Square wave Pulse Induction

[Davor]

Coil compensation integrator seem as a nice idea. The compensation transformer may operate well below any saturation effects so this may actually be very good. You also adopted several weighting functions for discrimination. I gave up when I found out there is no single function that works in favour, but this may actually work!

[deemon]

Coil compensation integrator seem as a nice idea. The compensation transformer may operate well below any saturation effects so this may actually be very good.

Yes , of course , but this is not the only advantage .... we also can radically decrease the transformer ratio on the secondary coil , thus we can give much more signal current to the preamp - without any noise and distortions , so it gives us a plenty of extra amplification ( and less turns to wind ) You see , in the first two schemes we need to calculate the transformer ratio so that preamp might "eat" all the transformed coil current without overload .... but now we can feed the preamp only by pure signal - because all the idle and useless square wave is compensated by our integrator . And another bonus is that we can now watch the signal even during the flyback interval - and this is with mono coil ! So we have all advantages both of PI and CW devices in one ...

You also adopted several weighting functions for discrimination. I gave up when I found out there is no single function that works in favour, but this may actually work!

Yes , I had tested this correlation approach in my previous device , and it works fine . The entire idea had been inspired by Taylor's series .... but I think that we can test and use different sets of basic functions . All what we need it their "linear independence" - I mean that they must have zero cross-correlation . And also a zero integral on the working interval - in order to suppress the DC component of the signal .

[Davor]

I actually thought of taking several samples and performing DCT on them, or something of that kind. There are several transforms that can be tried. My error in reasoning was that there must be some weighting function that would perform some kind of pre-calculation so that the frequency content is not spoiled by flyback, but in vain, mostly because of offset problems. Your approach fixes it elegantly by subtraction. Guess you fix most of the ground balance with integrator too. I like it.

[deemon]

Now , before I'll show the signal processing circuits , we must clearly understand what kind of signals we can expect on the power chain output . As I told before , the output current wave ( TR2 secondary coil , on the power chain circuit ) does have a "square" shape with 4 intervals in each period - 2 intervals of constant current ( working intervals ) with opposite polarity , and 2 flyback intervals between them , with half-sine pulse shape . In my working prototype flyback pulse has 8 uS duration , and working interval duration can be set between 500 uS , for 1 KHz frequency , to 42 uS , for 10 KHz frequency . So we can easily receive and process a target response signal from a different kind of targets , and we also can use the optimal frequency in the different cases - for example , if we need to find a very big and deep target - we can set lower frequency ( 1-2 KHz ) , but when we need to find only a little gold nuggets , or little coins - we can push the frequency up in order to obtain the maximum sensitivity . And what is interesting - power consumption rises very slow with frequency , for example , the circuit consumes about 65 ma current from 12 V battery on 1 KHz working frequency , but when I set 10 KHz - current is about 120 ma . This transmitter is pretty power-efficient , especially on high frequencies - because we don't need to charge the coil again and again , but only reverse the current 2 times a period , and by the way , when this device runs on 10 KHz - it's equivalent to conventional PI device at 20 KHz PRR ( because we have 2 flybacks on the period ) .... it's not bad , anyhow

But what we can see on the scope when we bring the metal target to the coil ? As I explained before - each current reverse induces the decaying eddy current in the target body , and this exponentially decaying current process , due to Lenz's law , makes the search coil current to decay too - it occurs just on the working interval .... and this decay repeats in every period - so the average coil current goes down and down - because the next period starts with the current the previous period finished with . So , if we measure the average coil current , and notice a sudden current drop - we can use it as a sign of a metal target presence , and the first algorithm of signal processing is based on this fact .

And if the current goes down on every period when a metal is near the coil ( more precisely speaking - on every half-period ) , what does prevent it from totally disappear , from decaying to zero ? This preventing factor it the current rise caused by reduced power chain energy loss . As we can see on the attached picture , in the stable state we have a kind of "equilibrium" - the supply voltage ( Usupply ) is completely equal to the sum of the voltage losses ( voltage drops ) through the whole chain . These losses are - the voltage drop on the power switch ( diodes and transistors ) , and the voltage drop on the coil wire , due to the current flow , according Ohm's law . So the stable state condition is defined by Kirchoff's voltage law - http://en.wikipedia.org/wiki/Kirchhoff%27s_circuit_laws - "the sum of the emfs in any closed loop is equivalent to the sum of the potential drops in that loop" . And if we haven't a metal near the coil , we does have a stable state - an "idle process" , producing an idle square wave - graph (1) on the picture . When we bring the metal to the coil - the coil current begins to decay ( graph (2) on the picture ) , and after a several periods it can loose quite a noticeable part of its initial magnitude . But when the current become lesser - the voltage drops becomes lesser too .... and with the constant Vsupply voltage - the sum of potential drops becomes lesser than EMF , so the coil current becomes to rise , with almost linear law ( graph (3) on the picture ) .

