detection depth

[Ferric Toes]

Interesting test. Wondering if you could add a couple data points the next time you test.

1964 US silver quarter$ - IACS 87% - Total decay time 1500uS - Optimum TX width 550uS - Pulse rate 736.

Saturation time at optimum Tx pulse width _______ and Tx pulse width when Rx amplitude is 1/2 saturation voltage at the recorded saturation time ________.

Example:

1964 US silver quarter$ - IACS 87% - Total decay time 1500uS - Optimum TX width 550uS - Pulse rate 736- Saturation time 960usec - Tx width(Rx 1/2 saturation volts @ recorded saturation time) 185usec

I expect you must have mentioned this previously, but what saturation time are you referring to? There is the TX saturation of the target when the eddy current have diffused such that there is a single exponential. This is represented in your log plots by the commencement of a straight line decay. In the case of my tests, this should roughly correspond to the 'optimum TX pulse width', as beyond that, the decay is invariant.

The other saturation is when the coil current has stabilised to a steady value and limited only by the total coil circuit resistance. This happened at 300uS in the case of the above tests and which can be reduced to 200uS by an additional 4.7 ohms in series with the TX circuit. The TX transistor is an IRFP250N with 0.075 ohms on resistance.

When I have refined the thing a bit and added sampling and an integrator, I will be able to get some measuring points and plot graphs.

Eric.

[Ferric Toes]

On a (13kHz) VLF machine, clad 25c and 90% silver 25c coins read almost identically, with a time-constant TC = 100 usec (f_c = 1600Hz). The Hocking machine runs at a freq of 250kHz, I recall, so predominantly reads the skin, less so the bulk core.

Will Eric's next experiment require a tungsten wire coil encased in ceramic to stop it melting? It could be handy underwater - the film of boiling water vapour on the coil surface would reduce drag when sweeping....

Yes, my Hocking is a 250kHz unit, so I am sure that must be the explanation.

Underwater the coil would be fine, but I had the idea of building the coil into a type of diver's glove for searching freezing cold water at night, or with zero visibility when you can't see a normal search coil at the end of a shaft. At least, when you got a signal it would be right under your warm hand.

Eric.

[green]

I expect you must have mentioned this previously, but what saturation time are you referring to? There is the TX saturation of the target when the eddy current have diffused such that there is a single exponential. This is represented in your log plots by the commencement of a straight line decay. In the case of my tests, this should roughly correspond to the 'optimum TX pulse width', as beyond that, the decay is invariant.

The other saturation is when the coil current has stabilised to a steady value and limited only by the total coil circuit resistance. This happened at 300uS in the case of the above tests and which can be reduced to 200uS by an additional 4.7 ohms in series with the TX circuit. The TX transistor is an IRFP250N with 0.075 ohms on resistance.

When I have refined the thing a bit and added sampling and an integrator, I will be able to get some measuring points and plot graphs.

Eric.

The optimum TX width was determined by examining the time that the RX came out of saturation with the target in place. Above the optimum time, any increase in TX width had no effect. Below the optimum time the saturation time got shorter, indicating that the received signal was decreasing in amplitude.

It's the saturation time from your test

.

The time when Rx comes out of saturation. The Tx width at half saturation volts is the Tx time that gives a Rx voltage 1/2 of saturation at the recorded saturation time. Do your measurements plus record time that Rx came out of saturation _________ at Optimum Tx width. Decrease Tx width until Rx voltage is 1/2 saturation volts at the just recorded time and record Tx width.


I'm wondering if your (total decay time minus the added Rx saturation time)/4 or 5 wouldn't be close to the target TC. Thinking the added Tx width(1/2 signal) could be used with the coil diameter, amplitude vs detection charts to predict lose of distance if the lower Tx width was used.

[Ferric Toes]

The optimum TX width was determined by examining the time that the RX came out of saturation with the target in place. Above the optimum time, any increase in TX width had no effect. Below the optimum time the saturation time got shorter, indicating that the received signal was decreasing in amplitude.

It's the saturation time from your test

.

The time when Rx comes out of saturation. The Tx width at half saturation volts is the Tx time that gives a Rx voltage 1/2 of saturation at the recorded saturation time. Do your measurements plus record time that Rx came out of saturation _________ at Optimum Tx width. Decrease Tx width until Rx voltage is 1/2 saturation volts at the just recorded time and record Tx width.


I'm wondering if your (total decay time minus the added Rx saturation time)/4 or 5 wouldn't be close to the target TC. Thinking the added Tx width(1/2 signal) could be used with the coil diameter, amplitude vs detection charts to predict lose of distance if the lower Tx width was used.

