Bipolar PI work in progress

[mickstv]

Decided it was better to start a new thread for the Bipolar Pulse PI I'm building. If anyone has any comments and ideas, they will be appreciated.

The diagrams should be correct, this is the first time I've used any software to generate them.

I haven't built the output section yet, the project is a work in progress ok.

With the testing I've done so far the circuit perfectly cancels out EF on both the Target sample circuit and GB sample circuit. So now all I have to work on is the Target/GB cancelling circuit. If anyone has any ideas they want to put forward on ways to achieve this comments will be appreciated.


With the circuits shown below and just monitoring the Sample channel output directly using a CRO I can see the results listed below. The CRO was set to 0.2v per div, any lower and the interference from everything running in the house makes the output too unstable on the CRO screen, this is to be expected though.

Australian One Penny at 45cm
Australian two dollar coin at 38cm
One gram lead weight at 22cm
Ten gram lead weight at 29cm

[6666]

Good work Mick.

[Carl-NC]

Comments...

As I recall, you are not using the differential coil response because it turned out to be slower. You may want to revisit that with a better opamp; it's a shame to toss out half your signal, though it may be a wash in SNR.

It may improve turn-off slew by replacing the 3906 in the MOS driver with a Darlington PNP.

I doubt the 200-ohm pots are really necessary... they should only compensate for gain error in the inverter, and that should be easy with 1% resistors.

Why 2 integrators?

- Carl

[mickstv]

Hi Carl, the cause of the differential slow down was because of the addition of the second input resistor and diode protection, they slowed the flyback decay by a least 2us which then takes the minimum sample out past 10us not really good for small targets, although I could have used higher gain amps but then this leads to more noise which may hide smaller targets.


Your right about the 200ohm multiturn pots they only have a minor effect on EF, like you said I've only put them in to take care of any slight imbalance issues.


Why the separate integrators, it's just how I decided to build it. The output of each integrator will be low pass filtered and ac coupled to a second set of gain stages. The target amp will have a set gain whereas the GB amp will be variable. The output of both of these will go into a diff amp and hopefully the circuit will Ground Balance.



Mick

[Carl-NC]

Why the separate integrators, it's just how I decided to build it. The output of each integrator will be low pass filtered and ac coupled to a second set of gain stages. The target amp will have a set gain whereas the GB amp will be variable. The output of both of these will go into a diff amp and hopefully the circuit will Ground Balance.

Be mindful that in using separate channels the taus of the two channels need to be matched. This probably means hand-matching cap values and using 1% resistors. You can do variable GB using one channel, where all 4 samplers feed one integrator (and meaning that you really only need 2 sample switches). The gain of the GB is varied by changing the sample width of the GB samples; see my post "Calculating integrator gain" to see why sample width affects gain.

The TDI uses the (matched) 2-channel approach in order to get separate audio tones for high/low targets. But I also once did a 1-channel TDI and it worked just as good, but without the 2-channel audio, which didn't bother me at all. In fact, the 1-channel approach might be a little less noisy.

- Carl

[Qiaozhi]

The gain of the GB is varied by changing the sample width of the GB samples; see my post "Calculating integrator gain" to see why sample width affects gain.

The TDI uses the (matched) 2-channel approach in order to get separate audio tones for high/low targets. But I also once did a 1-channel TDI and it worked just as good, but without the 2-channel audio, which didn't bother me at all. In fact, the 1-channel approach might be a little less noisy.

- Carl

Which presumably shows the advantage of using bipolar pulsing to cancel the EF, as this makes it much easier achieve GB (by changing the sample width) and also not having to take a second EF sample.

Therefore: A1*(Target-EF1) - A2*(GB-EF2)
is simplified to: (A1*Target) - (A2*GB)

Thus only requiring 2-channels to implement.
Nice.

[mickstv]

Be mindful that in using separate channels the taus of the two channels need to be matched. This probably means hand-matching cap values and using 1% resistors. You can do variable GB using one channel, where all 4 samplers feed one integrator (and meaning that you really only need 2 sample switches). The gain of the GB is varied by changing the sample width of the GB samples; see my post "Calculating integrator gain" to see why sample width affects gain.

The TDI uses the (matched) 2-channel approach in order to get separate audio tones for high/low targets. But I also once did a 1-channel TDI and it worked just as good, but without the 2-channel audio, which didn't bother me at all. In fact, the 1-channel approach might be a little less noisy.

- Carl

Thanks for the info Carl. I'll keep this design as it is for the moment. Then once I see it working out in the bush, I'll start modifying it and try some of your ideas.



I decided to try your GB width ideas and I setup the following to do some testing.

I LPF and AC coupled the two integrators the Target side I used a opamp with a gain of about 10 and the GB side setup a opamp with a minimum gain of 10 but variable upto 100. The output of both of these then into a diff amp.


Now as I don't have any bad ground to test this with I used various ferrite cores and a couple of red bricks.

Using 15us sample width for both target and GB I was able to cancel out the bricks and ferrite cores, and of course some targets of certain sizes canceled out.

Then using a 15us target sample and 30us GB sample I was able to use less gain in the GB circuit to obtain the same result.

More testing is needed though.



