Schumann resonance receiver

[FrankBuss]

I found an interesting article here: https://scispace.com/pdf/a-new-portable-elf-schumann-resonance-receiver-design-and-otp6s1oi1n.pdf and want to rebuild such a receiver. Unfortunately they don't provide a full circuit diagram with component values. What values do I need for the pre-amplifier? They write the input voltage is in the tens of nanovolts. They show the circuit on page 5, but it might be not complete, there are some gaps in the annotations, like R4 is missing. Gain is 35.5. I came up with these values:



This should have an output highpass filter of about 0.2 Hz to block DC. But not sure about R1 and R5, and if I should maybe lower R7 and R3 for less noise.

[moffy]

I am suprised that in the article they didn't go for just a differential input, though they do mention an instrumentation amplifier, as the antenna according to the block diagram looks to be differential in its output. A differential approach with shielded twisted pair, like in audio and the XLR connector, should provide better environmental noise rejection. Then a good low noise differential preamp, like a microphone preamp would provide low noise e.g. THAT1512.

[FrankBuss]

The THAT1512 looks interesting, even lower noise than the OPA376. And yes, the article has some inconsistencies. The coil looks indeed like an antenna, with like a center tap. This would make sense for a differential amplifier, then common mode noise would be eliminated, and the output would be 2x.

I don't understand what they write with the mu metal and clamp between the coils. Would it work to just use one long rod, and wind the left side in one direction, and the right side in the other direction for the center tapped version with differential output? Then connect the center tap to GND.

[moffy]

The THAT1512 looks interesting, even lower noise than the OPA376. And yes, the article has some inconsistencies. The coil looks indeed like an antenna, with like a center tap. This would make sense for a differential amplifier, then common mode noise would be eliminated, and the output would be 2x.

I don't understand what they write with the mu metal and clamp between the coils. Would it work to just use one long rod, and wind the left side in one direction, and the right side in the other direction for the center tapped version with differential output? Then connect the center tap to GND.

I think the Mu metal sheet/clamp is a cylindrical pipe clamp made of Mu metal to join the cores. Not sure about winding direction, but a single core would avoid any air gaps and improve coupling, but it looks like a lot of turns. Where would you a obtain such a Mu metal rod, it would cost a little bit?

[FrankBuss]

My plan is to use these rods, 50 cm long, 6 mm diameter

https://de.aliexpress.com/item/1005005721336378.html?spm=a2g0o.order_list.order_list_main.11.ba8d18020iiibx&gatewayAdapt=glo2deu

Claude AI thinks this could be even better than their solution with the two 25 mm diameter rods, because with my rod, the high aspect ratio (83:1 vs 12:1) gives me a demagnetization factor N = 0.000028, which is about 78 times better than from their paper, and it would more than compensate the smaller cross-section, if the material is really high permeability, like in the range of μr = 50,000, which would be close to mu metal. I'll measure it with a small winding when I get the rods.

But what do you think about the idea to use a differential setup with the center tap? I would guess any capacitive noise couplings (e.g. from mains) would be eliminated with it, because they would increase the voltage of both coils in the same direction, while magnetic changes increases it in different directions, if I understand the physics correctly.

BTW for building it, I plan to 3D print some bobbins, so that I don't have to wind the 50 cm rod. It will then have some gaps between the bobbins, but shouldn't affect the performance much I hope. Will try to calculate it all first, and then compare to the actual result, which would be a nice task by itself for me.

[moffy]

It's a pity they don't specify the nickel content, if they are high permiability like Mu metal, let us know, it could come in handy sometime. The center tap can be connected to ground, sounds reasonable. Because of the size of the coil any stray magnetic fields won't be even across them, so a low noise environment, like an open field might be an advantage, but it depends upon your environment and a simple matter of orientation could give the lowest induced noise.

[moffy]

Since there are a limited number of frequencies of interest, one could use a band pass filter for each frequency and that would solve most of the noise and spuriously induced emf problems.

[RoGeorge]

After reading this wikipedia page https://en.wikipedia.org/wiki/Schumann_resonances it looks like the expected levels are very, very small.  Doesn't look like a single opamp will cut it.  Typical schematic for measuring nanovolt levels will have many paralleled amplifiers, so not just one opamp, but as many as the budget allows.

Another observation, the shielding seen in the pics in that article (a single thin metal cookies box) won't be too effective at such low frequencies (single digit Hz range).

- 1.  How does the authors of the portable ELF receiver knew they were receiving Schumann oscillations, and not just random noise that happen to also have some expected frequencies in it?

- 2.  Assuming the receiver would be quiet enough, and sensitive enough, what would be the meaning of receiving such Schumann signals, what would be good for to have such a receiver?

Never mind those questions, meanwhile took the time to read the portable ELF receiver paper.  It is still not clear, did they use one amplifier, or six in parallel?


