Simpler PI...

[Beenthereuk]

http://videobam.com/jITBK


Waikiki, my circuit uses an NMOS FET with a VGSth = 2V. It needs a threshold that is 1V or more above ground. BUT, the incoming pulse is positive in mine and hence it is inverted and is negative when it hits the transistor's gate - if your pulse is in the opposite direction, it will not pass through the threshold and trigger the nFET. Either way, it is so simple that you should be able get it to work by using a PMOS FET at the output if you have to.

This is a really crap video I made of mine working. Detects small bars of Al and Cu at upto 12 cm or so. Adding another tranny for a speaker driver will seriously boost the sound..

[Waikiki_Sweep]

Yep. Video is slow. But it detects! Very good for first try!
Probably some improvements will make it detecting a coin from same distance.
Also it is good to build a pinpointer.

Also move your targets side by side not up to down because it imitates sweeping coil over ground and keep targets in horizontal position because it gives strong signal. Detecting coin or ring that stays on edge reducing range 2-3 times with any detector.

I have P-MOS with very low threshold J176 so I will try. Not tried it yet because weather getting better and I am sitting in water with my SurfPI all the day.

Audio volume is not much important for using headphones.
Headphones are better than speaker because often if you go outdoor with your detector noise is too high to use speaker anyway.

[Beenthereuk]

You're waddling in the Black Sea this time of year ?!

[Waikiki_Sweep]

You're waddling in the Black Sea this time of year ?!

No. I am not that crazy. Last eight years I live in Honolulu.
And weather here looks much better especially for waddling at winter. Google "Waikiki".
But also a competitive place - one detectorist on ten tourists probably.

[Beenthereuk]

Oh wow!, the fishing must be excellent there too. You're too lucky

[Waikiki_Sweep]

The NJM2068 is around 3 nV/sqrt(Hz) input referred voltage noise density.

In any official datasheet you can see only that statement:

Equivalent Input Noise Voltage for NJM2068 - 0.56 μV

May be it is equal 3 nV/sqrt(Hz) may be not.
Too much work for me to check it anyway.

If you have your calculator ready:

http://www.nikhef.nl/~jds/vlsi/noise/Amplifiers.PDF

[Aziz]

In any official datasheet you can see only that statement:

Equivalent Input Noise Voltage for NJM2068 - 0.56 μV

May be it is equal 3 nV/sqrt(Hz) may be not.
Too much work for me to check it anyway.

If you have your calculator ready:

http://www.nikhef.nl/~jds/vlsi/noise/Amplifiers.PDF

Hi Waikiki_Sweep,

the specified 0.56 µV noise energy is the integrated noise energy density for 20 Hz to 20 kHz.
Divide this by 141 and you get approx. the voltage noise density in nV.
en = 0.56 µV/sqrt(20000 Hz - 20 Hz) = 0.56 µV/141 = almost 4 nV/sqrt(Hz)

We know, that the low frequency and the 1/f noise contribution is much (20 Hz .. 1kHz). If we neglect this low frequency noise (we are doing the PI at PPS rates and won't really see this low frequency noise), then we are at approx. 3 nV/sqrt(Hz) region for f > 1kHz.
The typical measured noise is even better than the maximum specifed noise voltage in the data sheet.

I know that the NJM2068 is an excellent device. I have tested and seen its noise spectrum compared against several op-amps (NE5532 for instance). It's really a bargain.

Cheers,
Aziz

[Waikiki_Sweep]

Thanks! That is really simple explanation.

NJM2068 where to buy:

Tayda Electronics - $0.40
Mouser - $0.60
Digikey - only wholesale 2000 :-(
eBay - crazy shipping charges
Also more offers are here: http://octopart.com/parts/search?q=NJM2068

[green]

[We know, that the low frequency and the 1/f noise contribution is much (20 Hz .. 1kHz). If we neglect this low frequency noise (we are doing the PI at PPS rates and won't really see this low frequency noise]

If I replace the amplifier with a signal generator, (10hz, 1mv) output and read the voltage with an A-D or a sampling integrator at 1khz sample rate what should I expect see?

[Carl-NC]

the specified 0.56 µV noise energy is the integrated noise energy density for 20 Hz to 20 kHz.
Divide this by 141 and you get approx. the voltage noise density in nV.
en = 0.56 µV/sqrt(20000 Hz - 20 Hz) = 0.56 µV/141 = almost 4 nV/sqrt(Hz)

You need to divide by 177, not 141, for a single-pole noise BW. This results in 3.2nV/rtHz.

[Aziz]

You need to divide by 177, not 141, for a single-pole noise BW. This results in 3.2nV/rtHz.

Thanks Carl, the noise density figure becomes more attractive now.
Aziz

[Aziz]

[We know, that the low frequency and the 1/f noise contribution is much (20 Hz .. 1kHz). If we neglect this low frequency noise (we are doing the PI at PPS rates and won't really see this low frequency noise]

If I replace the amplifier with a signal generator, (10hz, 1mv) output and read the voltage with an A-D or a sampling integrator at 1khz sample rate what should I expect see?

The low frequency amplifier noise will be filtered out by the demodulator stage (acting like a high-pass filter). For the pre-amp, only the noise density above the pulses per second rate (PPS) is relevant.
Aziz

[green]

The low frequency amplifier noise will be filtered out by the demodulator stage (acting like a high-pass filter). For the pre-amp, only the noise density above the pulses per second rate (PPS) is relevant.
Aziz

Thinking after my reply, The example wasn't random. I think I need to understand random noise better.

[Carl-NC]

The low frequency amplifier noise will be filtered out by the demodulator stage (acting like a high-pass filter). For the pre-amp, only the noise density above the pulses per second rate (PPS) is relevant.
Aziz

Be careful, our demodulators are not like analog mixers which do frequency translation. Rather, they are more like sample-and-holds which fold all the spectra into the Nyquist band. The target response gets dropped right on top of 1/f noise.

[Davor]

Oh, but they are - you have a bit nasty effect, charge injection, that in fact provides LO injection into the demodulated signal path. Problem may be very small in a balanced mixer, but in most VLF-s it is not so. PIs with EF subtraction in effect have an equivalent of a balanced mixer, because the same charge gets injected at target sample, and subtracted at EF sample, in effect balanced out.
The charge injection is unavoidable with CMOS switches, there are only some that give off smaller charges, but the only way to reduce them completely is a balanced mixer.

As for 1/f noise, we are dealing with targets responses centred around 10Hz, and it is as nasty as it gets. Choppers for DC amplifiers are not exactly commonplace, but may be completely avoided if sampled signal is amplified, and autozeroed, prior to integration.