TGSL Tuning & troubleshooting

[pebe]

Simon,
An 0.022µf cap has an impedance of 517Ω at 14.5kHz.That's very low compared with 470k.

Will you satisfy my curiosity and try removing C13 and C16 to see how much that increases gain?

[mikebg]

Simon,
An 0.022µf cap has an impedance of 517Ω at 14.5kHz.That's very low compared with 470k.

Will you satisfy my curiosity and try removing C13 and C16 to see how much that increases gain?

Gentlemens! The stages build by U103 are bandpass amplifiers (filters) of demodulated signals. The gain of each stage is 470K/4K7=100 or 40dB.
Capacitors C13, C14, C16 and C17 are not designed properly. The frequency band of demodulated signal is between 0,6Hz and 6Hz. That means C13 and C16 should are 56nF for upper cutoff frequency 6Hz, and C14 and C17 should are 56uF for lower cutoff frequency 0,6Hz. No need C14 and C17 to be bipolar because they process negative signals only. Positive signals are suppressed by D8 and D9.

[pebe]

The FETs TR4 & TR5 switch the received signal through, ON for a half cycle and OFF for half cycle. That gives an output at the switching frequency, ie. 14.5kHz.

So where did you get the figures of 0.6Hz and 6Hz from?

[simonbaker]

Simon,
An 0.022µf cap has an impedance of 517Ω at 14.5kHz.That's very low compared with 470k.

Will you satisfy my curiosity and try removing C13 and C16 to see how much that increases gain?

pebe, mikebg:

I have done a few studies of these filter stages using LTSpice. You can read about it at the beginning of this thread:

http://www.geotech1.com/forums/showthread.php?t=14655

Those capacitors and resistors are carefully chosen to create a bandpass filter with a center frequency of about 10 Hz.

The FETs TR4 & TR5 switch the received signal through, ON for a half cycle and OFF for half cycle. That gives an output at the switching frequency, ie. 14.5kHz.

So where did you get the figures of 0.6Hz and 6Hz from?

I agree with mikebg that the demodulated signal is very low, 0 to 6 Hz is believable.

The signal is our sweeping of the coil over the target. It makes a very slow pulse. The JFets act like a diode to the detected signal because they are synchronous.

So it acts like a diode radio where the audio signal is removed from the carrier signal. In this case, the "audio" signal is our sweeping of the coil and picking up the target signal.

Capacitors C13, C14, C16 and C17 are not designed properly. The frequency band of demodulated signal is between 0,6Hz and 6Hz. That means C13 and C16 should are 56nF for upper cutoff frequency 6Hz, and C14 and C17 should are 56uF for lower cutoff frequency 0,6Hz. No need C14 and C17 to be bipolar because they process negative signals only. Positive signals are suppressed by D8 and D9.

I believe the designers of the TGS wanted a fast response MD. That is why the bandpass center is 10 Hz.

It is certainly possible to lower the bandpass center and slow down the response, and you can do it by choosing different capacitors and resistors -- but you must be careful to choose them all together, or you will make a very poor filter. You can't just change one capacitor or resistor. Well, you can, but it won't be optimal.

If you do use a slower filter, you may find deeper targets more easily -- but pinpointing will be difficult, because the beep will be delayed considerably as you sweep across the target.

However, I do agree this is an interesting area to experiment with, as you can see from my link, it was the first thing I was interested in. So I do plan to continue with this and make a slower response version of the TGSL, but it requires some careful simulation and design first.

No need C14 and C17 to be bipolar because they process negative signals only. Positive signals are suppressed by D8 and D9.

I don't think so. Depending on how you null your coil, the non-inverting input of U103 LM358 may have a positive or negative bias. Because of the negative feedback around U103, the inverting input will assume the same bias voltage I believe.

Regards,

-SB

[mikebg]

Spectrum of demodulated signal

Spectrum of demodulated signal

The ancient method is to calculate period T of demodulated signal because we know diameter D of search coil, its velocity V and target size d.
The modern method is to connect a spectrum analyzer to output of demodulator.

[Vladimir]

This is my method for solving the problem of noise.
Try, it is very effective.
Greeting

[lunamay]

Vladimir what depth did you achived?..can you post some pictures of your work?
best regards

[simonbaker]

This is my method for solving the problem of noise.
Try, it is very effective.
Greeting

Interesting, I'll try to check it out.

