felezjoo PI(the best pulse induction metal detector that I made until now)

[ivconic]

Yes, that's very interesting. As far as I can tell from your videos (especially the last one) the pulse width doesn't seem to change, but there are discrete gaps between the pulse chains. I'm not sure whether this has any significance, or simply that the TX oscillator goes off while the micro is processing data.

That's why you hear audio response also; i tested to see how it will behave with detection signal from coil too.
I might be wrong but it's like there is a difference in pattern?
Things are moving fast and my hands are occupied holding the "wires" and camera...
Without coil there is no tx, but i didn't recorded that.
Obviously it's not just any dull tx.. yet it's not neither the ML waveform...
Right now i can't use both channels, don't have two probes yet.
Ordered them and awaiting to arrive...

[green]

Interesting thread. Reply#9, (sample rate, 200pps) (gain, 10,000) Reply#576 has gain at 2000. Not a big deal but how do you calculate gain? The amplifier has a 5pf cap and a 1 meg resistor in the feedback(32kHz cut off) 10 volts p-p at 32kHz has a slew rate of about 1 MHz. Does it need 50v/usec slew rate? The low sample rate and low amplifier cut off frequency goes against what a lot of the other threads suggest. The detector seems to be working fine, so I'm interested in understanding why. I've charted a lot of different targets in the past. I'm wondering if we can select a few that would be available to all to use as test targets. I charted some cut from an aluminum coke can plus another cut from a different can. The odd can didn't chart quite the same, so I'm not sure what to suggest for targets. Some in the low time constant range would be good. I like the 10x10, 20x20, and 40x40mm aluminum can plus maybe a coin or two. Time constant data from a smaller diameter coil.

[Qiaozhi]

The gain is 2000, not 10,000. That was a error made early on in the investigation.
From the GBP of 20MHz, this puts a limit on the cut-off frequency of 20E6/2000 = 10kHz.
I have run a simulation with the LF357 opamp, and with the AD745. Although both opamps have a GBP of 20MHz, the LF357 still settles slightly quicker due to the faster slew rate. The AD745 is a rail-to-rail device, so the signal swing is larger, but although they both come out of saturation at exactly the same time, you can readily see that the LF357 reaches steady state in a shorter time. The NE5534 is not even in the same ballpark. So slew rate seems to make a difference. What this equates to in practical terms, we'll have to wait and find out.

[ivconic]

The gain is 2000, not 10,000. That was a error made early on in the investigation.
From the GBP of 20MHz, this puts a limit on the cut-off frequency of 20E6/2000 = 10kHz.
I have run a simulation with the LF357 opamp, and with the AD745. Although both opamps have a GBP of 20MHz, the LF357 still settles slightly quicker due to the faster slew rate. The AD745 is a rail-to-rail device, so the signal swing is larger, but although they both come out of saturation at exactly the same time, you can readily see that the LF357 reaches steady state in a shorter time. The NE5534 is not even in the same ballpark. So slew rate seems to make a difference. What this equates to in practical terms, we'll have to wait and find out.

Now is proper moment for one of your previous claims to become very interesting and actual!
Now... let's try to imagine how to replace LF357, the way it is and as you described it; with much faster stage!
Look back through your posts and you will dig it.

[Qiaozhi]

Now is proper moment for one of your previous claims to become very interesting and actual!
Now... let's try to imagine how to replace LF357, the way it is and as you described it; with much faster stage!
Look back through your posts and you will dig it.

I've already considered that, but we need to investigate the current design as published before making any modifications.

[ivconic]

I've already considered that, but we need to investigate the current design as published before making any modifications.

Agree.
But since it is based mostly on coding and we don't have source... tough!
Only option we have is to widely experiment.

[ppodg]

lm318?
+an222, bcm847 instead of lm394
see also an006831

[Qiaozhi]

lm318?
+an222, bcm847 instead of lm394
see also an006831

Yes - the LM318 has a GBP of 15MHz and a slew rate of 50V/us.
In fact there is an LM318N on the Farnell website [146-8915] with a slew rate of 70V/us.
I put an LM318 model into the simulation, and it looks like an excellent replacement for the LF357.

[ppodg]

It only needs good decoupling

[ivconic]

Page 13, post #312...

[ppodg]

ivconic,
it looks like aliasing, try faster tb.

[ivconic]

ivconic,
it looks like aliasing, try faster tb.

Since it actually is real dsp it may be liable to aliasing if analogue part is not filtered well. (Actually we can only talk on analogue part until we get source somehow.)
You said "decoupling" and i would like to hear more detailed explanation.
Explain "tb"...excuse me for my English!

[ppodg]

Page 13, post #312...

sorry - i've missed it... but look at figs 3&4@an006831

[ivconic]

Analyzing wider context; it becomes obvious what is all about.
"Plain and ordinary" coil will not do, due large inter-capacitance between windings.
So to dsp works accurately in this case; it demands perfectly clear signal and fast recovery at front end.
All we can do here now is to focus on those things.
More better filtering is good idea.
Coil with less inter-capacitance between windings is first step.
My first 3 coils sucks, all made on "traditional" way. (windings are close, large inter-capacitance)
I made my 4th coil from stranded wire (less inter-capacitance) and it is working "satisfactory", actually not bad at all, but i feel it can do much better.

[ivconic]

sorry - i've missed it... but look at figs 3&4@an006831

...