IGSL

[Jerry]

Thanks for posting your Mods and corrections Davor. However I do have one nit to pick

U8b is shown with the inputs shorted together. I think you intended for pin 6 to go to Rser2, the 6.8k resistor.

Still have not digested the whole thing but will give it some study.

Jerry

[Davor]

Sure I did. The largest was a jar lid. The rest of them were various pieces of galvanized sheet that is used for roofs etc. Nothing of value. And difficult to dig. We have strong Bora here. It often blows harder than 180km/h. So even if you keep your garden tidy some strange things find their way in anyway.

Regarding filtering ... it is important to filter your signal in order to minimise noise, but what noise exactly? There are white noise sources and a single filter section will fix that. For impulse noise you are just increasing pulse duration with added filtering. For EMI, OK, you might get a little better off, but at a price of worse pulse response and odd delays. That particular capacitor creates lots of problems for strong signals, and negligible benefit for weak signals, so out it goes. Besides, targets appear much sharper than the sinewave, and for that the target response will have a bit more spiky appearance which is good for detection ... unless its sharpness is ruined by the filters. Unlike the original, and TGSL, etc. this solution maintains the most of the duration of the response in both Geb and Disc subchannels, and throughout the complete dynamic span. It is less confusing this way.

The rest of the circuitry is best if observed using a simulation.

[simonbaker]

Here comes the scheme for IGSL with my mods, set as a delta against the original. I did not include my mods of the preamp stage, as there is a minimum benefit if you do nt make your own coil with a center tap etc.
Here you'll find some details I did not discuss on the forum yet, but as they are tested and obviously work fine, and I do not intend to change them any time soon ... that's about it.

Ikebana correction of a Fe channel Disc works perfectly. However, I had to solder some wires directly on chip pins, swap them and insert them into a socket. It did not work the first time I did that. It's action is best perceived with weak signals, and it enables joining the controls of Sens and Thrash. into common ones.

If you follow everything through, you'll find loudspeaker reproduction a bit too quiet, and that's because I favour headphones.

I replaced LM358 with TL062, but beware, it works only with a comparator stage modded as shown - it does not work well if inputs are forced too close to the rails.

I also replaced LF347 with TL064, mainly for lower power consumption. LF347 has somewhat better noise and lower 1/f corner frequency, but noise is dominantly defined by the preamp, so I didn't care. Besides, I don't have LF347s at hand so TL064 there is. I might replace them with LF347 eventually but I don't expect much improvement. LF347 are better, but are they worth the milliamps they consume - I don't know.

There is no LF351. It is not doing anything different than U1B does, so the potentiometers are just wired together. Works fine.

Fe channel's GEB subchannel does not do anything different from Non-Fe channel's GEB, and it could be used for all metal indication. My inspiration ran down on that one, so if some inspired soul has some idea, please step forward.

You'll find diodes only in the second stages of the gain blocks. That provides better response for strong signal. There are no capacitors in those stages because they only ruin timing at the subsequent comparators. There are two LPF stages prior to this one, so those capacitors will not be missed. You'll see that in Disc subchannels there are no resistors in feedback, and the reason is near 20 dB gain that ensures sharp discrimination triggering up to the discrimination edge, and less signal strength dependance on discrimination. Now I can be sure the Al foil will give double tone from faintest to the strongest signals - as seen in the first video. Increasing Geb subchannels gain did not make sense because of the noise.

U7 circuitry is pure Feng Shui. We may discuss that, but it obviously works better than the original and all other Tesoros.

Just ask.

Thanks for schematic. Your mods seem to make some useful changes; in particular, the mods affecting target beep duration and response speed are interesting to me. I see you kept the 22N cap and added Rcomp1 in the "chatter filter", which I'm guessing helps stretch a detected pulse; it also changes the "floor" voltage level that the comparators can pull the drive voltage down to, so everything has to be adjusted to get the desired delay. I have wondered if a very small schmitt trigger interval could be used also to improve the chatter quality and target beep.

