Preamp

[green]

If we can turn that into a reality, then the problem of thin gold not being detectable is cracked. That amp is F A S T:......!

Is the noise figure one you put in or a calculated one?

I can but marvel at what you are doing as this depth of analogue design is outside my sphere of current abilities. But I think we have a winner in the input amplifier stage design. My only comment is that I think a gain of 1K is a little high, in order to use this in Oz it will have to be a lot lower.

Now something else not mentioned here, what about saturation and recovery? From what I see there is no problem with this with respect to this design. Does this mean it ois NOT a problem now?

Peak coil current is part of the gain equation. One amp with gain of 1000 same as ten amps with gain of 100. If thin gold is included in what we are looking for what would peak current and gain should we try for?

[Teleno]

One amp with gain of 1000 same as ten amps with gain of 100.

Oly if the flyback decay rate is the same, which means 10x higher flyback voltage for 10 amp.

Your compromise is to barely reach reach the breakdown voltage of the MOSFET that drives a given coil.

Cc = coil capacitance

Co = Mosfet output capacitance
Vbr = MOSFET breakdown voltage

L = coil inductance.

This is the maximum I you can work with for a given coil.

For small gold depth is not as important as a fast flyback, therefore you'd go for faster coils (less L and less Cc) and faster amplifiers which would normally have to be designed around discrete components as in my example above.

Peak coil current is part of the gain equation. One amp with gain of 1000 same as ten amps with gain of 100. If thin gold is included in what we are looking for what would peak current and gain should we try for?

[Teleno]

Peak coil current is part of the gain equation. One amp with gain of 1000 same as ten amps with gain of 100. If thin gold is included in what we are looking for what would peak current and gain should we try for?

We're dealing only with the amplifier. Thin gold does not require higher gain per se, but a faster transient to capture the short time constant.

A fast amplifier allows for earlier sampling, which also means more signal for the same gain. Implementations around NE5534 introduce great delays which are unacceptable.

CORRECTION: The previous formula was wrong, the value of the peak current for a critical damping resistor R is:

where R is

Example:

L = 300 uH (typical PI coil), with Cc = 150 pF.

The drive MOSFET has a diode in series with the coil, therefore the capacitance seen by the coil is the diode's. Co = 100 pF

The damping resistor R should be under 547 ohms

The MOSFET breakdown spec. is 500V, then the maximum current is 1.24 A:

If shorter time constants are not interesting for you (hunting for larger objects) then you can increase Imax by adding capacitance to the coil resonant circuit, for example, by removing the series diode in which case Co is the output capacitance of the MOSFET, in the order of 1 nF:

But then you get flyback twice as long as in the first example.

[Sean_Goddard]

EEK..MATHS!!!

But I see where you are going..I think.

That amp seems fast enough. I'm thinking to target 9ct thin gold rings. If we can hit those at a good depth (6" or more) then I think we will have a machine which is good for small nuggets too. My reasoning being is that I know the GPX4000 is a good gold machine, but look how much it costs. If we can make soomething which is comparable, maybe even better, then this will be a good project to build. I also want to keep it simple with no complex set up or calibration.

I was quoting the low gain front end as it seems that high gain ones suffer in Australian goldfields, now I'm NOT looking to compete with ML,I would simply like a machine which had good performance on wet sand and costs less that $100 to build. We can do this can't we?

I want a few more people to comment on your amp designs Teleno. I am happy with the latest one. Have you run a component value variance sweep (MonteCarlo) on it to see how it performs? I envisage all values in the final build will be 1% resistors, but if the thing still works with 5% then we can wrap this section up and move on if all agree.

FYI I have acquired an ML Etrac with an 18X15 SEF coil (plus standard) and on the beach it simply ROCKS. Deep, quiet and it discriminates, but I KNOW I'm missing those deep and thin rings, THAT is why I want to make a good PI. The Cscope 4PI just can't cut it in terms of the deep finds (sadly).

[Teleno]

I was quoting the low gain front end as it seems that high gain ones suffer in Australian goldfields

I have posted my novel GB scheme in another thread, should be tested in Aussie goldfields. If it works then gain shouldn't be a problem.

I want a few more people to comment on your amp designs Teleno. I am happy with the latest one. Have you run a component value variance sweep (MonteCarlo) on it to see how it performs? I envisage all values in the final build will be 1% resistors, but if the thing still works with 5% then we can wrap this section up and move on if all agree.

The final design will be MCU based. I'm thinking ATtiny. The amplifier will be self-adjusting by two DAC outputs directly controlling the gates of J1 and J3. Tolerance is of no concern.

The current at Q2, which is the most critical parameter, is temperature compensated by D3.

, now I'm NOT looking to compete with ML,I would simply like a machine which had good performance on wet sand and costs less that $100 to build. We can do this can't we?

Definitely.

The following components should be added:

1 x ATtiny84 MCU (evt. ATMega family)
1 x MCP4728 quad DAC 12 bit
1 x standard 5V regulator.
1 x BS170 for driving the MOSFET's gate.

EEK..MATHS!!!

Alright I'll do the maths and the electronics design, you do the PCB layout and prototyping. Deal?

