TGSL Experiments

[simonbaker]

Unnecessary Roughness

(for amusement)

Don't do this...

I just spent an hour on my newest TGSL PCB extracting the soldered LM7808 like a rotten tooth and trying to replace it. (I have a crappy little desoldering bulb that sucks everything but the solder while I cook the parts until they steam...)

Problem was the LM7808 measured 10.2 volts output instead of 8. Figured I'd better get it out of the circuit and test it.

It didn't want to come off the PCB. Really made a nice mess trying to work it out lead at a time. After finally dragging it kicking and screaming from a scarred PCB, it indeed measured around 10 volts in a test circuit, so I figured nothing to do but replace it.

Picked another from my nice little collection I got cheap from an online parts store. Measured it -- also about 10 volts!

Now I'm thinking I got ripped off with a whole bag of out-of-spec parts. I start testing the whole bag, and they're all about the same. I'm ranting and raving and mentally figuring how to contact the online store and give them heck.

I take the one from my first prototype breadboard. Also 10 volts! I'm even thinking of desoldering one from my previous PCB, which I remembered always measured about 8 volts. Then I decide to try another meter just for the heck of it.

Of course, you know it... it was my first meter that was wrong -- low battery, even said so on the display. In spite of all the evidence, I couldn't get my thick skull to question that the meter was wrong.

I guess I'm glad I spent some money for that extra meter. Just need to remember to use it before hacking apart my PCB.

Don't even ask how many wrong screws I took out trying to change the battery...

Back to troubleshooting and tuning the PCB...

[lunamay]

hi
i started to mesure components..a noticed that they are a lot off out of range. For eg: capacitors a specialy polistyren are much higher than they supost to be..0.22n has for eg. 0.28..not big but it all can produce problems with tgsl..it seems that we need more exact parts and less noisy..i noticed also that lenth of pots wiring has also matter..and allso grounding a casing of pots can give better parameters..this all is obvious but we must remember about it

best regards to all
lunamay

[simonbaker]

hi
i started to mesure components..a noticed that they are a lot off out of range. For eg: capacitors a specialy polistyren are much higher than they supost to be..0.22n has for eg. 0.28..not big but it all can produce problems with tgsl..it seems that we need more exact parts and less noisy..i noticed also that lenth of pots wiring has also matter..and allso grounding a casing of pots can give better parameters..this all is obvious but we must remember about it

best regards to all
lunamay

That's true, certain parts you should measure, particularly ones affecting TX frequency and RX resonant frequency. I find capacitors are often very far off spec.

On some PCBs, Ivconic provided a space for a "trimmer" capacitor for RX coil to adjust in case C6 .015U cap is off spec. You can do the same with capacitor C2 .022u in the TX oscillator if frequency is not correct.

Many other capacitors affect the pass band of the amplifier section and thus the responsiveness of the MD.

So good point -- use meter to check your parts.

Of course, maybe a better MD can be made if the parts are slightly changed anyway, that is part of experimenting.

-SB

[ODM]

Radio and audio equipment are good examples of building practices, which really matter as much as the choice of parts does! If there are shared ground leads and "floating" metal parts, they risk of poor performance. Plenty of things to pay attention toward, far more than can be listed in a single posting! I've been watching a good while at the various layouts found in designs here, and while some designers make electrically good looking layouts, some boards look like they risk performance with careless layout.

It's not something that is often spoken about. I will gather more useful resources and make a post out of them, material like these application notes can be found through google but it takes time to filter out the most relevant ones:

http://www.fairchildsemi.com/an/AN/AN-389.pdf
http://www.anadigm.com/_doc/AP020800-U202.pdf

There are often remnants of ringing or slow risetimes in the scope readings posted here, which hint toward problems in designs or bad measurement connection. A circuit that operates at 15kHz is not free from design rules regarding tens of MHz if it has digital signal edges (comparators etc) that rise as fast as that!