The mods with the feedback diodes also affect the target beep duration and response speed. I guess you are OK with throwing away one LP filter; I have wondered if we actually need a higher order filter to make a sharper high-end cutoff so there is less HF noise to make chatter, and we can relax or get rid of the "chatter filter" delay circuit. Don't know.

Enjoyed the videos. I hope you dig some of the good beeps in the trashy area to see what they are.

-SB

[simonbaker]

I think you both have it a wee bit wrong. It is the specific tau (L/R) of various metals that are in play. There are various views of the tau, but in all cases it is a time constant by which we have a phase shift (VLF) or current decay (PI). But basically it is the same thing. Copper tau is ~150μs so you can continue from there.

Being in essence a first order filter, it will never shift a phase by more than 90°.

The next thing to clarify is a famous motion frequency response. First thing to make right is a frequency shift: it does not exist. There is only an amplitude modulation of the somewhat suppressed carrier wave, and every phase shift is due to the taus of various metals and changes very little due to the target geometry etc. That is very good news because earth response will mostly be within a very narrow phase span and that's the reason ground balance works regardless of the height above ground.

What kind of amplitude modulation? Hop to your kitchen and grab a grater and a carrot, strike a carrot against the grater - that kind of amplitude modulation. While sweeping the ground with your coil same thing happens - every metallic object, will produce a bump in an envelope of a received signal. So in fact the motion filter is just enforcing a limitation to how fast the bumps can come to pass them through. So if you are sweeping as a lumberjack with an axe, you'll miss a lot.

"All metal" channel gives you only positive spikes in a GEB channel, which is a channel with suppressed ground signal, your grater. "Grate" faster and you'll get a response rich in higher frequencies.

Testing a Rx with a signal which is not closely related to a Tx is a bad idea. It would produce arbitrary phase shifts, rotating phases ... something similar to EMI. Metal detectors use cross-correlation to get meaningful results. This is good because both Rx and Tx coils are sharing the same medium, and both are affected by same temperature drifts etc. Even if Tx is heavily phase modulated it will produce sensible result. Why? Because Rx is in effect a lock in amplifier. That's why multiple-frequency metal detectors work at all. Even with a FM modulated Tx it would work just fine. As long as Rx switchers are driven by constant tau phase shifters locked tightly to a Tx.

Yes, I agree with all that. I remember now my question about more phase shift with higher TX frequency was satisfied one day when I looked at some of Aziz's spectral graphs and I played with LTSpice. And sweep speed won't affect phase/frequency unless you sweep near the speed of light I would imagine! Which could be fun!

But maybe Golfnut is referring to a different design, or something more like a BFO type of MD.

-SB

[Davor]

Here comes the scheme for IGSL with my mods, set as a delta against the original. I did not include my mods of the preamp stage, as there is a minimum benefit if you do nt make your own coil with a center tap etc.
Here you'll find some details I did not discuss on the forum yet, but as they are tested and obviously work fine, and I do not intend to change them any time soon ... that's about it.

Ikebana correction of a Fe channel Disc works perfectly. However, I had to solder some wires directly on chip pins, swap them and insert them into a socket. It did not work the first time I did that. It's action is best perceived with weak signals, and it enables joining the controls of Sens and Thrash. into common ones.

If you follow everything through, you'll find loudspeaker reproduction a bit too quiet, and that's because I favour headphones.

I replaced LM358 with TL062, but beware, it works only with a comparator stage modded as shown - it does not work well if inputs are forced too close to the rails.

I also replaced LF347 with TL064, mainly for lower power consumption. LF347 has somewhat better noise and lower 1/f corner frequency, but noise is dominantly defined by the preamp, so I didn't care. Besides, I don't have LF347s at hand so TL064 there is. I might replace them with LF347 eventually but I don't expect much improvement. LF347 are better, but are they worth the milliamps they consume - I don't know.

There is no LF351. It is not doing anything different than U1B does, so the potentiometers are just wired together. Works fine.

Fe channel's GEB subchannel does not do anything different from Non-Fe channel's GEB, and it could be used for all metal indication. My inspiration ran down on that one, so if some inspired soul has some idea, please step forward.

