JSPI: using transmitter as discriminator

[Anonymous]

ref: Robert Hoolko's post approx. 10 Jan 02, in which he describes target responses during the transmit pulse (apparently) observed in an induction balance. He said that nonferrous goes up then down, whereas iron goes down only. (sorry, relying on memory here)
In a jumpstart pulse induction system, it is possible, if one wishes to do so, to jumpstart the coil at a high current, and let it decay to a much lower current before flyback.
If you get the timing and decay on everything just right, then iron and/or high-conductivity stuff and/or maghemite will be nulled or discriminated out, because the response to flyback won't be strong enough to erase the response left over from jumpstart.
Robert's observation that nonferrous goes up, then down, whereas iron stays down, is (if this was done in an induction balance) partly due to the permeability coupling between transmit and receiver, an effect which is irrelevant to eddy current response which is all that PI cares about. Nonetheless, there is a possibility that with the right timing and decay parameters, iron could be brought outside the nonferrous response region and isolated by itself. It is also possible that the same could be done to maghemite.
--Dave J.

 

[Anonymous]

Dave
I think you are quoting Dave Emery's reply to my post.
http://www.findmall.com/metal/school1/config.pl?read=2809
What he seemed to be saying is the opposite of what I would expect. But I am verbally challenged and need to see pictures to understand.
The plot below is what I would expect to see.
The black curve is coil current, and the blue curve is the response of a non ferrous target.
An iron target would also have a reactive response proportional to di/dt. The red curve is the sum of the non-ferrous response plus the reactive signal. I chose an arbitrary magnitude for the reactive signal, it would depend on shape and orientation.
I think if you took a couple of samples during the on time the conductive target would always be negative. The ferrous target could read positive then negative depending on shape and orientation.
For convenience I have ignored the different decay shapes of ferrous vs. non-ferrous.
Robert

 

[Anonymous]

[No message]

 

[Anonymous]

Hi Robert and Dave,
Fortunately, I still had a PPD1 in the museum to do some tests on. The PPD1 uses a stacked coaxial coil i.e. transmitter in the middle and a receiver coil either side. The coil looks a bit cumbersome, but this configuration has a simple response above and below the coil. The TX pulse is 350uS wide and the current has a normal exponential growth, just reaching maximum before cut-off. The scope was connected to the output of the first amplifier in the discrimination channel and the waveforms sketched for a range of objects. The top plot just shows the situation with no object and the time span displayed on the scope. The next plot down is with a ferrite rod brought near the coil. The rod is one that I use in probes and has no viscous signal, hence no response after TXoff. The curve is just the result of the permiability coupling and is the derivative of the TX current. With this coil arrangement, if the rod were placed near the other coil face, the responses would reverse in polarity. The next plot is a pair of steel pliers. Here we can see the downward signal from the permiability coupling plus a modified front edge due to switch on eddy currents. Also now apparent is the normal eddy current and viscous decay after TXoff. The fourth plot down is a silver ducat, which is a large, highly conductive coin. Now we see the response after TXon going the other way, but it dips negative about half way along the TX pulse. After TXoff we see the long eddy current decay from this type of object. The fifth plot down is a UK 50p cupro-nickel coin with much less conductivity. After TXon there is a fast decay that dips negative in about 75uS. After TXoff is the normal fast eddy current decay. The last plot is an aluminium ingot about 3in square and 1ft long. The measured time constant is 8milliseconds. This shows a pure conductive response after TXon, decaying to zero at TXoff, and a much lower amplitude, but very long, decay from then on.
I can photograph these from the scope if anyone wants to see them more accurately depicted.
It is obvious that where you sample during TXon has a great effect on the result from the discriminator channel. Particularly, low conductivity non-ferrous objects show as ferrous if you sample too late.
Have fun, Eric.

