TESORO GOLDEN SABRE

[simonbaker]

The more I look at it, the more I'm intrigued with the U103 filter, as it has zero phase shift at peak frequency, and zero at DC and high freq. Sort of interesting to ponder...

[dfbowers]

Hi Don!

Sorry, that wasn't enough overkill!

I thought it is actually better if all graphs were to same scale, so it makes better visual sense and you can feel the contribution of each stage better.

So, one more time (with feeling)...

-SB

(*Of course use whichever you want)

Awesome!

Thanks..

[simonbaker]

Hi Don!

Sorry, that wasn't enough overkill!

I thought it is actually better if all graphs were to same scale, so it makes better visual sense and you can feel the contribution of each stage better.

So, one more time (with feeling)...

-SB

(*Of course use whichever you want)

[simonbaker]

Simon,

Thanks for all your input.. Much appreciated! Would it be OK to to use your freqency plot with the document I have been working on?

Also, you have two plots on the graph. What is the one with the dotted line?

Don

Hi Don:

Dotted line is phase shift vs. frequency, values on right vertical axis. Left axis is gain in dB for the solid line.

So you have choice (since I'm an overkill guy), I'm attaching the filtering characteristics of each separate stage in the amplifier, including the SD capacitors, and the total combined characteristic of all the stages.

Cheers!

-SB

[dfbowers]

Your descriptions are great and may well make more sense to builders than my way of looking at it. I've been trying to make sense of that filter section for many moons and still working on it.

It does seem to have two main purposes - 1) amplify weak signals, and 2) determine the "response" of the detector to a target pulse.

The response is really determined by the total filter, including the C15, C12 smoothers and the LM358 and LM308 sections. The total bandpass is actually centered on about 8 Hz I think.

I actually wondered if the response was a little too fast for finding deep targets, and I've been playing around with lowering even more (which also helps get rid of some noise). I think that's what eduardo and vladimir tried also. The downside of slowing the response is you get a longer delay before the detection beep, so it has a more sluggish feel. But theoretically it might help improve signal-to-noise (depends on noise sources too).

What would be great would be to find a neat little formula where we can pick the resistors and capacitor values to get whatever response we want. With some grind we could probably come up with some rough rules of thumb. I'm aware there are math ways to analyze it, but it's pretty complicated so I just fool with LTSpice for now to get filter shapes that look desirable.

The other half of the problem is: what response do you want? I'd like to take a detector into the field where I could "dial in" different filter responses and see how they work for different kinds of hunting conditions. This would be a perfect fit for a microprocessor-based MD, with software filtering, which is why I'm getting interested in that. What you learn from playing around with it could then be used to pick component values for our analog designs (which I find more fun to build and use).

In the end, I think your guide excels by keeping it simple without too much overkill on these technicalities, so that people have the best shot at building a good detector that works.

-SB

Simon,

Thanks for all your input.. Much appreciated! Would it be OK to to use your freqency plot with the document I have been working on?

Also, you have two plots on the graph. What is the one with the dotted line?

Don

[simonbaker]

Thanks for the analysis. That's exactly the kind of feedback I'm looking for! Sometimes (like in this case) I can visualize what is going on in a circuit but cannot alway describe things in proper terms. For purposes of the TGSL 101 document I may just rewrite that section and utilize the "bandpass" terminology.

Of course this does bring up even more questions:

Why 11 Hz? I do remember the discussion in the thread a while back where part of the TGSL circuit is tuned to match the approximate sweep speed of the loop and metal passing under it... But 11 Hz seems awfully low. Maybe it's just wide enough to include that frequency that is critical for sweep speed. From your frequency response plot I can see where sweep speed can be quite important!

Also, I need to catch up with Eduardo and find his rationale for doubling the values of C14 in his design. Maybe he was just visualizing them as an "anchor" as well.

Don

Your descriptions are great and may well make more sense to builders than my way of looking at it. I've been trying to make sense of that filter section for many moons and still working on it.

It does seem to have two main purposes - 1) amplify weak signals, and 2) determine the "response" of the detector to a target pulse.

The response is really determined by the total filter, including the C15, C12 smoothers and the LM358 and LM308 sections. The total bandpass is actually centered on about 8 Hz I think.

I actually wondered if the response was a little too fast for finding deep targets, and I've been playing around with lowering even more (which also helps get rid of some noise). I think that's what eduardo and vladimir tried also. The downside of slowing the response is you get a longer delay before the detection beep, so it has a more sluggish feel. But theoretically it might help improve signal-to-noise (depends on noise sources too).

What would be great would be to find a neat little formula where we can pick the resistors and capacitor values to get whatever response we want. With some grind we could probably come up with some rough rules of thumb. I'm aware there are math ways to analyze it, but it's pretty complicated so I just fool with LTSpice for now to get filter shapes that look desirable.

The other half of the problem is: what response do you want? I'd like to take a detector into the field where I could "dial in" different filter responses and see how they work for different kinds of hunting conditions. This would be a perfect fit for a microprocessor-based MD, with software filtering, which is why I'm getting interested in that. What you learn from playing around with it could then be used to pick component values for our analog designs (which I find more fun to build and use).

