TGSL Experiments

[Dennis the Mennis]

Hey Danny -

I was a little hasty with this request. I think the outputs of the LM308s are not really what we are interested in.

We really want to see something closer to the output of the SD mosfets with very little filtering, just enough to show the shape of the target phase pulse as the target passes our coils.

That would require a special circuit I think, although perhaps just looking at the voltage on C15 and C12 with a sensitive scope (probably needs broadband amplification) might be close enough.

Regards,

-SB

Hi Simon ,

Allready made some clips on the outputs of the LM308's Sundaynight or must I say Monday morning ...

I will post them asap , I'm on a short break right now and the GPRS Internet connection
on my provider is terrible slow , I can't upload them with this kind of speed.

I will try to catch some of the other signals when I'm back home.

kind regards ,

Danny

Dennis the Mennis

[simonbaker]

Hey Danny -- I am wondering if you can help us with gathering data on target pulses using your USB scope. I'm interesting in seeing the outputs of the LM308's for a typical target (like a Euro) at different depths.

It would require a very consistent sweep. Maybe the easiest way is a "swing" with a long string that you can release to swing the target over the coil at an exact speed. Then change the distance to the coil for different depths.

I'd like to actually see both the time domain and frequency domain of the pulse (at different depths) -- in other words, the oscope traces and some kind of FFT or power spectrum. I think your scope software does that.

Of course noise is the eternal problem -- you'd need to find a pretty quiet place to do the test.

Just something to think about doing some day.

Cheers,

-SB

Hey Danny -

I was a little hasty with this request. I think the outputs of the LM308s are not really what we are interested in.

We really want to see something closer to the output of the SD mosfets with very little filtering, just enough to show the shape of the target phase pulse as the target passes our coils.

That would require a special circuit I think, although perhaps just looking at the voltage on C15 and C12 with a sensitive scope (probably needs broadband amplification) might be close enough.

Regards,

-SB

[golfnut]

gain of igsl stages

trying to understand and maybe - if required optimise the IGSL - maybe it is optimal now? (I know the filter/amp following the detector can hoot)

I make the

pre amp Det/mixer post mix filter amp 2nd post mix amp

33dB -6dB 50dB 30dB


Ive been trying to work out the pre amp Noise Figure in dB - bit lost really
Im GUESSING its somewhere between

12 and 20dB (which I think is a bit poor really) Im not good with op amps


Im going to make a quiet 20dB (LME49990 or some such) preamp and mount at coil end and bring the single ended signal in to the existing preamp (with its gain set to ~10dB)


OTHER.
the Disc threshold pot should span the -ve and +ve rails , currently one side is GND ???

The reason the sens pots dont do a lot is they are feeding a digital input - so they either work or they dont (not like the TGSL where the ots feed a Jfet gate - you do have a control range)

I saw some old silver coins today a clipped and some big old ones with the writing worn off.

I gotta get more from the IGSL - or like everyone else it gets lobbed in the garage never to re emerge

S

[simonbaker]

job for Dennis the Mennis?

Hey Danny -- I am wondering if you can help us with gathering data on target pulses using your USB scope. I'm interesting in seeing the outputs of the LM308's for a typical target (like a Euro) at different depths.

It would require a very consistent sweep. Maybe the easiest way is a "swing" with a long string that you can release to swing the target over the coil at an exact speed. Then change the distance to the coil for different depths.

I'd like to actually see both the time domain and frequency domain of the pulse (at different depths) -- in other words, the oscope traces and some kind of FFT or power spectrum. I think your scope software does that.

Of course noise is the eternal problem -- you'd need to find a pretty quiet place to do the test.

Just something to think about doing some day.

Cheers,

-SB

[simonbaker]

Simon ,

I've noticed same behaviour before watching the output of LF353 vs Tx , comparing the USB and Belden 8723
At first I tought I reversed the leads but I had no double beeps , still reversed the leads resulting in double beeps ,so I was right in the first case.

I'll get back on this thing.

My ordered connectors have arrived so I can finish my 2 new cables : new usb cable and a 4 wire cable with each wire seperatly shielded ( braided ) all together with a twist .


kind regards ,

Dennis the Mennis

Ok, interesting. I'm pretty sure my coils are not backward, in answer to der_fisherman's good observation (but I'd better check again). Especially because with the USB cable the phase moves the other way.

