VLF MD with digital signal processing : Bee-Buzz 1

[Aziz]

Hi all,

Android systems with USB-C interface support seems to be very interesting now. USB Audio Class 1 (UAC1) is already supported. Maybe USB Audio Class 2 (UAC2) too or very soon. Newer Smartphones and Tablets support now USB-C.

This is becoming very very interesting. Android systems are cheap. Damn cheap. You can use a smartphone or 6 inch Android tablet. You have a lot of processing power. An external USB sound card will be required for signal processing. The internal sound of the smartphone or tablet should be used for beeping. You have incredible possibilities. This is it.
This is very very interesting now.
Aziz

[dbanner]

Maybe Fisher approach was " we've got surplus bags of thousands of transistors in the stock room, what pray can we do with them?"
Opamps were pricey in those days?
Although matching transistors to a level of precision would be a little time consuming.

[Aziz]

Hi Aziz,

I wrote "very low noise", not "ultra low noise". Explanations for R8,C4,R14,C6,C1 and R1 are more than obvious. This part of schematic is connected as noninverting OpAmp, not as differential amplifier. For me will be interesting answer from Moodz at my question, sorry!

You have misunderstood me. The discrete RX front end is a standard discrete Op-Amp design. With the differential input stage, buffer, and driver output stage at the end. This is a differential amplifier design such as all standard op-amps are.
Aziz

[dbanner]

CubeMX CubeIDE is c?, c++?
MX generates dot C file for chain?

[moodz]

Hi Aziz,

I wrote "very low noise", not "ultra low noise". Explanations for R8,C4,R14,C6,C1 and R1 are more than obvious. This part of schematic is connected as noninverting OpAmp, not as differential amplifier. For me will be interesting answer from Moodz at my question, sorry!

Some designs seem to have focussed on low noise preamps however there is point at which the noise contribution from the amplifier will not be the significant factor in the signal chain. With VLF detectors we are generally talking about a demodulation system that is trying to extract magnitude and phase changes in one or more frequencies and this will involve the use of a mixer to demodulate the target caused amplitude change and phase change. Also we could note that most of the noise contribution in opamps is below 1 khz generally .. so if we have a transmit / recieve frequency of 20 khz and we only desire a bandwidth of +/- 20 Hz ( target modulation BW ) then the noise for most opamps is going to be fairly low and its going to be random ( in this bandwidth ). This can be proved by measuring the noise in a 20 hertz bandwidth @ 20 Khz on a spectrum analyser that can do this ( eg high spec audio bode analyser ).

So the question is "why did they use these low noise frontends ?" ... (and I was not in the lab looking over the designers shoulder when they made these decisions ) but I will put my money on "mixer leakage" and where they have used a simple "chopper" style demodulator which most designs use ... the presence of low frequency noise ( ie down to DC ) which is within the target demodulated BW at the input to the mixer will pass through the mixer due to low "port isolation" and will appear in the target signal. Of course if you have ever worked with radio designs that is one of the reasons why they invented double and triple balanced mixers to reduce this effect amongst other things.

With an ideal mixer where the input desired frequency is Fc and there is some undesired out of band low frequency noise Fn and we are using a local oscillator of Fo then the ouput of the mixer will be (Fc - Fo) + (Fc + Fo) + (Fo + Fn) + (Fo - Fn). From this we see that the low frequency noise is now translated onto the Local Oscillator carrier frequency as Fo +/- Fn . Eg if our noise frequency is 60 hertz and our Lo is 20 Khz then the ouput of the mixer will be 19.940 khz and 20.060 khz.
Using an "ideal" mixer these two noise frequencies are easily removed with a low pass filter. If the port isolation of the mixer is low then the 60 hertz will appear at the op of the mixer and this is a problem.

Where can you get an "ideal" mixer ... well you can come pretty close in a direct sampling scheme where you sample with an ADC then do the mixing "digitally". You can do it in the analog world but you would need more sophistication than a simple chopper switch tyep demod.

[moodz]

Below is a diagram to help explain ... a wanted signal at 1 Khz ( much lower than other unwanted signals ) is mixed with an LO at 1 Khz and the wanted signal is translated down to near DC and the unwanted signals are translated to near 1 Kz or above. A low pass filter can now be used to remove the unwanted signals.

[moodz]

CubeMX CubeIDE is c?, c++?
MX generates dot C file for chain?

[dbanner]

Ok. thanks moodz, I hadn't installed the IDE yet.

