DEEPER PI DETECTION DEPTH

[simonbaker]

@simonbaker, I'm so glad you should ask this.
First off, pulse duration is determined solely by L and C time constant. Second, there is no stress on any active components - all switching happens at 0A, and third, over 80% of energy is conserved, just like with VLF tanks.
Pulse shape is near perfect cosine, not some square abomination.
Impedance at coil terminal is constantly low.

A way to improve this even further would be an H-bridge connection for the capacitor, so that the conserved energy could be recycled.

The circuit I'm so cryptic about uses a BJT equivalent of unijunction transistor, but when I squeeze it for efficiency I get narrow tolerances and instability problems - a no free lunch problem.

Rewiring it as I did here just gave me an insight that switching can be made incredibly simple and non-critical, and I can even get a H bridge polarity reversal for the capacitor to make it even better.

To see this rig pulsate replace Vctrl pulse statement with: PULSE(-1 1 8u 0 0 70u 10m), replace .tran command with .tran 0 50m 0 1u , and change R8 value to 10k - it will not benefit from energy conservation in this configuration, but just shows that it can be done.
Also if you wish to play with inductances up to 10mH change Vctrl pulse statement with: PULSE(-1 1 8u 0 0 200u 10m), otherwise it switches off to early.


So you think it is a PI exciter

Ok, thanks. Yes, sounds a lot like some energy recovery designs from other threads I've seen. It does seem like the right direction to go, assuming you get as good a target stimulus as conventional circuits.

-SB

[moodz]

I'd say challenges, not flaws.
A voltage multiplier can take care of the capacitor charging, and the charging current will be very small.
A not-too-complicated H-bridge with break-before-make arrangement switches would add only twice the Ron resistance to the current path, but no switching occurs during current flow so I'm cool.
What I'm proud of is a self-maintained pulse duration that is constant even with wild variation of Vcc. Switching off just happens, and you don't have to push it. This can have some impact on PI Rx timing. As a byproduct of low impedance all the way the off transition is immaculate - try playing with the dumping resistor.

What I am looking forward is a way to interpretation of PI signal reception in VLF style. I think I'm onto it. A single coil VLF with PI excitation - how crazy is that

Davor .... been there and done that .... the half cosine or sine pulse you describe has already been patented by Whites ( ask Carl ) .... you cant use it ... however you can use the sinc pulse ( my proposal ) .... the spectrum of a sinc pulse is mathematically complete and balanced whereas the half cosine has discontinuities both in amplitude and phase.

read this thread.

http://www.geotech1.com/forums/showthread.php?t=15518

note particulary spectral plots in this post ....

http://www.geotech1.com/forums/showpost.php?p=91710&postcount=8

The problem is generating a sinc pulse with sufficient power density .....
The Asinc pulse shows promise in this post .... it is easy to generate with very high power density and spectral power distribution.

http://www.geotech1.com/forums/showpost.php?p=118177&postcount=60

ps the plots shown are coil current not voltage plots.

moodz

[Davor]

Hi Davor,

thanks for sharing your TX circuit. It has minor flaws however:
Tank capacitor Cap voltage reversal (voltage at Cstore +100V -> almost -100V)!!!

You need a bipolar voltage drive (+/- 100V) and have to select the appropriate recycling path to avoid the voltage reversal. So you need two tank capacitors (+100V, -100V).

Ok, if you solve this issue, your TX circuit is a good one. It is in principle feasible.

Aziz

I'd say challenges, not flaws.
A voltage multiplier can take care of the capacitor charging, and the charging current will be very small.
A not-too-complicated H-bridge with break-before-make arrangement switches would add only twice the Ron resistance to the current path, but no switching occurs during current flow so I'm cool.
What I'm proud of is a self-maintained pulse duration that is constant even with wild variation of Vcc. Switching off just happens, and you don't have to push it. This can have some impact on PI Rx timing. As a byproduct of low impedance all the way the off transition is immaculate - try playing with the dumping resistor.

What I am looking forward is a way to interpretation of PI signal reception in VLF style. I think I'm onto it. A single coil VLF with PI excitation - how crazy is that

[Davor]

@simonbaker, I'm so glad you should ask this.
First off, pulse duration is determined solely by L and C time constant. Second, there is no stress on any active components - all switching happens at 0A, and third, over 80% of energy is conserved, just like with VLF tanks.
Pulse shape is near perfect cosine, not some square abomination.
Impedance at coil terminal is constantly low.

A way to improve this even further would be an H-bridge connection for the capacitor, so that the conserved energy could be recycled.

The circuit I'm so cryptic about uses a BJT equivalent of unijunction transistor, but when I squeeze it for efficiency I get narrow tolerances and instability problems - a no free lunch problem.

Rewiring it as I did here just gave me an insight that switching can be made incredibly simple and non-critical, and I can even get a H bridge polarity reversal for the capacitor to make it even better.

To see this rig pulsate replace Vctrl pulse statement with: PULSE(-1 1 8u 0 0 70u 10m), replace .tran command with .tran 0 50m 0 1u , and change R8 value to 10k - it will not benefit from energy conservation in this configuration, but just shows that it can be done.
Also if you wish to play with inductances up to 10mH change Vctrl pulse statement with: PULSE(-1 1 8u 0 0 200u 10m), otherwise it switches off to early.


So you think it is a PI exciter

[Aziz]

Yes, if both signals same frequency and phase, which they should here, thanks. Just have to deal with interesting pattern of two coils...

Is there some way we can make a single coil "oscillate" at multiple frequencies....????

-SB

Yes. Very trivial.

Impedance matching circuit, which has two resonant frequencies.
Look at the induction heater principles.

Aziz

[Aziz]

Would you explain again how this is different from the normal PI circuit?