So something happens differently depending on the cable. I'm thinking maybe there are two basic sources of capacitive cross-talk -- shields and cables. And the Belden cable has such small cross-talk that the shield cross-talk, which is opposite polarity, is able to dominate. But that's just a convenient model for now.

So we'll see what new things you find.

Cheers,

-SB

[Dennis the Mennis]

Still pounding this experiment with null voltage vs phase into the dust...

Latest experiment surprised me, as the results looked radically different from the previous.

This time I tried a Belden 8723 cable with kitchen foil shields...

The nulling voltage vs phase curve was "inverted", meaning the phase moved in the opposite direction as I moved through the null point.

To try to understand what this meant, I tried the theory that the null voltage was comprised of two components as follows:

Asinx - .02sin(x + 90)

where A is varied from 8 volts to -8 volts as the coils pass through the null point.

The results seemed to match quite well the experimental data. The difference between this theory and the previous one matching the other data was simply the sign of the constant component "- .02sin(x + 90)", being negative in this case and positive in the previous cases.

This suggests that with the foil shields and Belden 8723 cable, there is a "non-magnetic" component that is still in quadrature with the Tx signal but 180 degrees different than the other cases.

Not sure what this means if anything. Perhaps the foil creates a small quadrature coupled signal that only shows up with the Belden 8723 cable, and it is overwhelmed by a larger quadrature signal that is opposite sign (180 deg phase) when the USB cable is used.

Well, having fun...

-SB

Simon ,

I've noticed same behaviour before watching the output of LF353 vs Tx , comparing the USB and Belden 8723
At first I tought I reversed the leads but I had no double beeps , still reversed the leads resulting in double beeps ,so I was right in the first case.

I'll get back on this thing.

My ordered connectors have arrived so I can finish my 2 new cables : new usb cable and a 4 wire cable with each wire seperatly shielded ( braided ) all together with a twist .


kind regards ,

Dennis the Mennis

[der_fisherman]

Simon, I have only seen similar to this when I accidently reversed one coil to the other.......
regards
Andy

[simonbaker]

nulling, foil shields, belden cable, theory

Still pounding this experiment with null voltage vs phase into the dust...

Latest experiment surprised me, as the results looked radically different from the previous.

This time I tried a Belden 8723 cable with kitchen foil shields...

The nulling voltage vs phase curve was "inverted", meaning the phase moved in the opposite direction as I moved through the null point.

To try to understand what this meant, I tried the theory that the null voltage was comprised of two components as follows:

Asinx - .02sin(x + 90)

where A is varied from 8 volts to -8 volts as the coils pass through the null point.

The results seemed to match quite well the experimental data. The difference between this theory and the previous one matching the other data was simply the sign of the constant component "- .02sin(x + 90)", being negative in this case and positive in the previous cases.

This suggests that with the foil shields and Belden 8723 cable, there is a "non-magnetic" component that is still in quadrature with the Tx signal but 180 degrees different than the other cases.

Not sure what this means if anything. Perhaps the foil creates a small quadrature coupled signal that only shows up with the Belden 8723 cable, and it is overwhelmed by a larger quadrature signal that is opposite sign (180 deg phase) when the USB cable is used.

Well, having fun...

-SB

[simonbaker]

I never thought about depth altering the swept velocity like that.

I guess if coil is v close to target as you sweep it you get a crisp fast signal.


For deep targets the approach of the coil is more slow, due to triangulation. - Tangent effects.

So for deep targets, more gain is required at lower frequencies.

Or sweep fast for deep stuff!

As we all know not sweeping on a motion detector gives a target signal at DC. The frequency of the blip is propotonal to swept speed and depth. Like a doppler radar.

S

Yes, definitely the sweep speed determines the target signal freq as well as the depth. Most people probably have a habitual sweep motion, but it is certainly something you can control, so it is a "skill" factor we can employ in detecting.

The ground composition and target depth determine the pure relationship of the ground signal to target signal, regardless of sweep speed.

