Missing Link between pulse and continuous waveform detectors ?

[Aziz]

Hi all,

the software is extended for sinc burst's now. I made a quick test to drive the coil with ultra low power using the stereo output line of the sound card. Result: The transmit output power is not big enough to give satisfactory results so I have to use a high power amplifier next time.

Below is the time domain signal and frequency domain response (power spectral density) of different transmit sinc burst's with different frequency limits. Can you see the brick-wall rolloff of the power spectral density?

Aziz

[Aziz]

Hi all,

the power amp of my Hi-Fi station got really hot. I pumped the center tapped TX coil with differential signals to double the output power (100-150W, don't know the amplifiers power specification exactly).
The preamp was also driven in full differential mode using the low noise inst. amp (SSM2019) with gain of 100.

Coil configuration:
Induction balanced DD coils,
Total coil diameter: 20 cm
RX coil: 2 x 50 windings, 6.8 Ohm
TX coil: 2 x 15 windings, 1.6 Ohm (impedance to power amplifier not matched, should be 8 Ohm in my configuration, anyway)

Here are the results:
sinc up to 1 kHz: coke tin at roughly 50 cm
sinc up to 6 kHz: less than 50 cm (30..40 cm) why? answer: less coil current

Changed the output wave form to pure sine wave of 12 kHz frequency.
Result: coke tin at 70..75 cm. why? more samples within the transmit frame are analyzed (box-car integrator principles).

Note: the coils are not in resonant mode!

Bad: very very susceptible to magnetite and magnets
Good: phase changes could be detected to discriminate but it is shape, size and orientation variant.

Aziz

[moodz]

Hi all,

the power amp of my Hi-Fi station got really hot. I pumped the center tapped TX coil with differential signals to double the output power (100-150W, don't know the amplifiers power specification exactly).
The preamp was also driven in full differential mode using the low noise inst. amp (SSM2019) with gain of 100.

Coil configuration:
Induction balanced DD coils,
Total coil diameter: 20 cm
RX coil: 2 x 50 windings, 6.8 Ohm
TX coil: 2 x 15 windings, 1.6 Ohm (impedance to power amplifier not matched, should be 8 Ohm in my configuration, anyway)

Here are the results:
sinc up to 1 kHz: coke tin at roughly 50 cm
sinc up to 6 kHz: less than 50 cm (30..40 cm) why? answer: less coil current

Changed the output wave form to pure sine wave of 12 kHz frequency.
Result: coke tin at 70..75 cm. why? more samples within the transmit frame are analyzed (box-car integrator principles).

Note: the coils are not in resonant mode!

Bad: very very susceptible to magnetite and magnets
Good: phase changes could be detected to discriminate but it is shape, size and orientation variant.

Aziz

Hi Aziz .... good investigation .... however your results seem to indicate that this method wont work very effectively.
I hope you are not using the amplitude of the received pulse as an indication of sensitivity ... it is essential that the analysis is done in the frequency domain .... because the real 'pulse' is being transmitted in the Freq domain not the Time domain. A square pulse in the time domain will have a sinc characteristic in the Frequency domain.
Note also that no windowing is required for the FFT as there is no discontinuity at start / end of samples.
The lockin is essential also unless you can hardware synchronise your sound card.
I am very busy in a new job at the moment but will fire up the test bed and post results ASAP.



moodz

[moodz]

clarification

blah blah blah ... A square pulse in the time domain will have a sinc characteristic in the Frequency domain.
blah blah
moodz

Clarification .... traditionally we transmit a square pulse in the time domain and analyse in the time domain .... however the square pulse in the time domain has a sinc reponse in the Frequency domain.

Using sinc method we transmit a square pulse ( or spectral pulse ) in the frequency domain ( ie sinc pulse in time domain ) and analyse in the frequency domain ... a square pulse in the frequency domain has a sinc response in the Time domain.

This is why it is important to NOT analyse the amplitude of the Rx sinc pulse. ( Actually you could analyse the shape ... but not just the peak amplitude )

Regards,

moodz.

[moodz]

This is wild cool stuff... just watching and enjoying...