[Detectorist#1]

Hi Moodz,
Thank you for this detailed explanation! I understand that using of balanced mixers in analog treatment of the signals eliminates the need of special low noise OpAmps in frontend stage of VLF detectors. Using of direct sampling allows more simple hardware solution with help of firmware after ADC stage again without need of low noise amplifiers in frontend stage.

This discussion is pointed on VLF MD, but In case of PI detectors this analysis also is interesting. I know your post for "Filters - low-pass, high-pass or band-pass" in PI designs. In this discussion you also mention that useful signals are near of the TX frequency. Is this valid only in the case of single sample unipolar TX pulses. What happens in the case of two samples (main and EFE) in PI designs? I see many PI designs with high-pass filter for 15.9Hz ( 0.1uF and 10K) after front-end stage. If EFE sample have 100us delay after Main sample - filter with 1ms time-constant will change the condition for right substracting and sure elimination of Earth's Field signals or I'm not right?

[moodz]

Hi Moodz,
Thank you for this detailed explanation! I understand that using of balanced mixers in analog treatment of the signals eliminates the need of special low noise OpAmps in frontend stage of VLF detectors. Using of direct sampling allows more simple hardware solution with help of firmware after ADC stage again without need of low noise amplifiers in frontend stage.

This discussion is pointed on VLF MD, but In case of PI detectors this analysis also is interesting. I know your post for "Filters - low-pass, high-pass or band-pass" in PI designs. In this discussion you also mention that useful signals are near of the TX frequency. Is this valid only in the case of single sample unipolar TX pulses. What happens in the case of two samples (main and EFE) in PI designs? I see many PI designs with high-pass filter for 15.9Hz ( 0.1uF and 10K) after front-end stage. If EFE sample have 100us delay after Main sample - filter with 1ms time-constant will change the condition for right substracting and sure elimination of Earth's Field signals or I'm not right?

Yes you are right but consider that the preamp ( before sampling - which is demodulation ) is either driven to saturation during flyback or switched away .... this is a form of modulation .... so this causes convolution of low frequency noise ( like mains Earth field etc ) to be modulated onto the "carrier" which actually is the pulse reptition frequency. So each sample you take is a separate demodulation process. ( whether one or many ). Considering that any waveform can be deconstructed to sine waves ... the filtering is altering the phase and amplitude. So in the MAGPI project the aim was to remove the EF and low freq noise by feeding the deconvoluted signal back to the input as negative feedback causing it to be cancelled. Easy to show in a simulator like LTSPICE ... but works in practice too.

[Detectorist#1]

Hi Moodz,
Thank you for the fast answer! What are advantages of MAGPI project with remove the EF and low freq noise in comparison with bipolar TX PI with inherently eliminating of EF and low freq noise of front-end stage? Maybe saturation process in frontend stage involving additional modulation make the difference?

[Detectorist#1]

Hi dbanner,
Yes, you are right. If low noise OpAmps was pricey in old days, now LT1028 (maybe lowest noise amp for VLF designs still year 1992) haves price of 13USD. Interesting - how was the price of LT1028 at the time of designing of Fisher Gold Bug?

[Aziz]

Hi all,

I have found my ultra-low-noise preamp breadboard (PNP-Version with 3 x BD176 in parallel). Gain = 1 + 470 Ohm/4.7 Ohm = 1+100 = 101. Approximately 40 dB Gain. It should rise the signals by 40 dB.

Installed the sound card driver on the Windows 7 PC. Measured the noise floor without any input on the sound card. Plugged the preamp with coil -> Noise, a lot of noise! Shorted the input of the preamp -> down to noise floor.
Checked the initial noise floor of the sound card without any input connectors. Connected the ultra-low-noise preamp with shorted input: No rise of the noise floor. At least not measureable. Maybe less than 0.5 dB rise.
This is a good preamp. A super-duper-ultra-low-noise preamp.

It will take some time to make some real measurements. I have not found everything. Everything (connectors) have been corroded. I need my soldering iron.
Cheers,
Aziz

[Detectorist#1]

Hi Aziz,
As Moodz said in #244, the problem with noise in VLF MD isn't in front-end stage. Ultra low noise solution not helps in real life of VLF MD - but if you have different results - all we will be happy to know this!

[Aziz]

Hi Aziz,
As Moodz said in #244, the problem with noise in VLF MD isn't in front-end stage. Ultra low noise solution not helps in real life of VLF MD - but if you have different results - all we will be happy to know this!

We will see, what is possible.
I will measure some resistors connected to the input and measure their noise at +40 dB gain.