Ground Balance Theory

[Tepco]

Main issue with op-amp\diode based log-amp is not log-linearity, but open loop BW limitation of amplifier used, limiting it's dynamic range. Tried using LM318 and diodes at 32kHz and get stucked. Multistage designs are complex and usually high precision components are needed (tried to rip-off design from old portable Tek spectrum analyzer time ago, still have hardware, worked). Nothing can be achieved without some silicone. Remembering AD606 obtained for 125E (around 160$, over ten years ago!), i just dismissed any idea to use something similar for hobby. Today, useful chip like AD8307 is 12.5E (reichelt.de), acceptable for design. Just to figure out how to use it optimally, rules are different, band limiting, no need for DC coupling, extreme dynamic range to deal with, chips are log-amp\detectors giving absolute values... If you think you have no problems at all in your design, just add one, situation will change dramatically. I get used to spectrum analyzer log display, but observing time domain process on scope this way is something one have to get used to...



What i'm doing here is to just try to utilize log-amp to extract some parameters from test setup, but for actual use in MD one criterion must be met: it most solve more problems compared to what it can produce, so i'm far from strongly advocating log-amps in actual MD design. May be changed...

[Altra]

The log-amp idea seems to have potential, maybe not practical. But worth investigating. I found LOG102s for sale on ebay US, for $4.00/ea. Not sure if they are suitable? There is a data sheet and spice model on the link below. They appear to output the log of the input current.



http://www.ti.com/product/log102

[Ferric Toes]

What I hope to try is a VCA810 in the feedback of a normal front end (LME49990, or even a NE5534). Scroll down on the 810 datasheet and you will see a wide band log amp in the applications. I have a few VCA610, which is the earlier version, that I bought on ebay a while back. Just waiting to find them as they are hidden away somewhere in one of my many parts boxes.

http://www.ti.com/lit/ds/sbos275f/sbos275f.pdf


Eric.

[Davor]

Main issue with op-amp\diode based log-amp is not log-linearity, but open loop BW limitation of amplifier used, limiting it's dynamic range.

My point is that you don't need as much bandwidth anyway, as it gets buried in noise - true log amp or not. As a benefit of this setup you gain access to sampling a few us sooner. Besides, opamp/diode setup is easy to model in spice, so knowing your target dynamic range you can adjust everything else accordingly.

[Aziz]

A banned person wants to know something from Eric. I have tried to post the answer below in that forum but there are problems right now. The answer might interest some people here too so I post it here.

[quote author=Doug link=topic=6529.msg28087#msg28087 date=1366109592]
Can anyone suggest a reason as to why Eric (on geotech) wants to try a logarithmic op amp("VCA810 in the feedback of a normal front end (LME49990, or even a NE5534)” in an experimental? Pi. Pi.Would this make it easier to implement new GB methods or algorithms?
Today, 05:52 AM #228
http://www.geotech1.com/forums/showt...Theory/page10&
doug ::419::
[/quote]


Well,
I don't see a benefit of using a log-amp. That's the opposite.
It only will show you, whether a target has a single time constant (a straight line in the log/lin graph). The derivation of it ( d(log(f))/dt) will reveal the time constant.
Proof:
Target response = f(t) = A*e^(-t/TC)
d log(f(t))/dt = d log(A*e^(-t/"TC")) / dt = -1/TC
(see http://www.wolframalpha.com/input/?i...%22%29%29%2Fdt)
And if you make the "-1/x" - operator to the result you get the TC.


Well, we would need two sampling windows to make the GB. If we want to cancel the static magnetic field/EMI cancelling too, we would need additional a late sampling window. Total 3 sampling windows can make a good GB.
There is no need for a log-amp.

Aziz

[Ferric Toes]

A banned person wants to know something from Eric. I have tried to post the answer below in that forum but there are problems right now. The answer might interest some people here too so I post it here.

Well,
I don't see a benefit of using a log-amp. That's the opposite.
It only will show you, whether a target has a single time constant (a straight line in the log/lin graph). The derivation of it ( d(log(f))/dt) will reveal the time constant.

Well, we would need two sampling windows to make the GB. If we want to cancel the static magnetic field/EMI cancelling too, we would need additional a late sampling window. Total 3 sampling windows can make a good GB.
There is no need for a log-amp.

