VLF MD with digital signal processing : Bee-Buzz 1

[Aziz]

Hi all,

RX current sensing would produce much TX - RX coil interaction. Regardless of wether good or not good. My observation is: no good.
In case of RX voltage sensing, there is almost no current through the RX coil and hence almost no interaction with the TX coil.
So I tend to do the voltage pre-amp next time.
---
My choke L1 for the dual resonant frequency VLF/LF is getting too difficult to wind with the thick wire. Anyway, it must be done. 500 g 1 mm copper wire will be split for two different chokes. As tight as possible. As compact as possible for max. inductance per weight. This is the opposite way one would wind the TX coil of a PI detector.

I want to see the dual frequency response.
Cheers,
Aziz

[moodz]

No, it's not. It's practically shorted to virtual ground. Here's a circuit to consider:

What's the gain of this circuit? How does it work? (Hint: you already have the answer.)



The RX coil is capable of producing either a voltage or a current. In the case of the Lee amplifier, the RX coil current is converted into an output voltage, therefore it's a transimpedance amplifer (G=V_out/I_in). It's a perfectly legitimate way to make an RX preamp whether the ultra-low noise is a benefit or not.

Also, maximum power transfer only matters when you want to transfer maximum power. In the kinds of designs we do, we avoid that because it's just wasteful.

When a coil has a current flowing because there is some impedance across it and a changing magnetic field it it forms a magnetic circuit with other inductors / susceptibities in the vicinity.
When a coil has no current ( ie we just measure the voltage across it ... there is no magnetic circuit formed ( because there is no current ).
This is the key decision.

At the rate this simple "Kiss" project is progressing we will never get to the main detector.

[Carl-NC]

The RX-coil is practically shorted to ground.

No, it's not. It's practically shorted to virtual ground. Here's a circuit to consider:

What's the gain of this circuit? How does it work? (Hint: you already have the answer.)

An ideal current amplifier will have an input impedance of zero ohms. A well designed current amplifier will have a input impedance well below your 1 ohms.
You are designing a voltage amplifier with a low input impedance ( which is counter intuitive and possibly wrong ).

And theres no free lunch because maximum power transfer occurs at source impedance = load impedance.

The RX coil is capable of producing either a voltage or a current. In the case of the Lee amplifier, the RX coil current is converted into an output voltage, therefore it's a transimpedance amplifer (G=V_out/I_in). It's a perfectly legitimate way to make an RX preamp whether the ultra-low noise is a benefit or not.

Also, maximum power transfer only matters when you want to transfer maximum power. In the kinds of designs we do, we avoid that because it's just wasteful.

[moodz]

Hi all,

I have revisited the Richard Lee's pre-amp and have achieved an input voltage noise density down to 0.18 nV/rtHz. Pure discrete solution operated at 3 V supply voltage with output buffer.
But!
The input impedance of the pre-amp gets down to effectively almost 1 Ohm (actually 0.96 Ohm).
The RX-coil is practically shorted to ground.
I have never thought, that it is so much low.

Even the circuit simulation get these values:
Rin = VT/IE,
Rin = input impedance
VT = thermal voltage (= 26 mV)
IE = emitter current = 13.5 mA (taken out of the circuit simulation)
Rin=26 mV/13.5 mA = 1.93 Ohm for single transistor, but I have two, Rin=1.93 Ohm || 1.93 Ohm = 0.96 Ohm.
see https://en.wikipedia.org/wiki/Common_base

This is no good. All the RX-coil sensitivity gets lost. But the pre-amp is the quietest pre-amp I have ever built.

Next time, I will try the other pre-amp with much higher input impedance (10 kOhm). This will take some time as I have to travel to Berlin this week.
Cheers,
Aziz

An ideal current amplifier will have an input impedance of zero ohms. A well designed current amplifier will have a input impedance well below your 1 ohms.
You are designing a voltage amplifier with a low input impedance ( which is counter intuitive and possibly wrong ).

And theres no free lunch because maximum power transfer occurs at source impedance = load impedance.

Its an interesting question ... should the coil amplifier be a current amplifier or a voltage amplifier ?
Until you answer that question .. then you can move to designing the amplfier.