Little Britain fan????

-SB

I am the only engineer in the village.

[Aziz]

Hi moodz,

Hi Aziz .... good investigation .... however your results seem to indicate that this method wont work very effectively.
I hope you are not using the amplitude of the received pulse as an indication of sensitivity ... it is essential that the analysis is done in the frequency domain .... because the real 'pulse' is being transmitted in the Freq domain not the Time domain. A square pulse in the time domain will have a sinc characteristic in the Frequency domain.
Note also that no windowing is required for the FFT as there is no discontinuity at start / end of samples.
The lockin is essential also unless you can hardware synchronise your sound card.
I am very busy in a new job at the moment but will fire up the test bed and post results ASAP.



moodz

I have analysed this in the frequency domain. Interestingly, the target response on induction balanced DD coils behave like VLF technique. You can measure the phase shift and amplitude change.

I have looked to the frequency response. It is not clearly invariant to same targets. They depend on many factors. So you get different frequency responses.

My software digital lock-in amplifier is synchronized from the laptop. The output and input of the sound card is taking the same reference clock in the hardware. This allows me quite accurate phase and amplitude measurements of AC signals. The FFT wasn't windowed (the sinc itself is a good windowed function). I even removed phase leaps between the bursts so there is a perfect transition from one to another transmit burst (important on low frequency bursts).

I also had to use a much higher power amplifier.

But I couldn't see a significant benefit yet. The IB DD coil is very susceptible to ferromagnetic minerals (magnetite, ferrite, ..)

Anyway, looking forward to your observations and findings.

Regards,
Aziz

[simonbaker]

I am the only engineer in the village.

ya, i know...

[moodz]

PI pulse spectrum.

The magnetic pulse in the PI monocoil is related to the current in the coil.

So we have the simple PI circuit.

The current in the coil for 100us pulse.

The spectrum of the coil current ( what the target sees )
The dip at 600 Khz is the coil self resonance.
At the red circle we have 3/4 of the spectral power in 12 Khz or lower.
This will be more significant when we examine the performance of the sinc pulse in later posts.

moodz

[bbsailor]

Moodz said: The magnetic pulse in the PI monocoil is related to the current in the coil.

How the pulse kicks the eddy currents in the target is also based on the time constant of the target. The speed of current turnoff or the time constant of the turnoff should be 5 times faster than the TC of the target to fully energize the eddy currents in that particular target.

In a mono coil, the coil discharge TC is governed by the combined value of the damping resistor (Rd) and the parallel value of the input resistor (Rin) to the first amplifier stage down to the bias voltage of the clamping diodes about 0.7V. While the clamping diodes are conducting, Rin is effectivly in parellel with Rd. Below 0.7V the value of Rd is used alone and the lower portion of the coil discharge TC will be a little steeper. Use the MiscEL to graph the coil discharge TC using the example below to see how all these things interact.

Example: If you have a 300uH coil with an Rd of 680 ohms and an Rin of 1000 ohms you have a coil discharge TC of 300 divided by 405 ohms (parallel value of Rd and Rin) or 0.741 us. This means that it will energize target TCs down to 3.7us (5 X 0.741).

For those looking to seek out very small gold targets with low TCs (below 3.7us), you need a fast, low capacitance coil, short coax cable and a low COSS MOSFET to get the parallel value of Rd and Rin one hundred or two hundred ohms higher to have a faster coil discharge TC and also allow faster sampling.

Remember, that the target TC discharges about 63% of the eddy currents in the first TC so unless you fully energize the target and sample fast you may not be optimised for these small, fast TC targets.

If you know the TC of the smaller targets you wish to detect, you can reverse engineer the PI coil discharge TC and pulse characteristics to be optimized for that class of targets, however at the expense of being optimum for locating longer TC targets.

That is why when you start designing PI circuits, you need to state your assumptions about the TC range of your targets so that everyone stays on the same page and is working with the same assumptions.

I hope this helps?

bbsailor

[moodz]

TC or not TC that is the question.

