Preamp

[Sean_Goddard]

I will do that this evening. Sorry about the size of the drawing, I'll have to redraw it as I didn't save the original *face-palm*. Values are largely irrelevant at this stage and will need tweaking.

Regarding the noise Teleno, can we activley damp the coil and will this help reduce it? I know another manufacturer cites avalanche noise from the FET as a problem which is why they use a snubber. Also those front end diodes look like a nice noise source too, can we eliminate those, maybe by actively switching out the high voltage portion of the Rx signal, and only switching through a voltage which will not either saturate the front end amp and is low enough to process directly.

I think the stage which can take the high voltages, like a common, base arrangement (nice low input capacitance) suggested earlier, along with active dampling and only switching the required section of the signal through to the front end amp is the way to go. Unless anyone has any better ideas, can we pursue this line of design please?

[Monolith]

Monolith had some good questions about noise in reply #39. To start leave the coil stationary(no EF signal). Target signal probably less than 10 Hz. Including a scope trace of integrator out sweeping a US nickel across the coil. The bottom width varied from 100 to 200 milliseconds sweeping the coin at speeds I thought would simulate sweeping the coil. The PI post amplifier circuits I've looked at roll of at 10Hz or less. Think we need to look at post amplifier out when looking for microvolt noise at the coil. Still trying to answer some of Monolith"s questions.

This looks like a perfect target response. No noise visible.
Now, if we could compare this signal with the signal at extreme distance from the coil, we would see where the signal disappears into the noise.
What are the frequencies included in that noise? The frequencies will give us hints of where the noise comes from.

By the way, I like to hang my target on a string from the ceiling. The target swinging like a pendulum past the coil gives the near equivalent of a steady sweep speed. A target in the hand is strongly influenced by the response signal of the hand itself.

The "standard" sweep speed for de-mining, is 1m/second. We use this as a very approximate standard as the ideal sweep speed is also related to the coil diameter.
Considering a time lag of 100ms from the peak target detection to the audible signal, would be a distance of 10cm. Spelled out, this means we hear the indication of the target when the center of the coil is 10 centimeters (4") past the target. Still OK for pin-pointing with a 30cm diameter coil.
A large time lag makes pin-pointing very difficult.

[green]

Yes Teleno, EXCELLENT and Monolith, I agree 100%. So get those ideas out there.

Now back to the noise. Have we identified where it is coming from? Has anyone put a 'scope on the preamp output to see what the signal is like from there?

My first thought is winding movement in the coil. Can someone pot a coil in resin in such a way that the winding CANNOT MOVE then we can see if that helps. After all we are dealing with microvolts here, less in some cases. I have to ask others to do these things as I work away from home and do not have access to my workshop or I would be doing the work myself and posting results on here.

I feel the better we can get the Rx signal the better depth we can get, remember, garbage in, garbage out. I have attached a "first thoughts" schematic on how to remove some noise with a short and long sample, but in the first instance we need to reduce the front end noise to a level where everything else is just polishing the corners.

Teleno, my maths is terrible but I see how you are doing the ground balance. Two samples with different contents. Since the ground is "constant" (in theory) then it should be possible to "null" it out?

What about bipolar pulsing to negate the ground? The theory being that a target will have little residual when non ferrous, but WILL have some residual when ferrous (BH curve) due to the RELUCTANCE of the target to have its magnetic polarity rapidly changed. Get the idea?

Monolith had some good questions about noise in reply #39. To start leave the coil stationary(no EF signal). Target signal probably less than 10 Hz. Including a scope trace of integrator out sweeping a US nickel across the coil. The bottom width varied from 100 to 200 milliseconds sweeping the coin at speeds I thought would simulate sweeping the coil. The PI post amplifier circuits I've looked at roll of at 10Hz or less. Think we need to look at post amplifier out when looking for microvolt noise at the coil. Still trying to answer some of Monolith"s questions.

[Teleno]

Now back to the noise. Have we identified where it is coming from?

Mainly fron this:



The hgher the resistance, the higher the noise (increase follows a root square law)

[6666]

Sean any chance you could repost a larger schematic please, when you zoom in to read values the picture pixelates

[Sean_Goddard]

Yes Teleno, EXCELLENT and Monolith, I agree 100%. So get those ideas out there.

Now back to the noise. Have we identified where it is coming from? Has anyone put a 'scope on the preamp output to see what the signal is like from there?

My first thought is winding movement in the coil. Can someone pot a coil in resin in such a way that the winding CANNOT MOVE then we can see if that helps. After all we are dealing with microvolts here, less in some cases. I have to ask others to do these things as I work away from home and do not have access to my workshop or I would be doing the work myself and posting results on here.

I feel the better we can get the Rx signal the better depth we can get, remember, garbage in, garbage out. I have attached a "first thoughts" schematic on how to remove some noise with a short and long sample, but in the first instance we need to reduce the front end noise to a level where everything else is just polishing the corners.

Teleno, my maths is terrible but I see how you are doing the ground balance. Two samples with different contents. Since the ground is "constant" (in theory) then it should be possible to "null" it out?

What about bipolar pulsing to negate the ground? The theory being that a target will have little residual when non ferrous, but WILL have some residual when ferrous (BH curve) due to the RELUCTANCE of the target to have its magnetic polarity rapidly changed. Get the idea?

