field test unit no 001 "model T"

[Tinkerer]

I have never succeeded in making a balanced coil system for a PI that didn't show a transient during the flyback time. Your flyback appears to be without any avalanche phenomena, so it is a cleaner transient and therefore your resulting RX trace is different to what I have observed. However, I do not get the long tail which stretches out to mid screen in your RX scope picture. Could this be due to the RC time constants that you have pre the clipping diodes - are they matched?

I have tried DD arrangement and coaxial stacked coils on a PI, and the coaxial stacked gave the best balance as the geometry has the best symmetry. However, each coil has to be individually shielded, taken via separate cables and carefully grounded at the electronics end. Shielding and where it is grounded is very critical and I dont know if this can be replicated on a simulator. Most commercial DD coils seem to have minimal overall shielding for both coils and is no good at minimising the transient. Having three cables is a pain and I suspect that it is possible to get it down to two by having screened twin and a good differential front end.

I would expect to see a transient during the switch off period but for the RX to flatline as least as quick as that obtainable with a mono coil.

By the way another thing to look out for is eddy currents generated in the coil connector, if it is metal bodied and has parallel pins. This gives a long tail as in your waveform due to the TX pins acting like a small one turn loop close to the metal body. I got rid of it by using separate connectors for RX and TX. For a mono coil a coaxial connector is best.

Eric.

Thank you for the valuable input and help.

On the TX coil current picture you can see the switching transients at TX ON and OFF. On the RX input, there are LP filters that reduce this noise. Since the input is differential, there is a common mode LP filter and a differential mode filter. These filters have a cutoff frequency above the smallest target TC.

The diodes do not clip. They are only there for protection against gross problems with induction balance. Therefore we see a "pure undistorted" full cycle wave form.

The coils are in a co-planar concentric configuration. As shown, not shielded, noise is not a problem in the extremely noisy lab, but when I get into high gain, the coil is sensitive to my hand. When using a solderless breadboard, any bad contact in the input increases the "hand sensitivity" considerably, so I think the sensitivity may be related to unbalance in the differential mode. I have not tried a stacked configuration yet.

I think that what you call a long tail, is caused by the TX energy being re-conducted to the battery or storage capacitors. At 10k PPS and 4A peak coil current, the power consumption is about 400mA.
This reverse current in the TX coil eliminates the need for a damping resistor, as it also eliminates the oscillations.

Another effect of the reverse current, is that the field it builds, kills the existing eddy currents in the target. After about 50us, all eddy currents , even for very long TC targets, are nulled. An interesting effect is that (at least in simulations) the amplitude does not seem to matter, only the TC.

In other words, it seems to be possible to measure the time it takes to bring the eddy currents to the zero point and thus get the TC of the target, instead of measuring the exponential slope of the decay, to define the TC of the target.

[Ferric Toes]

Thank you for the valuable input and help.

On the TX coil current picture you can see the switching transients at TX ON and OFF. On the RX input, there are LP filters that reduce this noise. Since the input is differential, there is a common mode LP filter and a differential mode filter. These filters have a cutoff frequency above the smallest target TC.

The diodes do not clip. They are only there for protection against gross problems with induction balance. Therefore we see a "pure undistorted" full cycle wave form.

The coils are in a co-planar concentric configuration. As shown, not shielded, noise is not a problem in the extremely noisy lab, but when I get into high gain, the coil is sensitive to my hand. When using a solderless breadboard, any bad contact in the input increases the "hand sensitivity" considerably, so I think the sensitivity may be related to unbalance in the differential mode. I have not tried a stacked configuration yet.

I think that what you call a long tail, is caused by the TX energy being re-conducted to the battery or storage capacitors. At 10k PPS and 4A peak coil current, the power consumption is about 400mA.
This reverse current in the TX coil eliminates the need for a damping resistor, as it also eliminates the oscillations.

Another effect of the reverse current, is that the field it builds, kills the existing eddy currents in the target. After about 50us, all eddy currents , even for very long TC targets, are nulled. An interesting effect is that (at least in simulations) the amplitude does not seem to matter, only the TC.

In other words, it seems to be possible to measure the time it takes to bring the eddy currents to the zero point and thus get the TC of the target, instead of measuring the exponential slope of the decay, to define the TC of the target.

Hi Tinkerer,

Thanks for the explanation. Looks like you have some novel features - no damping resistor for instance. Presumably this works for a wide range of coil inductance/resistance/capacitance, plus cable capacitance and the resulting resonant frequency?

My question still remains though: where is the noise that is troublesome? A transient can be avoided as noise if it occurs outside the sampling window and is synchronous - which this one has to be. Any long or stretched transient that strays into the sampling pulse will be fixed and of constant amplitude and can be easily nulled out.

Eric.

[Tinkerer]

Hi Tinkerer,

Thanks for the explanation. Looks like you have some novel features - no damping resistor for instance. Presumably this works for a wide range of coil inductance/resistance/capacitance, plus cable capacitance and the resulting resonant frequency?