[simonbaker]

Radio and audio equipment are good examples of building practices, which really matter as much as the choice of parts does! If there are shared ground leads and "floating" metal parts, they risk of poor performance. Plenty of things to pay attention toward, far more than can be listed in a single posting! I've been watching a good while at the various layouts found in designs here, and while some designers make electrically good looking layouts, some boards look like they risk performance with careless layout.

It's not something that is often spoken about. I will gather more useful resources and make a post out of them, material like these application notes can be found through google but it takes time to filter out the most relevant ones:

http://www.fairchildsemi.com/an/AN/AN-389.pdf
http://www.anadigm.com/_doc/AP020800-U202.pdf

There are often remnants of ringing or slow risetimes in the scope readings posted here, which hint toward problems in designs or bad measurement connection. A circuit that operates at 15kHz is not free from design rules regarding tens of MHz if it has digital signal edges (comparators etc) that rise as fast as that!

Thank you, it is always good to know about good ways to lay out PCBs. Of course the proof is in the pudding and people have made nice working MDs with all these PCBs, although I have not made a really good one yet.

I sometimes worry about the long ground trace in some PCBs which ends at the speaker. The speaker makes a big current pulse, and it seems like that would lift up the ground trace. I have wondered if it would be better to connect the speaker ground directly to the point where the battery ground attaches. Sort of a mini "star" configuration.

I'm sure double-sided boards would make better layouts easier. But since people can make fine MDs with these one-sided boards, is it needed?

It is kind of amazing how these "ringing" and various spikes that show up on the scope don't seem to hurt the MD operation; I think this is because the synchronous detector just turns any signal that is in phase with the TX oscillator into a constant voltage that is ignored by the amplifier section. The amplifier section has a bandwidth of about 10 Hz, which helps ignore most stray signals.

A particulary scary looking waveform is at pin 1 of the LF353 in the discrimination circuit. If the regulating JFet is removed from the oscillator (which most TGSL designs do), the TX voltage is higher and I think overloads the LF353 input, causing clipping which rings with big jagged oscillations. Yet people make good MDs with it!!! Even the original Tesoro TGS showed some ringing there (not sure why though).

Anyway, I am experimenting with a small modification (voltage divider) to prevent the overloading. It works, but I'm not sure if it causes some phase shift that causes other problems. These are the kinds of things I'm experimenting with in this thread.

My difficulty is so much noise where I experiment that hard to tell what is going on. But will keep trying.

-SB

[ODM]

Of course the proof is in the pudding and people have made nice working MDs with all these PCBs, although I have not made a really good one yet.

True! And even today it's a manufacturing economic choice to produce single sided vs dual sided or even multiple layer boards. If it works well enough, it can be shipped to the customer.

A relatively slow ramp of the edge can make for erratic switch timing in the demodulator section. Also there seems to be some overhearing of various digital signals in different stages of demodulation and detection. It works and it's likely the original IBs worked equally well, the performance gain might be marginal - but still it's something that could be cured.

...Also writing about this stuff takes far less time than working on a schematic & layout design program, I need to clap my trap and try to contribute more

I sometimes worry about the long ground trace in some PCBs which ends at the speaker. The speaker makes a big current pulse, and it seems like that would lift up the ground trace. I have wondered if it would be better to connect the speaker ground directly to the point where the battery ground attaches. Sort of a mini "star" configuration.

These kinds of little things can sometimes make a big difference and frustration. Sometimes they're tough to trace even with a scope, requiring careful choice of ground spot for the probe.

[pebe]

TGSL

Not a reply - just a question that I cannot seem to find the answer for.

What does TGSL mean?

[hobbes_lives]

What does TGSL mean?

Tesoro Golden Sabre Light, I believe. If I'm not mistaken, it's a project to create a scaled-down version of the Tesoro Golden Sabre. I'm sure others with more knowledge of the project will provide a better explanation than I.

