field test unit no 001 "model T"

[moodz]

WARNING : THIS CIRCUIT COULD KILL OR HARM YOU .... DANGEROUS VOLTAGES CAN BE GENERATED.

... having said that ... the fully protected circuit.

[ATTACH]19206[/ATTACH]

well what do ya know ....active damping without resistors is a reality

[jakub]

I was looking at a tiny, cheap DSO (DSO 203) with a function generator..

http://www.diyertool.com/download/DSO203_Guide_RevA.PDF

I know of the benefits of using 24 bit super duper sampling, and this is only 8 bit, but it is fast and does have some very good features.

If the gain is adjustable on the front end of the circuit moodz put up, would 8 bit be enough?

Such an approach would make this whole thing much more accessible.

Also you can get firmware upgrades that allow other useful things like fourier...

No sound but visual interpretation may be viable.

[moodz]

I was looking at a tiny, cheap DSO (DSO 203) with a function generator..

http://www.diyertool.com/download/DSO203_Guide_RevA.PDF

I know of the benefits of using 24 bit super duper sampling, and this is only 8 bit, but it is fast and does have some very good features.

If the gain is adjustable on the front end of the circuit moodz put up, would 8 bit be enough?

Such an approach would make this whole thing much more accessible.

Also you can get firmware upgrades that allow other useful things like fourier...

No sound but visual interpretation may be viable.

If I had more time in a parallel universe ......

You can hack the scope into a detector if you are good with code ... the unit you mentioned comes with source code ....

http://iteadstudio.com/application-n...ource-release/

So if you are a programmer you can do it ... and probably add a 24 bit converter somehow.

moodz

[ashkelon]

hello moodz
i am working with this schema
http://www.geotech1.com/forums/showt...t=17287&page=2
http://www.geotech1.com/forums/showp...1&postcount=35
i could not understand which is the damping control pulse.
i see sample and TX pulse but not damping control pulse.
All metals have same direction flyback movement. thats mean all metals are ferrous , there is not ferro-non ferro disc.
Thanks for your help.

[moodz]

hello moodz
i am working with this schema
http://www.geotech1.com/forums/showt...t=17287&page=2
http://www.geotech1.com/forums/showp...1&postcount=35
i could not understand which is the damping control pulse.
i see sample and TX pulse but not damping control pulse.
All metals have same direction flyback movement. thats mean all metals are ferrous , there is not ferro-non ferro disc.
Thanks for your help.

Hi ... I clearly demonstrated disc in this post.... http://www.geotech1.com/forums/showp...&postcount=113

others have been able to duplicate see post ...
http://www.geotech1.com/forums/showp...&postcount=129


The damping control comes from pin 24 ...

Remember this is an experimental thread ... lots of ideas and variations but no "recipe" or "color by numbers" yet

Moodz.

[chudster]

Like to get involved

Hi,

I have joined the forum and I am looking to get involved.

My aim: I want to have/ build 4 good detectors for my family so we can have some fun camping and have a good chance of finding something. I can't afford 4 $$$$$ detectors but I also aren't in a hurry to do this next week.

Having looked through, this project seems to have the potential to do that. Have I got that right or might there be a better project somewhere else?

I have a lot of engineering background. Some analog and DSP experience and lots of microcontroller (especially real time programming) experience.

Chudster

[chudster]

Hi ... I clearly demonstrated disc in this post.... http://www.geotech1.com/forums/showp...&postcount=113

others have been able to duplicate see post ...
http://www.geotech1.com/forums/showp...&postcount=129


The damping control comes from pin 24 ...

Remember this is an experimental thread ... lots of ideas and variations but no "recipe" or "color by numbers" yet

Moodz.

Moodz,

I decided the way to help and to come up to speed was to go carefully through the posts and provide a 'review' of sorts, comments, opinions and questions. I hope that its useful and something can be taken away from them to help the project even if I am a naiive newbie to it .

I have to say that having gone through, I am impressed and fantastic work is being done by very talented people- more talented than me!