[Aziz]

Hi all,

I have revisited the Richard Lee's pre-amp and have achieved an input voltage noise density down to 0.18 nV/rtHz. Pure discrete solution operated at 3 V supply voltage with output buffer.
But!
The input impedance of the pre-amp gets down to effectively almost 1 Ohm (actually 0.96 Ohm).
The RX-coil is practically shorted to ground.
I have never thought, that it is so much low.

Even the circuit simulation get these values:
Rin = VT/IE,
Rin = input impedance
VT = thermal voltage (= 26 mV)
IE = emitter current = 13.5 mA (taken out of the circuit simulation)
Rin=26 mV/13.5 mA = 1.93 Ohm for single transistor, but I have two, Rin=1.93 Ohm || 1.93 Ohm = 0.96 Ohm.
see https://en.wikipedia.org/wiki/Common_base

This is no good. All the RX-coil sensitivity gets lost. But the pre-amp is the quietest pre-amp I have ever built.

Next time, I will try the other pre-amp with much higher input impedance (10 kOhm). This will take some time as I have to travel to Berlin this week.
Cheers,
Aziz

[Aziz]

These are VLF detectors .... Bandwidth is less than 20 hertz at each frequency of interest. The amps Aziz are talking about might work on a PI preamp ...but async noise is almost meaningless in synchronous detection and processing ...this is not PI land.

Interference has to be synchronous, be phase modulated and the phase modulation has to be within the 20 hertz bandwidth at the transmit frequency eg 19.17456 khz.
Security through obscurity ... Not much noise there or on Mars LOL.

Hi Paul,

if you are setting your detection bandwidth < 20 Hz, sure you will get low demodulation noise. I will get it at any case too. As I said, it is only a matter of demod bandwidth. Regardless of using Goertzel, Lock-in or FFT. There is absolutely no difference. Goertzel may produce more numerical errors, if the block size gets (very) large. Whereas the Lock-in amp and FFT are using precalculated internal sin/cos reference in digital demodulations.

I think I won' get much EMI noise. The most dominating noise will be white noise (thermal noise). At least on my sound card rather than on the pre-amp.
Aziz

[pito]

https://www.experimentalistsanonymous.com/diy/Datasheets/NE567.pdf

[moodz]

These are VLF detectors .... Bandwidth is less than 20 hertz at each frequency of interest. The amps Aziz are talking about might work on a PI preamp ...but async noise is almost meaningless in synchronous detection and processing ...this is not PI land.

Interference has to be synchronous, be phase modulated and the phase modulation has to be within the 20 hertz bandwidth at the transmit frequency eg 19.17456 khz.
Security through obscurity ... Not much noise there or on Mars LOL.

[Riss]

The electromagnetic "smog" is getting thicker - from year to year. The need for super-low noise systems would make sense on the planet Mars, to search for artifacts from the Martians who self-destructed millions of years ago (we will soon turn the earth into a lifeless desert). However, let's see how you will deal with the Telluric currents, which unfortunately are below 2 hertz and are strongest in July-August north of the equator ... gain 120 dB ...

[moodz]

Hi Paul,

well ok, you are obviously using hw demods (hardware implemented lock-in amp).

But this is not the case for me. To make things simple, I use the real digital direct sampling (software) demod. I have no other way as to amplify the signal and noise. That's the reason, why I need the pre-amp to go down the noise floor level of the adc system. It is surely not perfect if I compare it to the hw demod.

I could get similar results with hw demods too. But this would make the detector controller much more complex:
RX demod -&gt; x-ch, y-ch (complex) -&gt; AC modulator -&gt; adc (sound card) -&gt; decode mag and phase.
What about more channels for more frequency demods? Well I'm limitted here or it would make the AC modulator more complex.
But this project wouldn't be a KISS project anymore.

You guys take the hw demods and I take the pre-amp and doing the sw-demods.
Cheers
Aziz

Aziz ... for a smart guy you sure are getting sucked in by fake news ( audophile amplifier schemes )

The RX coil connects directly to the ADC in my scheme except for some passive protection and impedance protections to stop the front end of the ADC being destroyed by EMI spikes.
So I dont use HW demods.

By adding amplification to a high resolution ADC ... its the output noise of the amplifier ... not the input noise that matters more.