You'll find diodes only in the second stages of the gain blocks. That provides better response for strong signal. There are no capacitors in those stages because they only ruin timing at the subsequent comparators. There are two LPF stages prior to this one, so those capacitors will not be missed. You'll see that in Disc subchannels there are no resistors in feedback, and the reason is near 20 dB gain that ensures sharp discrimination triggering up to the discrimination edge, and less signal strength dependance on discrimination. Now I can be sure the Al foil will give double tone from faintest to the strongest signals - as seen in the first video. Increasing Geb subchannels gain did not make sense because of the noise.

U7 circuitry is pure Feng Shui. We may discuss that, but it obviously works better than the original and all other Tesoros.

Just ask.

[Davor]

I think you both have it a wee bit wrong. It is the specific tau (L/R) of various metals that are in play. There are various views of the tau, but in all cases it is a time constant by which we have a phase shift (VLF) or current decay (PI). But basically it is the same thing. Copper tau is ~150μs so you can continue from there.

Being in essence a first order filter, it will never shift a phase by more than 90°.

The next thing to clarify is a famous motion frequency response. First thing to make right is a frequency shift: it does not exist. There is only an amplitude modulation of the somewhat suppressed carrier wave, and every phase shift is due to the taus of various metals and changes very little due to the target geometry etc. That is very good news because earth response will mostly be within a very narrow phase span and that's the reason ground balance works regardless of the height above ground.

What kind of amplitude modulation? Hop to your kitchen and grab a grater and a carrot, strike a carrot against the grater - that kind of amplitude modulation. While sweeping the ground with your coil same thing happens - every metallic object, will produce a bump in an envelope of a received signal. So in fact the motion filter is just enforcing a limitation to how fast the bumps can come to pass them through. So if you are sweeping as a lumberjack with an axe, you'll miss a lot.

"All metal" channel gives you only positive spikes in a GEB channel, which is a channel with suppressed ground signal, your grater. "Grate" faster and you'll get a response rich in higher frequencies.

Testing a Rx with a signal which is not closely related to a Tx is a bad idea. It would produce arbitrary phase shifts, rotating phases ... something similar to EMI. Metal detectors use cross-correlation to get meaningful results. This is good because both Rx and Tx coils are sharing the same medium, and both are affected by same temperature drifts etc. Even if Tx is heavily phase modulated it will produce sensible result. Why? Because Rx is in effect a lock in amplifier. That's why multiple-frequency metal detectors work at all. Even with a FM modulated Tx it would work just fine. As long as Rx switchers are driven by constant tau phase shifters locked tightly to a Tx.

[golfnut]

Simon, I was talking about generating a signal for the Rx to see.

If on the IDX the machine is in non motion - the IF has no frequency components.
If the machine is motion - the IF has a frequency component and it is propotional to sweep speed primarily. - a good swoosh on a long unit may give 10Hz frequency component signal at the IF.

If the Rx sig was simply phase shifted from the Tx (which it isnt its frequency shifted also i.e. more than a cycle of shift) you could have put the Tx signal down a roll of coax to get a phase shifted signal for the Rx, or another RC delay.

SInteresting parallels . How do you conclude there is more than a cycle of shift, at the frequency we typically use (14.5 kHz)? I don't see that. However, I have wondered if we used a much higher frequency, would we see greater shifts? Simple question to answer with LTSpice maybe -- it depends though on the target model you assume.

[Davor]

TX power is negotiable. I never really checked up to which level IGSL will tolerate it.
I guess; up from some level, more power will cause unpredictable phase shifts following frontend saturation.

It is more likely that it will get "touchy" for quality of soil and that will be your limit. I had a severe problem with audio section, and thanks to Petravka's keen eye it is fixed now. The 1000uF was drawn in reverse on a silk screen, and I failed to notice that. So in fact It worked as a bad resistor instead. It affected sound quality in a crackly way so I guess that must have had something to do with the power rail, and influencing the oscillator. Anyway, it is fixed now.

Increasing a Tx power can be done exactly as this Musketeer coil does it, but push the concept a bit further. If the coil inductance is lowered a bit, and a coil is wound with thicker wire, or maybe litz wire, you'll get it to resonance with a larger capacitor, and you'll get more H per applied voltage, while power consumption will relate solely to the coil Q factor. Guess it is worth a try.