 

[Anonymous]

Thanks, Eric.
Anyone looking at the graphs with an intent to analyze them should realize that in the receiver waveforms at TR-OFF, there is actually a big negative-going spike not shown on the graphs which is relevant to analysis.
Your post suggests using the signals during the transmit-on time for discrimination by demodulating in the time domain, which is conceptually similar to VLF reactive discrimination. It requires an induction balance loop. In the past several weeks, there have been posts by several people suggesting what can be (and already has been) done by demodulating transmitter on-time signals to extract reactive components. The graphs Eric posted should help guide thinking in this matter.
Just to make sure nobody gets confused by this, the discrimination scheme I proposed in my post "JSPI: using transmitter as discriminator" on 13 Jan 02, is of an entirely different nature.
The basis of that system is to manipulate the shape of the transmit waveform such that after flyback, low conductivity targets will respond in one direction and iron will respond in the other direction. Whether it would be possible to isolate iron from high conductivity targets, I don't know yet. I doubt that anyone's target simulation software is good enough to answer the question: it would probably be necessary to actually build one and play around with it a while to optimize it, and then see what you have.
I'm somewhat doubtful that this scheme, by itself, represents a good choice, even if it works. I proposed it just to see if anyone else would see something in it that I missed.
If the scheme does separate all nonferrous from most iron, then it could be incorporated into another system which I will disclose later this evening.
--Dave J.

 

[Anonymous]

Eric,
Now blow everyone away and tell them what year you designed and sold the PPD1?
As with most everything else to do with PI's, I have consistently found that you did it first!
As to the coaxial coil, How do they compare to the good old Dual D or other coils?
I have made and used them on a couple of home made VLF's about 20 years ago. As I recall, they were very wide scan and worked well in bad ground. Dave. * * *

 

[Anonymous]

The only coax stacked coil I have actual experience with, is a 5 inch (or so) coil that Fisher made back in the 1980's for the 1260 & 1220, which used loops electrically different from those used on later 1200's. Its air sensitivity was within about an inch of the standard 8 inch coplanar loop.
I'm under the impression that coaxial stacked coils used to be a lot more common than they are now.
Coaxial stacked loops tend to be hard to align and keep stable, although I suppose with practice and ingenuity those problems could be licked. They're pretty much limited to the smaller sizes because a 9 inch coax stacked loop would have to be several inches thick. It'd be like having a huge loaf of bread out on the end of the rod.
--Dave J.

 

[Anonymous]

Hi DJ,
Yes, they would be like a loaf of bread for a VLF. Most PI's use a lot less wire and should be a lot lighter. Maybe litz wire would be a good idea? Candy has a patent on using litz wire ???
When you say that the depth of the Fisher coaxial coil was within an inch of the regular coil, was it an inch short or an inch deeper? I remember an article by Charles Garrett where he stated that a stacked coaxial coil was about as optimum a coil that it is possible to make.
The two that I built were for my home brewed 5KHz VLF/TR. They were indeed very heavy and as you said, they were a real problem to null. They did go deep and I was pleased with them.

 

[Anonymous]

The manufacturers of Litz have always said that the primary use for the stuff, was for winding coils. To say that the Candy patent fails the obviousness test seems a terrible understatement.
Litz has occasionally been used in the metal detector industry since before anyone ever heard of Minelab. So the Candy patent fails the prior art test.
Most patents are not defensible. Not Candy's, and not mine either.
Patents usually get awarded because the examiner (and often the inventor) was not aware of relevant prior art, or because the examiner was not sufficiently well versed in the subject matter to reject it with confidence. We would like to believe that nothing gets past the examiner, and a lot doesn't. But in the end, it's no problem for the examiner if an unenforceable patent is awarded-- it's a problem for the inventor/assignee, who spent money for a patent which will cause them to lose in court if they ever try to defend it.
Some examiners, when they find that an inventor intends to wear them down, will cave in just to get the case closed, laughing all the way home, because the inventor thought he won, when in fact he lost.
--Dave J.

 

[Anonymous]

Hi Dave and Dave J.
The coaxial coil I used was 8in outside diameter and 1.75in thick. Weight 16oz; only 2oz heavier than my current 8in coil. Charles Garrett kindly supplied me with the parts - injection molded outer shells and a very rigid IM internal former. Whole thing filled with polyerethane foam. No problems with stability, although fine balancing was continuous via feedback.
As to Litz wire, this has been used for upwards of 30 years in industrial high frequency balanced coil detectors.
Eric.

 

[Anonymous]

Hi Dave,
The PPD1 was in production from 1980 - 1983. I have designed a later version that uses coplanar concentric coils and that works OK. Never tried a DD but there is no reason why it should not perform well. In fact both coplanar and DD should be less affected by ground signal than the coaxial.

