multi-point sensing

[chris]

If multi-point sensing, as used in the SD2200, appear to be the "ultimate" locating method, why is everybody so quiet about it?
As a novice, can anybody explain to me how it works???
Thanx
CHRIS

[Reg]

Re: multi-point sensing

Hi Chris,

Multi-point sensing on an SD? This is obviously one of ML's advertizing hypes for the basic techniques used. That is why everybody is so quiet about it.

Basically, the SD 2200 operates like all the SD's which are really nothing more than PI metal detectors that are just a little different in the fact they have a form of ground balance using a little different technique than that used by Eric Foster a few years before the SD's were built.

This means ML takes a couple of different samples while using two different pulse lengths to try to obtain a ground balance with fewer targets ignored, and then processes the sampled info basically like any PI.

When ground balance is obtained by subtracting a later sample from an earlier one, there will be certain targets that will be eliminated just like the ground signal is eliminated. These targets have basically the same characteristics as the ground signal.

Now, if two different pulse lengths are used and are sampled, the targets ignored in the first sample are not the same as those ignored in the second because the ground signals supposedly change with different pulse lengths. So, by combining both pulse length samples, supposedly, no targets will be missed.

Reg

[Robby_H]

Re: multi-point sensing

As Reg said but ML compares an early short sample with a later long one in a way that ignores a large number of minerals with different amplitude and decay times. This is done by firstly comparing the early long pulse sample and short pulse sample with the same later sample and the result from the two different pulse lengths is then further compared to attain one channel that subtantially ignores a very large spread of magnetic mineral effects.
With the math used the operator or machine doesn't need to compensate for a lot of changes in ground conditions whereas just subtracting a later sample from an early one leads to constant adjustments being needed and also the gains need to be changed quite a large amount between different types of ground resulting in the math changing to a much greater degree.

http://www.countermine.minelab.com/default.asp?pageid=product-info&template=content&sid=26

Gives a rough idea of the principle being used (which relies on rectangular tx waveforms) and I don't expect to ever see a hobby circuit published for this particular idea as the electronics needed for it to successfully quieten wildly varying ground are way too precise.

[Robby_H]

Re: multi-point sensing

Just to clarify a bit, the detector relies on the comparison between the result from the two different pulse lengths to give an audio response as it is at this time that it calculates the difference between a nugget and the ground signal as the decay from minerals is dependant on pulse length.
The SDs have a second channel that samples later and is less sensitive to the ground and smaller objects but otherwise the processing is the same for it.

[Sidley]

Re: multi-point sensing

Minelab, more s**t than a herd of bulls!

Quite how this company gets away with the stuff they put in their adverts is beyond belief. Someone somewhere should point out to the advertising standards people.

Intel are at it now too, with the Centrino. Making out you can broadcast from a Roman amphitheatre with no electricity. Funny, in my experience you have to be within 50m of a connection node.

Seems like Intel and Minelab should go into producing metal detectors together.. OH, no maybe not, the thing would overheat and crash for a passtime and then fall apart at the first useage ;-)

[chris]

Re: multi-point sensing

THANX A MILLION TO REG, SIDLEY AND ROBBY FOR THE DETAILED EXPLANATION ON MULTI-POINT SENSING.
IT IS GOOD TO KNOW THAT THERE ARE GUYS WHO REALLY KNOW THERE STUFF
THANX AGAIN
CHRIS

[Ismael]

Uninformed

Just to reply to some obvious retorts in this thread, if it is so easy to do the ground balancing why is it that no other detector has been able to be competitive in Australia, could it have anything to do with not being able to ground balance any non-minelab detector in our mineralised soil?

Just my 2 bobs worth...

Regards Ismael

[Reg]

HI Ismael,

I don't think anyone said it was so easy to ground balance. The question was more of how it was done and what is multipoint sensing.

As for ground balance on a PI, Corbyn had some suggestions in his design of many years ago.

Eric Foster actually designed and built the first ground balance PI available to the public in the 1980's. The machine really never caught on, so the ground balance feature was dropped.

Minelab, later refined the process and brought out the SD 2000 in the 1990's.

The fact is, it wasn't a matter that no one could develop a competitive dettector, but rather no one felt it was economically practical to develop one. A few years ago, I personally had conversations with a couple people involved with the development of new detectors for a couple of manufacturers and their feeling was the market simply wasn't big enough to justify the expense of developing a competitive high powered PI machine.