Some cornerstones of what I understand:

1) The MAX32 with the PIC32 is a chosen direction for development. Once development is 'finished' a lesser and cheaper micro might be used.
2) The frontend is fairly stable and people are happy with its performance. Ground effects and interference are greatly(??) solved by the differential coil approach. (Although this is development and the aim is to always improve so we should never really be satisfied).
3) Typical waveforms to be processed are as in the quoted posts.
4) The current sampling is low bit rate (<32Khz) via a high accuracy A/D (24 bit). Three options are being looked at for future sampling- higher rate but striving for accuracy.
5) Development is ongoing and probably always will be. At some point a practical design (or numerous) will spin off.

At this point I ask what are we trying to measure/ determine?

I am sure that much has been written about the various theories and folks here know more than me but....
A) There are different targets to detect. Their electromagnetic properties (conductivity and permeability) give different responses as in (2).
B) The mass, depth and dispersion of the target are variables that influence the response.
C) Metal detectors traditionally have used analogue techniques. From other innovations, movements to a digital domain is a game changer.

What is common to both analogue and digital techniques is that we hit the target like a blind bat with an electromagnetic chirp and analyse the echo to figure out whether it is tasty or not.

The 'ideal' system would definitively tell us.
1) The mass of the target
2) The depth (Z axis) of the target
3) The material of the target
4) The X-Y axis location of the target (I envisage a laser beam deflected to pinpoint it to make it shine )

There are complications. There could be multiple targets of different material types and ferrous materials in particular might mask non-ferrous targets, however detecting practice to scan, find remove and rescan solves this.

However, with one coil in the Z axis we are unlikely to have all the information needed and in particular mass and depth may not resolve completely as the amplitude/ characteristic of the echo is almost certainly proportional to both.

So, an obvious ideal 'scientific' solution would be to analyse the 'echo' and exactly calculate the conductivity and permeability of the material, look up a table and tell the user what is there. To do this we could use scientific models of sorts. I did say ideal ...... and this is approximately what analogue systems try to do.

The key advantage of digital systems over analogue ones (I have been part of a revolution in one industry in a former life....) is their adaptability. Different and non linear strategies can be thrown at a problem which is difficult to do in the analogue domain. So the design needs to have that flexibility for digital control of key parameters (I think you already know this).

The key to this is to make sure that the information of the waveform is correctly recorded in the digital domain with a minimum of processing/ distortion so that the advantage of adaptive strategies can be thrown at the problem. This is especially true for a development system. (From posts I think you also know this).

With those comments I make the following points for discussion:

1) Analogue integration (and differentiation) of a signal do provide information about it but with an accurate digital representation these can be done digitally. Frequently filtering and other analogue processing is used to overcome sampling limitations or as an analogue technique to process to the answer of (1) (2) (3) (4).

You mentioned three possible sampling regimes:

2) Looking at the signals in the waveforms I have seen where the target information is contained in a signal typically 5 uS to 25uS long the sampling rate should be at least in the mega sample per second range. Note that Midas' scope is set at the 75 MSPS range so imagine what we 'see' in the detector in comparison.

3) Processing and sampling at key points in the waveform is (in my view) pandering to an analogue strategy and view of the world.

4) High accuracy (ie 24 bit) ADC may not be the best strategy here. Less is more, that is higher sample rates would seem to reconstruct the waveform information better. The serial clock rate of a serial ADC is limiting on the sample rate and it is unlikely that an adequately high sampling rate can come from such a chip. A high accuracy waveform might help reconstruct some of the ringing (artifacts??) I can see in Midas' scope but do you want these other than to improve the electrical integrity/ coil design? To see these too you need high sample rates.

5) High accuracy and sample rate (and cost) are trade-offs in ADCs. While it is great to have both, we should particularly question how sample rate and resolution helps us in our quest for (1) (2) (3). Key to this is to ask questions like whether the small ringing perturbations shown in (for example) in Midas' scope add information to (1) (2) (3) or are something else that we should ignore. I don't know the answer to this (they could be undiscovered/ unrecognised 'R&D' info that helps us) but suspect they are tuning artifacts.