[Teleno]

My proposal for ground balance: http://www.geotech1.com/forums/showt...718#post208718

[Monolith]

Just an idea, but what if we take TWO simultaneous samples, one long and one short then process them (probably by subtraction) that would give us N(oise) + s(ignal) from which we could then average one to get S(ignal) then maybe subtract (invert one signal and add)?

Just a suggestion. I have yet to work out the semantics of how this could be best done.

Dave, I would like to use the 9ct gold ring as the target reference. if we can get a good response on that then we should get good results overall. I'm aiming to 12" in wetsand on that target. I think we can get that if we solve the stable target sample issue as the rest is just a matter of gain in the later stages IMHO.

I have also been looking at this problem from a power supply point of view. Has anyone else monitored the supply lines of the front end amp to see if there is "bounce" on them? This would imply we have to use a front end with a high PSRR.

One final thought, eveyone else samples VOLTAGE, why not sample TIME? What I mean is that setting a threshold and measuring the time taken to decay back to that (simple comparator). I wonder if there is any mileage in that as a discrimination method as we know dv/dt curves are a function of conductivity . Again, just a thought.

One final thought, eveyone else samples VOLTAGE, why not sample TIME? What I mean is that setting a threshold and measuring the time taken to decay back to that (simple comparator). I wonder if there is any mileage in that as a discrimination method as we know dv/dt curves are a function of conductivity . Again, just a thought.
The way to do that is by using the time to define the TC of the target.
Once we use the TC of the target instead of the amplitude, we have a GROUND BALANCE.

Explaining: When we use the amplitude of the target response, changing the distance of the ground or/and target, changes the amplitude of the response.
When we define the target as to it's TIME CONSTANT or TC, the distance does not change the signal response. Therefore, we have a fundamental GROUND BALANCE.

After all, we keep saying our PI technology is in the time domain. So why not use the time to define our target?

WARNING:We need to discuss this GROUND BALANCE METHOD EXTENSIVELY, so that everybody knows that it is in the public domain and can therefore not been patented.

[Sean_Goddard]

Just an idea, but what if we take TWO simultaneous samples, one long and one short then process them (probably by subtraction) that would give us N(oise) + s(ignal) from which we could then average one to get S(ignal) then maybe subtract (invert one signal and add)?

Just a suggestion. I have yet to work out the semantics of how this could be best done.

Dave, I would like to use the 9ct gold ring as the target reference. if we can get a good response on that then we should get good results overall. I'm aiming to 12" in wetsand on that target. I think we can get that if we solve the stable target sample issue as the rest is just a matter of gain in the later stages IMHO.

I have also been looking at this problem from a power supply point of view. Has anyone else monitored the supply lines of the front end amp to see if there is "bounce" on them? This would imply we have to use a front end with a high PSRR.

One final thought, eveyone else samples VOLTAGE, why not sample TIME? What I mean is that setting a threshold and measuring the time taken to decay back to that (simple comparator). I wonder if there is any mileage in that as a discrimination method as we know dv/dt curves are a function of conductivity . Again, just a thought.

[daverave]

yes i can see the waveform...maybe once ive put circuit in metal box the threshold stability should be better.

[Monolith]

my test items are US nickle...9ct gold ring....and UK old half crown coin...if i went down to say 5pf cap the noise on the wave form was still there but at 10pf a lot better....i have no frequency in my circuit above 4 khz.

The Nickel has a short TC, but it is a relatively large target, good for testing.

Do you see the waveform on the Scope?

[daverave]

my test items are US nickle...9ct gold ring....and UK old half crown coin...if i went down to say 5pf cap the noise on the wave form was still there but at 10pf a lot better....i have no frequency in my circuit above 4 khz.

[Monolith]

i just used the normal 1 mega ohm feedback resistor....if i used a 15pf cap across the feed back resistor the threshold was really nice and stable but again i had to move the sample to different position to around 30 uS delay and the detection depth was fair but not as good as without the capacitor.

The cutoff frequency of the 1M-15p LP filter is about 10kHz. At this frequency it also integrates short TC targets.
What kind of test target do you use?

You say the threshold was stable. This means that there are several additional stages of signal processing up to your "stable signal". But the noise is less with the filter on the pre-amp, ergo, the noise is present at the pre-amp stage.
So lets see if we can define the noise frequency.

Do you have anything running at a frequency of above 10kHz in your circuit?
Is there anything that could produce harmonic frequency noise in your circuit?

Could the noise be from outside? like fluorescent lights?

Could you try 2p, 4p 8p 10p to see at what frequency the noise really starts?

[daverave]

i just used the normal 1 mega ohm feedback resistor....if i used a 15pf cap across the feed back resistor the threshold was really nice and stable but again i had to move the sample to different position to around 30 uS delay and the detection depth was fair but not as good as without the capacitor.

[Monolith]

i tried a small 10 pf capacitor across the fed back resistor of the first amp...made the threshold more stable but i had to change the sampling position to get any signal...so i just now removed the 10 pf capacitor and put it back to normal...the noise from the first amp seems to be the major problem of instability of the threshold...hope you find some way around this problem.

That is good and useful information. What is the value of the feedback resistor? Then we can calculate the frequency of this low pass filter.