Personally, I think this is why Garrett devolped the Infinium the way it did. The present design would allow it to sort of compete with the ML's but more importantly, it would find its own nitch as a water machine.

Even today, the GS 5, Eric Foster's new PI is not designed to compete directly with the ML's, but is more of a hybrid made to be used both for gold hunting and beach hunting. Will people try to compare the GS 5 to the ML's? Yep, that will happen. It is probably not the best thing to occur, but it will.

The GS 5 has the added flexibility to be more at home on the beach as well as the in the hills. With the ability to adjust the delay, once can sort of taylor the machine to be what they want it to be.

Nobody is really knocking the ML's. They are presently the king of the hill when it comes to nugget hunting in the US and Australia. They have been at it now for 10 years or so and we have seen several design evolutions to get to where they are today. Since they have been able to continue to bring out a new and improved model, one thing is clear, the best hasn't been made yet.

Reg

[Muntari]

Multi-point sensing

Hi Reg,
It is interesting reading yours and others explainations and feelings on the ML and other PI detectors. I am glad to see that you gave some balance on the subject of ML too.
I believe that developing any product that has a small market essentially demands a preminum price,especially so when the market is looking for the features available in the product.
Like any other manufacturing sector, metal detector manufacturers worth their salt would do the market research and sell to that market. If a person wants to seriously detect for gold here in Australia, then they will need to pay serious dollars for a detector and that's pretty much it I think. ML are not the only units to choose from, it's just a matter of what one can live with in use. As to over the top claims or stretching the truth? Hmmm.... buyer beware same as any other product out there in consumer land. We have laws to protect the buyer from unsrcupulous dealings and wild claims but they must also be proven to be so too. Just my 2 bobs worth!

On a different note.
I have been detecting for about 20 years, designed and built a few detectors, including Corbyn's beast and I enjoy the challenges involved. About 18 months ago I started my own design from scratch to prove my ideas and I am still just at the second prototype revision. It will take me another 12 months to have enough data for any form of public release but hey, it's good fun for me! I have no intention of marketing a product. I just enjoy what I do.
The design is based on a TI DSP front end and a Sharp ARM module for the human interface. I am not trying to compete with anyone or anything, just building a design and working through the issues involved. My idea is based around spacially detecting and identifying a target. My colleagues say it's over the top and/or won't work....now that's like a red rag to a bull to me!! 5 years ago I would have believed them but today where DSPs are less than $20 and analog ICs with noise figures down around 1nV and a 6 layer PCB can be tooled for sub $700.. .I tend to think I have a chance to prove them wrong as the cost is all but taken out of the equation.
Noise has been my biggest problem. I expected it and thought I had it mostly licked but another board layout will still be required (DSP clk track layout). I will post to the forum when I am satisfied that the design is (1) stable (2) usable by others. I was thinking of puting together a website for such a design and maybe open source it but I do not have the time just yet. Hardware has cost me around $800 in 1 off qtys so I am happy with that. The main aim is to keep the hardware design modular and easily programmable.

Thanks for your posts Reg, I enjoy reading them.

Regards

Brian K

[Largesarge]

ML's MPS is similar in a way to radars MTI (Moving Target Iindicator.) In MTI any target that can't get out of it's own way is canceled out. They do this by splitting the return period in two and putting one thru a delay line and then invert it. When the next returns comes in the undelayed signal is compared to the delayed signal of the first return period. Because a positive will cancel a negative any targets that appear in the same place are canceled. Any target that has moved will not be canceled and passes on to be amplified and displayed.


ML does virtually the same thing but within the same overall sampling period.
Anybody could do it and wouldn't be that expensive.

[_Raz_]

ML does virtually the same thing but within the same overall sampling period.

Actually, you need to do almost the opposite.

Since the ground response varies according to pulse duration, an MPS system needs to eliminate the changing component of the response and retain the non-changing component.

I guess the procedure isn't too different, as once you have the isolated the ground response, it would be fairly simple to subtract that from the original, leaving you with the desired (non-changing) component.

[_Raz_]

About 18 months ago I started my own design from scratch to prove my ideas and I am still just at the second prototype revision.

