Waveforms + Measurements

[WM6]

The way I see them :
A=> start of flyback (end of TX)
B=> saturation during flyback
C=> coming out of saturation during decay
D=> ending of decay
F=> quiescent level (fully decayed)
G=> start of TX pulse
H=> TX pulse duration

Kind regards,
J. L. King

Thanks, even clarifying expert J_P cannot give me better guidance. Hope I am back in phase now.

So possible problem with present status of discussed construction is in saturation of RX amplifier during fly-back which take too long time (we wish to get 10us or less instead of present ~30us).

What is then term "pulse delay"? I this the same as H=> TX pulse duration, or it is equal to point of sampling (or sampling delay)? Last question, I hope, in field of signal components terms.

[KingJL]

Thanks, even clarifying expert J_P cannot give me better guidance. Hope I am back in phase now.

You are welcomed

So possible problem with present status of discussed construction is in saturation of RX amplifier during fly-back which take too long time (we wish to get 10us or less instead of present ~30us).

Problem isn't the saturation. The differential op-amp in question has excellent recovery charactaristics. There is not really a problem (depends on what you goal is). The coil (as described and verified in the actual measured waveforms) has such a large inductance that the decay is extended considerably. Also contributing to extended decay is an abundance of capacitance. Like I said, it is not a problem if your goals are achieved.

What is then term "pulse delay"? I this the same as H=> TX pulse duration, or it is equal to point of sampling (or sampling delay)? Last question, I hope, in field of signal components terms.

Not sure of your context of "pulse delay". Usually in reference to PI designs, the term refers to the delay of the sample pulse after TX turn-off. In some designs (multi-pulse) it refers to the delay between the multiple TX pulses.

[J_Player]

Thanks for explanation KingJL.
It would be great if you and others from design group join this great moodz development.
With faster decay you mean something like this?:

Hi WM6,
See picture to understand what we want the pulse to look like.

[Mrand]

The request to Moodz. Before the measurement always place the oscilloscope track exactly on the zero. Himself will avoid like this misunderstanding.
I have simple solution of this problem which I use for a long time. This is short DC correction pulse in the end flyback. I will send schematic when I will draw him.
Mrand

[moodz]

The request to Moodz. Before the measurement always place the oscilloscope track exactly on the zero. Himself will avoid like this misunderstanding.
I have simple solution of this problem which I use for a long time. This is short DC correction pulse in the end flyback. I will send schematic when I will draw him.
Mrand

Thanks and noted .... I am used to to just positioning it anywhere and taking relative ... however in this case note that the output does not decay to zero volts. It decays to analogue ground which in this circuit is the battery +ve 12 or so volts. I dont have the fancy differential probe so I can reference the signals to ground. However I will try to be more attentive to initiial positioning of the trace.

Regards,

moodz.

[moodz]

Eureka ! I am surprised and delighted at this result.

Hi WM6,
See picture to understand what we want the pulse to look like.

I removed the 220pF bypass cap and all the 1n4148 diodes in preference for a fast switching heaver current type ( 32 amp impulse / 10 pF reverse ). I have been "tuning" for want of a better word the damping response of the circuit and the traces below are the results.

Below is the output from the circuit ... note the "dip" after the flyback.

Now lets expand that a bit below.

Theres kind of some wave form in there ??

Now see below the waveform response for my trusty dollar coin centred in the coil.
The purple reference trace is NO TARGET
The white trace is TARGET.
Over two volts of response !!!! ( amp gain is 100 )
This response is rock solid and repeatable. I can move the coil and the response changes then restores when moved same position again. There is no jitter on the waveform ... it is rock solid with respect of the timing of the tx pulse / flyback. So presumably a detector will only have to listen in this small window. Of course I have not tested for ground effect etc etc however this result is amazing !!




moodz

[moodz]

Possible Explanation ....

