Tweet
@Carl:
-
-
The next thing I tried today (more for the sake of order) ...
and with this I finish with larger diameter coil attempts; is this, it also works very badly!
So, the only solution I can think of is a coil under 20cm...
So far I have done many PI coils for Delta Pulse, FelezJoo PI, Barracuda, SurfPi, recently a few for Pulse Star 2...
I have NEVER had a problem with it working perfectly fine.
Obviously this is a whole different "beast"... a whole new area for me...
Too large horseschmitt... if you ask me!
Comment
-
I don't know why you insist on this coil = I am preparing myself to build my first PI detector, starting with building the best coil for it, so far I'm not satisfied even that I have 350V, compared to your coil 140V .
I'm trying to get on oscilloscope respond to the coin on the output of the first amplifier (pin #6 of U11), but I have nothing.
I have respond to 10x10 cm iron plate but not to the coin.
Your wire is about 1 mm in diameter and have resistance higher than my 0.5 magnet wire ?Comment
-
Haha I wasn't sure that resistor trick would work, just a guess. Worked great!
It looks like the TX is 200us (maybe a little more) at 500Hz. For high conductor coins the pulse width is good, but I would heat sink the FET and double the pulse rate.
The sample delay looks to be 30us which is way too high. You should be able to easily run 10-12us and with a good coil down to 6-8us. This will help sensitivity, especially for lower conductors but even high conductors.
You can also do better on the GB timing. Here is what I would try:
TX period = 1ms
TX width = 200us
Tgt delay = 10us
Tgt width = 20us
Gnd delay = 30us
Gnd width = 80us
The EFE spacing will need adjusting and R52 may need to be shorted.Comment
-
The EFE spacing will need adjusting = EFE is a signal from the Earth's magnetic field when the direction of movement of the detector coil changes, but this sampling is not synchronized with the movement of the coil, so the chance that the pulse will appear when the direction of movement of the coil changes is very small.
Between changing directions EFE = 0V
Earth's magnetic;
When moving the PI coil in the air, I did not notice any voltage on it due to the Earth's magnetic field, the coil has too few turns, my magnetometer coil has 30,000 turns (Magnetometer with a rotating coil)
Comment
-
Your guess is better than good!Originally posted by Carl-NC View Post
Even 3 pulses can be "caught" at the same time with the help of a trick.
And now as I watch this... I could try the others on the second channel, sync on the first channel, align both channels...
and get the timings as one oscillogram! I'm not sure it will work; but i will try.
But it's still good like this, you can clearly see the times. Your estimates are correct.
Yes the sample delay is 30uS, quite correct. And that is the minimum on the potentiometer.
And you're right again, when I turn the potentiometer and increase the sample delay; the detection of coins drops rapidly until
the moment when the coil no longer sees a coin at all. (Rough but good enough method to define TC of targets!!!... Wow! Ideas just keep flooding!)
For the purpose of searching for such small targets; the need for a potentiometer is lost, I can put a simple jumper and omit the potentiometer.
But if I change the timing, according to your "scenario", it might be better to keep the potentiometer for later.
Or...later when I establish a few "swetspots" around the delay; in imitation of Pulse Star 2: I can put a switch with a couple of precise fixed resistances.
I even like that a lot more. Because fixed resistors are more reliable than potentiometers. And there is no need for more than just a few states. There are 4 on PS2.
A good trick is an additional toggle switch for quick checking between low and high conductors. Some kind of pseudo-discrimination!
I like the "scenario" and will try to do so!
TX period = 1ms = 1kHz = 1000 PPS
TX width = 200us = stays the same
This is simple NE555 tweaking, alright.
So far easy.
But this:
Tgt delay = 10us
Tgt width = 20us
Gnd delay = 30us
Gnd width = 80us
... will need the tweaking on all 4 monostables.