Hi Muntari,

I was very interested to read about your research, as I have been doing something similar here, albeit on a simpler scale. A friend and I designed and built a couple of prototype machines, using a simple 2 channel ground balance system, using an Atmel MPU for timing purposes, but otherwise totally analog.

Results were excellent on the ground and beachs down here where I live (Mornington Peninsula), but very disappointing in the goldfields, the ground was just too mineralised.

We have since retrieved a large(ish) soil sample from the goldfields and are currently designing a better detector using 6 receive channels.

Like yourself, we are only doing it for fun, and perhaps to learn a thing or two along the way. Anyway, it keeps us busy.

Raz

[Reg]

Hi Brian,

Your project sounds interesting. Good luck with it.

I have built a ground balance system couple of years ago that works quite well any place I have tried it, including many areas in AZ and Colorado where gold is found.

Now, I have avoided using a uprocessor because of the noise factor. Since I have been trying to obtain a very high level of sensitivity plus a very short delay (6 usce to 8 usec and be able to alter the delay on extremely short increments), the uprocessor didn't seem the way to go. Unless scyncronized, it is all too easy to introduce noise with a uprocessor.

As for general noise, using the subtract method for GB introduces noise, simply because of the increased sensitivity necessary to obtain the required level of signal for subtraction. Obviously, amplifiying the signal also amplifies the noise.

Now, I am using a technique whereby I limit the subtract signal to the level just necessary for by GB. This limits the noise also, Since noise is of short duration, a signal of greater voltage is needed to generate a strong signal, thus limiting it has a greater effect on the noise than on the subtract signal itself.

Anyway, I also shorten the time frame between taking the main signal and the subtract signal. This will have a negative effect on overall sensitivity since some of the gold signals will also be reduced along with the ground. However, this appears to be offset by the lack of noise, so hearing the weak signals is much easier.

Now getting back to ML, this company has gotten a really bad reputation here in the US because of its advertizing hype. You have to realize, we went through similar hype many years ago with American manufacturers, so it was like a rerun for us. It was the type of rerun that we didn't want to see. (BTW, I have been at this detecting thing since the late 1960's so I have pretty well seen it all when it came to advertizing hype.)

As an example of what bothered people was this, when ML claimed the SD's would go 3 times deeper than a VLF, people were very quick to test this claim and found out it wasn't true, or wasn't true in any of the places they tried. In fact, in many tests, it was hard to see much of a difference at all. Such advertizing did little for ML's reputation, and that will take quite a while to change.

The patent hype is another factor. ML appears to be in a patent frenzy. Patenting the use of Litz wire is a joke and, like most people, I serously doubt such a patent would hold up in court. Yes, American manufacturers patented a lot of things in the early years, but never really tried to enforce them. It was more of an advertizing thing, with, in my opinion, Garrett being the worst.

ML, on the other hand has tried to bully others. I recall a discussion with an engineer from another company who nform me the ML threatened to take them to court for using a uprocessor, claiming they were infringing on a ML patent. This didn't go anywhere in the long run but did cause a few hard feelings.

As for specific patents, lets take the fact that pulse duration alters the ground signal, something uses in their PI patent. Well, take a look at the first patent to state that basic fact. No, it wasn't a ML patent but one by George Payne in 1978. Actually, George's patent didn't discuss signal duration as much but simply stated that pulsing the ground with a negative response and different size pulses would alter the ground signal because of the residual effects because of pulse polarity and duration. So, when one expands the concept of the earier patent and the intent, the pulse duration change would be implied plus the use of a negative pulse. At least, it does in my opinion.

The point I am trying to make is patents for metal detectors are nice, but with enough money and time, most can be challenged by previous patents.

Getting back to ground balancing, there are many techniques that can be used even when using a subtract signal. As an example, one could take a much later sample, say at 40 usec and subtract it. If more subract signal is needed, there are multiple ways this can be done. One can simply amplify the signal more, one can shift the sample time to a shorter time such as 30 usec, one can shift the primary sample time to a later time such as from 10 usec to 15 usec, or one could simply reduce the primary signal gain to meet the requirements. Thus, any of the above techniques will work. (Now, the above sentence should give some of those trying to design a ground balance system some different ideas.

The big technique being tried now is to use a uprocessor and sample and hold the ground signal in memory and introduce it as needed.

