A Ground Balance Circuit (build and field tested)

[Tim]

Hi digger. Did you do the standard MPP or the KRinAZ version? I have the board here, and was planning on the KRinAZ version. It would be interesting to know how your mod goes on it.
And, yes please, a board pic with descriptions would be helpful for us technologically retarded types.

[Digger429]

For be it from me to suggest you follow the exact method I used, because my MPP is a tangle of prototype modifications and changes including several that KRinAZ mentioned. I've recently even bought a new box so I can completely redesign the mess it is currently in. But since you asked, I'll show.

First the outside, showing knobs and the (see different thread) small box halfway down the shaft with the preamplifier. I've tried various coils, but the one I have at the moment is a spiral wound (from outside to inside, like a vinyl record) 250 uH with no shielding. I tried shielding in the past for other coils but didnt see any benefit, only reduced performance. The added Ground Balance circuit is put in the only available space I had left being directly under the lid. The GB circuit has two controls - the gain control that I will eventually remove and replace with an on/off switch, and the GB width control (on the box side, with no plastic knob).



Inside the box you can see where I have made the new Veroboard based GB circuit and put it under the lid. It only has the two MC14538 chips plus associated resistors and capacitors, and also a header socket for the wires for power and signal in/outputs.



The top and bottom of the GB circuit board. It is two or three times as large as it really needs to be:



The underside of the MPP circuit board, showing how I attached the two new FETs in parallel with the existing ones:



There also a lot to ignore. I havn't found any great difference in performance with Main Sample Pulse rate or width, and the DIN plug just has a bunch of signals (eg. test points) attached to it for ease of access. I'll get rid of all those when I transfer the mess to a fresh new box. Also, the onboard speaker wasnt loud enough, so I have a bigger one temporarily attached via the headphone socket. With any luck, I can also transfer the FM radio transmitter to be inside the box.

Also, still haven't found any gold yet, so don't go thinking this GB circuit is a miracle worker. The last two times I went out in the field were firstly to Ballarat (Victoria, Australia) where I was only there for an hour or so before putting the MPP back in the car because a friend had rented a Minelab machine. The mineralisation was so low I didn't even use the GB circuit. Then secondly to a place near Wedderburn where the mineralisation was so high I had the GB circuit on full. That reduces the performance a lot, and it was a stinking hot day so again I was only detecting for about an hour or so.

---

Edit - Nothing to do with ground balance, but heres some quick air tests a few weeks ago with a Fisher Gold Bug Pro (5 inch coil) and a Minelab GPX 5000 (large coil, perhaps 15 inches? in size) against my 8 inch coil...

0.2g Aluminium : GoldBug = 9cm, MPP = 11cm, Minelab = 17cm

10mm square Aluminium : GoldBug = 10.5cm, MPP = 11cm, Minelab = 14cm

Maybe a lot of the performance can be attributed to coil size, the MPP performs well in terms of distance per dollar spent.

[Tim]

Great stuff digger. Geez, she's a beast! The pics have really helped me understand your processes.
I have 10 Hammerhead boards here, and once I get one functioning, I will try and add a 3rd sample on a joey board. This will be my first foray into using my oscope. I bought a Hantek 6022BE 18 months ago, and haven't had it out the box.
Yeah, Wedderburn's pretty harsh ground. However I found specks in conglomerate out at Beggary Hills first trip out with my GMT E series. I have an 11"Coiltek and a 10x5 Commander for my SMpro-that 11 thumps 14 inches on my 1.7g flour gold sample on an air test, and I am yet to optimise damping. The soil here at Morgan (although no gold) is also quite harsh, so good for experimenting. Heck, even my Baracuda with the home-made 10x5 mono will see it at 10 inches. Just luck I s'pose.
I asked for and got the info here (thanks everyone) to put a Vsat on the SMpro, which may also help ride the harsh ground. I like to have all the adjustments, including an audible threshold, at hand.