My question still remains though: where is the noise that is troublesome? A transient can be avoided as noise if it occurs outside the sampling window and is synchronous - which this one has to be. Any long or stretched transient that strays into the sampling pulse will be fixed and of constant amplitude and can be easily nulled out.

Eric.

I think you refer to my post #745, where I mention switching noise with traditional PI.
The circuit above is the solution I found to the coil, cable and shield capacitance. It does indeed work over a very wide range and the resonance makes it very power efficient.
It also makes it possible to increase power very much, without avalanche.
There is no damping resistor on the TX, but on the RX there is a damping resistor. R1 +R7, shown in the preamp circuit. I add the screen shots of the differential signal, as it comes on the cable, before the filters and without damping.

There is a feature on these 2 pictures that I find interesting. The change in the waveform is clearly seen between the NO target and WITH target. But I think it tells us something more although I have not yet deciphert what it says.

[Aziz]

BTW guys,

the TEM TX is a wideband and continious current wave (CCW) TX. As you have an IB coil configuration (separate RX), you can sample the response at any time (or frequency response). Why don't you take the physical fact of the frequency dependent susceptibility to solve the GB issue?

Aziz

[Aziz]

Again the avalanche mode noise:
If you want to beat the best, just avoid it. And I persist on avoiding this. At all costs.
Aziz

[moodz]

Again the avalanche mode noise:
If you want to beat the best, just avoid it. And I persist on avoiding this. At all costs.
Aziz

I agree ... stay away from the avalanche noise ... its bad practice just from an EMI compliance point of view to generate unecessary hash. The optimal PI will avoid avalanche issues .

[moodz]

The TEM pix above indicate severe switching noise.

[Aziz]

I agree ... stay away from the avalanche noise ... its bad practice just from an EMI compliance point of view to generate unecessary hash. The optimal PI will avoid avalanche issues .

Aziz

[Mechanic]

I agree ... stay away from the avalanche noise ... its bad practice just from an EMI compliance point of view to generate unecessary hash. The optimal PI will avoid avalanche issues .

And to add to that, high frequency noise can pass straight across an analog switch(demodulator) depending on the capacitance of the switch even when not taking a sample. ADG1234 has some pretty impressive capacitance specs

Cheers Mick

[Aziz]

And to add to that, a low impedance & fast high voltage input protection circuit.
Low impedance (during sampling) -> low noise.
High voltage phase: high impedance

Ok, another tip:
Keep the mosfet cool. (That's not a heater.)
*LOL*

[Mechanic]

And to add to that, a low impedance & fast high voltage input protection circuit.
Low impedance (during sampling) -> low noise.
High voltage phase: high impedance

My gift to you all for Christmas...... IRF5801 I have pushed these up to I think 220v, possibly 240v before avalanche and it survived Very low capacitance and on resistance.

Cheers Mick

[dougAEGPF]

My gift to you all for Christmas...... IRF5801 I have pushed these up to I think 220v, possibly 240v before avalanche and it survived Very low capacitance and on resistance.

Cheers Mick

Is the RDSon of 2.2 Ohms of any concern?
Merry Christmas to all .
dougAEGPF

[Mechanic]

Nope, I would not think so, 2.2R = very low impedance input path = low noise source, much less than the preamplifier.
Its a bit of a juggle trying to find a mosfet with low capacitance and on resistance. As the resistance goes down the capacitance goes up. Also the maximum current causes the cap spec to go up. The irf5801 will handle 500mA, so all round it is a very good front end fet. Higher capacitance = slower switching speeds and slower settling so it is important to find one with the lowest capacitance possible, whilst also satisfying all of the other requirements. I have been using these for at least 2 years now and never blown one with the exception of doing some silly things with them!

Cheers Mick

[Tinkerer]

switching noise

The TEM pix above indicate severe switching noise.

Any time you hard switch 10A current, you will get considerable switching noise. The question is, how to see it and how to mitigate it.

On post 748 you see the RX signal after the preamp. Looks smooth enough, but when you look closely, you can see the switching noise of the sample switch pulses. However, there is no trace of the Mosfet switching noise.

[moodz]

...been slaving over a hot soldering iron all day here LOL ....the first version of the detector is finished and I am conducting field tests ....so far so good .... I have one small problem that is on the verge of being solved .. the current detector auto ground balances however it is very sensitive to magnetic fields ( eg earth field and magnets of course ).. With an 18 inch mono coil the unit picks up a handheld magnet from 12 feet away which is not good unless you want to detect magnets ...however the plus side is that this means my signal processing channels are working well.

update .... earth field sensitivity is now solved without impact to the the signal processing chains or auto ground balance. The detector only shows a slight response to the ferrite of the magnet not the magnetic field itself now ... even when held right to the coil. Thanks to Urban Fox for stimulating a line of investigation to retry an idea I had previously dismissed. Field testing commences again.