In the end, I think your guide excels by keeping it simple without too much overkill on these technicalities, so that people have the best shot at building a good detector that works.

-SB

[dfbowers]

I think I used to understand this stuff around 30 years ago - maybe not -- but this is my current "understanding":

U103 is basically just a slightly peculiar bandpass filter - peculiar because it has unity gain at zero and infinite frequencies (rather than zero gain) with a nice fat gain (36 db) at the center frequency of about 11 Hz. (Yes, capacitor C14 is a crucial player in the filter band -- it is not just a shock absorber -- the value is critical in shaping the filter.)

The unity gain means that the output exactly tracks (output equals non-inverting input) any DC bias voltage (by nature of the negative feedback loop). It also exactly tracks high frequencies. Frequencies around 11 Hz are amplified greatly and so become relatively more important, thus the low-bandpass nature.

So, whatever DC bias appears at the non-inverting input will also be present in the output. This is not a "feature" I think, just an artifact that is tolerated because the next filter stage (LM308 ) removes the DC bias and rolls off the high frequencies. The Synchronous Detector (SD) "summing" capacitors C12 and C15 also roll off the high frequencies.

I believe it is the "bandpass" nature of the overall filter (removal of DC bias) that makes the TGSL a "motion detector" (i.e. if you leave the coil over the target, the signal will die out).

Those DC biases that appear at the non-inverting inputs of U103 are due to our "null" signal passing through the SD and accumulating on capacitors C12 and C15. They will be different for the DISC and GB channels because of the different sync pulse phases. But ultimately those DC biases are simply ignored -- their main significance, I feel, is that it means we have to choose a "null" signal phase that does not forward bias the JFets in the SD and screw up the works.

So bottom line is I just see U103 as a bandpass filter stage with an unfortunate DC gain of 1 that is a tolerated tradeoff for the advantage of using the non-inverting input as a high impedance input (to be easy on the SD).

-SB

Thanks for the analysis. That's exactly the kind of feedback I'm looking for! Sometimes (like in this case) I can visualize what is going on in a circuit but cannot alway describe things in proper terms. For purposes of the TGSL 101 document I may just rewrite that section and utilize the "bandpass" terminology.

Of course this does bring up even more questions:

Why 11 Hz? I do remember the discussion in the thread a while back where part of the TGSL circuit is tuned to match the approximate sweep speed of the loop and metal passing under it... But 11 Hz seems awfully low. Maybe it's just wide enough to include that frequency that is critical for sweep speed. From your frequency response plot I can see where sweep speed can be quite important!

Also, I need to catch up with Eduardo and find his rationale for doubling the values of C14 in his design. Maybe he was just visualizing them as an "anchor" as well.

Don

[simonbaker]

Simon,
Back to your response on block #6, I can see where C13, C16, C18 and C21 determine the upper and lower cutoff frequencies thereby creating a "bandpass" filter of sorts. But since the inputs to pins 2 and 5 are not necessarily at 0V, the OPAMPs have to have some reference voltage all the time on pins 2 and 6. You could just replace the 10uF caps with pots and provide a reference voltage but you would be constantly adjusting them to be slightly lower than the DC value on pins 3 and 5.

My take, for example, was that the DC value on pin 6 would always track the voltage on Pin 7 and any change on Pin 5 would be compared to what was on pin 6.. At least until the caps (C14) began to charge or discharge and caught up to the voltage level on pin 7 again.

Are you suggesting that C14 also has a function in the bandpass filter as well?

I think I used to understand this stuff around 30 years ago - maybe not -- but this is my current "understanding":

U103 is basically just a slightly peculiar bandpass filter - peculiar because it has unity gain at zero and infinite frequencies (rather than zero gain) with a nice fat gain (36 db) at the center frequency of about 11 Hz. (Yes, capacitor C14 is a crucial player in the filter band -- it is not just a shock absorber -- the value is critical in shaping the filter.)

The unity gain means that the output exactly tracks (output equals non-inverting input) any DC bias voltage (by nature of the negative feedback loop). It also exactly tracks high frequencies. Frequencies around 11 Hz are amplified greatly and so become relatively more important, thus the low-bandpass nature.

So, whatever DC bias appears at the non-inverting input will also be present in the output. This is not a "feature" I think, just an artifact that is tolerated because the next filter stage (LM308 ) removes the DC bias and rolls off the high frequencies. The Synchronous Detector (SD) "summing" capacitors C12 and C15 also roll off the high frequencies.

I believe it is the "bandpass" nature of the overall filter (removal of DC bias) that makes the TGSL a "motion detector" (i.e. if you leave the coil over the target, the signal will die out).

Those DC biases that appear at the non-inverting inputs of U103 are due to our "null" signal passing through the SD and accumulating on capacitors C12 and C15. They will be different for the DISC and GB channels because of the different sync pulse phases. But ultimately those DC biases are simply ignored -- their main significance, I feel, is that it means we have to choose a "null" signal phase that does not forward bias the JFets in the SD and screw up the works.