Out of interest there is detectors on Ebay presently called TX850 that are a DIRECT copy of the GBpro with the 12x8(?) coil, for $279.
As far as ML's, I can only afford an XT17000 (only the box), but I put a new Coiltek 6" on that. Considering an F1A4 (at the right price) to mod myself also.

[6666]

I've put a project on the back burner untill the new year but it sort of goes like this , leave the 555 inplace and remove the sample timing chips and replace with small plug in micro to do the sample timing , thats if I can crack the code to track the 555 , if I can get that to work then it might be possible to produce the 4 sample pulses to switch the 2 fets, problem is the integrator design is not ideal for GB, anyhow will try and give it a go in the next few weeks, I aint no expert at writing code.

[6666]

Been tweaking a little bit of code, it seems to track 555 pulse, but I will not be able to devote more time to it for now.

[Digger429]

Keep going 6666, a micro is the way to go given that about 50% of the circuitry inside the MPP is merely generating signal pulses.
If I was to invent my own metal detector circuit I'd scrap all those 555's and 14538's and use a micro of some sort to generate them.

My only concern is the speed of the micro not being able to generate accurate pulses to within the microseconds time frame currently used. eg. If I have a 6uS main sample delay pulse, that is very consistent and stable with analog circuitry, but if use a micro, is that 6uS going to be consistent or will it vary a lot eg. between 5us and 7uS? And if so, does it matter?

Edit - just did a count. With my GB circuited MPP there are 10 circuit blocks that generate signal pulses, and 7 bits of circuitry that dont. 59%.

[waltr]

If the code is well written the timing will not vary. There is code (Carls's HH II) on the forum here that is well written and I posted the PIC16F88 code I am using on an HH detector (see the HH II forum, "My take on the HH2" thread). The timing is rock solid. The PIC's timing has 200nsec resolution.

Yes, a processor eliminates a lot of circuits and makes timing exact, doesn't drift, and allows changing timing pretty easy.

I am planning on experimenting with the 'triple' sampling to do GB. I have code written to do the GB but have not yet tested it.
Just got the complete detector build this past weekend and only spent a few minutes outside since it was -14°C. I want to have a reasonable base-line without GB and knowing the ground is causing an issue before trying the GB code.

[Tim]

waltr I am planning a third sample also for the HH1 RevD. My thoughts are that although micro's save space, and are interesting to learn, I would rather have a fully analog circuit. It seems to me that technology is relied upon too much in this modern world, and we are leaving ourselves open for disastrous failure and dependence on others. Held to ransom by technology, so to speak. SMD and micro's....Na. It is far better to intimately understand and implement the basic building blocks that lead to these "improvements", and have something that is "hands on". Miniaturisation has its place. But it is not the be all and end all. Quite the contrary. There is a good reason why vinyl records and electric cars are making a comeback. CD's are prone to disastrous failure, and we are held to ransom by fossil fuel engined cars. Ahhh..technological advance...(?). Remember the old engineers adage - KISS. Beauty and functionality in design will ace the premature introduction of technology every time. Indeed some roads taken cannot be reversed, and that in itself can lead to arriving at the wrong destination.

[Chiv]

You been swimming in the Murry Tim and swallowed some of that water . wether its didgi or analog every thing has a cycle life. just have to learn about it to repair it I guess, been reading this post because the mineral sand here plays havoc with pi, if you can just write a code/script to help the detector then this must be the way to go
. Being able to right a code to make something do exactly as you want has a lot of advantages (quicker and less parts to replace) just to change parameters, even minlab has realized that I think as you can down load source code for gpz and ctx.

[Tim]

I do NOT appreciate your personal remarks. I no longer contribute at PA because of this mentality.