[Davor]

I like the choice of tones they are very clear. Even the mixed high/lows.

I played the part where you said how far the tape measure was extended several times, but was not able to make it out for sure. Is it 35 Cm ?

The tip of the meter is at 35 cm which is the distance I have positive indication for 5 kuna. Without a meter it goes a bit further, but only a cm or two, so I didn't bother. Large things like the iron pliers go much further etc.
The clean mixture of tones is a direct consequence of fixing the Fe channel, ikebana style. Now both channels are following each other's dynamic and the sound quality changes negligible with distance and sensitivity setting.

So I have 3 bands of discrimination, and that is a real bonus.

[Davor]

Before I get to answering previous posts, here are two more videos. Live action. I did all my stunts myself
http://youtu.be/Kqbee8gYWM8 This one is about discrimination of small things in a string. Perhaps a bit too close, but you'll get the idea anyway. This time I shifted discrimination to exclude kuna tokens from Fe tone and there is only a high pitched tone on kunas. Lipa (small tokens) that are iron cored sound just as they should: iron. Al foil is on double tone.
http://youtu.be/FZRSgCuzyWI is a small hunt in a back yard. Full of "ancient" trash. Tried to clean it, but in vain.

[ivconic]

@ivconic, yes, you noticed it right, I adjusted a somewhat larger overlap in discrimination of Fe and Cu in order to include a wide variety of metals in a double tone area. I could have left a narrow band just to show how cool is that on Al foil, but it is the actual setting I am getting used to for hunting. Iron - a clear cut low pitched tone. Silver, copper, bulk aluminum, brass - clear high pitched tone. Alloys with nickel, galvanized sheet iron and aluminum foil - double tone with varying quality. Remember the iron cored 20lipa? You can't miss that. A can of spam which is galvanized sheet steel sounds just about the same, only stronger, and with finishing high pitched clicks on the edges. It helps me sift through trash.
Did you notice how exact tone I get at low scanning speed, with no delays etc.? And how the tone scratches through chatters at the edge of reception? It can be perfected, but only a tiny bit so. You already hit the right spot with the gain (more gain gives only more noise and results in more chatters), so the only improvement will be in fighting noise, and maybe increasing a Tx power.
I'll make another video with rig mounted on shaft so you'll see some live action. I think there are very few rigs with such dead-on tone and swift reaction.
In short: it works...

Yes; overlap between blocks is exactly what i planned at the beginning.
I can't measure it because i don't have proper instrument but i guess it is some 30 degrees at maximum overlap.
Fe block is the one that overlaps Nfe "area" up to +30 degrees.
And best part in this is in fact that user can define it with Fe and Nfe Disc presets (as you did).
Your choice of setting it up is telling me that you understood my idea very well.
I wanted exactly that to achieve: to have "area" with both tones by which i can comprise all the "bad" alloys which still are not pure Fe signals.
Because i noticed that majority of the most precious small finds are placed exactly in that "area". Funny but very true in real life.
Example can be small golden coin covered with mineral or hot rock. It will never produce clear high tone in such situation.
At conventional designs as it was TGSL, such coin will not be detected due hard masking. IGSL will detect it and indicate it with mixed combination of tones.
Tone mix will depend on exact coin position and exact coil movements at that moment.
About delay, yes there is no delay and it shouldn't be anyway. If you hear delay on my video than it is about bad video quality i made and additional bad compression i probably done when preparing it to upload on Youtube. In reality there is no delay at all too. I guess i missed to match proper video/audio codecs when compressed files.
Yes tone chatters at edge of detection if Treshold trimmer is adjusted high.
Lower it to half and less and you should get clean and decaying response at edge of detection.
In case you can't: than i guess you should experiment bit with values there at U8. Try with different values for R36 and R56.
TX power is negotiable. I never really checked up to which level IGSL will tolerate it.
I guess; up from some level, more power will cause unpredictable phase shifts following frontend saturation.
Will be good to investigate that more.

Cheers!