6) Thus the 1MSPS 10 bit ADC built into the PIC might be a better choice than applying sample and hold/ integration/ analogue type techniques. You can augment the accuracy with some kind of programmable amplifier and adaptively change the gain depending upon the echo to maximise it while avoiding clipping or saturation. That is start out looking for the weak target and then adaptively drop gain to better determine what you have if it clips.

7) Maybe consider strategies to do better than the 1MSPS rate in the PIC although getting such high speed data into the PIC may be difficult (20MHz max serial clock???).

If the PIC (and its ADC or the ability to get high sample rate data) is the limitation then an alternative processor with faster ADC (if such a beast exists) might be warranted.

I know development is a lot of effort so I hope these comments provide a review point and stimulate some discussion. My key comment is that of your 3 directions for sampling you need to place more weight on sample rate even if that costs you resolution. Said another way, if you hope to do good freq domain processing you need good freq sampling relative to the signal to be analysed.

Please don't stone me if I have been naive! I really hope my comments are useful.

Chudster

[straton]

Hello to everyone. My thoughs... (just saw this design.) I have been using the that1510/1512 for 3 years now. Very quiet chip intended for audio appl. and not DC amplification, with some major flaws. Extremely temperature sensitive! I have mounted mine on a large heatsink to make it temperature "steady/stable". Offset on this thing is close to 40uV/C. Datasheet does not mention these cause of audio appl in mind. That means do not amplify a lot, or you get close to supply rails and saturate the output...or if oscil/on. occurs, now you know why!!!

[moodz]

Moodz,

I decided the way to help and to come up to speed was to go carefully through the posts and provide a 'review' of sorts, comments, opinions and questions. I hope that its useful and something can be taken away from them to help the project even if I am a naiive newbie to it .

I have to say that having gone through, I am impressed and fantastic work is being done by very talented people- more talented than me!

Some cornerstones of what I understand:

1) The MAX32 with the PIC32 is a chosen direction for development. Once development is 'finished' a lesser and cheaper micro might be used.
The MAX32 is easy to get, cheap, small and "relatively" easy to program.
2) The frontend is fairly stable and people are happy with its performance. Ground effects and interference are greatly(??) solved by the differential coil approach. (Although this is development and the aim is to always improve so we should never really be satisfied).
The differential coil largely eliminates EMI .... audio and radio technologies have been using them for years ... ground effects are another story....
3) Typical waveforms to be processed are as in the quoted posts.
The published waveforms clearly indicate that differential coils do work.
4) The current sampling is low bit rate (<32Khz) via a high accuracy A/D (24 bit). Three options are being looked at for future sampling- higher rate but striving for accuracy.
The sampling rate will be lower if sampling demodulated signal and higher if direct sampling ... I am looking to use 24 bit sampling with rates from 5 KHz, 192 Khz and 2.5 Mhz.
5) Development is ongoing and probably always will be. At some point a practical design (or numerous) will spin off.
Practical designs are being worked on however some commercial companies are trying to tie up the PI market with Patents ...we are fighting back with our own patents before publishing new designs.

At this point I ask what are we trying to measure/ determine?

I am sure that much has been written about the various theories and folks here know more than me but....
A) There are different targets to detect. Their electromagnetic properties (conductivity and permeability) give different responses as in (2).
B) The mass, depth and dispersion of the target are variables that influence the response.
C) Metal detectors traditionally have used analogue techniques. From other innovations, movements to a digital domain is a game changer.

Right on all three points .... digital is the future however analogue techniques can be used as inspiration.

What is common to both analogue and digital techniques is that we hit the target like a blind bat with an electromagnetic chirp and analyse the echo to figure out whether it is tasty or not.

The 'ideal' system would definitively tell us.
1) The mass of the target
2) The depth (Z axis) of the target
3) The material of the target
4) The X-Y axis location of the target (I envisage a laser beam deflected to pinpoint it to make it shine )

You missed the automated hole digger function ...

There are complications. There could be multiple targets of different material types and ferrous materials in particular might mask non-ferrous targets, however detecting practice to scan, find remove and rescan solves this.