So bottom line is I just see U103 as a bandpass filter stage with an unfortunate DC gain of 1 that is a tolerated tradeoff for the advantage of using the non-inverting input as a high impedance input (to be easy on the SD).

-SB

[porkluvr]

Hi everyone
I must be the last person on here to be making one of these I got the board done , mounted alot of components, I been looking for BF247A, apart from J107 does anyone know of any other fets that will do the job as they were discontinued years ago, and the only sources are exspensive.
Many thanks in advance

Try 2N4393, 2N5484, J113, BF245A.

If you can't find one of those, alternate choices are 2N4392, 2N5485, J112, BF245B. Use one of the first four if you can.

Some of those have reversed pin connection. Either use a multimeter to find the gate pin or else get that from a manufacturer's data sheet.

[dfbowers]

So that's what you've been up to Don!

Man, what a nice job, showing all the functional modules! Anyone coming to this forum will have a real crash course in the TGSL and much better odds of completing the project. Very nice gift to all of us. And I know Danny Dennis the Mennis is putting a labor of love into a similar project. Thanks to both of you (and heck, to everyone who contributes to these forums and into Don's opus).

I like the way you managed to communicate the ideas economically and make the points clear -- I'll need to read some more to really appreciate.

I don't quite get this point though:



I just thought of it as a kind of bandpass filter, but I'll think about it some.

Thanks for super resource!

-SB

Simon,
Back to your response on block #6, I can see where C13, C16, C18 and C21 determine the upper and lower cutoff frequencies thereby creating a "bandpass" filter of sorts. But since the inputs to pins 2 and 5 are not necessarily at 0V, the OPAMPs have to have some reference voltage all the time on pins 2 and 6. You could just replace the 10uF caps with pots and provide a reference voltage but you would be constantly adjusting them to be slightly lower than the DC value on pins 3 and 5.

My take, for example, was that the DC value on pin 6 would always track the voltage on Pin 7 and any change on Pin 5 would be compared to what was on pin 6.. At least until the caps (C14) began to charge or discharge and caught up to the voltage level on pin 7 again.

Are you suggesting that C14 also has a function in the bandpass filter as well?

[dfbowers]

Don, Going into the role of Dummy I ask you: "Why these phasor diagrams seem different; why they have different titles; where is the scale of conductivity and how it explains discrimination?"
Here is the correct orientation:
http://en.wikipedia.org/wiki/Complex_plane

I probably should have cleaned up page 22 before posting as that diagram is not accurate. Maybe reuse the one depicted in the original Gifford patent or remove it all together for page 22? I would agree with you that the orientation is not correct on either one.

I will have a go at redrawing the diagram from the Gifford patent.

Thanks for the suggestion.

[simonbaker]

Don, Going into the role of Dummy I ask you: "Why these phasor diagrams seem different; why they have different titles; where is the scale of conductivity and how it explains discrimination?"
Here is the correct orientation:
http://en.wikipedia.org/wiki/Complex_plane

I think such diagrams are qualitative anyway, just to give a flavor. Actual phase will depend on the frequency you are using, doesn't it? Also it can be misleading to say ferrite has a phase -- it is really phaseless, just amplitude-modulates whatever your null phase is (ref: Qiaozhi) (from a received signal standpoint). I think that's the case. (Well, maybe you can say ferrite is zero phase if you consider it as having its own magnetic field component -- out of my physics depth there).

It would a nice experiment to use a TGSL or modified circuit to actually verify some target phases.

Regards,

-SB

[mikebg]

Don, Going into the role of Dummy I ask you: "Why these phasor diagrams seem different; why they have different titles; where is the scale of conductivity and how it explains discrimination?"
Here is the correct orientation:

http://en.wikipedia.org/wiki/Complex_plane

[dfbowers]

So that's what you've been up to Don!

Man, what a nice job, showing all the functional modules! Anyone coming to this forum will have a real crash course in the TGSL and much better odds of completing the project. Very nice gift to all of us. And I know Danny Dennis the Mennis is putting a labor of love into a similar project. Thanks to both of you (and heck, to everyone who contributes to these forums and into Don's opus).

I like the way you managed to communicate the ideas economically and make the points clear -- I'll need to read some more to really appreciate.

I don't quite get this point though:



I just thought of it as a kind of bandpass filter, but I'll think about it some.

Thanks for super resource!

-SB

Actually.. I was hiding in the Bahamas with the wife for a while!!
You may be right about the bandpass filter thingy. Maybe I misused the term "filter"? Let's ponder a bit and maybe I can clarify the explanation.

[satdaveuk]

I have been making some progress on writing a "TGSL for Dummies" manual . Here it is for review! Anyone that has some time, please let me know your thoughts for improvement.

I need suggestions for:
- Additions
- Corrections
- Deletions
- General formatting
- Content in general
- Understanding from the viewpoint of the novice

Plus.. I may be competely off base on my understanding of some concepts!

But, one thing I might ask. I have a specific goal in mind, stated on the first page. I will not entertain any ideas if not in line with my goal statement. AND.. this is TGSL 101. Two hundred level is probably next winters project.

Don

You done a nice job there mate, very profesional
Many thanks