I might need a help here...Comment
-
I might need a help here...= all of these pulses can be easy produced by stm32f103Comment
-
The signal created from EFE has a frequency exactly equal to the swing rate of the coil, which might be 1m/s. During that meter of movement, the RX has sampled 500 times so each sample span sees what looks like a DC offset. This offset is what gets cancelled by the EFE sample, so it's not important to synchronize to the sweep rate. If the pulse rate was far lower (like 5Hz) then synchronizing would be important.Originally posted by pito View Post
Comment
-
The first thing to do is reduce the main sample delay. This is controlled by R18,C15 and the 2 pots. Set both pots to zero and reduce R18 if needed to achieve 6us, then trim it up to 10us with RV2. Then, if you wind an aggressive coil you can trim it back down to 6us.Originally posted by ivconic View Post
There are 4 sample pulses, I call them TGT, TGT-EFE, GND, and GND-EFE. (TGT is the Gold channel, GND is the Iron channel)
U5a sets the delay of TGT with R18*C15
U5b sets the delay of TGT-EFE with R22*C17
U7a sets the delay of GND with R15*C4
U7b sets the delay of GND-EFE with R21*C16
U8a,b sets the widths of TGT, TGT-EFE
U9a,b sets the widths of GND, GND-EFE
Eric runs all the samples with the same pulse width. This is set by (R19+R20)*C19 and triggers U8,U9 through diodes D4-D7. It is important that the width of TGT and TGT-EFE match, and the width of GND and GND-EFE match. But TGT and GND don't need to match. My preference is to run the GND width much wider than the TGT width, this reduces the noise of the GND channel. To do this, disconnect D5,D6 from the other diodes and add a new RC to from U9-pin1 to ground with D5,D6 connected to the C, exactly in the same manner as R19+R20,C19. That will give you independent control of the GND pulse width. A wider GND pulse effectively increases its gain so that R52 may not be needed.
If you really want an education, set up all of these RCs with pots so that you can dynamically vary them and try different settings, seeing how various timings affects GB, target hole, and sensitivity. That was the purpose of the Hammerhead project, and it's easy to do here as well. You can do the same with frequency and TX width by partially replacing R7 with a pot. Be very careful with frequency and TX width as there are combinations that will burn up the FET. Also, it is possible to increase the frequency so that the EFE pulses end up in the TX-on cycle or even beyond.
I think with the resistor trick you can monitor all 5 pulses, since none of them are on at the same time. I will have to use this myself as even with a 4 channel oscope I often find myself wanting to monitor multiple timing pulses, plus some analog signals. I can put all the timing on one channel, and have 3 left for analog.
Comment
-
During that meter of movement, the RX has sampled 500 times so each sample span sees what looks like a DC offset. = sample every 2 mm , that is good but most of the time EFE is 0V so DC offset is not precise. ?
Comment
-
AHHHHHHHHHHHHHH! KT315, did not have that one schematic. YOU HAVE MY RESPECT, we know each other don't we???Originally posted by kt315 View Post
MELBETAComment
-
Originally posted by Carl-NC View Post
Will do!
Quite educational post, thanks!
I will go slowly, step by step.
Once I got it; I will report here.
First thing I do now is adding a heatsink, also replacing those 2 parallel 1k damper with one 470 / 5W (noticed those 2 get hot in time, 470 one is first that came handy), and 2R2 with 5W one too.
Now, I understand the role of 2R2, more like a "fuse" and to add safety resistance to coil... but from the photos of the original GS4 I see Eric put "inductive" resistor in ceramic case???
That's strange to me, because I learned long time ago to avoid "inductive" (wire) resistors on such places.
Also 2R2 could be omitted; providing that coil resistance is high enough. Speaking this in a sense of the noise reduction. But for now I'll keep it, S/N issues will come later, probably when everything else is done fine.
Carl.. what to say? Thanks! Indeed very educational post, in so few words; lot of useful things are said!
Cheers!Comment
-
If the Earth field is exactly uniform, and you are able to keep the coil exactly flat and at a constant height, then yes, the induced voltage should be zero. But things are not so perfect, and there is a slowly changing induced voltage, with most of it being at the ends of the sweep where people have a tendency to lift the coil slightly. Between each pair of TGT and TGT-EFE samples there is very little change in the induced EFE voltage (even at the sweep ends) so the 2 samples are close to equal (almost DC), and subtracting them will mostly cancel the EFE.Originally posted by pito View Post
EFE is only noticeable in a PI detector because it often has a lot of gain. If you disable EFE and very carefully sweep the coil, you might find it's not too bad. Until you get in an area with magnetic rocks.Comment
-
You can completely omit the 2R2 and make a 0-ohm coil if you like. The purpose of making the coil non-zero ohms is to somewhat flat-top the current pulse, which reduces the target reverse eddies during TX turn-on. This is especially helpful with high conductors. See this thread, especially posts 11 & 12. Also, the 2R2 doesn't contribute noise because it is removed from the RX path. Even coil resistance doesn't hurt noise, as R27 dominates the RX noise.Originally posted by ivconic View PostComment
-
I think I understood what you wrote.Originally posted by Carl-NC View Post
Yes, the R27...someone suggested earlier, and I agree, that it's better to put a 1K there.
And if I put anti-parallel diodes, then the value of R27 could be slightly lower.
Your opinion? And the difference in two diodes in the same direction and two antiparallel ones?
Is it related to the fact that the "ground" is at a potential of 14.4v?Comment
Comment