Archaeology Resistivity Meter

Interesting ideas so far.  I meant to chime in earlier, but things have been pretty busy for me lately (building a deck and entertaining my Grandchildren again).

Several years ago, I was asked to sit in with some friends of my sister that work at a geotech company.  They were looking to build impedance measurement devices for soil surveys.  I came across this amazing looking chip from Analog Devices that looked like a great way to measure impedance (including a complex component).

Here is a snippet from the specification:

The AD5934 is a high precision impedance converter system solution that combines an on-board frequency generator with a 12-bit, 250 kSPS, analog-to-digital converter (ADC). The frequency generator allows an external complex impedance to be excited with a known frequency. The response signal from the impedance is sampled by the on-board ADC and a discrete Fourier transform (DFT) is processed by an on-board DSP engine. The DFT algorithm returns a real (R) and imaginary (I) data-word at each output frequency.Once calibrated, the magnitude of the impedance and relative phase of the impedance at each frequency point along the sweep is easily calculated using the following two equations:Magnitude = 22IR+Phase = tan−1(I/R) A similar device, available from Analog Devices, Inc., is the AD5933, which is a 2.7 V to 5.5 V, 1 MSPS, 12-bit impedance converter, with an internal temperature sensor, available in a  16-lead SSOP.

This might be of some help in your planning.

Good luck and let me know if you need any help on this project.

Hi Gene,

It's a super-interesting chip, I was keen to use it a few years ago for plant soil purposes, and for hydroponics - to try to see if the soil or liquid has nutrients. The idea being to have a signature of known good soil or water by sweeping through the spectrum. I never got to try it though sadly, the project moved on to something else.

It was felt that it could have had a lot of merit because then you could publish the signature, so others could try to replicate a yield (it wasn't going to be for farmers, more for home use), and to not waste nutrient. But, I have no idea in practice if the result would have been usable, or too inconsistent/variable.

I wish I'd done some work on it at the time, since it could have been useful for other purposes too.

The proposed design so far is one half of the impedance measuring system, but with digital processing. In theory it could be converted to an impedance measuring system with no additional hardware change, just a software upgrade, since the frequency will be know, and there will be some sync pulse from the FPGA, we just need to internally multiply with a 90 degree out of phase signal from that sync pulse too.

Paul, Mike, Shaz et al:

paul_d_arch  wrote:

...The problem is - as you have said - is you have no experience of using this equipment in a wet field and those who have (may not have any electronics experience)  are telling you about the commercial equipment they are using....

just to explain where I'm coming from - Physics first degree, then microelectronics. Career in IT then rugged embedded electronics, including designing carrier-phase GPS, inertial nav, high-vibration-deployed kit, etc. Long term archaeologist, arch geophys for at least 20 years.

Dave

Frank,

Copying may have been was the wrong word - duplicating/replicating  may have been better. You have a valid point, why not take advantage of new technology to move forward.

The 137Hz is, I suspect, a leftover artefact from the1920's and early hand wound equipment. There is a mathematical reason for 137 but the need for it may have long gone or other frequencies may be better suited to modern equipment.

Paul D

137Hz reversal is, I believe, chosen to be the lowest frequency that is clear of 50Hz and 60Hz mains and their harmonics, that gives just long enough in most cases to take a reading. Higher reversal rate negligible or no benefit for most use cases.

Hi Dave,

It can be demonstrated with (say) MATLAB. I'm tied up today, but it's possible to show that when switched at the source frequency, this is a mixing or multiplying action, and you end up with a steady fixed value measurement, regardless of sine or square excitation. The only difference is that there is more non-DC content to filter off with square excitation.

Although it seems to be a DC measurement, it isn't, it is a very slow AC measurement that can be (sometimes) better understood as DC, but when looking at it from how it is measured, it can be seen it really is AC.

I have a test instrument that uses square wave excitation. However I have another test instrument that uses a sine wave. Both provide similar results (almost near-identical, because they are both well-designed instruments). Both use averaging.

In any case, it can be tested for real too. The proposed design can handle the square wave and synchronised measurement method, i.e. there's no hardware difference to do either method.

The sine-wave method can be tested to see if it makes any difference, good or bad.

Frank,

There's no doubt that AC excitation will produce results of some kind, and may even reveal something different to current techniques.