However, with one coil in the Z axis we are unlikely to have all the information needed and in particular mass and depth may not resolve completely as the amplitude/ characteristic of the echo is almost certainly proportional to both.

So, an obvious ideal 'scientific' solution would be to analyse the 'echo' and exactly calculate the conductivity and permeability of the material, look up a table and tell the user what is there. To do this we could use scientific models of sorts. I did say ideal ...... and this is approximately what analogue systems try to do.

The key advantage of digital systems over analogue ones (I have been part of a revolution in one industry in a former life....) is their adaptability. Different and non linear strategies can be thrown at a problem which is difficult to do in the analogue domain. So the design needs to have that flexibility for digital control of key parameters (I think you already know this).

The key to this is to make sure that the information of the waveform is correctly recorded in the digital domain with a minimum of processing/ distortion so that the advantage of adaptive strategies can be thrown at the problem. This is especially true for a development system. (From posts I think you also know this).

You are right ... there are many benchtop detector designs that use an oscilloscope as the main user display. LOL

With those comments I make the following points for discussion:

1) Analogue integration (and differentiation) of a signal do provide information about it but with an accurate digital representation these can be done digitally. Frequently filtering and other analogue processing is used to overcome sampling limitations or as an analogue technique to process to the answer of (1) (2) (3) (4).

You mentioned three possible sampling regimes:

2) Looking at the signals in the waveforms I have seen where the target information is contained in a signal typically 5 uS to 25uS long the sampling rate should be at least in the mega sample per second range. Note that Midas' scope is set at the 75 MSPS range so imagine what we 'see' in the detector in comparison.

3) Processing and sampling at key points in the waveform is (in my view) pandering to an analogue strategy and view of the world.

4) High accuracy (ie 24 bit) ADC may not be the best strategy here. Less is more, that is higher sample rates would seem to reconstruct the waveform information better. The serial clock rate of a serial ADC is limiting on the sample rate and it is unlikely that an adequately high sampling rate can come from such a chip. A high accuracy waveform might help reconstruct some of the ringing (artifacts??) I can see in Midas' scope but do you want these other than to improve the electrical integrity/ coil design? To see these too you need high sample rates.

5) High accuracy and sample rate (and cost) are trade-offs in ADCs. While it is great to have both, we should particularly question how sample rate and resolution helps us in our quest for (1) (2) (3). Key to this is to ask questions like whether the small ringing perturbations shown in (for example) in Midas' scope add information to (1) (2) (3) or are something else that we should ignore. I don't know the answer to this (they could be undiscovered/ unrecognised 'R&D' info that helps us) but suspect they are tuning artifacts.

There is lots of stuff in the signal that is not known in the art .... DSP is revealing it ... again it depends if you sample pre or post demodulation

6) Thus the 1MSPS 10 bit ADC built into the PIC might be a better choice than applying sample and hold/ integration/ analogue type techniques. You can augment the accuracy with some kind of programmable amplifier and adaptively change the gain depending upon the echo to maximise it while avoiding clipping or saturation. That is start out looking for the weak target and then adaptively drop gain to better determine what you have if it clips.

The UNIPI samples at 12 bits on chip and integrates with SAT function ...sensitivty is adequate for basic to medium performance. On the other hand I have some AD7760 chips and they will do 2.5 MSPS at 24 bits ....

7) Maybe consider strategies to do better than the 1MSPS rate in the PIC although getting such high speed data into the PIC may be difficult (20MHz max serial clock???).

There are faster CPU chips but $20 of programmable logic allows you to keep your favourite microprocessor.

If the PIC (and its ADC or the ability to get high sample rate data) is the limitation then an alternative processor with faster ADC (if such a beast exists) might be warranted.

I know development is a lot of effort so I hope these comments provide a review point and stimulate some discussion. My key comment is that of your 3 directions for sampling you need to place more weight on sample rate even if that costs you resolution. Said another way, if you hope to do good freq domain processing you need good freq sampling relative to the signal to be analysed.

Absolutely right ...

Please don't stone me if I have been naive! I really hope my comments are useful.