Here is a little simpler method has not been discussed as far as I know. Simply take a constant voltage and amplify it as necessary to cancel the ground signal. The key is to design the gain curve of the ground balance circuitry signal to correspond with the ground signal curve. This technique should minimize the ground response pluse minimize any noise characteristics common with the more normal methods presently used. With this method no second sample or memory of the ground signal is necessary.

Reg

[Muntari]

Hi Reg,
Thanks for the post.
Yes I do agree with your point on the hype wagon which was similar here in the 80s.
It was like a feeding frenzy until people wised up to the hard work involved in finding gold. I can also see why US consumers would be wary,it must have been pretty off-puting. ML I am definitely not defending. I certainly don't like the Patent pushing that goes on these days either, just glad to see a little balance on subjects else it tends to get way off beam. That's what I like about this type of forum too. It has a way of correcting myths, exposing frauds and showing facts,simply by the sheer number and variation of interests of people from all walks of life who read and contribute, understand and ask questions....all good stuff.

Your ideas and experiments are very interesting and somewhat along the lines of what I am trying to achieve using digital techniques. I liked Corbyn's approach to ground balancing but thought I could improve the application of it. I use lookup tables and last known target information 'bins' as I like to call them. The ground signal is sampled at different intervals and averaged then held in memory until a dynamically set trip point is exceeded (just above the system noise floor). This happens concurrently on 4 channels at different sampling rates and if after averaging, the content of each (after some maths)is within a preset acceptance range which is also set dynamically, this data becomes the ground signal threshhold which is then excluded from future samples in the stored target bins (using 2 different algorithms). While this is going on, another 2 channels are working on the earlier target bin data (pipelined data) and a further 3 channels of ADC are collecting search coil position,(x,y,z) and velocity (ADXL2... accelerometer and Gyro mounted up the shaft). I am looking at spacially detecting an object, it's possible size and it's possible shape. Sounds a little bizzare I know but basically, if you can determin where your coil is relative to the ground and target (sort of dead reckoning) and if this sampling is synchronized to a variable frequency/duty cycle transmit pulse train quite a bit of interesting data can be obtained. Still early days but it's looking ok for now. I can for instance, log the positon relative to the ground left to right and back again (swing action) and log any target signal and where it occurs within that swing. I can tell how wide a target appears to the coil as I have velocity information as well so it does not matter if you swing fast or slow (within available ranges of course) Everthing is relative. I can't yet tell how deep a target is but I can graph it's relative signal strength. This code is not yet optimized however I am working on it. It is far from a perfect method and perhaps a little over the top but I continue to have fun doing it and as long as I am enjoying doing it I shall push on. As you would be aware, temperature drift and self-induced circuit noise contribute a large portion to the noise floor in signal paths and I liked the idea of using a very short analog signal path consisting of a low noise differential front end, controlled gain stage ( I use a VCA8613 from TI here),then oversampling ADC with decimation. At least this way I have a method to spread the noise and recover low level signals. So far I have been able to produce an acceptable front end using this method. I missed a couple of things on the board design around the DSP Clock lines and also the 1.8v core supply regulator bypass capacitor. Simple mistakes but when you moonlight these things it's bound to happen! Luckily I have a friend in Western Australia who has recently decided to help with the PCB side of things and wised me up to my now obvious mistakes! As a PCB designer, I make a good plumber! At least now I can concentrate on circuit schematics and algorithms knowing the board layout is in good hands.

Anyway must go. Waffled on enough.
I will perhaps tidy up my flow charts and post to the forum it might make more sense...just need time to do it!

Cheers and good luck with your methods too Reg.

Brian

[Muntari]

Hi Raz,

Your project sounds interesting.
Good to hear someone in Victoria is designing and experimenting!
Used to live in Latrobe Valley and not far from the Thompson River and a place called Tanjil Bren was our favourite gold hunting area. I think it is all catchment area for the Thompson River dam these days...Long way from Mornington Peninsula I know but closer than me these days...If you can still get up there might be a good little spot. It's up the back of Moe. Found some small nuggets on a trip some time ago.
With your design, are you using a differential front end and DD coils or another arrangement? How do you find the Atmel MPUs?

Keep up the fun factor and good luck with your design.

Regards

Brian

(Muntari)