Universal PI Micro

[bugwhiskers]

Hi okantex,

There are now only 2 samples required. I started with 64 thinking it was going to be very complicated and it ended up being very simple.

The sampling can start almost anywhere in the knee and away from the saturation, it's the positioning of the second in relation to the first that is important.

The pivot point is set to be the point where the user wants the signal to change from negative to positive. The change to negative with Steel and Lead is far more pronounced than the very slight change from Copper to Silver.

The attached .PDF is a rough schematic of the circuit to achieve the discrimination. The 2 samples are positioned so that with Aluminium as the target the final Op amp output is slightly positive. Steel and Lead will then cause a very strong shift negative and Copper and Silver more positive.

regards
bugwhiskers

[okantex]

is the pivot point in the first 6.4us of knee.
according to delay graf . it looks like at 150us .

and for bars did you divide 64 samples to two 32 samples. than average of two group makes two different bars.
Am I right.?

[bugwhiskers]

Hi oktanex,

The attached .PDF will help explain the processes.

Things to note:
Sample 2 is amplified by 2 and the position (pivot point) is adjusted so with an Alum/Gold target the output of a differential amplifier fed with samples 1&2 produces a zero or slightly positive output. Any metal type with less conductivity than Alum/Gold will then produce a negative voltage.

The labelled curves are indicative of the flattening of the decay curve with increasingly conductive targets.


regards
bugwhiskers

[bugwhiskers]

Hi okantex,

It's very late here, tomorrow I will make a drawing and post it.
The signal is related to the conductivity of the target with Lead and Steel being at one end and Silver at the other with Aluminium and Gold at the pivot point as set by the positioning and weighting of the samples.
It sounds terribly complicated but in reality is ridiculously simple.

regards
bugwhiskers

[okantex]

BTW
did you chabged screen.
you are just talking aabout two vertical bars.
according to your explanation. I see almosst everything clear accept silver.

how do you make up this bars. can you give formula
cause your silver discription looks like what expect from copper.
for silver I was thinking seceond bar much lower than iron.

still wonder what about decay time measuring to see delay.

[okantex]

Hi Bugwhiskers
I can not understand term signal strenght.
what is it and how do you measure it.
is it integral of a certain(preset) period.

do not you need to get samples to see full delay time to have discrimination.

and one more question.
I could not understand term saturation for opamp.
but I see you are takeiiing samples from the begining of knee(yield point)
how do you decide to start sampling,I mean how do you understand the time for sampling occurred

[bugwhiskers]

Hi okantex,

I tried to take some photo's but they were blurry due to the noise. What they showed were 2 vertical bar graphs. The left graph represented the early sample and the other bar the later sample. With steel as the target the left graph went much higher than the right bar (about double) with Lead slightly less high. With Aluminium and Gold the bars stayed equal and rose together with increasing signal strength. With Silver the right bar went slightly above the left. In software it is easy to ignore any signal where the left is higher than the right leaving only Alum, Copper Silver and Gold to produce an audible signal.
Many people would be happy to just eliminate Lead (bullets) and Steel (nails etc).
The board I am working on at the moment has 2 OP amps buffering the 2 samples with the later sample being amplified by 2. Carefull adjustment of the second sample position so that the 2 outputs when fed to a Differential amp will produce zero volts will go negative for any signal where the first sample is greater than the second sample and a positive voltage when the second sample is greater than the first sample. Another differential OP amp will remove the buffered late ground signal sample.

Anyone that has a dual channel CRO could lash up the circuit and feed each sample to each of the channels and observe the phenomena.

regards
bugwhiskers

[okantex]

do you have photos of effects of different metals .
I think there is a misunderstood that I will return baack after a search.

[okantex]

do you have picture of different metlas effect on LCD.

[Alexismex]

Very good work Bugw, and clear explanation of the procesus, it is a greatttt idea to switch capacitors secuencialy ,
shure you will have a good aproach to discriminate very well and with a K.I.S.S process ...bravo keep up good working,
Alex
PS:and for negative post came from negative and destructive mind, don't worry

[bugwhiskers]

Hi oktanex,

The samples are only taken in the knee of the curve, depending on your pre-amp's characteristics this may only be the first 10 uS after coming out of saturation. The information about the metal type via it's "conductivity" can be gleaned from that area.
Essentially, if the metal is a poor conductor ie Lead and Steel then the early part of the curve is more pronounced and if the metal is a good conductor ie Aluminium or Silver then the curve is flatter. Using this knowledge and getting a sample of the early part and comparing it with a slightly later sample will give info about the curve shape and therefore the metal type.

