PI limit to early sampling

[eclipse]

Yes I'm probably going to need that too, thanks for the suggestion!
I will add simple op amp level shifter to keep current scheme as it is.

[Ferric Toes]

I noticed 1 resistor is missing on the schematic 1K between C2 and GND.

I used to select C2 for <1uA leakage at 3V to avoid significant voltage drop across the 2Meg resistor. I suspect today's capacitors are better. A later version of the cct. had a separate IC as a level shifter to get the -1.8V bias for the Jfet. This leaves the existing IC output to run at 0V for no signal to drive meter/audio.

Eric.

[moodz]

I did go on with the development of an auto-zero system using the overall feedback and this was incorporated in a detector called AUTOPULSE 2. The output of the final d.c. amplifier controlled a jfet which injected a small portion of the first sample pulse back into the preamp. The AP2 had normal dual sampling and differential integrator as most basic designs do today.

[ATTACH]38200[/ATTACH]

Further developments used this overall feedback idea to auto-balance a coaxial coil (RX1 - TX - RX2) for Fe/non-Fe discrimination. What I was not aware of at this time was the serious effect that viscous mineralisation could have on a PI detector. Most areas of the UK, except North Wales and Scotland, were free of soils developed on basaltic rocks which require ground compensation. Only after I visited Australia in 1982 did I really appreciate this problem, and, to keep this on topic, the necessity of early sampling

Eric.

I will dig up some CRO shots for you showing the effect of ferrite / iron on the early X samples.

[eclipse]

I noticed 1 resistor is missing on the schematic 1K between C2 and GND.

[eclipse]

I will definitely try out this modification, thank you Eric, I was looking for this type of arrangement.
It looks better than the simple "compensating pulse" which takes the signal level after the first amplifier, while this graphic suggests taking signal level after it's soil effect is being eliminated.

[Ferric Toes]

I did go on with the development of an auto-zero system using the overall feedback and this was incorporated in a detector called AUTOPULSE 2. The output of the final d.c. amplifier controlled a jfet which injected a small portion of the first sample pulse back into the preamp. The AP2 had normal dual sampling and differential integrator as most basic designs do today.



Further developments used this overall feedback idea to auto-balance a coaxial coil (RX1 - TX - RX2) for Fe/non-Fe discrimination. What I was not aware of at this time was the serious effect that viscous mineralisation could have on a PI detector. Most areas of the UK, except North Wales and Scotland, were free of soils developed on basaltic rocks which require ground compensation. Only after I visited Australia in 1982 did I really appreciate this problem, and, to keep this on topic, the necessity of early sampling

Eric.

[moodz]

Looks like I should have continued my investigation in the early 1970's

Eric.

[ATTACH]38199[/ATTACH]

...yes you are so close. I am still impressed your thinking was along those lines so long ago and no one has come up with much since ;-).

[Ferric Toes]

Hi Eric ... the XR sampling is not really illustrated in the previous schematic as they were mainly intended to convey the noise reduction capability ... See attached schematic illustrating more clearly below ...

[ATTACH]38197[/ATTACH]


There are pending patent(s) ..

Looks like I should have continued my investigation in the early 1970's

Eric.

[moodz]

Hi Paul,
In your NEW2.png cct it shows M4/M3 driven by TX and M11/M10 as SAMPLE. In post 139 it shows two close samples. Is the TX the X sample?

Are aspects of this arrangement patent protected?

Eric.

Hi Eric ... the XR sampling is not really illustrated in the previous schematic as they were mainly intended to convey the noise reduction capability ... See attached schematic illustrating more clearly below ...





and the timing diagram ... note the sample duration / timing can be adjusted as required by the user / control circuit. Balance/ gain control is achieved by doing this.




Basically you set up a X sample then a R sample ... to balance both samples are advanced or retarded in time ( ie slide back and forth in time ) till balance is achieved. Gain is set by varying overlap of the samples when they are fully overlapped the input difference to the diff integrator is 0 so gain control is achieved without varying amplifier gains.

Of course the samples should only be confined to the RX window ... sampling during TX ON time is not good.


There are pending patent(s) ..

[Ferric Toes]

Hi Paul,
In your NEW2.png cct it shows M4/M3 driven by TX and M11/M10 as SAMPLE. In post 139 it shows two close samples. Is the TX the X sample?

Are aspects of this arrangement patent protected?

Eric.

[moodz]

Sure .... it could be done with a non digital cct but a micro is the easiest way to go ... It has been field tested.

[WM6]

.... the samples must be accurate and adjusted in duration and start stop times to balance the diff integrator summation.

Great solution moodz, but ... is this practically realizable in accurate way?

[moodz]

Not what I was expecting. Does it look right or do I have a problem somewhere?

Its fine ... the loop forms a type of oscillator however we are preventing it from oscillating by interrupting it with TX pulses :-) .....

If you like try varying R2 from high to low ( eg 1K to 10meg ) to see what effect on the noise cancellation is ....

You can also vary C3 say from 10p to 500p to see the what effect also.

The BIG THING is the new way of sampling .... we DO NOT sample late in the RX waveform anymore. The two samples for the differential integrator inputs are as below ...

This gives us ground balance and ferrite rejection ;-) The samples are typically equal in length and 1 or 2 microseconds wide but can be varied .. in phase and duration with respect to each other even overlapping. ( that is how gain control is done .... when they equally overlap the o/p from the diff integrator is zero with respect to target ;-) )

Because we are taking all our samples early in the RX period the RX period can be shortened somewhat leading to higher PRF for the detector. :-) .... its a free lunch LOL

Note that the first sample ... samples the "edge" or X SAMPLE of the flyback and the second sample samples the amplitude R SAMPLE .... the samples must be accurate and adjusted in duration and start stop times to balance the diff integrator summation.

[green]

Not what I was expecting. Does it look right or do I have a problem somewhere?

[6666]

...about the same time .... you dont need to run it for that long ... 200 ms to 1 sec is fine ... i only run long times to gather fft data and check for instability.

Ahh thanks for the tip, made it various lengths in ms to test and only takes a few seconds now, so learnt something cheers