complete project

[6666]

Tinkerer your schematic seems to list the pic as being a PIC16F690-x-p
is that some generic label ?
I can get PIC16F690-I/P will that work ?
and do you have any pictures of your coil ?
thanks
6666

[Tinkerer]

Reminder:

Anyone made the measurements, I have raised recently?
It would be very interesting to show the effect.

Single freedom of parameter (a simple threshold) isn't enough to GB or to discriminate.
Aziz

I am not sure what you want to demonstrate with the test you propose. But here is some data that is pointing more or less in the direction you indicate.

The response is different for different targets.

Now, there are many possible differences with in targets. Actually millions of variations. If We want to differentiate the targets, we need to start somewhere. We could start by dividing the different responses into a few categories and then classify the targets within these categories.

One way is to take many samples along the RX signal wave form and look for differences in the response at the different sample times.

The more different samples we take, the more information about the target we can gather.

Here is a small example:

Taking 4 samples at different times, with a target of gold, FE or iron, and a ceramic magnet. The ceramic magnet was actually used for the purpose of representing the Earth magnetic field.

Sample 1, gold down, FE up, magnet down. Representing in binary 0 1 0
Sample 2, gold up, FE down, magnet down. 1 0 0
Sample 3, gold up, FE down, magnet down. 1 0 0
Sample 4, gold down, FE up, magnet down. 0 1 0

We see that we could add sample 1 to sample 4 and get more gold signal amplitude, but still have the EF response within the total response.

If we invert one sample and subtract, we get a different result. What I mean to say, is that using a mathematical formula, we can manipulate the result, to discard the part that we do not want or enhance the part that we want.

This does not take into account the size of the target that would be represented by signal amplitude, nor does it take into account the conductivity of the target, that would be represented by the TC of response decay.

Each one of these characteristics changes the signal amplitude at a certain spot along the wave form.

The same happens with the ground response, which is just a large target with specific characteristics.

So let's say we take 10 samples along the RX signal wave form. The time of each sample is chosen to give the most information about one specific aspect. We will then end up with a lot of numbers to manipulate.

We can further enhance or attenuate the response of each sample by manipulating the sample width.

Tinkerer

[Aziz]

Reminder:

Anyone made the measurements, I have raised recently?
It would be very interesting to show the effect.

Single freedom of parameter (a simple threshold) isn't enough to GB or to discriminate.
Aziz

[Midas]

The general idea is to keep all noisy parts together and away from the analog, where the noise is really not desired.

On the above circuit I have decoupled the audio transistor and added the capacitors to keep the hard switching square wave noise out of the supply.
In fact, I see no reason why the whole audio should not be on the TX board.

Tinkerer

I'd like to discuss this a little bit more because there is a lot of different factors to consider:

1. PIC on RX board near front end amplifiers - bad
2. PIC on TX board near coil inputs - also bad
3. PIC on TX board with ADC input far from final stage amplifier output on RX board- bad, maybe not as bad

Is the TX board the right place for the PIC once it starts performing late stage signal processing? In this configuration it has two negative qualities but maybe being away from the front end amplifiers makes it worth it ? Maybe on the RX board is better, who knows ? Not me...

Midas

[6666]

Thanks pelanj, 16F690 is ok, there are tons on fleabay
cheers

[Aziz]

Hi Aziz, I can see what your talking about.
Mick

Excellent Mick.

You can put a gold ring nearby the iron plate in the "iron orientation response". This should show the "target masking effect". It is similar to a hot iron mineralized ground, masking out the target.

If you can show some CRO pics, others could see it too.

Cheers,
Aziz

[pelanj]

I use what I have at my hand right now - that is 16F690. All I need is ADC at the front and PWM output. Some free digital I/O pins could be used to select different modes, use as mute or something like that. This one should be powerful enough with enough memory as well - but once the program is ready, we can try to squeeze it into some lower model if 16F690 would be overkill.

[6666]

pelanj

what type of PIC are you using ?

[mickstv]

Ok Mick, take a flat iron plate and change it's orientation to the coil. Should show the same effect.