[ivconic]

Actually, I have no one else to ask about these things - I set the discrimination such that it gives off double tone for the contemporary money that is all rich in nickel, and to give me a clear indication for the small change tokens. But maybe I'm completely wrong. There are many more experienced hunters that know it better. So - is it just a preference thing, or you have some experience-based setups that work better than the others? And why?

Thanks.

Yes it is matter of preference. I didn't know that 5 Kuna is rich with nickel. If so; than i guess it is ok to be half "rejected" as on your video.

[Jerry]

Lo and behold, my first ever video on Youtube:

http://youtu.be/AX6oQ4UDNtk

It is an air test, a short video.
More to come, as soon as I prepare a fixture for a camera on a shaft.

Very nice video! Good Work!

I like the choice of tones they are very clear. Even the mixed high/lows.

I played the part where you said how far the tape measure was extended several times, but was not able to make it out for sure. Is it 35 Cm ?

Actually, I have no one else to ask about these things - I set the discrimination such that it gives off double tone for the contemporary money that is all rich in nickel, and to give me a clear indication for the small change tokens. But maybe I'm completely wrong. There are many more experienced hunters that know it better. So - is it just a preference thing, or you have some experience-based setups that work better than the others? And why?

Thanks.

I think it is a preference thing and those preferences will change a lot depending on what type of hunting you do. For example, I am mainly a coin hunter and much of what I find are modern U.S. clad coins. One machine I have has the upper break point set right on a zinc penny so when one passes under the coil it will give a distinctive combination of high and mid tones. I can set the low tone break point but not as much as I would like. However, the IGSL has a lot to offer since you can choose both the upper and lower tone break points pretty much anywhere you want. I think choosing a specific target conductivity on both ends would be a good place to start. The three combinations of tones will give a lot of target information.

Jerry

[Davor]

I expected clear high tone response on it but it is responding with mixed tone!?

Actually, I have no one else to ask about these things - I set the discrimination such that it gives off double tone for the contemporary money that is all rich in nickel, and to give me a clear indication for the small change tokens. But maybe I'm completely wrong. There are many more experienced hunters that know it better. So - is it just a preference thing, or you have some experience-based setups that work better than the others? And why?

Thanks.

[Davor]

@ivconic, yes, you noticed it right, I adjusted a somewhat larger overlap in discrimination of Fe and Cu in order to include a wide variety of metals in a double tone area. I could have left a narrow band just to show how cool is that on Al foil, but it is the actual setting I am getting used to for hunting. Iron - a clear cut low pitched tone. Silver, copper, bulk aluminum, brass - clear high pitched tone. Alloys with nickel, galvanized sheet iron and aluminum foil - double tone with varying quality. Remember the iron cored 20lipa? You can't miss that. A can of spam which is galvanized sheet steel sounds just about the same, only stronger, and with finishing high pitched clicks on the edges. It helps me sift through trash.
Did you notice how exact tone I get at low scanning speed, with no delays etc.? And how the tone scratches through chatters at the edge of reception? It can be perfected, but only a tiny bit so. You already hit the right spot with the gain (more gain gives only more noise and results in more chatters), so the only improvement will be in fighting noise, and maybe increasing a Tx power.
I'll make another video with rig mounted on shaft so you'll see some live action. I think there are very few rigs with such dead-on tone and swift reaction.
In short: it works.

@simonbaker, I was about to answer you about futility of scaling down the resistors that ends in overloading of one side of the coil, but you found it yourself already. Regarding the cross-coupled balanced amplifier, it is actually a less fancy half of a very fancy Graeme Cohen double balanced microphone preamplifier that was among the first to break a 1nV/rtHz barrier. Today you can buy off-the-shelf THAT1510 that follows that particular philosophy, yet the Cohen's solution is still better.
There is nothing spectacular about the cross-coupled thing - you can see it in the attached sim. It is actually an anti-parallel bond of two garden variety differential amps, but it fixes many problems at once, including impedance symmetry for both common and differential modes. It has a balanced output, which is just perfect for supplying some balanced mixer, say 4053.

I don't know about dfbowers, I'm new at this.