The reason for maintaining continuity is that the physics is (relatively!) well understood and that current processing software can deal with it (particularly if inverting a many-probe section).

Crucially though we need to consider why such equipment is used - it is not just to say "there's an anomaly in that part of the field", the plots are interpreted to try and postulate what may be beneath the soil. That is done on the basis of experience and by reference to known plots which have been ground-truthed (i.e. survey with a particular technique and then excavated).

So, for general archaeological fieldwork it needs to use the same measurement principles.

Alternative excitation strategies could well be an interesting research project but would not be relevant for those who, like Mr MacDonald who made the original posting, were looking for a more usable version of a low-cost instrument.

Dave

In any case, the design supports the same square-wave method.

There's no hardware cost or effort difference, which is good.

Thanks great Shabaz, what worried me was that a recent message said "the intent is to use sine waves",

cheers

Dave

I thought it might be helpful to followers of this thread to walk through the operation of the synchronous detector concept used in the Robert Beck design.

This can be done with the aid of LTSpice (which is a free download from www.analog.com).

I'll follow it up with a similar modelling of a phase sensitive detector -  which works on sine waves, and explain why any differences in results are all to the good of the PSD.

As every one seems to agree: we are using AC because of electrolysis problems with DC, and we have to have some way of removing interference signals

which are commonly at mains frequency and harmonics.

This is a reasonable model of the key parts of the synchronous rectifier in the Beck design.

I've given it a nice friendly input of a 1V pk sine wave at 50Hz (to represent interference) and a +/- 5V

square wave for the signal. The signal frequency is not quite 137Hz. The node on the schematic is

labelled IN

This is the signal at UNF (for unfiltered) and is what you get once the modulator has done its work. The

spikes are caused by the finite switching speed of the FETs and amplifier. (We'll come back to them later.)

And final the output of the filter - on the Beck design this goes to a digital DC voltmeter and you record the reading

by hand. Note the different scale.

The filter has made a good job of getting rid of the AC components and the output is very close to what we would hope.

There are some important points to note:

1) the unfiltered signal never 'settles' because it has chopped up bits of the 50Hz interference added to it.

2) in real life the interference signal may be larger than the wanted signal

Synchronous detectors like this using square wave modulation work by assuming that the detector and input signal are perfectly in phase.

What happens if the are not ?

I added a delay of 73uS, equivalent to 3.6degrees phase error to the input signal, the output has changed by about 4%.

The point is not that this amount of variation in timing is likely or not but that this kind of detector is sensitive to edge timing.

Square wave signals have worse problems than that - as we know a square wave is composed of s series of sinewaves, the fundamental

and an infinite series of odd harmonics, amplitude decreasing with frequency. The 7th harmonic has an amplitude of 14.3% of the fundamental.

Lets put a low pass filter on the input:

R6/C2 cuts the frequency response by 3dB at 1kHz and 6dB/octave thereafter.

And makes for an error of nearly 10% in the output.

I need my sleep so there isn't time to model a phase sensitive detector tonight - I'll try tomorrow but I

may have to use MATLAB because LTSpice isn't good at maths.

SO you can run your own experiments here's the model:

Version 4
SHEET 1 1048 680
WIRE 432 -48 256 -48
WIRE 544 -48 512 -48
WIRE 576 -48 544 -48
WIRE 704 -48 656 -48
WIRE 256 -16 256 -48
WIRE 320 48 304 48
WIRE 320 80 320 48
WIRE 320 80 256 80
WIRE 480 160 368 160
WIRE 368 192 368 160
WIRE 704 192 704 -48
WIRE 80 208 48 208
WIRE 256 208 256 80
WIRE 256 208 160 208
WIRE 336 208 256 208
WIRE 544 224 544 -48
WIRE 544 224 400 224
WIRE 336 240 240 240
WIRE -288 288 -384 288
WIRE -80 288 -208 288
WIRE 48 288 48 208
WIRE 48 288 -80 288
WIRE 80 288 48 288
WIRE 240 288 240 240
WIRE 240 288 160 288
WIRE 368 288 368 256
WIRE 480 288 480 160
WIRE -384 320 -384 288
WIRE 240 384 240 288
WIRE -80 400 -80 288
WIRE -384 448 -384 400
WIRE 320 448 288 448
WIRE 384 448 320 448
WIRE 544 448 448 448
WIRE 320 464 320 448
WIRE 544 464 544 448
WIRE -384 560 -384 528
WIRE -80 560 -80 464
WIRE -80 560 -384 560
WIRE 240 560 240 480
WIRE 240 560 -80 560
WIRE 320 560 320 544
WIRE 320 560 240 560
WIRE 368 560 368 368
WIRE 368 560 320 560
WIRE 480 560 480 368
WIRE 480 560 368 560
WIRE 544 560 544 544
WIRE 544 560 480 560
WIRE 704 560 704 256
WIRE 704 560 544 560
WIRE 240 592 240 560
FLAG 240 592 0
FLAG 544 -48 UNF
FLAG 704 -48 FIL
FLAG 48 208 IN
SYMBOL njf 304 -16 M0
SYMATTR InstName J1
SYMATTR Value 2N3819
SYMBOL njf 288 384 M0
SYMATTR InstName J2
SYMATTR Value 2N3819
SYMBOL OpAmps\\UniversalOpamp2 368 224 R0
SYMATTR InstName U1
SYMBOL res 176 192 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R1
SYMATTR Value 100k
SYMBOL res 176 272 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R2
SYMATTR Value 200k
SYMBOL res 528 -64 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R3
SYMATTR Value 100k
SYMBOL res 672 -64 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R4
SYMATTR Value 10k
SYMBOL cap 688 192 R0
SYMATTR InstName C1
SYMATTR Value 10µ
SYMBOL res 304 448 R0
SYMATTR InstName R5
SYMATTR Value 100k
SYMBOL diode 384 464 R270
WINDOW 0 32 32 VTop 2
WINDOW 3 0 32 VBottom 2
SYMATTR InstName D1
SYMATTR Value 1N4148
SYMBOL voltage 368 272 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V1
SYMATTR Value -15
SYMBOL voltage 480 272 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V2
SYMATTR Value 15
SYMBOL voltage -384 432 R0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V3
SYMATTR Value SINE(0 1 50)
SYMBOL voltage -384 304 R0
WINDOW 3 -193 100 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V4
SYMATTR Value PULSE(-5 5 0 1u 1u .00365 .0073 500)
SYMBOL voltage 544 448 R0
WINDOW 3 -149 149 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName V5
SYMATTR Value PULSE(10 -10 0 1u 1u .00365 .0073 500)
SYMBOL res -192 272 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R6
SYMATTR Value 1
SYMBOL cap -96 400 R0
SYMATTR InstName C2
SYMATTR Value 1.6e-4
TEXT -184 -8 Left 2 !.tran 0 1 0

Have fun.

MK

Hi Dave,

I understand the need for assurance that the capability of older instruments and knowledge built on them is not lost.  Michael is doing a comparison of square wave with sinusoidal excitation.  And as Shabaz has pointed out the outlined design is quite capable of putting out a square wave and synching with it.  That should allow the value of a flexible design with additional capability to be assessed.

Frank

AS promised - now for the phase sensitive detector.

I couldn't easily model this in LTSpice, which is no great surprise because it needs multiplication and square roots.

I used Simulink in MATLAB - which is very much a paid for tool (although they do offer very good discounts to home users.)

The model is fairly self explanatory:

The left hand plot shows the input signal without interference added.

The input low pass filter is disconnected.

I've added some phase shift (pi/5 radians).

The amplitude of the output is 3.183V

With the filter engaged you can see that its doesn't have a huge visual effect on the signal.

The output voltage is 3.158 so the change is 0.0785%

This is a sine wave input with no filter, the output is 2.500 V

This is the sine wave input with the filter, you can see that it introduces a discernible phase change.

The output is 2.48 V so the change is 0.08%.

So we've demonstrated that the phase sensitive detector is not affected by changes in the system bandwidth

(except where they affect the gain at the tuned frequency).

And it makes no difference if the excitation is a sine wave or a square wave.

(You may be wondering why the sine wave amplitude is 2.50V and the square wave 3.183V, that's because both

are 5V pk but the first harmonic in a square wave has an amplitude of 4/pi  * pk = 1.273,

and 2.500 * 1.273 = 3.183

But we saw in the previous series of simulations of the simple synchronous detector that the introduction of

the 1kHz filter cause a 10% error in the reading.