Aziz

Hi Aziz, I can see what your talking about.



Mick

[pelanj]

I will test how the audio responds to a pot on the low pin count demo board. And you are right, even low precision sine is better than square. If the PIC is able to do a high precision sine generator (I have seen some projects like that). Once I have the sound generation right, it can be used instead of the hardwired VCO.

[Tinkerer]

I am still working on my audio PIC - it will use either one tone, which would shift up or down with the voltage at the ADC input or a two tone, low for iron, high for non ferrous mixed together. I think about using the PWM and some 32-64 level lookup table for the sine. It will be programmed using XC8, as that one is the easiest for me. I cannot promise any dates, but one day, it will come. Once I have the first version, I will post the code for comments and improvements.

Thank you for the update.
It sounds good. Probably it is a good idea to try any execution as soon as possible to see how it works with the complete circuit.

The sine wave does not need to be very perfect, anything is better than a square wave.

Tinkerer

[pelanj]

I am still working on my audio PIC - it will use either one tone, which would shift up or down with the voltage at the ADC input or a two tone, low for iron, high for non ferrous mixed together. I think about using the PWM and some 32-64 level lookup table for the sine. It will be programmed using XC8, as that one is the easiest for me. I cannot promise any dates, but one day, it will come. Once I have the first version, I will post the code for comments and improvements.

[Tinkerer]

IMHO you miss a capacitor in series with the noninverting input circuitry of U4a to fix all the offsets from the previous circuits.
Yes, that's the whole point of a step voltage source - it does not have a short lived exponential fall as PI does. I observed this very behaviour in my simulations right when I joined this forum, and wondered - how come that no one uses it ?!? My simulations used low voltage and symmetrical Tx supply with equal pulse durations on + and - side.
The catch is that targets present themselves as a loaded transformer, where a Tx coil's resistance/load is somewhat added to the target resistance/load. The damping resistor does that - it is transformed from a Tx coil to the target and appears as an additional series resistance, hence shortening the target tau. With Tx coil being attached directly to a voltage source, its resistance is the very resistance of copper wire, and that is not much.
Otherwise you can distinguish iron from, well, everything else. At best.

I believe this is a surefire approach to better detectors with discrimination.

Davor, thanks for the tip of the AC coupling of U4a.

There are other areas of improvement possible.

For example, the integrator- difference amplifier at U1d. The inputs are uneven because of the low impedance of the minus input and the high impedance of the plus input.
Maybe using a 2 opamp instrumentation amplifier design would bring some improvement.

Or divide the function into 2 parts, separate integrators and then an instrumentation amplifier to remove common mode noise.

What do you think?

Tinkerer

[Tinkerer]

Thanks Tinkerer your a gent

Regards

Dave

Remember that this is just a simple basic design. You notice that the motion version is nearly the same as the non motion version.

Now we can take this basic design and adapt it for a specific purpose, for example a pin-pointer.

There are lots of improvements that can be done on a pin-pointer design.

And many different choices. For example, it can be made autonomous, everything incorporated in a stick. No cable. Or it can be made as an attachment to a normal detector. Or it can be made with a short cable and attached to the upper digging arm.

Many variations are possible.

Then there are the possible variations of target indication. For example a vibrator, like cell phones use.
It needs some thinking of how to make the indicator with an autonomous just a stick design.

So, basically lots of thinking and figuring out to do. Too bad that only few people have a thinking hat.

Thinkerer

[Tinkerer]

all major pi fails for the audio circuit, we say that 3 o 4 years ago in the forum , we put a lot of efforts in the front end etc...circuits ..... but a very poor design audio , 50 mV too much ...first use a correct VCO you need a specialized circuit like a ad 654 and mix it with the treshold audio classic ....it is the best for me for audio....

Alexismex, thank for the feedback.

The AD654 seems to be a step in the right direction.


What about using a Digitac Direct converter, like the AD9837. It produces a sine wave that is easy on the ear and frequency modulation.

Tinkerer