Antilog amp = linear ground signal ???

[Anonymous]

Hi
To avoid having to do log arithmetic on a PIC uC (latency issues) would an antilog amp give a near enough linear ground signal???
If so, the arithmetic would simplify to V = -tx + k
where V is signal amplitude, t is sample time, x is the initial sample time divided by a latter, and k is the first sample amplitude.
After two samples, a 3rd can be predicted. If it fits within a pre-determined window then only ground components are being detected.
Another point. The samples would be held on integrating sample and hold amps. Would a sample width of 5uS be sufficient to achieve a good s/n ratio. I guess that's like asking how long is a piece of string when you can't see my integrator design.
Cheers
Kev.

 

[Anonymous]

Hi Kev,
I don't know about others but I can't quite get the math (I seldom do). Are you saying V is the ground signal, t is the over all sample time? I am confused by the meaning of "time" and "divided" in "t is sample time, x is the initial sample time divided by a latter". Please forgive the dumb question but a diagram would be nice at least for the likes of me.
5 usec samples aren't all that short in a pi design provided you use a short enough duty cycle and this can still be quite long.
Rob.

 

[Anonymous]

Hi Robby,
Sorry about that, it is quite ambiguous, I agree. I'm miserable at math too!
I've added a graph and you will likely recognise this from Bosnar's patent. He uses sets of sample ratios to determine if the curve fits the magnetic mineral model, or metallic one. I want to stay well away from log tables and floating point calculations, as the frequency rate will drop way down, which can only be compenstated for by higher power pulses and of course this leads to beefing up the front end.
I'm looking at V2 the centre sample amplitude. If the signal to noise ratio was very good at t3 it might be better to look at V3 and compare this with an extrapolation from v1 and V2. If you imagine a straight line drawn through V1 and V2 on the metallic response curve, and then see how far away from V3 it would be....It would be very hard to fit a magnetic response curve over a metallic one. I don't know, it all depends upon linearising the magnetic response?
It's been boring here lately, thought it was time to get thinking.....
<img src="http://img62.exs.cx/img62/9375/Graph_Bosnar.jpg" alt="" />

 

[Anonymous]

Hi,
Should have mentioned that t1 = 0
thus t2 and t3 are the elapsed time from t1
Also looks like a little floating point division will be required.
Cheers
Kev.

 

[Anonymous]

Hi Kev,
I think I get it. This idea and similar appear to be the only way of getting some form of universal GB without using high current, multiple, rectangular tx waveforms although the Infinium maybe uses a high resistance coil circuit? Anyone seen the patent?
ML and Garret are the only manufacturers offering any form of universal mineral exclusion at the moment.
Including t in the formula is confusing but I think I can grasp what he is up to.
I don't know if you have read it but patent US6586938 (Geotech patent page), column 7 in reference to fig 2 might help you a bit more. I'm afraid I'm a bit useless with the pic side as my programming ability is limited to generating all my pulses and sample timings and the only input to the device is to force it to change to another program.
I know a guy that doesn't know anything about detectors but assures me he has spent years comparing waveforms in other fields using a micro processor and he seems keen on having a go but we haven't got our heads together on this as I'm still pretty well lost in an analogue attempt at doing much the same.
Rob.

 

[Anonymous]

Hi Robby,
Thanks for taking the time and showing some interest in my proposal.
I have studied that patent you mention. Most of his solutions would require continued adjustment for varying ground composition, or background noise. His 3rd embodiment though, does have a tracking facility, but here again, logarithms rear its teadious head.
And I say tedious, because I think back, and I suspect you'd remember too Robby, the time before calculators, and before the teacher would trust you with a sliderule, the hassle of having to fumble through grubby dog eared log tables to do floating point multiplication and division etcetera .....woe is me.....isn't life short... must get busy heaps of gold to find, but first I need to tailor my detector to the conditions.
So I really want to do as much analog arithmetic as possible so that the uC can concentrate on timing and curve evaluation.
I had hoped someone may have had a good method of linearising the magnetic decay curve, and felt like sharing it, but it looks like I'll need to learn how to use Matlab, or similar and try to model these things.
Oh well
Good on ya mate!
Kev.

 

[Anonymous]

Hi Kev,
Conversely, why not linearise the largely exponential decay of metallic conductors, which leaves the ground signal as decidedly non linear. A wide band log amp for the front end does this nicely.
Eric.

 

[Anonymous]

Hi Eric,
I'm a little slow, but the tortoise always gets there in the end.
OK, log e-x roughly equals x / 2.302... a straight line, near enough.
The magnitude of the misfit will determine the metal/mineral matrix. The acceptance level of the misfit can be tuned to accommodate varying ground conditions. I'll look into it.
Cheers
Kev.