I can do the programming as well unless we have a more experienced volunteer.

[eclipse]

Using a dedicated mosfet driver is better than a transistor pair. As for the preamp op amp consider the AD8620(x2) or AD8610(x1) - lowest noise, high speed, etc.

[Teleno]

Using a dedicated mosfet driver is better than a transistor pair.

It is when you need both fast turn-on and fast turn-off.

In this case, however, fast turn-on is not required, only shorting the gate quickly to ground. A single BS170 does the job.

As for the preamp op amp consider the AD8620(x2) or AD8610(x1) - lowest noise, high speed, etc.

AD8610 is 6 nV/√Hz, my amplifier is 3.6 nV/√Hz.
AD8610 is dual supply, my amplifier is single supply.

The purpose of my design is to do away completely with op-amps, be faster and achive lower noise. Look again at the circuit (download the simulation file). It's based on discrete components and has a gain of 1000 x Input noise is 3.2 uV (can be made 2.5 uV with emitter degeneration on Q4 and Q5).

AD8610 might be an option as a second stage though, not as preamplifier. I'll consider it and thanks for the sugestion.

P.S. LT1037 has similar characteristics and is usually cheaper.

[eclipse]

The LT is not JFET type like the AD (not sure?!). Descrete solution like cutting edge standalone op amp offers "repeat-ability" - it's all there in the package.
If you take 10 different op amps they will perform the same.
While the other solution may give unexpected results - lots of connections may pick up some noise long the way, component tolerance, etc.
It will be interesting to see a real working example, if it will prove better / worse.

[green]

We're dealing only with the amplifier. Thin gold does not require higher gain per se, but a faster transient to capture the short time constant.

A fast amplifier allows for earlier sampling, which also means more signal for the same gain. Implementations around NE5534 introduce great delays which are unacceptable.

CORRECTION: The previous formula was wrong, the value of the peak current for a critical damping resistor R is:

where R is

Example:

L = 300 uH (typical PI coil), with Cc = 150 pF.

The drive MOSFET has a diode in series with the coil, therefore the capacitance seen by the coil is the diode's. Co = 100 pF

The damping resistor R should be under 547 ohms

The MOSFET breakdown spec. is 500V, then the maximum current is 1.24 A:

If shorter time constants are not interesting for you (hunting for larger objects) then you can increase Imax by adding capacitance to the coil resonant circuit, for example, by removing the series diode in which case Co is the output capacitance of the MOSFET, in the order of 1 nF:

But then you get flyback twice as long as in the first example.

What is the advantage of reducing circuit resonance vs using a snubber to keep the voltage below avalanche volts? The coil decay time is faster with a snubber.

[Teleno]

The LT is not JFET type like the AD (not sure?!). Descrete solution like cutting edge standalone op amp offers "repeat-ability" - it's all there in the package.
If you take 10 different op amps they will perform the same.
While the other solution may give unexpected results - lots of connections may pick up some noise long the way, component tolerance, etc.

The discrete solution does not require tight tolerances because an MCU will adjust the 2 offsets required (via gates of J1 and J3). That's all.

If you look at the circuit, all the connections except the input one are low impedance, carrying currents in the order of mA. Parasitic noise stands little chance.

It will be interesting to see a real working example, if it will prove better / worse.

Here's a version with LT1034 as a second stage. Noise-wise is the same, but the op-amp version is a bit slower to settle and the output swing is limited. Then there's the added overhead of a negative rail.

[Teleno]

What is the advantage of reducing circuit resonance vs using a snubber to keep the voltage below avalanche volts? The coil decay time is faster with a snubber.

Once you surpass the MOSFET's breakdown voltage the decay rate is constant no matter what you do, the L is as good as short-circuited.

A snubber has to reduce R anyway in order to avoid reaching the breakdown voltage. I believe there's no tradeoff.

[green]

I believe it would be the same, after all a snubber is reducing the R as well.

The snubber effects the decay at clamp volts and then drops out. I/T=E/L, 500volts/300uH equals 1.67amps/usec. Snubber adds to decay time. [But then you get flyback twice as long as in the first example, reply #108]

[Teleno]

The snubber effects the decay at clamp volts and then drops out. I/T=E/L, 500volts/300uH equals 1.67amps/usec. Snubber adds to decay time. [But then you get flyback twice as long as in the first example, reply #108]

A snubber makes no sense because it's better to go above the MOSFET's breakdown voltage, which then becomes a faster snubber (higher voltage). Just make sure you don't surpass the MOSFET's avalanche ratings which are quite generous.

[green]

A snubber makes no sense because it's better to go above the MOSFET's breakdown voltage, which then becomes a faster snubber (higher voltage). Just make sure you don't surpass the MOSFET's avalanche ratings which are quite generous.

I agree, but am told the avalanche is noisy and it's better to snub the volts. Not much difference in time, 470 snub volts vs 500 avalanche volts.

[Teleno]

I agree, but am told the avalanche is noisy and it's better to snub the volts. Not much difference in time, 470 snub volts vs 500 avalanche volts.

No problem, at the time of sampling the avalanche is long gone together with its noise.