Mick

[ZED]

G`day Mick

You might wanna consider different demod chips,4053 chips instead of the 4066 chips for instance,these can be controlled directly from the mcu without the need for level shifters providing the demods and mcu share the same ground plus there are high spec 4053 chips that have very low leakage as well which is also an important consideration as in the MAX4053A.

You may also like to think about dedicated mosfet driver chips that can be controlled directly from the mcu.

Zed.....

[mickstv]

Hi Zed, Thanks for the advice, I'll order a couple ready for version 2, I'll also going to look at a better Micro for the next version, so I can add an LCD and have all the timing menu driven.


This incarnation is nearly complete, The GB circuit seems to be functioning as I think it should. also with the extra gain stages added the EF is still canceling out which is good.


Hopefully by mid week I'll post up the last schematic.


I do have some TC4422 drivers in the parts bin so they'll come in handy for the next version.



Cheers
Mick

[ZED]

Seeing as your in Victoria you will be encountering some highly minerlized ground in which case I would recommend having a preamp stage with a total gain of no more than 40,

Zed

[mickstv]

Seeing as your in Victoria you will be encountering some highly minerlized ground in which case I would recommend having a preamp stage with a total gain of no more than 40,

Zed

Hi Zed, Thanks for the extra info, I might have a look at some sort of digital variable resistor perhaps. Or better still just lock the gain at less than 100 for the two stages, then this will enable an even earlier first sample time. Then make the gain back after integration.


Cheers
Mick

[mickstv]

ATM the design revolves around Pch Mosfets

Just an update. I decided to use Pch Mosfets for the moment because it was easier to work on IE no messing with level shifting for the samples etc.

This circuit uses the Moodz twisted pair coil, I went away from using the THAT1510 IC for the frontend at this stage due to a couple of issues. The circuit also doesn't show the TX drivers or processor section which I'll upload later once drawn.

Also the output section isn't finished still working on it, But I will post up what I've done so far shortly.

The timing in use at the moment is.....

TX1 50us, delay 10us, Sample 1 A 15us then delay 10us, Sample 1B 15 us (but can be variable width) then a delay of 200us to TX2
TX2 50us, delay 10us, Sample 2 A 15us then delay 10us, Sample 2B 15 us (but can be variable width) then a delay of 200us to TX1

Sample 1A - Sample 2A = Target - Earth Field
Sample 1B - Sample 2B = Ground Balance - Earth Field

With the timing as shown I can cancel out various bigger targets whilst retaining smaller fast decaying targets IE small gold under 10g. One test done was to cancel out a pair of pliers whilst still being able to detect a small nugget at the same time. Just mucking around.

I can see holes in detection but from all my reading this is a normal issue with this method of ground balance.


If anyone has any comments and or improvements please speak up. This will be staying a Pch fet version.

[mickstv]

Had to reload circuit accidentially deleted in imageshack.

[Midas]

Just an update. I decided to use Pch Mosfets for the moment because it was easier to work on IE no messing with level shifting for the samples etc.

This circuit uses the Moodz twisted pair coil, I went away from using the THAT1510 IC for the frontend at this stage due to a couple of issues. The circuit also doesn't show the TX drivers or processor section which I'll upload later once drawn.

Also the output section isn't finished still working on it, But I will post up what I've done so far shortly.

The timing in use at the moment is.....

TX1 50us, delay 10us, Sample 1 A 15us then delay 10us, Sample 1B 15 us (but can be variable width) then a delay of 200us to TX2
TX2 50us, delay 10us, Sample 2 A 15us then delay 10us, Sample 2B 15 us (but can be variable width) then a delay of 200us to TX1

Sample 1A - Sample 2A = Target - Earth Field
Sample 1B - Sample 2B = Ground Balance - Earth Field

With the timing as shown I can cancel out various bigger targets whilst retaining smaller fast decaying targets IE small gold under 10g. One test done was to cancel out a pair of pliers whilst still being able to detect a small nugget at the same time. Just mucking around.

I can see holes in detection but from all my reading this is a normal issue with this method of ground balance.


If anyone has any comments and or improvements please speak up. This will be staying a Pch fet version.

Congratulations Mick, sounds like its all coming together well. I'm curious, what were the issues you had with the THAT1510? Because I have a design in the works that uses it... Also how much time have you spent optimizing the timing schedule? Would pulsing faster improve sensitivity? Because I was thinking one of the advantages of bipolar might be the ability to use a faster pulse rate, (since it doesn't have to take the late EF sample).

Midas

[mickstv]

Hi Midas, for some reason when using the THAT1510 on high gains if I had larger targets near the coil the EF would become unbalanced, if I then ran the 1510 at a lower gain it cured the EF balance with bigger targets in the field but it lost response to smaller targets. I went back to the NE5534 because I knew it worked, Also if I decide to try the 1510 again I only have to move a couple of parts around to install it.

The timing previously posted was just a quick mock up. I have experimented with the GB samples at approx 40us width but still the same delay from the primary sample and this changed the detect hole. So eventually I'll have the processor setup to adjust the minimum decay target samples and decay + width of the GB channels.

Yes quicker pulsing does improve sensitivity.

Also I small typo on the circuit I actually run the circuit on about 8 volts.

Mick