Aziz

Aziz,
I must say that I dislike this cross forum posting! If Doug breaches the guidelines here and gets himself banned, then that is his problem. I will not post on his forum because of all the verbal battles and anti Minelab propaganda that is generated there. If you choose to post on both forums then that is your prerogative. However, you are wrong in your assessment that a log amp has no benefit. All the clues are there in what I have posted, but so far you appear to have missed them. That is the value of doing practical experiments and looking at real signals on a scope. Maths and calculations can only take you so far, then you have to get a chunk of ironstone, pulse it, look at the result and see what the maths has missed. I am certainly going to delve further into log amps and encourage others to do the same, if they are so inclined.

Eric.

[Aziz]

Aziz,
I must say that I dislike this cross forum posting! If Doug breaches the guidelines here and gets himself banned, then that is his problem. I will not post on his forum because of all the verbal battles and anti Minelab propaganda that is generated there. If you choose to post on both forums then that is your prerogative. However, you are wrong in your assessment that a log amp has no benefit. All the clues are there in what I have posted, but so far you appear to have missed them. That is the value of doing practical experiments and looking at real signals on a scope. Maths and calculations can only take you so far, then you have to get a chunk of ironstone, pulse it, look at the result and see what the maths has missed. I am certainly going to delve further into log amps and encourage others to do the same, if they are so inclined.

Eric.

Hi Eric,

but you like to lurk regularly there. Don't you?


BTW, Doug didn't deserve to be banned. UFox ok, he really did deserve it as he is regularly discrediting us in other forums too. I would like the admin to unban Doug.

Regarding the log-amp:
I stand by my opinion.

Cheers,
Aziz

[Ferric Toes]

Hi Eric,

but you like to lurk regularly there. Don't you?

quote Wiki - Lurkers make up a large proportion of all users in online communities. Lurking allows users to learn the conventions of an online community before they actively participate. Lurkers are referred to using many different names, including browsers, read-only participants, non-public participants, legitimate peripheral participants, or vicarious learners. No problem with that surely?

Regarding the log-amp:
I stand by my opinion.

Cheers,
Aziz

Each to his own.

Eric.

[Altra]

What I hope to try is a VCA810 in the feedback of a normal front end (LME49990, or even a NE5534). Scroll down on the 810 datasheet and you will see a wide band log amp in the applications. I have a few VCA610, which is the earlier version, that I bought on ebay a while back. Just waiting to find them as they are hidden away somewhere in one of my many parts boxes.

http://www.ti.com/lit/ds/sbos275f/sbos275f.pdf


Eric.

Thanks for the lead.

I just ordered a few of the VCA810s. They look very useful. I'm looking into ferrous/nonferrous
disc, which nobody seems to care about anymore. A lot of energy has gone into GB methods for a
couple of remote hot spots in the world. The Holy Grail for PI in the other 95% of the world is reliable disc.

[Ferric Toes]

Thanks for the lead.

I just ordered a few of the VCA810s. They look very useful. I'm looking into ferrous/nonferrous
disc, which nobody seems to care about anymore. A lot of energy has gone into GB methods for a
couple of remote hot spots in the world. The Holy Grail for PI in the other 95% of the world is reliable disc.

Agreed. Reliable disc in PI that works in any soil would revolutionise the detecting world. However, even in medium mineralised soil many PI mine detectors struggle to perform adequately, so you have to factor in GB as well. What happens in a country that is remote for many of us may seem not to be relevant. But if a detector design works well there, it stands a better chance of doing a good job in any other country. Luckily, the 1/t decay for magnetic non-conductive ground applies everywhere and the Oz material gives much more signal to work with when doing development and testing. I can set up a GB circuit accurately with a 0.5kg lump of ironstone, whereas I would need a bucketful of a weaker soil.

Keep us informed of how you get on with the log amp. I must find my VCA610s that are hiding somewhere. These are 8pin DIL and easy to experiment with. I see no reason not to tailor the gain of the main amp to prevent the huge amount of amplified noise that you get with no target signal. Or, run the amp linearly until a certain threshold has been reached and then switch in the log circuit.

The other benefit of a log amplifier is that it compresses the target signal right at the front end and reduces the large range of signal that has to be processed by the following circuitry.

Eric.

[Aziz]

Bugger me!

Hi guys,

you are getting huge complexity with the log-amp re the GB issue.
Try to solve the integral of the log(signal responses):
Integral log (Exponential induction decay + Magnetic induction decay + Const signal) dt =
Integral ( log( A*e^(-t/"TC") + B/t + C) ) dt =

(If you are doing the integration windows as usual in a PI configuration of course.)

And you get incredible noise issue and temperature dependency too. The temp. compensated log-amps do cost a fortune.

