related to weakness of signal or the far/small target.

Kind regards,
Max

Interesting point, I not even imagine operate in linear region! But wait -- I would think noise at that stage is greater than linear region anyway. Any signal small enough to be in linear region totally swamped by noise. Yes, we adjust sensitivity control to be above noise, and deep coin could move signal into linear region, but the noise would come with it. I think very low probability it stay in linear region because of noise -- but I'm just guessing!. But still you are right because it all adds up and linear region is less volts for sure.

I will keep it in mind for design experiments.

But pulse width very important too -- I keep watching signal from comparitors on scope, very low sweep. If disconnect output from rest of circuit, pulses very square (binary), even skinny ones.

Reconnect, and pulse leading edge rounded off by capacitor (trailing edge still sharp, as predicted). Very quick pulse never rise too high, less voltage. That is good I think - makes "chatter" pulses quieter. Maybe that is why trailing edge designed to drop fast, so many chatter pulses will not combine into one big false pulse.

This makes me think: maybe improvement if leading edge rises even slower so skinny chatter rejected even more -- increase C25 and increase R40? Just something more to play with, even though Tesoro is always right! (but they must design for mass production, we can tune.

Regards,

-SB

[Max]

When you look at circuit, you wonder: why is audio proportional? Comparators U106 make a binary signal, original signal amplitude is forgotten -- right?.

(Also look at gain of final U107a LM358 op amp - about 50 in latest schematic - very high considering input is 10 volt pulse from comparators! Earlier schematic has gain of only 3 -- Who changed it and why? But let's see...)

So, if audio volume is proportional, it must be due to pulse width from comparator output, not strength of RX signal. So deeper coins must be making quicker pulse as coil passes over.

Capacitor C25 will make very skinny pulses into a lower voltage due to rise time through 100K pull-up resistor R40. But notice, fall time is very fast because pulled down directly by comparator outputs through diode. Quick pulses never get very big or stretched, pulled down quickly.

I watched pulses at U107a non-inverting input for a long time on my oscilloscope. Mostly as expected when a square pulse charges capacitor -- rise quickly and start leveling off. Then drop suddenly with trailing edge of pulse because diode makes shortcut to V-.

This is basically a good shape because it is like logarithmic amp -- shorter pulses charge faster relative to their width, longer pulses don't add that much extra per time.

But I think maybe problem is on trailing edge - pulse drops too fast, so quick pulses don't have time to sound in speaker. We want to stretch short pulses. Maybe C25 needs a slower discharge path. The 4K7 resistors R47, R55 make a slower discharge, should make audio tone longer, especially for quick pulse. Note that R47, R55 (mystery resistors) are not on PCB.

I think some tuning can be done with these resistors and capacitor 25. Gain of U107a may need to be adjusted if resistors attenuate all the pulses.

There may be an idea with positive feedback through RC filter on U107a to hold pulses longer too...

-SB

(I love the smell of frying LM7808 in the morning...not! 33 ohm 10W resistor is now in series with battery V+ for boneheaded breadboarders...)

Hi,
"When you look at circuit, you wonder: why is audio proportional? Comparators U106 make a binary signal, original signal amplitude is forgotten -- right?."

No, it isn't. In theory any comparator has a binary output (+Vcc or -Vee) but it really isn't on practice: simply doesn't exist a comparator that give output at 2 discrete levels of voltage... not considering also that you have also output stage voltage drop... in real devices.

What happens really: you have an op amp that works in open loop at highest possible gain... that is similar in behaviour to an ideal comparator ...that doesn't mean you have really, always 2 discrete levels at output and no intermediate voltage.

When you consider gain you have the answer: consider +8 and -8V as supply voltages and open loop gain of 200000 (as typ of lm393)... how much differential input voltage is required to saturate device ?

deltaV+ = +8/200000 = +40uV
deltaV- = -8/200000 = -40uV

So... you have full saturation voltages out the interval +40 to -40uV at differential inputs.

But what you'll get inside the interval, what if you have e.g. +10uV there ?

V=10uV --> Vout = +10uV*200000= +2V

+2V , not +8 or -8V... so it's an "intermediate" voltage out of discrete limits.

Why I'm doing all this ? Simple... you actually HAVE 10uV and similar voltages there... expecially when you'll measure submicrovolt influence of a small coin at 30cm from coil bottom! So you'll get not full saturation at comparators... but intermediate values and then (sadly) intermediate sound... cracks.

So, when you consider solid signals... like coins at 15cm... you'll get full sound... and real saturation of device, but on weak signals you have linear amplification and intermediate values, thus resulting in poor sound.