The inputs to the differential amplifier are fed off either side of the differential coil which will + and - with respect to the centre tap analogue ground.
To investigate this I disconnected the negative input to the amp and connected it analogue ground. The amplifier was now only amplifying the signal at the + input side of the coil. I could see that the "huge" signal I am getting for the target is right at the base of the decay which is practically vertical voltage drop. The diff amp will try to amplify the difference at the inputs ... somewhat obvious. If the decay on one input is slight slower than the other there will be a relatively huge voltage differential this is what is being amplified ... a slight difference in the + and - decays caused by the target.
How is this .... ??? ..... the collapsing magnetic field of the TX coil will cause an equal and opposite flyback in the two sides of the diff coil. However the collapsing magnetic field of a target will be opposed to the main field ... so it will induce a very small decrease of field in the one coil and a very small increase in field in the other coil thus producing an ever so slightly different rate of decay in the coils and a shift in balance which the diff amp detects.
Because the waveforms at each input are falling at a huge dV/dt at this measure point it only takes a few nano seconds of shift to generate 10s of mv of difference between the two diff signals thus relatively small targets produce relatively huge ouputs. By varying the damping resistor in the way I had it connected I was effectively balancing the two coils which even a small coin would unbalance to produce a target response. If the two coils are each damped independantly the decays of each side may be balanced by varying one side to produce a balanced condition. This is what I inadvertantly achieved with my variable damper which had an offset resistance.

The required circuitry now is a slow autobalance which keeps the two sides balanced whilst still allowing relatively fast target "unbalance" to occur.

moodz.

[Aziz]

Hi Moodz,

you know, the input resistors before the diodes, should be 0.1% or better of tolerance. They should be hand matched to increase the CMRR. This is very important, if you have a high gain stage. The CMRR is usually very worse at high frequencies.

I recommend to make simple CMRR measurements first. Balance out the input impedance and increase the CMRR.

Aziz

[Mrand]

I found this old patent US 3,761,831 "Common Mode Rejection Means for differential circuits" http://www.freepatentsonline.com/3761831.pdf . Perhaps will be useable.
Mrand

[J_Player]

Things are getting more interesting. It appears the coil voltage flattens out shortly after 10 us has elapsed.

10 us is the magic number coil builders shoot for to start their target sampling. If you start sampling sooner than 10 us, you may be able to detect nearly invisible gold particles. Very early sampling can also cause problems with beach hunting where you begin to pick up salt water and wet sand. But we hear reports of some coil builders making very fast coils that allow them to start sampling at times approaching 6 us, which allows detection of very small gold nuggets that would otherwise not be detected (presumably they are not detecting tiny gold around salt water). These reports are from builders of analog PI detectors. The same principles will apply to digital sampling, But with digital sampling, we have an opportunity to look deeper than an analog circuit can efficiently.

And now We see in the recent differential coil oscilloscope photos -- the coil signal has flattened after 10 us.
But we also see that the target signal is clear and stable, and can be easily recognized before 1 us has passed!
This seems to be especially easy to recognize when using digital sampling. Of course, I would prefer to see a sharp square wave without oscillations and inductive artifacts, but even with this noise, the target signal comes through before 1 us as a voltage differential and a time differential. We also see the wave shape has changed. Best of all, the target signal is stable. It looks like we hit pay dirt.

Would it help to remove more of the noise?
Maybe. I am not sure about this but I would think so.
A big part of these artifacts come from details of the coil winding methods. We have some excellent tips on coil winding in the Geotech forums which address the basics that can improve any PI coil. BBSailor has given some of the best advice, along with others who show what worked in the field. See here for a very informative thread that tells many secrets for PI coils: http://www.geotech1.com/forums/showthread.php?t=11198
Midway through the thread you will find some of the best information you have read in awhile for reducing capacitance and other tips for fast coils.
Another great publication from BBSailor condenses most of his tips in one report here: http://www.geotech1.com/pages/metdet...s/FastCoil.pdf

I am sure much of the work for improving the wave form in the differential coil design will need to be done on the circuit board. Hopefully some coil tips will make the electronic work easier and will allow reaching better performance at a low cost.