So with a metal near the coil we does have a two simultaneous factors - exponential decay (2) and linear rise (3) .... and they both leads the system to the new "equilibrium point" ( another stable state (4) ) with less average coil current . As we can see in the picture - this average current drop ( Iidle-I ) depends on the target-induced input signal magnitude ( delta Iexp ) , operation frequency F and the coil time constant ( Lcoil/Rwire ) . The system behaves itself as it has a kind of inherent "negative feedback" , just like an OP-amp with negative feedback resistors does . Signal goes to the input (-) via the input resistor , moves the output , then negative feedback returns to the input wia the feedback resistor , and the system comes to the new stable state , with voltage on the output , amplified by the Rinput/Rfeedback ratio . And like the OP-amp with negative feedback , this power chain has it's own "amplification factor" , being defined by the Tperiod/Tcoil ratio , so if we watching the signal on the scope , we can notice that amplitude does drop much more than we can expect from the exponent decay during the only one working interval .

[deemon]

Here are the real pictures of the signals in the circuit . The first is the coil current versus the flyback voltage -



this is the same signal , magnified by the scope -



this is the coil current signal after the full-wave rectifier , so the both half-waves has the same ( positive ) polarity -



and this is the output signal with the metal target ( 0.33 coke can ) near the coil -




We can see a big amplitude drop with relatively low remaining exponent signal on the working interval , as I explained before .... and we can see also a power chain supply current drop ( from 63.8 ma to 48.7 ma ) .

[WM6]

Beautiful.

[techo_bob]

Hi deemon.

Following this thread since it started, your waveform's look too good to be true
No ringing any where just perfect wave forms, very well done!
Are you able to give any details of the coil transformer, turns ratios (No of turns) etc.
And are you still running a 10:1 inductance ratio between search coil and current transformer winding?

Any help is much appreciated

Techo_Bob..

[deemon]

Hi deemon.

Following this thread since it started, your waveform's look too good to be true
No ringing any where just perfect wave forms, very well done!
Are you able to give any details of the coil transformer, turns ratios (No of turns) etc.
And are you still running a 10:1 inductance ratio between search coil and current transformer winding?

Any help is much appreciated

Techo_Bob..

All waveforms are real , of course . What about turn ratios - I marked them on the power chain circuit ( n , n/4 , n/5 , 2.5*n , etc ) on the page 1 of the topic . N is a turn number of the primary coil - TR1.1 . Current transformer TR2 has 2 turns on primary and 500 turns on secondary side . And also I marked coil inductance of all transformers - it will be enough to make them . All what you need is to recalculate all windings for your core - because I used the cores that I have here , but your ferrite may have a different permeability and saturation point . By the way , these 2 pictures shows the "kick circuit" operation ( look the power chain schematic , test point TP3 ) - positive and negative transition .





Flyback pulse ( upper scope trace ) has 330 V amplitude , kick circuit pulse - about 20 V . Timescale is 2 uS/div .

[techo_bob]

Thank's Deemon

For your help and explaination

Bob..

[deemon]

By the way , I can write more info about TR1 transformer , that I'm using now in my experimental setup . For the first time I wound the primary coil - TR1.1 ( 160 turns , copper wire , 0.13 mm diameter ) , then wound TR1.2 ( 33 turns , Litz wire , 0.5 mm diameter ) , then TR1.3 ( 40 turns , copper wire , 0.13 mm diameter ) and after all - TR1.4 ( teflon insulated wire , 0.3 mm diameter ) . But here I found that this core hasn't enough space to wind TR1.5 - for kick circuit supply , so I made another transformer , especially for this supply rectifier . But of course , it's not critical for this circuit operation , so when we have a bigger core - we can wind all the coils on the same core , as shown on the power chain circuit . Why I use a teflon wire for TR1.4 - because this winding is connected directly to the "hot" side of our search coil ( with 300-500 V flyback pulses on it ) , and we cannot use a simple insulated copper wire here , because of electric breakdown risk . And don't forget that all turn numbers are suitable only for my core , and for another core type they must be recalculated .