Seen on EEVblog forum 2 topics that might help, the nanovoltmeter one, and the measuring of a LDO ripple.
https://www.eevblog.com/forum/metrology/nanovolt-design-challenge-build-and-show-your-own-nv-meter-in-256-days/
https://www.eevblog.com/forum/metrology/how-i-made-my-nanovoltmeter/
https://www.eevblog.com/forum/projects/shield-connections-for-low-noise-measurement-(an-159)/ (see the AN-159 and its bibliography)

The THAT transistors arrays were already mentioned, that's the lowest noise option I've seen so far without using cryogenic cooling, and they are still available.


About ferrite antennas, the best I've seen would use many ferrite rods (to build the walls of a pipe-like ferrite shape out of cylindrical ferrite rods - sometimes called ferrite sleeve antenna).  Sometimes it is enough to use such an antenna as a passive filter near a smaller receiving antenna, to "collect" the magnetic fields lines from a larger area, and to also make a high Q filter, then mutually couple with the receiver's coil.  Something like this:

8" Ferrite Sleeve Loop Antenna Demonstration
DXerGary


No idea if such builds are practical for 3Hz or so.  This is what I would do, or look for, but I have zero hands on experience at receiving such faint signals, and at such low frequencies, so don't put too much weight on my advice. 

[Maurizio1957]

 This is an interesting lecture about a receiver for vlf:http://www.vlf.it/immagine/minimal_E.html.
73, Maurizio

[FrankBuss]

It's a pity they don't specify the nickel content, if they are high permiability like Mu metal, let us know, it could come in handy sometime.

I got them. I tried it with a coil with 1000 turns, which has 12.5 mH without it, and 32 mH when measured with 1 kHz with my HM8018, and 17.7 mH at 25 kHz. Without the core just air, it is the same 12.5 mH for both frequencies. No idea, is it the high permeability as advertised? Looks like definitely not so good for higher frequencies.

[FrankBuss]

with 3 of the rods for the same coil, it gets to 59 mH at 1 kHz and 27 mH for 25 kHz.

ChatGPT thinks the permeability is about 20, so much for "high permeability"  it is probably just some cheap iron or steel.

[RoGeorge]

Claude AI thinks ...

ChatGPT thinks ...

No, of course it doesn't "think".  Is this some kind of late April's fool joke, in which everybody pretend to expect precise engineering answers from a LLM, only to entertain themselves when they see me arguing how such expectation can not be fulfilled?


This is what a LLM will produce, as far as I know, and this is why the answers are incorrect most of the time:

Google AI says "Observing a 10 microvolt (µV) peak-to-peak (Vpp) signal is extremely difficult on a standard Tektronix TDS series oscilloscope because the signal is likely below or at the instrument's inherent noise floor and minimum vertical sensitivity"

The attached screenshot ts_ch4_1MHz_10uVpp__Ch2_100kHz_10mVpp.png shows that in reality it is not so difficult.

That shouldn't be a surprise.

The so called "Artificial Intelligence" name is misleading.  AI is not "intelligent", and it does not return knowledge.  A LLM generates "words salad" responses, but with a given theme.  (LLM=Large Language Model, like ChatGPT, Grok, etc.)

Not kidding, this is how a LLM works.  A LLM starts from your "words", and returns other neighboring "words" that happened to be in the vicinity of your question.  Well, not exactly words, so let's call them "tokens".  The "distance" between the input tokens (what your requested) and the generated tokens (the constructed answer) is spiked with random noise, to avoid an identical answers, so it won't repeat itself if you reformulate or refine your request.  This is how LLMs normally work, this is not about "AI hallucinations". 

AI generated content only "looks like" it makes sens, but in fact the content is meaningless.  AI behaves like the perfect "impostor":  It replies with the vocabulary and the confidence of an expert, but the answer is not expert knowledge, the answer is "words salad", or "tokens salad".  Sometimes the words salad happens to be correct, but this is just a matter of luck.  It is not a proper answer, it's an illusion.


A LLM answer is nothing but a fairy tale story.  A fairy tale composed on the spot, for you only, about the theme you were asking.  The term for this kind of AI generated fairy-tales-with-a-theme is "AI slop".

All the AI answers, even the "good" ones, are in fact AI slop. 

(sorry for quoting myself, but I've just wrote that a few hours ago, won't copy-paste, and only quote instead)

Does everybody really expect correct engineering answers, or do you guys were just asking AI for entertaining purposes.  I'm not mocking anybody, but I've wrote about the same warning message as in the quoted text, wrote that more than a few times so far, in different forms and with different words, and no reaction.