[Chiv]

Sorry that I a fended you Tim, it was just tongue and cheek, and I do see where your coming from as regards to analog to digital. end of the day consumer makes the market as they say

[moodz]

...theres a complete PI controller chip in this thread http://www.geotech1.com/forums/showt...760#post235760 the firmware is there and a phone/tablet app for adjusting all the sample pulses and tx pulses.
MAGPI_SOAP.pdf

and firmware for chip is here ...

http://www.geotech1.com/forums/attac...4&d=1469107796

[Chiv]

Thanks Moodz for that

[Digger429]

As has been mentioned in various threads, one issue found with GB circuits in PI metal detectors is that you can get a hole in the response, depending on the object being detected. The signal created by the GB circuit with the integrator can drive the signal LOW instead of HIGH (as it does normally) making the metal detector go quiet when over metal instead of going beep.

A solution often mentioned is an absolute value amplifier, so signals going negative will be reversed and made to go positive as per normal operation. In the MPP, it would be placed after the first integrator (after TP4) but before the second integrator where the comparison to the threshold value is determined. So I built one... but its not a simple answer.

In the standard kit, the signal out of the integrator, TP4, is somewhere near zero when no metal is present, and continually rises as a larger chunk of metal is detected:


Case1: A small amount of Ground Balance effect.
If Ground Balance is switched on but set to a low level, the curve out of TP4 is basically the same except for very large or close objects, when the voltage starts to lower. In other words, an absolute value amplifier is useless because the detected signal is always above the no sample resting baseline.


Case2: A medium amount of Ground Balance effect.
With more Ground Balance, the drooping signal with larger objects is a larger proportion of the curve. The problem here is that the curve passes back down through the level at which there is no sample, rendering the detected signal quiet again regardless of what kind of amplifier you have. With an absolute value amplifier in circuit (dotted line) you get some signal back with larger objects, but its a hell of a thing to try and calibrate a sensible resting threshold value for.


Case3: A large amount of Ground Balance effect.
With the dials turned all the way up (large GB pulse width and small main sample pulse width) the curve is the flipped reverse of where we started, the signal increasingly negative going with larger samples being detected.

In theory this would be the best case situation because an absolute value amplifier would just flip it back to how the circuit normally operates. In practice it fails because it requires the resting (no signal) value of TP4 to be EXACTLY zero, and also have the absolute value amplifier to have no crossover distortion at zero volts. Neither of these are true in the real world.
Consider a MPP kit setup where TP4 reads slightly positive volts with no sample being detected, and this large amount of GB signal inserted. The chart is shown below with the dotted line representing the output from the absolute value amplifier. Because the curve crossed down then back up through the no signal level, the metal detector is unreliably useless (blue shaded area) with small targets.


The same problem occurs with a MPP kit setup where TP4 reads slightly negative volts with no sample being detected, because then the machine has the exact same problem when the GB level is low, as per Case 1.

So whats the answer? Buggered if I know.

[waltr]

So whats the answer? Buggered if I know.

No really good simple answer. However, read up on using the White's TDI.
In addition to an Absolute Amp (or rather a full-wave rectifier) the VCO is a 'dual tone' and uses the output of the SAT stage.

So basically- with your Case2&3 the sound goes to a Low audio Tone followed by a High audio tone on some targets, ones that reverse the Integrator output polarity or is above the GB Hole. There is still the 'hole' for some targets at some GB setting. By using the SAT output 'no motion' returns the signal to a 'preset' value (zero).

In the TDI user's forums/manual- the operator needs to listen to how the Tones go: high-low or low-high along with GB setting.
So it is the poerator's Ear and experience that 'discriminates' target types, size.

I have been experimenting with GB sampling and the responses obtained in this thread:
http://www.geotech1.com/forums/showt...228#post239228

Much of what I have reported is common knowledge to many others here and in many other threads.
It is from reading ALL these other threads that I have methods to explore. I do have the three sample GB code working but not a Dual-tone VCO except on protoboards.
That is the current project.
Also have other ideas I read about the I will be exploring.