This explains why the audio on TGS is proportional... after a particular depth/size combination... so input level threshold... the device will give you a proportional tone related to weakness of signal or the far/small target.

Kind regards,
Max

[ivconic]

I hope to experiment with audio because I agree that it should give as much information about target as possible.

I wish we could hear what Ivconic is hearing -- can you post audio or video of it?

As soon as possible i will record and post here. Need to fix some pc microphone first

First, there are questions about schematic that affect audio, nobody answers yet:

1. Schematic from "Gift Pack" shows R47, R55 on output of last LM393 comparators, but other schematics do not have. They must affect audio I think, but has anyone tested?

I tested those already. I havent spotted any difference than.


2. I have one schematic of TGSL which shows R40 connected directly from output of LM393 instead of after diode D12 -- but most schematics show it connected after diode. Which is correct?


Right way is after diode.

On speaker vs. earphone:

Maybe speaker is very different from earphone as follows: speaker has very large inductor. Audio signal is very narrow pulse, small duty cycle. Speaker coil stores energy from pulse and spreads it out, making rich sound. Earphone has less inductance, cannot spread pulse as well, therefore more loss goes into volume pot or other part of circuit (sucks away energy).

No. I reffered plug thing in situation ONLY when plug exist,connected in setup but speaker is used. Cracks appeared on speaker - not on ears?? Funny and strange? Once plug removed - speaker sounds perfectly!? Funny!?

It is sort of related to concepts in this link: http://www.tinaja.com/glib/msinprop.pdf .

But then Ivconic says speaker gives cracks on deep coin and adding resistor helps -- that seems opposite to my thought. Maybe resistor helps self-bias the Darlington -- I give up .

What bias? Resistor was in serie with speaker, usually 10 ohms. Looks like speaker forming some sort of "loopback" and distort something at previous stage. Not sure what is going there, but noticed various speakers do distort more or less.

Try removing Diode D3 on CD4024 Q3 output -- should make wider pulse, how does it sound?

Actually....yes. Removing one (or even 2) of 3 transients will add a bit of loudness, yet not enough. This is true. Amplifying one "clean" frequency is always easier than amplifying composite signal - less losts.


Maybe for very deep objects, output of LM358 is very small and you start hearing the voltage spikes from the V+ rail (yes, those pesky voltage spikes!).

You really hate those spikes!? Dont you?

You know, some Tesoro models have "all-or-nothing" audio, like my Compadre. All targets have same volume. Weak targets break up. Do we know how TGS was supposed to sound?

Ha! I have idea for you; try to locate and remove 1M resistor parallel to diode at last stage! (if there is any 1M resistor there in Compadre). Than you will get same audio as on TGSL.


However, looking at schematic, it looks like R45 from U107a LM358 output is the pull-up voltage for the audio, so audio signal voltage equals output of LM358 -- I would expect varying volume depending on output of LM358.

This is what i thought also. I experimented with R45....no significant differences....sheeeeeshh! That's why i suggested extra amplification, in my last post...

I keep working furiously to try to get a working TGSL so I can have fun with these problems...

You will have it in no time. Just do a proper pcb.

Cheers,

Double Cheers to you man!

-SB

REGARDS!

[ivconic]

"So, if audio volume is proportional, it must be due to pulse width from comparator output, not strength of RX signal. So deeper coins must be making quicker pulse as coil passes over."

This is true. Yet not explaining fact that same detector gives different audio (width,strenght and decay) with 2 "different" (almost the same) coils??
I hardly can claim 2 handmade coils "same". Almost impossible.
There is one more preference of coil needed to suit to make detector sing louder and wider, but which one?
The rest of your post is more than interesting. Now, with posts like yours and Aziz's this forum get real sence! Thanks much!
Man can learn here...finally!

[simonbaker]

Hi Simonbaker,



The difference between synchronous detector (in TGSL and most of the VLF MD) and lock-in amplifier is as follows:

Synchronous detector:

The transmit oscillator signal acts as a reference. This will be shifted through ground balance and/or discrimination pot. The sine wave is then converted to a digital rectangle wave form via a comparator with some hysteresis (or schmitt-trigger, zero crossing detector) which triggers an analog switch (FET). Only the half of the time-domain signal is passed through. The other half is not processed.
There are more possibilities for processing the synchronous signal:
- averaging over the gated window
- integrating over the gated window
- sampling over the gated window (last signal will be used)
Averaging or integrating over the gated window will increase the signal-to-noise ratio. TGSL uses the averaging method (=low pass filter, C15, R30)

Lock-in amplifier:
Lock-in amplifiers use the whole time-domain signal with some time constant (TC) back to past. The RX signal is multiplied (=mixed) with two reference signals Ref1, Ref2. Ref2 is a +90 degree phase shifted of reference Ref1. The mixed two channels are then low pass filtered. You will have a channel called in-phase (I) and another channel called quadrature phase (Q). The two channels represents a complex signal where the magnitude and phase can be extracted directly (computing, very expensive to implement this in analog hardware).
As the lock-in amplifier uses all the time-domain signal the signal-to-noise ratio will increase dramatically and will give a better sensitivity. You don't have a gated window anymore but a window of interest of time constant length. The more the time constant, the better the SNR.