The intriguing feature of this differential coil design is the digital sampling. If the sampling and signal processing was sophisticated enough, it may be able to provide some degree of target ID better than analogue designs have attained. I would not expect perfect target-ID because the size and shape of the targets have much to do with the wave form you will see. But then, we may be surprised at what we find after some field testing.

[Qiaozhi]

Things are getting more interesting. It appears the coil voltage flattens out shortly after 10 us has elapsed.

10 us is the magic number coil builders shoot for to start their target sampling. If you start sampling sooner than 10 us, you may be able to detect nearly invisible gold particles. Very early sampling can also cause problems with beach hunting where you begin to pick up salt water and wet sand. But we hear reports of some coil builders making very fast coils that allow them to start sampling at times approaching 6 us, which allows detection of very small gold nuggets that would otherwise not be detected (presumably they are not detecting tiny gold around salt water). These reports are from builders of analog PI detectors. The same principles will apply to digital sampling, But with digital sampling, we have an opportunity to look deeper than an analog circuit can efficiently.

And now We see in the recent differential coil oscilloscope photos -- the coil signal has flattened after 10 us.
But we also see that the target signal is clear and stable, and can be easily recognized before 1 us has passed!

Are you sure?
It appears from the oscilloscope photos that the coil voltage takes at least a minimum of 22us to decay, and in some cases up to 25us. The decay time is measured from the moment the mosfet is switched off until the start of the main sample pulse.

[WM6]

But we also see that the target signal is clear and stable, and can be easily recognized before 1 us has passed!

It seems to me that marked time distance on your picture is 1.25us not 1.6us?

And thanks for kind explanations, sometimes I am confused in technical English terms.

[J_Player]

It seems to me that marked time distance on your picture is 1.25us not 1.6us?

And thanks for kind explanations, sometimes I am confused in technical English terms.

Yes, WM6,
you are correct - it is 1.25 us. That was my error.

Are you sure?
It appears from the oscilloscope photos that the coil voltage takes at least a minimum of 22us to decay, and in some cases up to 25us. The decay time is measured from the moment the mosfet is switched off until the start of the main sample pulse.

In Moodz most recent tweaking session, I am referring to his first two photos which show the pulse stabilizing to fully flat at about 13 or 14 us. However, it seems to be nearly flat at 10 us, and hopefully flat enough to begin sampling targets at that time. Moodz final photos also show close up details of the first couple of us after the pulse is turned off. What is strange is that he found a usable target signal that could be sampled during the oscillations in the first us to identify the presence of his dollar coin.

Hopefully I am reading the time units correctly.

Best wishes,
J_P

[Monolith]

Yes, WM6,
you are correct - it is 1.25 us. That was my error.

In Moodz most recent tweaking session, I am referring to his first two photos which show the pulse stabilizing to fully flat at about 13 or 14 us. However, it seems to be nearly flat at 10 us, and hopefully flat enough to begin sampling targets at that time. Moodz final photos also show close up details of the first couple of us after the pulse is turned off. What is strange is that he found a usable target signal that could be sampled during the oscillations in the first us to identify the presence of his dollar coin.

Hopefully I am reading the time units correctly.

Best wishes,
J_P

Why can we not sample before the signal is flat?

Monolith

[J_Player]

Why can we not sample before the signal is flat?

Monolith

It seems we can sample before the signal is flat as long as the coil is not saturated so as to obscure the target signal. (Qiaozhi is correct, I was not measuring from the pulse switch-off time. The sampling was started after 23 us). I suppose it is better to have a clean looking square wave as a matter of s/n ratio when sampling early as possible. According to Moodz, his target signal is stable and repeatable even in the area where the signal is oscillating.

Best wishes,
J_P