I'm not kidding. It makes the task very complex.
But keep on doing this as we can see some nice log scope shots.

Aziz

[Tepco]

Variable pulse TX

My intention is not to dissolve this thread by posting pointless stuff, this well may go to off-topic or somewhere else, not directly related to ground balance, but indirectly can be.


Sorry for hand writing, this is from old scrapbook, i'm not too good for making documentation, but can be interesting. Built long time ago just for waveform observation, now something similar rebuilt to take some measurements, very interesting for experimentation.


Circuit TX part is based on constant peak current control of fixed inductance coil, turning it off when preset value is reached. Now just varying drive voltage, pulse width can be varied over very wide range, 50:1 or 100:1 is easily achievable, keeping energy stored in the coil and magnetic field during this time constant. Energy is 1\2 L x I sq. considering I is fixed, resultant pulse width needed to ramp up to predetermined value is I\ (1\L) x Vcoil. Circuit is built using UC3844 SMPS chip, driving power mosfet or IGBT. R\C constant at pin 4 sets operating freq. around 4.2kHz with given component values, but adjustable over wide range. Pin 3 is current sense input, with precision 1V comparator, sensing voltage across resistor from mosfet source to gnd to monitor peak current. Without coil or coil voltage applied, will produce 50% square wave drive, but when coil current ramp up and pin 3 voltage reaches 1V cycle is terminated. In this case, 0,5R resistance gives 2A peak.


This on picture is “high voltage” version built time ago, present setup is bit different, intended for lower voltage operation for convenience. Coil is 8in speaker wire flat spiral, about 160uH, 25 turns, connected using ordinary wire (same as coil is made of) not coax, but it consists of two identical halves, strongly coupled. Very convenient, RX side and dumping resistor can be galvanically insulated otherwise everything have to be floated on high voltage side. Value of dumping resistor must be adjusted for each pulse width, so this is still point by point method (some sort of dumping network will be needed, not resistor). With this values, 1uS is reached at somewhat above 110V, 3uS around 36V, 50uS at only few volts (i'm running out of PSUs). UC3843 is needed for 12V operation, original HV circuit was built using rectified tube filament voltage, and variac\HV rectifier to supply coil voltage (not good, ripple voltage can interfere with integrated measurements). Separating TX from rest of the circuit, just to contain damage in case of “big bang”.


I assembled another similar device, but for differential measurement, based on SG3525 chip and two coils from flat spiral, to generate one fixed long, and another variable pulse, to measure just difference between them. Can be connected for unipolar or bipolar pulsing, can be used for actual detector TX. Not finished timing and integrating part yet. Want to try another modification, “transformer coil” multifilar, say 5:1 turns ratio, most probably UTP, giving large inductance ratio to achieve two different pulse widths using single battery voltage. Naturally, field straight is lower for less turns, but fixed and can be compensated later. I wish if I have more time for this...

[Tepco]

Breaking the math

Hi guys,

you are getting huge complexity with the log-amp re the GB issue.
Try to solve the integral of the log(signal responses):
Integral log (Exponential induction decay + Magnetic induction decay + Const signal) dt =
Integral ( log( A*e^(-t/"TC") + B/t + C) ) dt =

(If you are doing the integration windows as usual in a PI configuration of course.)

A

OK, let's mess up mathematics even more. Inspired by VCA810 chip, capable of doing many unusual things. Maybe not best for front-end log-amp, but... Using it later, in integrator stage can produce integrator gain and time constant variable over sampling period. Considering linear or exponential ramp voltage is already present in RC circuit determining sample time, using it to modify integrator parameters over that time is easy. What math will look like then?
(Posting something truly stupid here is not forbidden by forum rules, but think twice about this proposal. May sound like reinventing analog computing in 21^st century, only for purpose)

[Kev]

The other benefit of a log amplifier is that it compresses the target signal right at the front end and reduces the large range of signal that has to be processed by the following circuitry.

Eric.

Hi Eric,
I'm really enjoying your contributions to this thread.
I wish I could fully understand the implications you are hinting at, I do though see the possible benefits to digitising early time samples with the log reduced dynamic range.

Thank you for sharing your results with us!
Best Regards
Kev.

[Ferric Toes]

Well, I have just found my long lost VCA610's, so let the log amp investigation begin. I see that the data sheet says that the log application is temperature compensated, 250kHz bandwidth and the device itself is low noise (2.2nV rt Hz). I remember once trying one of them as a gain controlled front end in a PI. Seemed to work OK, and BW in linear mode is flat to 15MHz at 40db gain if you want it.

Eric.