 

[Anonymous]

Hi Dave,
With regard to the negative going spike, I did not draw it in because I didn't think it had relevance to the signal. What I could have drawn was the gap of 20uS between the start of TXoff and the appearance of the decay waveform. This is the usual ramp down of the transmitter and damping to get rid of ringing. Sampling in the PPD1 was 30uS after start of TXoff. To get the responses shown, the metal targets were placed above the coil. Normally the targets would be under the coil, in which case all the responses would be inverted. This inverted signal is the one that the detector actually works on. Whether you place targets above or below the coil, the negative transition of the amplifier in the switch off time is always negative, and is just a result of capacitive and other imbalances in the coil system. Nothing to do with the target. This detector was in the days before shielded coils were necessary and no attempt was made to balance capacitive coupling. Ideally the voltage spike induced in each coil at TXoff should exactly cancel out, leaving maybe just a minor glitch. Perhaps you could sample then right into the off time.
Eric.

 

[Anonymous]

The sudden change in target current produced during flyback, induces in the receiver a corresponding voltage pulse. This pulse could get obscured by the (unwanted) capacitive coupling pulse which happens at the same time.
-Dave J.

 

[Anonymous]

Eric
From the shape of your plots there is obviously something seriously wrong with my simulation. Looking at your shapes, it appears that I am off by one derivative. That does not show up in the off time decay because the derivative of an exponential decay is also an exponential decay. So when all I had to look at was samples of off time decay the error was not obvious. It will be a day or two before I can look at the math and correct it.
Robert

 

[Anonymous]

Robert:
In the graphs you posted on 13 Jan 02 showing a simulated iron target, since you are graphing currents (field strengths) and not receiver voltages, the permeability component should follow the transmitter current rather than its first derivative.
In a post yesterday (don't remember which one) I suggest that the permeability component will always exceed the reactive part of the target response, during the transmit period. However, this is not the same as saying it will at all points in the transmit period exceed the total target response, as Eric's oscilloscope plots illustrate.
--Dave J.

 

[Anonymous]

Howdy folks,
I just looked up my lab notes. The waveforms I saw on my balanced Dual D coil are exactly the same as what Eric has posted. I never used a large aluminun ingot, but the rest are the same.
I sampled the receiver with a 2uS window, sample and hold which triggered on the positive edge of the transmit pulse. I used a second sample a little later. Dave. * * *

 

[Anonymous]

Dave
You are right, I was plotting the target current. I never finished the last step of converting it to coil voltage.
I have fixed the simulation spreadsheet. Now the curves look similar to Eric's sketches for conductive targets, and they agree with Dave Emery's description. I will redo the plot of a ferrous object, I think it will then look like Eric's sketch of steel pliers.
Robert

 

[Anonymous]

Hi Dave Eric
The resistance of the RX winding,does that have any bearing on the sensitivity of the coil.
Regards Frank Wallis

 

[Anonymous]

I had an error in the previous plot I added the reactive signal to the wrong thing. This is another attempt.
The black curve is coil current, and the blue curve is the response of a non ferrous target. The red curve is the sum of the non-ferrous response plus the reactive signal.
This now seems to agree with Dave Emery's description.
Robert

 

[Anonymous]

I'm not going to dig out the patent and reread it, but if my emf messed up memory, serves me well at all, Candy patented using up too a certain size of litz wire (the size of the small gauge wire making up the total litz).
The deal went something like this, and wasn't obvious at first. The PI, with a DD coil is going to induce currents in the cross section of the individual wire segments. Especially in the places where the two coils cross over. Makes sense. It will take time for these to decay. Nothing new so far. And whether you use silver coated copper wire , or just copper or aluminum or lead or steel wire, it won't make much difference --- as long as it don't change. (especially AC systems)(he didn't say all this, I am).
So what's the rub? Well I guess if you move the detector up and down and around and around different soils, and dielectic contants, and some hemymegamagnatight, then the magnetic transmit field is going get all distorted from the text book picture and this will induce new eddy currents in the receiver wire, which is right in your amplifiers face, so you will see this little change, unless you use litz wire and the right size litz wire for this extra special application, and thats what the patent part is about.
Least thats my guess.
JC