Regards,

-SB

The TX circuits impedance is very low (reduced to coils resistance and switch-on resistance). The reactance is zero due to LC resonant tank. Well, it's only resonating for a half period cycle.

To maintain a high current pulse, the source voltage is high (100V):
I = U/Z Z=almost coil's R
I = U/(R + switch-on R)
I -> high

Aziz

[Aziz]

obviously I didn't check for file types. Here it goes:

Hi Davor,

thanks for sharing your TX circuit. It has minor flaws however:
Tank capacitor Cap voltage reversal (voltage at Cstore +100V -> almost -100V)!!!

You need a bipolar voltage drive (+/- 100V) and have to select the appropriate recycling path to avoid the voltage reversal. So you need two tank capacitors (+100V, -100V).

Ok, if you solve this issue, your TX circuit is a good one. It is in principle feasible.

Aziz

[simonbaker]

obviously I didn't check for file types. Here it goes:

Would you explain again how this is different from the normal PI circuit?

Regards,

-SB

[simonbaker]

Phase rotation will not occur because stimulation source energy is locked.
Moodz

Yes, if both signals same frequency and phase, which they should here, thanks. Just have to deal with interesting pattern of two coils...

Is there some way we can make a single coil "oscillate" at multiple frequencies....????

-SB

[Davor]

Hi Davor:

The LTspice file did not get attached; can you try again?

Regards,

-SB

obviously I didn't check for file types. Here it goes:

[moodz]

Yes, it seems to me that it would be hard to air-combine the fields from two practical MD coils without them interacting. Moodz suggested separating them which works, but then the field vectors are probably adding at harsh angles and producing a lot of rotation rather than pure summing... which might work as well though for various reasons! Hope it is tried.

-SB

There are potential pitfalls however combining of fields has been done for years in MRI machines and electron beam scanning and focussing apps. Phase rotation will not occur because stimulation source energy is locked.
Moodz

[simonbaker]

This principle is called "air combining", and it is expected that antennas are mutually orthogonal, or at least suppressed to some practical degree. It is widely used in 802.11n and OFDM in general, as it improves peak to average power ratio, and hence efficiency. It is one of principles (ab)used in MIMO Tx, and everyone in that trade will tell you how MIMO is different from air combining (and diversity in Rx), while practical thinker will see it as one and the same.

Trouble is that it complicates designs a lot. I am not too convinced about the practical side of it in metal detection. Among other things because of very limited information rate that is obtained by MD-s, and high correlation of signal. Additional spectrum will give you minimum new and uncorrelated information - such complications just don't make sense.

Yes, it seems to me that it would be hard to air-combine the fields from two practical MD coils without them interacting. Moodz suggested separating them which works, but then the field vectors are probably adding at harsh angles and producing a lot of rotation rather than pure summing... which might work as well though for various reasons! Hope it is tried.

-SB

[simonbaker]

I know, but some things are better left unseen until they are ready.

Anyway, I am toying with an idea of PI-VLF meld using Tayloe mixer as a SAH device, and anti-Tayloe it to a desired audio. I know, it sounds futile, but that's the way I see it.

To obtain a good sample I'll need 2Pi sinus or cosinus excitation, cosinus current being a bit more practical. Then I'll sample it several times in total of Pi duration which is enough for full 2Pi reconstruction. Then I'll do some magic with trivial signal processing, and voila a Franken-Pi-VLF. What I expect to happen is a design with ~ 100 pulses per second, and a significant power conservation. That's the plan.

Now, the first step is the exciter with more or less accurate cosinus current. The LTspice design is attached herewith, and I'd say it behaves. I made already an implementation with transistors (to some degree), but due to some flaws I'll not put it here just yet. Instead you'll see an idealised idea for a positive pulse (0-Pi). You'll find incredible similarities with PI pulse, so enjoy. The best part is that there is no stress on any of the active components or a power supply. Each pulse conserves most of the energy.

Check for coil voltage and current, and a capacitor voltage, also see the pulse duration.

Hi Davor:

The LTspice file did not get attached; can you try again?

Regards,

-SB

[simonbaker]

Please note that it is a single-ended oscillator so it is inherently somewhat asymmetric. One side is always somewhat pointier than the other. Amplitude of a free running oscillator is affected by coupling with, well, anything. So yes, it is a PWM allright.

Yes, I see if DC level shifts, you get PWM. If the waveform is pretty symmetric and centered, then amplitude mod should not change zero crossings. But I guess you are saying, with an assymetric waveform, amplitude mod invariably produces DC level shift. Theoretically I would think there still might be some comparator level where the zero crossings are constant when amplitude changes, or quite insensitive to it.

Please note that phase shifting network is just propagating pulses by certain tau, so again, they just pass the error forward and do not help at all. Please note that zero crossing is situated at a very linear range for PWM, making it a near perfect small signal PWM modulator.

Well, I was saying that something that blocked DC and low frequencies produced by ground variations, in the TX signal feeding the sync pulse circuit, might help reduce the PWM that is caused by level shifting, keeping the waveform "centered" vertically and the zero-crossings more or less intact.

Anyway, PWM is a good point and is something worth actually investigating and measuring for the TGSL to see the impact. Another good experiment to try...

-SB

[Davor]

The TGSL oscillator is over-driven so it isn't too susceptible to amplitude modulation I think. I don't see how it would be DC shifted that much either; how does that happen?

Please note that it is a single-ended oscillator so it is inherently somewhat asymmetric. One side is always somewhat pointier than the other. Amplitude of a free running oscillator is affected by coupling with, well, anything. So yes, it is a PWM allright.
Please note that phase shifting network is just propagating pulses by certain tau, so again, they just pass the error forward and do not help at all. Please note that zero crossing is situated at a very linear range for PWM, making it a near perfect small signal PWM modulator.