[simonbaker]

Tesoro Golden Sabre Light, I believe. If I'm not mistaken, it's a project to create a scaled-down version of the Tesoro Golden Sabre. I'm sure others with more knowledge of the project will provide a better explanation than I.

Correct - TGSL is "Lite" version of Tesoro Golden Sabre (TGS), leaving out the notch circuitry and simplifying the oscillator and other small simplifications, changes to audio circuit to improve it, etc.

Main thread of discussion is called "Tesoro Golden Sabre", but mostly dedicated to TGSL designs.

-SB

[pebe]

Thanks to you both.

[ODM]

TX gain control circuit in original TGS - performance problem?

How does the TGS fare compared to TGSL, as the latter does not have a gain control circuit like the original golden sabre?

Gain control in the TGS transmitter varies the transistor's bias point, which varies its C-E capacitance, shifting the frequency of the TX as the external damping of the TX oscillator varies.

Does this generally balance out with the ground conditions changing TX inductance along the damping, or does the problem become worse depending on soil type?

Is a shift in TX amplitude or frequency a larger concern, with the low-Q but still tuned receiving coil and the RC circuit derived phase shift which varies its phase angle with frequency variation? Would this make ground balancing more difficult?

Then again, there are detectors which have a fixed switch for changing the TX frequency for using multiple detectors at the same time. They only switch capacitors in and out of the TX oscillator, so absolute frequency stability doesn't seem to be a concern. Then again, manufacturers often take the cost effective way out of things, while hobbyists can spend the extra bit to improve on the original if they wish (given how we're usually spending more than it would take to buy a similar product!).

--- edit: added text ---
I've noticed a builder using a smaller value discriminator pot to lower the amplitude of the TX osc away from clipping. This also hurts the transmitter's Q and makes it more prone to shift with external changes. If the frequency shifts caused by various effects are not a big deal, why not add a tuning potentiometer across the TX coil, adjusting it to shunt the TX until the waveform is pretty? The other way around would be adjusting the TX transistor bias or the capacitors, but not everyone has a wide selection of capacitors available to use for balancing circuits. :/

[simonbaker]

How does the TGS fare compared to TGSL, as the latter does not have a gain control circuit like the original golden sabre?

Gain control in the TGS transmitter varies the transistor's bias point, which varies its C-E capacitance, shifting the frequency of the TX as the external damping of the TX oscillator varies.

Does this generally balance out with the ground conditions changing TX inductance along the damping, or does the problem become worse depending on soil type?

Is a shift in TX amplitude or frequency a larger concern, with the low-Q but still tuned receiving coil and the RC circuit derived phase shift which varies its phase angle with frequency variation? Would this make ground balancing more difficult?

Then again, there are detectors which have a fixed switch for changing the TX frequency for using multiple detectors at the same time. They only switch capacitors in and out of the TX oscillator, so absolute frequency stability doesn't seem to be a concern. Then again, manufacturers often take the cost effective way out of things, while hobbyists can spend the extra bit to improve on the original if they wish (given how we're usually spending more than it would take to buy a similar product!).

--- edit: added text ---
I've noticed a builder using a smaller value discriminator pot to lower the amplitude of the TX osc away from clipping. This also hurts the transmitter's Q and makes it more prone to shift with external changes. If the frequency shifts caused by various effects are not a big deal, why not add a tuning potentiometer across the TX coil, adjusting it to shunt the TX until the waveform is pretty? The other way around would be adjusting the TX transistor bias or the capacitors, but not everyone has a wide selection of capacitors available to use for balancing circuits. :/

Here are some thoughts:

By gain control, do you mean the feedback JFet that keeps the transistor in the linear region and making a nice sine wave?

You probably have a point that the ground can change the TX frequency due to various effects. But I would think in any design where the search coil is part of the oscillator that the ground would change the frequency in some way -- ferrite would change the inductance, for instance.