However, it seems like our challenge is to either choose a sweep speed, or choose a filter characteristic, that helps separate out the two signals, presumably by orienting the target frequency at the peak of our filter and the ground signal at some lower point -- or actually whatever location provides max separation (without sacrificing S/N -- boy it gets complicated...).

It is probably an over-simplification to refer to the "target signal frequency", since the pulse will actually have some bandwidth and is not a pure frequency. More real data would be interesting.

-SB

[golfnut]

I never thought about depth altering the swept velocity like that.

I guess if coil is v close to target as you sweep it you get a crisp fast signal.


For deep targets the approach of the coil is more slow, due to triangulation. - Tangent effects.

So for deep targets, more gain is required at lower frequencies.

Or sweep fast for deep stuff!

As we all know not sweeping on a motion detector gives a target signal at DC. The frequency of the blip is propotonal to swept speed and depth. Like a doppler radar.

S

[simonbaker]

The existing amp has -3dB BW of 11Hz
The lowpass proposal in post 628 has the same bandwith. ..... So No more noise.


But 12dB more gain where we require it.

S

Ok, that would satisfy the noise issue. However, I suspect this: that targets actually create a "pulse" that is really higher than 1 Hz (often), and the purpose of the original filter is to separate the target from the "ground signal".

What I mean by "ground signal" is the slow variation of ground conductivity or mineralization. As we sweep our coil head over the ground, it will generate a signal which will depend completely on the nature of the ground. If the conductivity/mineralization varies quickly over distance, we'll get a higher frequency ground signal than with very even ground where the conductivity/mineralization is more constant.

My guess is that the ground can often create a significantly large signal in the 1 Hz region at times, but a target will still stand out as a higher frequency blip. So in a way, we want to tune for that target blip and try to suppress the lower frequencies which will come from the ground. It's like trying to detect the wart on your nose without detecting your nose (not your nose specifically...).

But there are probably times where the ground signal is naturally low because of smooth conditions, and maybe then we could take advantage of a slower response filter, or at least one that extends down into the lower frequencies.

I also suspect the commercial MD makers did a pretty good job at finding a filter that works. However, I'll also guess that the original TGS was oriented toward good stability, good pinpointing, good discrimination, a smaller coil than we typically make, and that utmost depth was not a priority.

I think an interesting question is the "frequency" of the target blip. I'm pretty sure it will change with depth. So that brings up a trade-off question for designing the filter. Perhaps we need some real data pretty soon where someone in a low noise, test environment can record some actual target pulses at different depths (I nominate dfbowers of course due to his contraption, but not sure he has any way to record the blips).

I'll also throw in an idea: just like the IGSL has two tones, one for iron, one for other metals, perhaps you could simultaneously have two different response filters, one for slow, one for quick, and a tone for each. That way you won't miss deep targets, and variable ground won't bury real targets.

-SB

[simonbaker]

I tried your double diode to reduce the crash time into the -ve rail... I see what your trying to do..


I used a diode with a lower turn on voltage - so the problem doesnt get so big - before you try to correct.


Yes I have the diode the other way - so you get a real solid +ve ID, almost no negative crash and so no dead time after a hit.

This is with a 100uF in there so we get 2 benefits - gain where we want it and no emarrassing 3 second recovery dead time.

I may try this as it looks promising.
S

Interesting. In my sim I tried a schottky diode for the added diode, and it seemed to ruin the action somehow. I thought perhaps it didn't allow enough lower initial swing. I'm thinking maybe we need a symmetrical swing at least or maybe we could lose a little depth in some cases??? I don't know, I'll be interested in your findings.

-SB

[golfnut]

I tried your double diode to reduce the crash time into the -ve rail... I see what your trying to do..


I used a diode with a lower turn on voltage - so the problem doesnt get so big - before you try to correct.


Yes I have the diode the other way - so you get a real solid +ve ID, almost no negative crash and so no dead time after a hit.

This is with a 100uF in there so we get 2 benefits - gain where we want it and no emarrassing 3 second recovery dead time.

I may try this as it looks promising.
S

[golfnut]

noise analysis

The original Amp has about 100uV of noise.

attached

Steve

[golfnut]

The existing amp has -3dB BW of 11Hz
The lowpass proposal in post 628 has the same bandwith. ..... So No more noise.


But 12dB more gain where we require it.

S