In the filtered square wave simulation in this post you can compare the square wave with its filtered version.

I've added a bit the the LTSpice model of the simple synchronous detector so you can do the same thing.

I'm hoping that by now I've convinced you that the phase sensitive detector really does work a great deal better

being much less susceptible to out of band interference or response changes.

To put it another way the simple detector is sensitive to the impedance of the earth up to at least 10x the nominal frequency.

Since the PSD only detects the fundamental frequency of the excitation it is obviously a waste of power to transmit the harmonics

it will ignore.

So sine wave excitation and a PSD detector, all other things being equal, is less susceptible to interference, less susceptible

to out of band changes in earth impedance and uses less power.

What's not to like

MK

AS promised - now for the phase sensitive detector.

I couldn't easily model this in LTSpice, which is no great surprise because it needs multiplication and square roots.

I used Simulink in MATLAB - which is very much a paid for tool (although they do offer very good discounts to home users.)

The model is fairly self explanatory:

The left hand plot shows the input signal without interference added.

The input low pass filter is disconnected.

I've added some phase shift (pi/5 radians).

The amplitude of the output is 3.183V

With the filter engaged you can see that its doesn't have a huge visual effect on the signal.

The output voltage is 3.158 so the change is 0.0785%

This is a sine wave input with no filter, the output is 2.500 V

This is the sine wave input with the filter, you can see that it introduces a discernible phase change.

The output is 2.48 V so the change is 0.08%.

So we've demonstrated that the phase sensitive detector is not affected by changes in the system bandwidth

(except where they affect the gain at the tuned frequency).

And it makes no difference if the excitation is a sine wave or a square wave.

(You may be wondering why the sine wave amplitude is 2.50V and the square wave 3.183V, that's because both

are 5V pk but the first harmonic in a square wave has an amplitude of 4/pi  * pk = 1.273,

and 2.500 * 1.273 = 3.183

But we saw in the previous series of simulations of the simple synchronous detector that the introduction of

the 1kHz filter cause a 10% error in the reading.

In the filtered square wave simulation in this post you can compare the square wave with its filtered version.

I've added a bit the the LTSpice model of the simple synchronous detector so you can do the same thing.

I'm hoping that by now I've convinced you that the phase sensitive detector really does work a great deal better

being much less susceptible to out of band interference or response changes.

To put it another way the simple detector is sensitive to the impedance of the earth up to at least 10x the nominal frequency.

Since the PSD only detects the fundamental frequency of the excitation it is obviously a waste of power to transmit the harmonics

it will ignore.

So sine wave excitation and a PSD detector, all other things being equal, is less susceptible to interference, less susceptible

to out of band changes in earth impedance and uses less power.

What's not to like

MK

michaelkellett  Michael, way cool, over my head as I don't do analog stuff, that is except with tubes for my stereo.

~~Cris

If you are into HiFi you will have phase sensitive detectors in your ears !

Mine used to be quite good but are less so now.

MK

michaelkellett Yes!

Thanks Michael, Shabaz and Frank.

With absolutely no disrespect to the collective expertise, and probably an indication of my lack of understanding of some of the posts, it still feels to me like this is aimed at measuring the resistance or impedance of a 'device under test', and I'm still a little uncomfortable that proposals may not match the actual requirement, so if I may I'll just recap.

Archaeological 'resistance' or 'resistivity' measurements are taken with fundamentally a four-probe setup. This is to reduce the effect of contact resistance (which can be up to maybe 40k and vary significantly between successive sample points) and the 'signal' which can vary by perhaps only single-digit of ohms or fractions of an ohm.

To achieve this, there are effectively two completely electrically independent circuits. One, which injects a constant-current between two probes or electrodes (C1 and C2) which may well be >50 metres apart (in some instruments it is a constant voltage rather than constant current injected). There is then a high-input-impedance voltage measurement between probes P1 and P2.

