Triangular Wave Technology

[mikebg]

VERY INEFFICIENT TECHNOLOGY

Frequency analysis shows that triangle wave technology is very inefficient for wide band (PI) metal detector.

[Aziz]

Frequency analysis shows that triangle wave technology is very inefficient for wide band (PI) metal detector.

No problem at all.

You can vary the pulse width (PW) of the step voltage source (for instance coming from an H-bridge output). You get two different time constant current slopes (switch-on/switch-off). E voilà, you can do some kind of spectrum response equializing (rising the high frequency parts). As the coil current isn't symmetric anymore, you even get the even frequency harmonics (2*f0, 4*f0, etc.).

This simple waveform gives you more target response for the long TC targets than a high end PI detector.

Aziz

PS: Try a PW of 0.3 or 0.8 and see the frequency spectrum.

[Tinkerer]

Attached is a simulation of a 40us TX pulse and the response of a target of 10us. We can clearly see that the maximum "target saturation" has been reached long before the end of the TX On time.
At switch OFF, the current in the TX coil reverses. The eddy currents in the target inverse too and reach a much higher amplitude because of the higher di/dt.

The green trace is the TX pulse.

The blue trace is the target response.

The coil is critically damped.

Any comments?

Tinkerer

[Ferric Toes]

This simple waveform gives you more target response for the long TC targets than a high end PI

High end relic and nugget PI's that are currently on the market would be more accurate. You can of course make a PI do what you want by changing the timing. Look at the geophysical PI's, many of which have TX pulses in the mS range and sample very late compared to the PI's we are involved in. I have a unit which is relatively poor on small coin size targets, but will be hard to beat on anything of golf ball size and upwards. Also there is another way of sampling which looks at the total field return from the target and this enhances long TC signals at the expense of the short.

Eric.

[Midas]

Yes, there are a few different ways of doing this hypothetical experiment. I chose the exploding magnet as a quick way of shutting off the field, even so I doubt if the magnet would be reduced to dust in 2uS. Quickly turning it in a fast enough time might prove even more difficult.

Eric.

2us, no problem. Some explosive can detonate at over 9km/s, so for a 10mm magnet that's just over 1us. But for the best bang for your buck what you really want is an explosively pumped flux compression generator (EPFCG). Good for a few millions of amps apparently

This whole debate reminds me of one we had about a year back about whether it was to ones advantage to let your TX flat top to allow the target eddy currents to subside before the flyback. I think the answer to that was the same as it is to this question now, it all depends on TC of the target your after and the compromises your prepared to make balancing efficiency with sensitivity. I believe Carl posted some sims that I've played around with and one thing I've noticed is that if your going for really big nuggets then high voltage fly-back is just a waste of power. You actually get a better signal from the coil charging period.

Midas

[Sergey_P]

(1) PI
field (Ф) – exponent (f(exp))
a) magnetization: dФ/dt - exponent (f(exp))
b) conductivity: ddФ/dt - exponent (f(exp))
response from the targets = a+b
----
(2) Triangular
a) magnetization: dФ/dt - constant (f(const))
b) conductivity: square wave electric field in the target - creates exponential signal targets
response from the targets = a+b
---
(2) - has a lower sensitivity to the magnetization and high sensitivity to the conductivity (good for finding gold nuggets )

[Aziz]

Zilch target response or huge target response - that's the question.

PI detectors are inherent optimized for small (and low conductivity) targets. Due to the fact, that TX coil = physical same RX coil, you have to damp all the coil energy very quickly to sample as early as possible (response signals die out quickly).

You don't use the TX on-time response. (1/2 total response energy wasted)
You don't use the TX off-time flyback period until the sampling begins. (big response energy wasted)
You are just sampling a tiny fraction of the whole response signal. And you have to amplify the response signal many many times (amplifying EMI noise, amplifier noise *LOL*).

If you do some sophisticated signal processing (using a DSP/µC) on the triangular/quasi-triangular continious wave current principle, you can get more target response than a PI could ever ever get.
And this principle is prior art, easy, cheap, not critical, very power efficient , ... a proven good old technology.
You really get more for your money or effort.