However, the concept of the synchronous detector is supposed to be fairly immune to slow frequency shifts, since the TX signal drives the Jfet switches, and only the phase of the RX signal relative to the TX signal is detected. A very large frequency shift would create an overall problem similar to the normal design problems if your oscillator is too far off the expected frequency of 14.5 KHz. But I think the circuit is fairly tolerant of frequencies +/- 250 Hz.

But you have an interesting observation, worth testing to see how much the ground changes the oscillator frequency. I think not much probably, but let's see.

So whether the feedback JFet is there or not probably is not important from the effect you mentioned (frequency changes). But I agree that the "stabilized" circuit (with the JFet) is in some ways less stable since it is dynamically trying to stay in the linear region of the transistor. It is even possible to have "limit cycles" in such designs, where the main frequency shifts up and down at a periodic secondary frequency. So in some ways the unregulated "over-driven" oscillator could be more stable, as well as having more amplitude (which we like).

The downside of the unregulated oscillator is that both the higher amplitude and the non-sinusoidal shape cause effects in the circuit. But the good news is that people seem to make very excellent MDs with this circuit -- I'm just amazed they work as well as they do.

Typically the discrimination signal at LF353 U101 pin 1 (Vout) is a total mess with the unregulated oscillator -- lots of jagged oscillations, causing the square wave at LM393 U102 pin 1 to be similarly messy. But it apparently works anyway!

As mentioned previously, I have tried to clean up the mess by using a voltage divider to reduce the TX oscillator signal at the input of of LF353 U101 (pin 2). But I have not made a good working circuit yet, my voltage divider may be shifting the phase around a little. Also, the nonlinear "bump" in the oscillator voltage gets magnified by the LF353 U101 "differentiator" circuit (C9 capacitor) and seems to distort the output waveform and reduce the phase shift range of the discrim pot because it warps the zero crossing point.

So I have a love-hate relationship with the stripped down oscillator at this point. The distortion of the waveform makes all kinds of unusual little glitches that look weird on the oscilloscope.

why not add a tuning potentiometer across the TX coil, adjusting it to shunt the TX until the waveform is pretty? The other way around would be adjusting the TX transistor bias or the capacitors, but not everyone has a wide selection of capacitors available to use for balancing circuits.

I have experimented a little with simpler ideas like that, including LTSpice simulations.

Using a potentiometer I think would not be very stable because any change in the transistor characteristics (e.g. due to heat) or coil-soil interactions would require a different pot setting. Same with TX transistor bias. Not sure how capacitors would help. Keeping the oscillator in a linear region actually requires a non-linear feedback element, like the JFet or a diode or a thermal resistor (like the lamp bulb in the original HP signal generator oscillator).

Also a pot across the coil could reduce Q and ideally it would be nice to have high Q to get an efficient use of power to generate the magnetic field (I believe).

Is a shift in TX amplitude or frequency a larger concern, with the low-Q but still tuned receiving coil and the RC circuit derived phase shift which varies its phase angle with frequency variation? Would this make ground balancing more difficult?

The TX oscillator frequency is fairly far from the resonant frequency of the receive coil LRC circuit and the Q is not too high in the receive circuit, so changes in TX frequency do not make much absolute phase change in the receive circuit (according to some LTSpice simulations I have done). And I don't think modest frequency shifts affect the GB vs Disc phase relationships too much. But until I have a good working TGSL, I won't say that for sure, because something causes some homebuilt MDs to work well and others not.

However, I have tried shifting the oscillator frequency and RX resonant frequency with capacitor banks and not found a magic sweet spot. But definitely found you don't want to get too near the RX resonant frequency with the oscillator frequency. I even found that you can get a working detector if you shift the oscillator frequency above the RX resonant frequency (normally it is below) and use a different nulling and maybe reverse RX leads - it was a little bizarre but worked to some degree -- maybe worth investigating.

The answer to most questions is to take a good working unit and try changing things and see how tolerant the design is for that change. And then see if there is a way to compensate with another change. There are an infinite number of designs that work; we have one that we trust here, but I'd like to know what variations will also work well too (and ultimately work even better).