Regarding pre-amp types, the prototype used an LF356 which is compensated and has a rather slow response which gave a nice curve rather than a steep drop to nearly nothing.
It is said that the decay curve for metals is exponential, I think the metal type must change the exponent.

regards
bugwhiskers

[okantex]

Hi Bugwhiskers,
thanks for yoru kind reply
but still there is something that I can not understand.

as I know longest delay is for silver and it is around 270 us .so we have to listen almost up to 300us for each Pulse.Am I right?

you say you take each sample at 100nS and at total 64 samples in 6.4us.
at the end of 6.4th us your caps wiill be charged.
but ie. you still have 300-6.4 =293.6 us delay period to be sampled.

if time between each sample is 100nS ,how can you sample all delay period in one cycle.
as I know ADC converters are not so fast to read 64 sample in 1us .
for each sample adc needs almost 20us.
64*20=1200us is needed to red 64 samples.which must be done inleisure time between tx pulses.
what is the wrong calculation in my ..........

take care.
okantex

[bugwhiskers]

Hi okantex,

I am not sure I understand your question so I will re-phrase the way the sampling works as a flow sequence.

START:
switch 4053 so pre-amp goes to caps
Transmit pulse
Delay (wait for pre-amp to come out of saturation)
sample to cap 1
.
.
.
.
sample to cap 64 (approx 6.4 uS at 100nS per sample)
switch 4053 so caps go to AtoD input
cap 1 to AtoD, digitize and store to RAM
.
.
.
cap 64 to AtoD ,digitize and store to RAM
Process samples
Display to screen
Go to START

Post #76 has a very detailed description of how the samples are processed.

The method relies on sensing the difference between samples that are a result of the coil passing over a target.

regards
bugwhiskers

[okantex]

Hi Bugwhisker
it really a good project.
I was trying to get samples at each 5us but cause of adc speeds I could not.
later saw your project .fast sampling and reading it at tx leisure tiem.really good idea.

this is a part from your previous posts
after this part I will have a question if you do not mind
""""
The circuitry following the preamp gets 64 samples (via switched capacitors) of the early part of the decay waveform. The time between each sample can be as low as 100nS. At that speed it is possible to capture 64 samples of the first 6.4 uS of the decay waveform. The capacitors are then switched in turn to the AtoD of the micro and the digitized result is stored in RAM.

One area of RAM holds an average of all the previous samples with user adjustable software lag and another area holds the most recent. The lag is necessary to prevent small changes being swamped by the averaging. Ground signal and preamp drift will become part of the average set of samples whereas any abrupt change will be seen as a target. Because the sampling is done very early in the decay it should be possible to discriminate.

What is unique about this approach is the the samples are gathered very quickly and then later processed at leisure between the TX pulses. This technique obviates the need for a super fast micro and a super fast AtoD.
""""
you ara talking about reading caps after delay.in leisure time.that is okay .ı understand it.
but
you also saying to take 64 samples at 6.4us .so all at the begining all your caps are charged even before entering to discrimination area.
you alo say averageing 64 samples.if I do this ,I have to read Ato D at the 6.4us of delay.reading 64 samples will take alot of time.

will I read another 64 samples of next 6.4us area in next rx cycle

and could not understand how to switch one cap to another at that speed.
btw I am not an electronist.do you use transistor which is triggered by PIc.
for this aim.

and what is the sampling speed of atmega16-64
can I suse 16f877 or 18f series of microchip.
their adc convertion time is around 20us .

[bugwhiskers]

Hi Guys,
Thanks for your continued interest and vote of confidence.

After tinkering with the ADUC845 I have found that to get the best performance the AtoD has to be run very slowly. Whilst this is a problem for taking lots of samples and trying to keep up a good TX cycle frequency there is more than one way to achieve the objectives.
The advantages of realizing gain through increased resolution are that the pre-amp gain can be kept low allowing earlier sampling and the advantages that brings to detecting small targets and also the fact that whilst keeping amplification of the signal low the amplification of noise is also minimised.

The bus board with the 64 samples allowed me to work out how to discriminate. At the end of the process I was able to discriminate with just two of the 64 samples strategically placed.

A late third sample would allow ground balancing and so the next step is to set up some OP amps to be fed from the well placed samples with the end result being one output that will be converted with 24 bit resolution and sent on to the audio stage.

As you can imagine, the layout of this board with appropriate ground planes is critical if the noise is to be kept low. All of this takes time so be patient as I have to be also having sold off both of my commercial detectors in anticipation of a worthwhile result.

regards
bugwhiskers