Thank you,; it is clear now. Yes, the RX amplifier goes into saturation during the period when the TX is switching. In the case of my circuit, the NE5534 goes into negative saturation at about -4.5V due to the diode clipped back emf from the coil. This, plus the recovery time of the amplifier results in a necessary delay of 20-25uS before any sampling is possible. Mine is an inverting amplifier, so the decays are from a negative value toward zero.

Before I explore further, I am going to build in a two stage preamp with higher spec IC's. The single stage preamp using a 5534 has a small problem with the recovery of the gain after saturation. The recovery waveform we normally look at after TX switch off, is the recovery of the dc level back to zero. if you inject a sine waveform of say 50kHz, it is apparent that the gain does not recover so fast. this will have the effect of distorting the early portion of the decay curve from an object. I suspect that to get the best detection range, early sampling is still a good thing to do. However, as the decay persists for much longer, we want to make use of that too; perhaps by the summation of more samples.

Eric.

[green]

Thank you,; it is clear now. Yes, the RX amplifier goes into saturation during the period when the TX is switching. In the case of my circuit, the NE5534 goes into negative saturation at about -4.5V due to the diode clipped back emf from the coil. This, plus the recovery time of the amplifier results in a necessary delay of 20-25uS before any sampling is possible. Mine is an inverting amplifier, so the decays are from a negative value toward zero.

Before I explore further, I am going to build in a two stage preamp with higher spec IC's. The single stage preamp using a 5534 has a small problem with the recovery of the gain after saturation. The recovery waveform we normally look at after TX switch off, is the recovery of the dc level back to zero. if you inject a sine waveform of say 50kHz, it is apparent that the gain does not recover so fast. this will have the effect of distorting the early portion of the decay curve from an object. I suspect that to get the best detection range, early sampling is still a good thing to do. However, as the decay persists for much longer, we want to make use of that too; perhaps by the summation of more samples.

Eric.

An attempt at doing a test similar? to yours. My bench tester is different than yours. Some things I could try without a lot of modifying. Tx rate_13pps, about 25usec ramp then .5A constant current. Preamp gain CH1 about 450, post amp(used for charting data only, not part of normal PI detector) gain x9, total gain CH2 about 4000. Rx two 8inch round figure8.

Procedure: increase Tx pulse width until CH2 saturation time(Tx off)stops increasing then decrease pulse width until I see a slight decrease in saturation time(about 100usec on the first scope trace)determines optimum Tx width for the target. Then decrease pulse width until CH2 amplitude is 1/2 saturated at previous saturated time(Tx pulse width about 80usec)determines 1/2 signal Tx width. Saturation time is about 1/2, maybe an option instead of 1/2 volts.

Recorded data and charted with Tx=230 and 80usec.

Was thinking (time CH2 signal crossed 1/2 saturation volts minus saturation time)/.71 would give target TC. First scope trace about(175-100usec) 75usec/.71=105usec, second scope trace about(105-50usec) 55usec/.71=77usec. Looks closer to 130usec the way I normally do it. I think if Rx saturation time was greater than the target TC the calculation might be close to the same. Need to increase signal to try, increase current or a smaller coil would be a couple options.

The charted data shows a decrease in amplitude(Tx=80 vs 230usec) of 1/2 after the target TC(130usec). Looks like signal amplitude is close to the same at the start(<10usec).

[green]

An attempt at doing a test similar? to yours. My bench tester is different than yours. Some things I could try without a lot of modifying. Tx rate_13pps, about 25usec ramp then .5A constant current. Preamp gain CH1 about 450, post amp(used for charting data only, not part of normal PI detector) gain x9, total gain CH2 about 4000. Rx two 8inch round figure8.

Procedure: increase Tx pulse width until CH2 saturation time(Tx off)stops increasing then decrease pulse width until I see a slight decrease in saturation time(about 100usec on the first scope trace)determines optimum Tx width for the target. Then decrease pulse width until CH2 amplitude is 1/2 saturated at previous saturated time(Tx pulse width about 80usec)determines 1/2 signal Tx width. Saturation time is about 1/2, maybe an option instead of 1/2 volts.

Recorded data and charted with Tx=230 and 80usec.

Was thinking (time CH2 signal crossed 1/2 saturation volts minus saturation time)/.71 would give target TC. First scope trace about(175-100usec) 75usec/.71=105usec, second scope trace about(105-50usec) 55usec/.71=77usec. Looks closer to 130usec the way I normally do it. I think if Rx saturation time was greater than the target TC the calculation might be close to the same. Need to increase signal to try, increase current or a smaller coil would be a couple options.