Implementing the true lock-in amplifier via analog hardware is quite complex. Particularly for two channel lock-in amps (I & Q). Most analog lock-in amps use only a single channel implementation to keep the hardware simple and cheap. Taking pure sine waves for the reference increases the SNR.

That's the main difference. I haven't seen yet a true VLF MD which implements a true lock-in amplifier.

Aziz

Thanks Aziz, you are the prof!

My thoughts:

Integrating and averaging are the same thing for square wave gate signal.

As mentioned in literature, difference between multiplying by square wave and sine wave is a square wave just has extra harmonics - with proper filtering, I think only adds 10% more noise.

TGSL does use quadrature I believe, two channels but not sure what phase difference is yet, maybe not exactly 90 deg.

Synchronous Detector also uses whole time domain because low pass filter is a Markov process and theoretically includes signal all the way in the past. Your digital lock-in amp has advantage in that it is much easier to make filter which is not Markov process (finite time sample), and can make really fancy filters of any kind. But how much advantage? Is it really that dramatic? I say TGSL JFets are not really gated window but rather integrating with a sine wave and many harmonics -- the time window is still the low pass filter time constant.

You point out an interesting design weakness -- TGSL "synchronous detector" only uses half the signal. I have thought about adding a second gate to each channel with a "low noise" inverter feeding the extra gate. But will the inverter add more noise than the extra gate takes out? A good question for experimentation.

So I stick with opinion that synchronous detector is very similar to a lock-in amp, maybe within 10 or 20 percent (maybe that was using hardware), and mathematically quite similar. But your software lock-in will win, no question about it.

Now back to frying parts...

Cheers,

-SB

[simonbaker]

Hi,
the audio of TGS is "proportional". Like in e.g. bandidoII ou get stronger audio signal with stronger signal from target... so shallow target are very easy detected but deep ones are usually listened as very small audio beep (most often "crack" using speaker).

...

Kind regards,
Max

When you look at circuit, you wonder: why is audio proportional? Comparators U106 make a binary signal, original signal amplitude is forgotten -- right?.

(Also look at gain of final U107a LM358 op amp - about 50 in latest schematic - very high considering input is 10 volt pulse from comparators! Earlier schematic has gain of only 3 -- Who changed it and why? But let's see...)

So, if audio volume is proportional, it must be due to pulse width from comparator output, not strength of RX signal. So deeper coins must be making quicker pulse as coil passes over.

Capacitor C25 will make very skinny pulses into a lower voltage due to rise time through 100K pull-up resistor R40. But notice, fall time is very fast because pulled down directly by comparator outputs through diode. Quick pulses never get very big or stretched, pulled down quickly.

I watched pulses at U107a non-inverting input for a long time on my oscilloscope. Mostly as expected when a square pulse charges capacitor -- rise quickly and start leveling off. Then drop suddenly with trailing edge of pulse because diode makes shortcut to V-.

This is basically a good shape because it is like logarithmic amp -- shorter pulses charge faster relative to their width, longer pulses don't add that much extra per time.

But I think maybe problem is on trailing edge - pulse drops too fast, so quick pulses don't have time to sound in speaker. We want to stretch short pulses. Maybe C25 needs a slower discharge path. The 4K7 resistors R47, R55 make a slower discharge, should make audio tone longer, especially for quick pulse. Note that R47, R55 (mystery resistors) are not on PCB.

I think some tuning can be done with these resistors and capacitor 25. Gain of U107a may need to be adjusted if resistors attenuate all the pulses.

There may be an idea with positive feedback through RC filter on U107a to hold pulses longer too...

-SB

(I love the smell of frying LM7808 in the morning...not! 33 ohm 10W resistor is now in series with battery V+ for boneheaded breadboarders...)

[ivconic]

Aziz thank you on splendid explanation!

The way you put it there, now, many important things said in a few simple sentences! Bravo!
Thanks again! I learned few things more!
Best regards!