Apparently people don't believe me.  Is it that everybody who is using AI, somehow becomes "hypnotized" by AI slop?  Or is it something that I don't know, or did I get it all wrong about what a LLM does?  What makes everybody keep asking AI, then struggling to make sense of a words-salad answer, then double checking if it really is correct and applicable or not, when it is obvious that many of the AI answers are as correct as tossing a coin. 


(Legit questions, not trying to be rude.  I dared to ask you all these only because you are the one who made me pay attention to AI many years ago, long before ChatGPT, when you told us about an AGI prize, and if AGI would be possible at all, remember?  It was  in a group chat from hackaday).

[ch_scr]

I think it's the "Newspaper effect", where when one reads something he doesn't know much about in the newspaper one tends to think "I suppose that sounds plausible enough".
IMHO, in these cases, one gets to "feel like an engineer" to "have calculated it before hand", while critically lacking the knowledge to judge or check the answer. And we thought wikipedia was bad...

[moffy]

I agree with RoGeorge about LLMs, it is a lie to call them AI, they are just pattern recognition/matching algorithms of dubious value.
The OPs method of measuring the permiability won't work because all he is measuring is the effective air gap. One would need a closed loop of the material to easily measure its permiability. And of course it is useless at higher frequencies because it is a solid core, not laminated, eddy currents will be horrendous at those frequencies.

P.S. Mild steel has very good permiability.

[FrankBuss]

I know, LLMs make horrible mistakes. I use Claude Code daily with programming, and while it is helpful, it shouldn't be trusted with anything and if someone uses it, it is best if they could do it themselves, then they can judge if something is wrong which it generated. Just in another posting I described a web page from Keysight, which was obvious AI slop, even to me:
https://www.eevblog.com/forum/beginners/measuring-inductance-475595/
And right, the skin effect probably makes such rods useless for higher frequencies. I tried to use 100 Hz with the Hameg, but the the value jumps around between 15 mH and 50 mH with the rod. Maybe I'm using it wrong.

But the VLF website Maurizio posted is very interesting. I think I'll just build a simple loop antenna, as described here: http://www.vlf.it/minimal/minimal.htm No need for any core, and even much less coil turns for a usable signal.

[moffy]

I know, LLMs make horrible mistakes. I use Claude Code daily with programming, and while it is helpful, it shouldn't be trusted with anything and if someone uses it, it is best if they could do it themselves, then they can judge if something is wrong which it generated. Just in another posting I described a web page from Keysight, which was obvious AI slop, even to me:
https://www.eevblog.com/forum/beginners/measuring-inductance-475595/
And right, the skin effect probably makes such rods useless for higher frequencies. I tried to use 100 Hz with the Hameg, but the the value jumps around between 15 mH and 50 mH with the rod. Maybe I'm using it wrong.

But the VLF website Maurizio posted is very interesting. I think I'll just build a simple loop antenna, as described here: http://www.vlf.it/minimal/minimal.htm No need for any core, and even much less coil turns for a usable signal.

Sounds good, the windings need to be tight against the core to avoid air gaps.

[FrankBuss]

But the VLF website Maurizio posted is very interesting. I think I'll just build a simple loop antenna, as described here: http://www.vlf.it/minimal/minimal.htm No need for any core, and even much less coil turns for a usable signal.

Sounds good, the windings need to be tight against the core to avoid air gaps.

There is no core for the loop antenna, if I understand it correctly, just big air coil.

[moffy]

But the VLF website Maurizio posted is very interesting. I think I'll just build a simple loop antenna, as described here: http://www.vlf.it/minimal/minimal.htm No need for any core, and even much less coil turns for a usable signal.

Sounds good, the windings need to be tight against the core to avoid air gaps.

There is no core for the loop antenna, if I understand it correctly, just big air coil.

Well that makes it easier.

[coppercone2]

stormwise sells high mu rods of large size, but you need to be willing to invest.

Its 'loopstick' material
https://stormwise.com/page26.htm

I have one, but have yet to build the project.

For the antenna built with a bunch of ferrite rods, there is a thread some where on this forum about a research paper that tested them. There is a weight savings at the cost of volume, and  a possible price savings. It won't give you performance that is drastically different from a large rod.

https://www.eevblog.com/forum/rf-microwave/ferrite-sleeve-antennas/

O\f course, if you go BIG, instead of trying to exceed the largest slug you can find, things get interesting. (100+ page research paper). Big will be expensive (but much less expensive then some kinda gigantic single ferrite). Like paint bucket+


also, this is interesting, whatever it is (looks like its from system shock)
https://aaroniausa.com/product/antennas/magnetic-field-antennas/magnotracker-elf-1hz-9khz/

seriously, what is that thing I thought it was a figure 8 winding, but it could be something like this


https://patents.google.com/patent/US6529169B2/en

https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=10668829

I still think its a figure 8 coil though, something about how only flux from the very middle excites it, making it very directional

but its spaced too much to just be a generic figure 8 coil. It would be like weird.