... hmmm I agree totally with your observations BB ... but I am not trying to design a PI ... I am experimenting with spectral pulses not time domain pulses ... I only posted to show the power spectrum of the PI decay pulse .... and having done some measurements you can show that a target increases the low frequency spectrum at the expense of the high frequencies ... which kind of makes sense as an air core coil is not going to decrease inductance / energy loss in presence of target.

moodz.

[Ferric Toes]

ah ... knew I had seen it somewhere

http://www.zonge.com/PDF_Papers/TEMposterAs.pdf

use of half sine pulses .....

err ... is this a secret division of Whites or are they violating your patent ?

... there is a spectrum of half sine pulses there .. not exactly like a sinc.

moodz

Just picked up on this thread. Barringer Research Ltd in Toronto had an airborne pulse prospecting system way back in 1962. These were 1/2 sine pulses with a width of 1.4mS and a repetition rate of 290 per sec. I built a 1/2 sine TX back in the early 70's by using a thyristor to discharge a capacitor into a coil. The coil becomes resonant with the capacitor for 1/2 cycle and the thyristor turns off automatically as the voltage reverses. L and C control the 1/2 period. You can charge the capacitor from 100 - 200V to get huge peak currents in the coil. The voltage reversal on the capacitor can be utilised for the next pulse which is of opposite polarity and requires a separate thyristor/diode arrangement. Never pursued it though, but it would be interesting to see how the response from metal targets differs between a 1/2 sine and rectangular pulse systems. Barringer had four RX channels of different delays and widths.

Eric.

[moodz]

Sinc pulse coming to FPGA soon

[Rov]

Yep! we watchin Moodz.time n effort?Luv it!
Encouragement Only!FPGA !
Regards Rov

[mikebg]

Just picked up on this thread. Barringer Research Ltd in Toronto had an airborne pulse prospecting system way back in 1962. These were 1/2 sine pulses with a width of 1.4mS and a repetition rate of 290 per sec. I built a 1/2 sine TX back in the early 70's by using a thyristor to discharge a capacitor into a coil. The coil becomes resonant with the capacitor for 1/2 cycle and the thyristor turns off automatically as the voltage reverses. L and C control the 1/2 period. You can charge the capacitor from 100 - 200V to get huge peak currents in the coil. The voltage reversal on the capacitor can be utilised for the next pulse which is of opposite polarity and requires a separate thyristor/diode arrangement. Never pursued it though, but it would be interesting to see how the response from metal targets differs between a 1/2 sine and rectangular pulse systems. Barringer had four RX channels of different delays and widths.

Eric.

Eric, the processing of received signal is described by figures and formulae in
Patent US 4,506,225 Mar. 19, 1985
Inventors: Arthur Loveless and Anthony Barringer.
Title: Method for Remote Measurement of Anomalous Complex Variations of a Predetermined Electrical Parameter in a Target Zone.
Filled Dec.28, 1981
Quote: "The invention has been described as applied to a primary waveform comprising Bi-Polar, half-sine pulses, but in general there is a wide latitude in the type of primary waveform that may be used. It is required that the waveform of the primary field be:
(a) known,
(b) time varying, and
(c) it should contain a sufficient number and range of frequency components appropriate for the electrical parameter (e.g. conductivity) to be measured."
For metal detecting, the term "electrical parameter to be measured" can be substituted by "spectral characteristic (frequency response) of target".
Mike-BG

[moodz]

Using DSP/FPGA it is quite easy to generate ideal sinc pulses however getting sufficient power is another issue.
In order to simplify the Transmit electronics greatly the ASinc pulse can be generated. A handful of analogue components produces better than expected results. It has all the advantages of the sinc pulse but much greater simplicity and can generate sufficient power in the TX coil.
It is possible to construct coils with over 900 amps of peak pulse differential current flow.

The spectrum of the ASinc pulse does not roll off as sharply as a Sinc pulse but it critically has a flat passband and phase relationship the same as a true Sinc pulse.
An additional diode and cap will produce a half sine pulse ! The ASinc is better because the half sine suffers from rolloff and it is patented ( worse ).

Pulse

Tx Spectrum

Demod Spectrum

Moodz