Chudster

with apologies to Monty Python ...
MANDY: Ohh, I hate wearing these beards.
BRIAN: Why aren't women allowed go to stonings, Mum?
MANDY: It's written. That's why.
HARRY THE HAGGLER: Pssst! Beard, madam?
DONKEY OWNER: Oh, look. I haven't got time to go to no stonings. He's not well again.
hee-haw hee-haw
HARRY THE HAGGLER: Stones, sir? MANDY: Naah. They've got a lot there, lying around on the ground.
HARRY THE HAGGLER: Oh, not like these, sir. Look at this. Feel the quality of that. That's craftsmanship, sir.
MANDY: Hmmm. Aah, all right. We'll have, uh, two with points and... a big flat one.
BRIAN: Could I have a flat one, Mum?
MANDY: Shh!
BRIAN: Sorry. Dad.
MANDY: Ehh, all right. Two points, ah, two flats, and a packet of gravel.
HARRY THE HAGGLER: Packet of gravel. Should be a good one this afternoon.
MANDY: Hehh?
HARRY THE HAGGLER: Local boy.
MANDY: Oh, good.
HARRY THE HAGGLER: Enjoy yourselves.

[moodz]

Hello to everyone. My thoughs... (just saw this design.) I have been using the that1510/1512 for 3 years now. Very quiet chip intended for audio appl. and not DC amplification, with some major flaws. Extremely temperature sensitive! I have mounted mine on a large heatsink to make it temperature "steady/stable". Offset on this thing is close to 40uV/C. Datasheet does not mention these cause of audio appl in mind. That means do not amplify a lot, or you get close to supply rails and saturate the output...or if oscil/on. occurs, now you know why!!!

The THAT amplifier is used for the following reasons ..

1. The bandwidth is over 3 Mhz at a gain of 1000 and very low noise ..... The layout and decoupling is absolutely critical as this is effectively an RF amplifier. I regularly run it at a gain of 60 db No Problemo.
2. It is a true differential amplifier and thus suited to the differential PI experiments for effective common mode rejection.
3. The amp is DC coupled at the input because it has to be ( read the data sheet very carefully ) though it is not being used as a DC amplifier ... PIs dont work that way.
4. The output is AC coupled to the ADC ... no point digitising DC offsets.
5. Temperature drift and DC offset ( less than 10 uv ) are solved by utilising a servo loop ... see below ... however I have never found this to be a problem.

[straton]

Hmm...servo loop is a "partial" solution since it will introduce some noise. Need careful design/layout etc. Temp drift according to my tests on 10 of these came out from 16uV/C up to 32uV/C. 60db of gain is typical but how about you use it with a follow up amplification so the total is 200KX to 500KX and let's see how well it works. We have to be realistic, 60db of gain is going to get you below average sensitivity. Again, I am not saying that this is a bad amp since I am using it too!!! but didn't agree with the " major PI sensitivity improvement " quote.

[Davor]

So why don't you build your own CFIA with discrete transistors and op amps? Unlike a THAT configuration that has unbal output you could make it double balanced, just like the THAT's inspiration - Cohen preamp had. You want to use it for AD converter, and those are happiest with balanced signal ... even a bit of bias would not bother you any more

Alternatively if you insist on THAT solutions, try THAT1570. It is basically 1510 without the output differential amp. The only downside is 0dB common mode gain (no cross-coupled diff amps to enhance CMMR), but for supplying AD converter WITH a balanced input - that's just fine.

[straton]

Thanks for the advice but my "THAT IC" testing period, is over. Yes,I have designed my own pre amp, with excellent results but it was a bit tricky and time consuming- not to mention the high cost of buying all precision components...have to congratulate you all for the time, effort and info you're given to all.

[sido]

Until i see proven design flaws, i am open to Moodz use of the THAT Ic. Looks interesting and I am all for it until proven otherwise.

Cheers Sido

[straton]

I don't see any design flaws because I have not! built this machine but at the same time I can pretty much "see" its capabilities and where this is heading. I gave advice and my opinion about some of its "flaws" and mentioned info that the datasheet does not have! It took me plenty of hours, for example, to do the set-up for the temp. drift. It's free info.