The charted data shows a decrease in amplitude(Tx=80 vs 230usec) of 1/2 after the target TC(130usec). Looks like signal amplitude is close to the same at the start(<10usec).

Looked closer at the data after I posted. Eric had a optimum Tx time of 550usec instead of 230usec for the quarter which I think made sense. Tried again, I have a 20turn trim pot for adjusting Tx time. Didn't notice Tx width had stopped increasing when Rx saturation time had stopped increasing(no stop on the pot). (earlier numbers)Rx saturation time 120usec(100usec), optimum Tx width 500usec(230usec), Tx width 1/2 signal 105usec(80usec). I'm thinking increasing the Tx width from 230usec to 500usec increased the signal120/100, not a large increase for the increase in Tx width. Maybe Tx width 1 or 2 times the target TC is all that's needed.

The calculated TC (time CH2 signal crossed 1/2 saturation volts minus saturation time)/.71 =115usec closer to the 130usec. Still need to try the small coil for more signal(Rx saturation time>TC). Maybe test for Rx saturation times at Tx width=1,2 3,4 and 5 times the target TC(125usec for the quarter). I'll try the 1oz copper coin I have to compare with Eric's 1oz silver coin data, should be close.

[Ferric Toes]

I have changed to a better preamp (LME49990) and have taken some pictures of the end of the decay. As previously mentioned it is inverted, so the waveform decays from a negative value back to the zero centre line.

The TX pulse was set to be exactly 1000uS in this instance.

The coins in order are, US 5 Cents..... US silver quarter..... US silver half..... US copper penny.

Eric.

[green]

I have changed to a better preamp (LME49990) and have taken some pictures of the end of the decay. As previously mentioned it is inverted, so the waveform decays from a negative value back to the zero centre line.

The TX pulse was set to be exactly 1000uS in this instance.

The coins in order are, US 5 Cents..... US silver quarter..... US silver half..... US copper penny.

Eric.

[ATTACH]41760[/ATTACH] [ATTACH]41761[/ATTACH] [ATTACH]41762[/ATTACH] [ATTACH]41763[/ATTACH]

Been trying your method, looks like an easier way to determine target TC than what I've been using. Couple questions. What is your amplifier saturation voltage? Have you gotten amplifier saturation time to be greater than the targets TC? TC=(time amplifier volts is 1/2 saturation-time amplifier volts is saturated)/.71. I've been using Tx off to trigger scope zero time. I need to try a smaller coil to get amplifier saturation time greater than targets TC so I'm not sure the calculation will give the right answer.

[Ferric Toes]

I hope this is somewhere near correct, but I am now looking at the start of the decay waveform. The saturation level of the preamp is slightly over + and - 3V. The graticule line just below the 50uS at top is zero level. This makes the middle line on screen to be half voltage. The position of the quarter in the first picture was such that it just touched the saturation level of the amplifier. The time to half voltage is measured as 50uS. which corresponds to 0.7 Tau.



Positioning the coin closer to the coil increases the amplitude so that the first part of the waveform is fully saturating the amplifier and we can now look at a later portion of the decay. Re-positioning the cursors to measure just coming out of saturation to the half voltage again, gives 84.9uS.



Repeating the above to get yet more signal and a longer saturation, results in a reading of 87.2uS, which is less of a change. These three results show that the first part of the decay has a faster rate due to the superimposed early time diffusion decays (the curved part of your log/linear plot) which settle down later to the single fundamental decay. You could start from the back end and work forward, but double the signal for each measurement.

This doesn't seem to match what I have done before, so I will look at the method again.



Eric.

[green]

Been trying your method, looks like an easier way to determine target TC than what I've been using. Couple questions. What is your amplifier saturation voltage? Have you gotten amplifier saturation time to be greater than the targets TC? TC=(time amplifier volts is 1/2 saturation-time amplifier volts is saturated)/.71. I've been using Tx off to trigger scope zero time. I need to try a smaller coil to get amplifier saturation time greater than targets TC so I'm not sure the calculation will give the right answer.

Been thinking amplifier saturation time needs to be greater than target TC. If scope over scale doesn't effect the displayed signal(saturate the scope), with a good scope zero the time between the scope trace crossing 2 and 4 divisions/.71 should equal target TC if time crossed at 4 divisions amplitude is greater than the target TC. Eric's quarter looks like 150usec/.71=210usec. Don't have a silver quarter but the newer quarter calculated 140usec with 100mv/div on my scope.