Aziz

[Davor]

Exactly. There is only one "advantage" of a classical PI over this kind of excitation: it is deaf to sea water (and small gold).
I'm still very new here, and the moment I started learning about metal detectors, a step voltage excitation draw my attention as a straightforward way of direct reading the targets' eddy current response.

As Carl already noted, the only restriction is your power supply high voltage capability, but otherwise this principle is far more energy, and time efficient than the classical PI.

In a way the approach with short and long step periods makes sense, as the bursts of short ones may "condition" the ferrous ground particles just as a CRT demagnetizer does.

[Aziz]

Are you guys interested in the frequency response of some triangular wave forms?
With varying pulsewidth too.
(I could quickly implement it and do a FFT)
Aziz

[Davor]

Go on, let's see what's on your mind

[Ferric Toes]

Attached is a simulation of a 40us TX pulse and the response of a target of 10us. We can clearly see that the maximum "target saturation" has been reached long before the end of the TX On time.
At switch OFF, the current in the TX coil reverses. The eddy currents in the target inverse too and reach a much higher amplitude because of the higher di/dt.

The green trace is the TX pulse.

The blue trace is the target response.

The coil is critically damped.

Any comments?

Tinkerer

That is exactly what I would expect to see. The total decay of the target trace post switchoff extends for about 50 uS, so as you say, it has a TC of 10uS. Looking at the negative decay it would likely extend for the same time, except that it is curtailed by the 40uS to switch off. However as the starting amplitude is lower, for the reason you give, it appears flatter. At 5 TC's though, there is only 1% of this smaller signal that subtracts from the switchoff decay, which is insignificant.

Eric.

[Aziz]

Go on, let's see what's on your mind

http://www.wolframalpha.com/input/?i=fft%28trianglewave[x]%29
(fugg, link doesn't work. Just enter in the edit line: fft(trianglewave[x]) )
(requires more computation time)

Damn it, I have to code it by myself.
I'll look for online FFT calculators before I start the coding..
Aziz

Wait: Look here... (for m = 3, 4, 5)
http://mathworld.wolfram.com/Fourier...angleWave.html

Wow, avoided a lot of coding work now..
Aziz

[Ferric Toes]

I would still like to hear an answer to this question - "Are you saying that if you pulsed a coil with, say, just a 2uS 500V pulse, then that is a good as a long low voltage (100uS or more) conventional pulse that generates a 2uS 500V flyback anyway?" Unless I am mistaken, that was what was implied a while back in this thread. We do need to get the basic fundamental facts right to be able to progress PI. The only use I can see for a high voltage pulse is to "kick start" the otherwise slow growth of current in the coil inductance. After that, the normal power supply voltage would take over until the switch off. Instead of wasting the stored energy in a damping resistor, it could be recycled to do this kick start on the succeeding pulse: which is easier to do on a bipolar design.

I still firmly believe that an appropriate length TX pulse is necessary prior to switchoff and flyback in order to set up the necessary field and energise the target. However, I am always ready to listen and learn.

Eric.

[Aziz]

Ok guys,

I have been told, that you all love math.
Check this zipped Excel sheet out. Just enter the PW field and see the spectrum distribution for up to n=8.

FFT-TriangleWave.zip
(I hope I didn't make a mistake.)

Cheers,
Aziz

[Ferric Toes]

Exactly. There is only one "advantage" of a classical PI over this kind of excitation: it is deaf to sea water (and small gold).

Classical PI can do the above. Use a delay of just 5uS or less (which is achievable with careful design) and you will detect both sea water and small gold. The trick is to eliminate the sea water response and still detect small gold. I'm sure it would not be too difficult to design a seawater balance circuit to do this. A TX/Fig8RX would be a good start. You could then hunt for fine gold chains on a flat seawater beach.

Classical PI as a principle is extremely flexible. i.e. you can sample in the ON time with a balanced coil: you can get the sensitivity to small objects up to a similar level as a Fisher Goldbug 2: you can optimise for large good conductors deeply buried. Early time short samples are no good for this. You want to start early and end late and this will remove the predisposition to small targets.

Eric.