In a typical twin-probe archaeological survey, P1 and C1 are located in a fixed location (at least for the duration of a grid) and mobile probes (P2 and C2) affixed to a frame are traversed by the operator to take readings at known points within a grid. The spacing between the probes P2 and C2 on the frame is chosen with an eye to the likely depth of feature expected on the site; the fixed probes P1 and C1 should be at a similar distance apart. As current flows through approximately a hemisphere with C1 and C2 points on the circumference, to avoid distortion due to the changing volume beneath the measurements, it is normal that the fixed probes P1C1 are at least 30x the probe separation distant from the grid edge.

The constant DC current is injected via a single wire core from the source in the instrument to remote injection C1, then flows via the earth (in bulk, not a skin effect) back to mobile C2 and then back to the source in the instrument box. The voltage gradient around the C probes is influenced by the underlying geology and man-made pertubation, and these changes in voltage measured at successive points (between fixed P1 and mobile P2 on the frame, via a second wire core) are what are plotted and can in due course hint at what lays beneath the soil.

The only connection necessary between the current injection 'C' circuitry and the voltage measuring 'P' circuitry is that in some instruments the start of measurements is held-off by a pre-programmed delay from the edge of the current pulse. The DC constant current is reversed periodically to avoid polarising the material around the C probes.

Readings start when the mobile probes complete the circuit, but care is needed as electrical contact can be made with, say, grass as the frame probes are plunged into the earth, therefore successive readings are taken, across multiple reversals, until they are sufficiently settled.

Dave

Well Dave, the whole point of this open source design malarky is to talk about stuff.

There is nothing unusual or special about 4 wire measurements - any decent DMM does them.

Our plan has addressed that from the start.

Beck's design does not work accoridng to the general scheme you describe, we know that for sure because the ciruit is in the public domain.

The Geoscan RM85 uses a phase sensitive detector.

As far as I can tell all the instruments available use AC excitation, primarily to avoid electrolysis effects. A soon as you use AC

the system is susceptible to AC interference so you need, in some way or other, a bandpass filter to allow only the excitation

frequency to be measured.

There are many ways this can be achieved, but two that are common are the synchronous detector (like Beck), which is fairly easy

to implement entirely in hardware, and a sin/cos type of phase sensitive detector which is extemely difficult to implement

in hardware but presents no problems to a hybrid design.

The 'delay from an edge to a single sample' method that you describe offers no discrimination against interference at all. This is

clearly shown in the second plot of my anaysis of the Beck design. If you were to average many readings taken in this way

it would behave like a bad implementation of the classical synchronous detector but would be even more susceptible to out of band

noise.

However, the proposed design is flexible, so it could be programmed to operate as you suggest. If we should get to the stage of

testing prototypes we could try it.

In passing I should point out that neither the synchronous detector nor the phase sesnitive detector offer any suppresion of interference

over a single cycle. They both rely on frequency shifting the inteference and using a low pass filter to remove it. The band width of the

detector is inversely proportional  to the cut off frequency of the low pass filter, the response time of the detector is inversely proprtional to

the band width.

It is common with PSDs to average over a fixed number of cycles, there is a Schlumberger paper with some nice graphs of

selectivity compared with number of samples averaged. Sorry I don't have a link to hand.

A filter implemented in software can easily change gear, using a fast response and wider bandwidth to detect ground contact and

then using a narrower bandwidth to take readings at the maximum accuracy.

One of the many advantages of the sine wave excited PSD technique is that all the signal power is used, both in transmission and

recpetion and that each reading represents the averageing together of hundreds or thousands of ADC samples thus allowing

a very large dynamic range and low noise.

MK

It will be exciting to see the performance - can't wait : )

The digital nature is opening up lots of possibilities to experiment and get the most out of any measurements. Hopefully it directly results in clarified imaging : )

Another nice thing is that correlated double sampling (CDS) could be used (if it was found necessary) where as well as the two quadrature signals, 180 degree out-of phase to both of them could be sampled too, and then subtracted (i.e. if there are any offsets anywhere pre-digital then they get removed too!).

I'm already thinking how the design could be later adapted/re-spinned to handle dual simultaneous ADCs off the FPGA, for different sensors that may not require dual sources. There was a very primitive magnetic gradiometer here (about halfway down: Portable Multi-Channel Recorder - Review  ), it has two 0-5V outputs.

Anyway that's major feature creep and very untested.. I only experimented with it briefly.