[mickstv]

Hi deemon, are you able to post cro shots of target responses at the output of your frontend amp stages.

[golfnut]

"we cannot use a simple insulated copper"

maybe if you put a layer of xfromer tape after each widinding - like on high voltage transformers.

http://www.aliexpress.com/item/4-5mm...727542550.html

5kV withstand

60um thick

S

[deemon]

"we cannot use a simple insulated copper"

maybe if you put a layer of xfromer tape after each widinding - like on high voltage transformers.

http://www.aliexpress.com/item/4-5mm...727542550.html

5kV withstand

60um thick

S

Of course , it may be a good solution . By the way , we don't need to use such a tape on the all layers , because the only winding needs to be thoroughly insulated is TR1.4 , it's connected directly to the "hot" side of the search coil . So we can simply do all the windings , then make a layer of this insulation tape , and then wind TR1.4 - it will be OK .

[deemon]

Hi deemon, are you able to post cro shots of target responses at the output of your frontend amp stages.

Yes , I have these pictures ..... one example we can see in the post 35 on this page - the bottom picture is the target response of a coke 0,33 can near the search coil , taken just from the preamp output . I use here a kind of differential preamp based on a dual op-amp , so I have both positive and negative signal waves in the same ( positive ) polarity , then I used a simple 2-diode rectifier and shot this picture . So this is a signal without any further amplification , just from the frontend . But the signal destination is different in a different signal processing algorithms ( see block diagram in the post 29 on this page ) .

In the simplest first variant of processing ( version 1.0 ) we feed this preamp rectified signal directly to the low pass filter ( about 15 Hz cutoff ) , in order to select its DC component , proportional to the average coil current . So we don't interest about the signal shape , we are catching only the amplitude drop when the target is near the coil . So we cannot see a detailed picture of the signal - we don't need to do it .

But in the second variant of processing ( version 2.0 ) , we have a goal to select and discriminate the signals from different targets with different time constant ( TC ) . So I use there a several gain stages , which makes an opposite operation - they discards the amplitude drop , but amplifies the signal on the top of the pulses . And after those gain stages we can obtain more interesting pictures . Here I upload a pictures of a signal after 2 gain stages , when the different targets ( with different TC ) were near the coil . You can see some RF noise on the pictures , it's because of amplification . Operation frequency is 2 Khz , time scale is 50 uS/div .

The first picture is an idle signal , without any target :






This is the signal from the thick copper ashtray :






This is the signal from a 0,33 coke can :






This is the signal of a 5 roubles Russian coin :





This is the signal of a thin aluminium foil :

[deemon]

By the way , when we look to these pictures attentively , we can notice the same characteristic in the all pictures . I mean that the start and the end point of the curves are virtually at the same height . And the same feature we can see on the post 35 bottom picture . But as we know , the target response is a piece of decaying exponent , so the start point of the function must be higher than the end point anyhow . But why they are equal here ? The explanation is in the post 34 - I told there that when we have a target near , this exponential decay causes the current amplitude drop , and the power chain , in order to prevent the current from total dissipation , "inserts" its own linear rising function to the current signal . In another words , the power chain tries to "equalize" the start and the end point of each period , maintaining the stationary operation of the circuit in every conditions .

And this is why we see not the target exponent itself , but the target exponent + linear supplement function . This not a problem for signal processing , because I use further a correlation procedure , with a set of a linearly independent reference functions ( square , tilt , parabolic , cubic , etc ) , and can easily separate this linear append from the other function components . But what if we wanna see the PURE target signal , without any additional stuff ? This circuit also can easily do this . What we need is to use an appropriate algorithm of the kick circuit operation control , in order to maintain now the constant amplitude of the coil current with or without the target near the coil . With this operation mode , the kick circuit control rules the voltage so that every amount of energy being "sucked" from the coil by the target , would be restored by the kick circuit before the next measuring interval begins . In another words , it's something like an ALC circuit in the oscillator . I tried this mode too , and took a picture of the typical target signal on the output of the second gain stage . The upper trace is the second stage output , the lower trace is the first ( the signal is much weaker on it ) :



This is a pure target signal , a classical exponent decay function . This mode can be useful in experiments , target TC measurements , etc . Maybe I will use it in a version 3.0 algorithm , but I will do it later , after finishing 1.0 and 2.0 .