Hi Eric, posted before I saw your reply#99. Looking at your previous quarter scope picture I read closer to 150usec to cross 2 and 4 divisions. Looks like your amplifier doesn't have a step response coming out of saturation. Maybe if you center zero, the amplifier will be out of saturation when the trace crosses -4div, 500mv/div scale.

[green]

Been thinking amplifier saturation time needs to be greater than target TC. If scope over scale doesn't effect the displayed signal(saturate the scope), with a good scope zero the time between the scope trace crossing 2 and 4 divisions/.71 should equal target TC if time crossed at 4 divisions amplitude is greater than the target TC. Eric's quarter looks like 150usec/.71=210usec. Don't have a silver quarter but the newer quarter calculated 140usec with 100mv/div on my scope.

Hi Eric, posted before I saw your reply#99. Looking at your previous quarter scope picture I read closer to 150usec to cross 2 and 4 divisions. Looks like your amplifier doesn't have a step response coming out of saturation. Maybe if you center zero, the amplifier will be out of saturation when the trace crosses -4div, 500mv/div scale.

Another test. Shows what I was thinking wrong. Tx_.5A constant current, Takes about 25usec to ramp up to .5A. Scope triggers on Tx off. Time scope is saturated(trigger to start of decay)is greater than time between 4 to 2 division decay in all scope traces. The TC increased with Tx on time increase. Thought if time at saturation was greater than decay time it wouldn't.

[Ferric Toes]

Been thinking amplifier saturation time needs to be greater than target TC. If scope over scale doesn't effect the displayed signal(saturate the scope), with a good scope zero the time between the scope trace crossing 2 and 4 divisions/.71 should equal target TC if time crossed at 4 divisions amplitude is greater than the target TC. Eric's quarter looks like 150usec/.71=210usec. Don't have a silver quarter but the newer quarter calculated 140usec with 100mv/div on my scope.

Hi Eric, posted before I saw your reply#99. Looking at your previous quarter scope picture I read closer to 150usec to cross 2 and 4 divisions. Looks like your amplifier doesn't have a step response coming out of saturation. Maybe if you center zero, the amplifier will be out of saturation when the trace crosses -4div, 500mv/div scale.

Hi Green,

Here is my amplifier trace at the end of the TX period. I'm using x10 timebase magnification to expand this portion. Full TX still 1mS.

First picture is for no object and the second picture is with Nickel to give a bit of saturation. The scope does not saturate if I increase vertical gain to 100mv div.

Eric.

[Skippy]

Just a minor technical detail I thought worth mentioning, about your time-constant calculations.
It appears you are calculating the time the waveform takes to decay from V volts to 0.5 * V volts, then using a scaling-constant to evaluate TC.
The correct scaling-constant for this calculation is: natural log (2) or -natural log (0.5) = 0.6931

So: TC = (50% decay time) / 0.6931
or: TC = 1.442 * (50% decay time)

Using 0.71 is going to result in a 2.4% error, not huge, but worth knowing.

[green]

Just a minor technical detail I thought worth mentioning, about your time-constant calculations.
It appears you are calculating the time the waveform takes to decay from V volts to 0.5 * V volts, then using a scaling-constant to evaluate TC.
The correct scaling-constant for this calculation is: natural log (2) or -natural log (0.5) = 0.6931

So: TC = (50% decay time) / 0.6931
or: TC = 1.442 * (50% decay time)

Using 0.71 is going to result in a 2.4% error, not huge, but worth knowing.

Thanks

[green]

Hi Green,

Here is my amplifier trace at the end of the TX period. I'm using x10 timebase magnification to expand this portion. Full TX still 1mS.

First picture is for no object and the second picture is with Nickel to give a bit of saturation. The scope does not saturate if I increase vertical gain to 100mv div.

Eric.

[ATTACH]41775[/ATTACH][ATTACH]41776[/ATTACH]

Hi Eric,

Wondering if you could try a test. Your scope, center zero, 500mv/div(amplifier saturates above scope screen FS). Test US copper penny, clad quarter, silver quarter and 1oz silver coin with 500usec Tx width and the 1oz silver coin at 2000usec Tx width. I'll try my test again with a US copper penny, clad US quarter, and 1oz copper coin to compare, don't have a silver quarter. I'll have to use 200mv/div because my amplifier saturates at a lower voltage than yours. I have compared a 1oz copper vs a friends 1oz silver coin and they were close. Skippy replied the silver and clad quarters were close on his VLF detector, interested if your PI bench detector shows a difference. Carl did a cache test using silver quarters, wondering if a clad quarter cache test would be similar. Wondering if your 1oz silver comes close to my 1oz copper test. Skippy corrected my multiplier for half scale, maybe we should multiply difference time by 1.44 to be the same.