-SB

[ODM]

That's exactly my point with the jfet based limiter, which works by shifting the bias point (and slightly off-tuning frequency along with it). Frequency shifts could impose a problem in designs which rely on a fixed RC constant for phase shifting - the phase shift is relative to frequency. Apparently the required precision is not that great, as less precisely nulled coils still seem to work. Are MDs based on a reference oscillator (crystal divider etc.) somehow better performers than tx-coil oscillators?

There are plenty of variables in circuit construction, so it's not unexpected to have some MDs made from the same plans fail while others work nicely.

I'm currently working on making a fiberglass mould for 9" cocentric coils, with 220mm and 110mm winding formers ready. Building as physically stable coil as possible will be my first step with IB detectors.

Experimenters working on TGSL will be able to confirm plenty of interesting things about the TGS design.

[simonbaker]

That's exactly my point with the jfet based limiter, which works by shifting the bias point (and slightly off-tuning frequency along with it). Frequency shifts could impose a problem in designs which rely on a fixed RC constant for phase shifting - the phase shift is relative to frequency. Apparently the required precision is not that great, as less precisely nulled coils still seem to work. Are MDs based on a reference oscillator (crystal divider etc.) somehow better performers than tx-coil oscillators?

There are plenty of variables in circuit construction, so it's not unexpected to have some MDs made from the same plans fail while others work nicely.

I'm currently working on making a fiberglass mould for 9" cocentric coils, with 220mm and 110mm winding formers ready. Building as physically stable coil as possible will be my first step with IB detectors.

Experimenters working on TGSL will be able to confirm plenty of interesting things about the TGS design.

I'm not sure what you mean by a fixed RC constant for phase shifting. The discriminator circuitry gives pretty much a 90 deg phase range regardless of frequency I think. The GB circuit is similar, although it is cruder and ends up with less range.

However, the targets in the soil probably give a different phase shift depending on frequency. So if you used a different TX frequency, you might discriminate aluminum at a slightly different setting of the disc knob, for example.

My gut feeling is that the change in the transistor junction capacitance by varying the bias slightly is tiny compared to the .022 cap which mostly determines the TX frequency. The biggest variable for operating frequency is our coil inductance -- we all wind different coils. And of course variations in our parts.

But I think if you get your TX frequency close to 14.5 kHz, it is not too critical, and the circuit tolerates freq offsets OK. And dynamic variations supposedly are ignored by the synchronous detector.

I am not sure what the effect of nulling is on the phase shift of a received target signal. Theoretically, I don't see any effect (to first order). The target sees a big magnetic field from the TX coil and doesn't care what the nulling of the RX coil is. The target re-radiates at the phase it wants to. The RX coil sees the re-radiated magnetic field from the target and responds in phase according to its natural resonance -- again, I don't see how the exact null point matters to the phase of the received signal. However, a second-order effect is the coupling between the RX and TX coils; as you move far from the null point, the coils become more coupled and thus the TX circuit becomes more part of the RX circuit. But near the null point, I don't know if that is a main effect. I tried to do some calculations, and maybe it's more than I think. But I'm not sure yet about the whole subject. We could answer it experimentally if we had a 360 degree discriminator to help us precisely determine the phase of the received signal, then vary the null point and measure the effect.

These are good issues you bring up and exactly the stuff I'd like to probe into with this thread and try to pin down with experiments.

-SB

[ODM]

I meant that a RC phase shift circuit will have a different phase shift for different frequencies, and thus the shift angle will change when the frequency changes.

Variations in ground response and differences between targets will likely matter more than that, at least it sounds likely.

Units built by different people will likely have different operating frequencies, component values vary with tolerance etc. which won't matter much after tuning the detector. It's the shifting in use after that tuning, or instability, which I thought could present problems.

And yes, the approximate ten picofarad shift in CE capacitance is likely to be insignificant compared to the 22 nanofarad capacitor. My bad.