1. You increase noise at the input to the ADC.
2. You reduce dynamic range.
3. You lose bit resolution .... reduction in effective number of bits.
4. Have to use differential amps ... single ended amps wont cut the mustard.

For 16 bits you may be able to maintain sensitivity ( ie effective bits ) but your dynamic range will be very small.

PS : you can bet the common mode rejection on your fancy amp will be crap ... so helllo wideband noise.

The main points below ( this assumes an ideal amplifier -- if you amp has any noise the figures will be worse )

[Aziz]

Hi all,

I will leave the simple NE5534 pre-amp. Too much thermal noise expected due to circuit simulations. I should not try it.

I will leave the discrete solution too. Not versatile use possible.

But 1x ZTX851, NE5534 as control loop and a transistor as driver to get RG1 down to 1 Ohm, it is possible to get 0.25 nV/rtHz at gain 100.
At the expense of high power consumption however (approx. 30 mA at 9V ). But with a stable gain and frequency compensation loop.

Anyway. This is just for clarification.
Cheers

[Aziz]

Hi all,

I will prepare two pre-amps with gain of 100 for the next measurements:
- simple NE5534 AC pre-amp (producing much more thermal noise)
- discrete ultra low noise AC pre-amp (en approx. 0.25 nV/rtHz)

This should definitely answer the question, whether we need a very low noise pre-amp.
At the same time to see, when the EMI noise is visible on quiet regions to define the gain of the pre-amp.

Back to the black board.
Aziz

[Aziz]

Any cancelling scheme would also cancel the target because the interferance you need to worry about is inband of your desired signal ..
Amplification only amplifies the wideband noise and the inband signal for no benefit unless you call making the noise bigger a benefit.
When the signal is under the noise then a preamp is more than useless.


So on my WORKING detector that does use a lockin amplifier ( read up on lockin amplifier reserve ) I have a real reserve of 112 dB ... which means it can pull a signal from 112 dB from below the noise level. Commercial lockin amps can pull a signal from 120 dB below the noise level so I am almost on par.
In the picture below I turn on the LED lighting in the lab for about 10 seconds then off again .... you can see the inband interferance coming from the lighting ... dont need a preamp on the ADC. The target signal ( 10 decimal digits in the upper left corner VA corresponds to 180 db of signal range but only the last digit is changing with low noise )

Hi Paul,

well ok, you are obviously using hw demods (hardware implemented lock-in amp).

But this is not the case for me. To make things simple, I use the real digital direct sampling (software) demod. I have no other way as to amplify the signal and noise. That's the reason, why I need the pre-amp to go down the noise floor level of the adc system. It is surely not perfect if I compare it to the hw demod.

I could get similar results with hw demods too. But this would make the detector controller much more complex:
RX demod -> x-ch, y-ch (complex) -> AC modulator -> adc (sound card) -> decode mag and phase.
What about more channels for more frequency demods? Well I'm limitted here or it would make the AC modulator more complex.
But this project wouldn't be a KISS project anymore.

You guys take the hw demods and I take the pre-amp and doing the sw-demods.
Cheers
Aziz

[Aziz]

LOL ....must be a real special amp that lifts the signal without lifting the noise.

Indeed, my old pre-amp seems to produce more (thermal) noise than expected.
I have taken very old tantal caps in my prototype. Guess what? ESR > 1 Ohm.

All measurements are obsolete. I'm preparing a better pre-amp with measured low ESR caps now. No more tantal caps!

Before making real coil measurements, I will measure first resistor noises: for RS = 0 (short input), 1 Ohm, 4.7 Ohm, 10 Ohm and open input (RS=input impedance).
I can then say, what is EMI noise and what is thermal noise.

[Riss]

LOL ....must be a real special amp that lifts the signal without lifting the noise.

,,The best amplifier is the antenna!" (I heard this wisdom when I was in kindergarten...) - it raises the signal level without introducing noise, respectively, it increases the signal/noise ratio. Last year I made and tested a prototype of a directional antenna system for metal detectors. To my surprise, it worked immediately! And in the way it was designed. However, I am like a Woody woodpecker - I experienced countless crashes with my experimental planes . I will be satisfied with this only if the directional antenna system shows advantages in real field tests. If not - I will laugh at myself like this – https://www.youtube.com/watch?v=iNxujJGnbB4