Regards,

-SB

[dfbowers]

This is my method for solving the problem of noise.
Try, it is very effective.
Greeting

Good suggestion. Looks like a good solution to higher frequency noise but probably not for noise from A.C. mains, etc. Anyone have access to a spectrum analyzer? Maybe get an idea of what some common noise sources are?

Anyway.. Installing a pair of caps is something easy to try..

Thanks!

[Vladimir]

Friends
Sensitivity to one euro coin is 40-42 cm in air from the coil DD27, old version of board(TGSL). Shematic with all the changes I've already posted to the forum. This additional change is only one capacitor as shown.
I'll try to upload a video file from the results.
Greeting

[pebe]

Simon,
So that’s how it’s done! I was wondering why the amps were designed with those components if they were for 14.5kHz. You can see I am not a metal detector man!

I’ll deal with the other points you raised later, but meanwhile:-

Looking at the filter components around U105, I calculate that the LF side of the passband rolls off with a 3dB point at 7.2Hz and the HF side at 7.23Hz. That’s hardly adequate for a variable hand sweep that goes down to zero rate of scan at the edges and maximum rate in the middle.

I have looked at your pics in the posting you referred to. In the 3rd pic, does V[Vout1,Vdet] - in red - mean the gain of the first stage? If so I cannot make sense of the figures. There is nothing at the input of U103a to limit the LF response yet the curve shows it to be rolling off to minus 56dB at F=zero.

I cannot make sense of the rest of the figures either. Stage gains are shown as:

First stage - 24dB loss

Second stage 2 - 3dB loss

Overall - 4dB loss

What am I missing?

[pebe]

OOPS! Sorry. Just spotted the LF cut R41, C14, C29 on U103a.

[simonbaker]

OOPS! Sorry. Just spotted the LF cut R41, C14, C29 on U103a.

But: you are still correct that the graphs I did are misleading.

I believe the reason they are confusing is because I basically plotted the voltage difference (Vout - Vin) across each stage, and also across both stages combined.

Let's call Vs the signal voltage, V1out the output of the first stage, and V2out the output of the second stage.

Because each stage has so much gain, you can see that (V2out - Vs) is not much different than (V2out - V1out). And V1out is the input to the second stage.

For example, let's say each stage has a gain of 100, and Vs is .0001 volts p-p.

Then V1out = .01 and V2out = 1 volts p-p.

So (V2out - Vs) = (1 - .0001 = .9999), and (V2out - V1out) = (1 -.01 = .99) .

So .99 and .9999 look like similar gains when you plot them on the frequency plot.

Instead of plotting (Vout - Vin), I think I should be plotting (Vout / Vin). Then your arithmetic should work out the way you expect.

In fact, I once had the same question and asked the LTSpice forum and they explained it, but it didn't sink in and I never got back to fixing it.

So good call, I'll post new curves on TGSL Experiments thread which I think make more sense. You can find it here:

http://www.geotech1.com/forums/showp...&postcount=250


-SB

P.S. I have probably made a similar mistake on other graphs, so watch out. However, usually both types of graphs reveal the same conclusions.

[simonbaker]

This is my method for solving the problem of noise.
Try, it is very effective.
Greeting

I think I'm correct that the filter made by this extra capacitor depends on the output impedence of the LM308 for the resistor in the RC filter. Therefore it may not produce consistent results depending on the chip you use for U104, U105.

I would rather see an actual resistor preceding the capacitor to make this more solid. (You have to be careful about the effect with R40 too.)

However, this filter does add another lowpass filter to the overall filtering, and therefore changes the response of the TGSL to targets, depending on what RC constant you choose. I think tinkering with the TGSL filter is a great area to experiment with, I'd just like to do it in a more thorough way to look at all the tradeoffs. But this may turn out to be a nice practical addition if done right and if the MD response pleases you.

Proof is always in the pudding.

Regards,

-SB

[pebe]

Simon,
Thanks for the new graphs. Now it makes sense.
Gain is now correctly shown as 38dB (1st stage) + 20dB (second stage) = 58dB (overall).