2
\$\begingroup\$

I am building an instrument intended to pick up weak (~100 picotesla) 25 kHz oscillating magnetic fields with Faraday induction.

As shown in the photo, I have a 4-layer solenoid coil for pickup followed by a two-stage preamplifier that amplifies ~1000x for frequencies above 10 kHz. The pickup coil is not tuned and the circuit looks like below:

enter image description here

enter image description here

Signal/noise after Fourier transform of the time-domain signal, 1500 nT field at 26 kHz. 500 ms acquisition time.

enter image description here

Presently the setup can detect oscillating fields down to 2 nanotesla. There is a significant change in noise floor when I connect/disconnect the pickup solenoid coil, which suggests sensitivity can be improved. A possible solution is tune the inductor to the frequency of interest. My question is what is the best way to approach this, and general comments about how to improve the sensitivity are also welcome.

EDIT Suggestion by Henry Crun: tune an LC circuit to 25 kHz and use as the pickup coil. I'm limited to an air-core inductor for pickup. So I attempted this with the following circuit

enter image description here enter image description here

And this was the result: a pickup coil tuned at 24 kHz. So, the question now is -- what is the best way to increase the Q factor of the coil, which would increase the sensitivity. I have a detection limit of ~100 pT/Hz^1/2 and want to improve on that by another factor of 10 to 100.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Are you only looking for near-\$25\:\text{kHz}\$? Or do you really need all of that wide bandwidth (and phase shift) I see in your Bode plot? \$\endgroup\$ – jonk Jul 3 '18 at 22:22
  • \$\begingroup\$ 25-30 kHz is the frequency region of interest, so no I don't need all of that bandwidth. \$\endgroup\$ – MichaelT Jul 3 '18 at 22:25
  • \$\begingroup\$ Noise is related to bandwidth. So the narrower your acceptance bandwidth (1st stage) the better. Your coil can be tweaked in this direction, too. There are whole books on winding, capacitance between windings, etc. The capacitance in the coil itself can act as a noise integrator and work for you before it gets into the electronics where it is harder to remove. Lots of gain with so much phase shift doesn't sound good, anyway. So you should see about driving gain back down -- bandpass instead of highpass? Do you need 60db per decade on the low end? \$\endgroup\$ – jonk Jul 3 '18 at 22:50
  • \$\begingroup\$ Your new arrangement L2/C9 is called an L match, which will be a peaky low-pass impedance raising network. R1/R12 is losing 95% of your signal. But an smd inductor will have no Q - its R will be too high to be any use BTW ADA4528 is not low noise. \$\endgroup\$ – Henry Crun Jul 5 '18 at 7:26
  • \$\begingroup\$ OK, for now I will try with only a 10-15 uF capacitor parallel across the pickup solenoid coil. A ~10uF MLCC ceramic capacitor should work ok? For now, I want to see an improvement in SNR - after that I can tune the resonance frequency to the one I am interested in. \$\endgroup\$ – MichaelT Jul 5 '18 at 13:05
1
\$\begingroup\$

I just homed-in on your detection-limit problem for AC magnetic fields in the higher audio frequency range. I would be highly interested to maybe add some useful suggestions, but before doing so (and say just stupid things) may I ask a few short questions.

  1. Am I right to see from the picture showing your coil with added numbers of 2.8 Ohm and 290 uH, that the underlying paper relates to NMR-type of experiments?
  2. Why are you so keen to use a very low DC (!) drift opamp like the ADA4528? For me 'only' AC properties matter here.
  3. Added to that: The ADA4528 is such low DC drift because it employs a chopper stabilizing scheme, running on an internal ~200 kHz oscillator. This is approximately the worst thing one can choose if AC detection is prime, as many interference products will turn up like Nxf_chop -/+ Mxf_sign.

Please look up the application note of the opamp, denoted as AN-1114: this could be very helpful.


Thanks for replying.

  1. Please notify my concern of the chopper 'rattler' contained in the ADA4528, and get rid of it a.s.a.p. In that respect I forgot to mention your beautiful adstruction to the interference issue: the FFT spectrum between 26020 and 26060, showing a nice regular beating of ~2Hz. I bet this is chopper-induced! So my first suggestion is to change to an opamp like LT1115 (powered by two 9V batteries), with < 2 nV/sqrtHz and < 1 pA/sqrtHz, starting from the adapted scheme which gave you 24 kHz res. and 100 pT/sqrtHz sensitivity.

  2. Secondly, please also add (as mentioned by others) a non-shorting Cu-foil sheet screen.

  3. Further you may remove quite some components in the latest scheme, i.e. R3 and R4, as well as R9 and R10 as bias is provided via the feedback resistors R7 and R11. Also capacitor C1 can be removed, as C7 and C9 ensure the DC isolation. All this should boost your system considerably.

Regards, and hope to hear from you.

\$\endgroup\$
  • \$\begingroup\$ Thanks for the interesting question. Yes, the hardware ultimately concerns NMR-type experiments where the opamp serves to preamplify the AC signal before it is digitized. I chose this opamp based on low cost, low noise and that it takes a single voltage supply. \$\endgroup\$ – MichaelT Jan 10 at 20:29
0
\$\begingroup\$

Reduce the bandwidth of your amplifier. Caps across R7,11 will roll off the gain at >30kHz. Choose C1,C5 to roll the gain off below 20kHz.

You have a very low source impedance from this coil (0.42 ohms@2.8uH@25kHz). This means low output voltage, and poor noise figure if you try to approach the noise floor

A ferrite core in the coil will push up L and therefore Z, and therefore the signal voltage. If L goes up 100x, V goes up 100x. (unless there is a strong magnetic field present, high temperature etc)

Making the coil a resonant circuit (C across L2) will increase the output voltage greatly, at the expense of bandwidth - you may have to put some shunt R to get the Q down and bandwidth. For +/-10% bandwidth, you can have a Q of 10, i.e. the signal voltage will increase 10x. You want more L (ferrite core) to make the C value more workable (instead of 15uF)

If you are sticking with an air cored coil, then a low noise bipolar power transistor input stage will lower the noise (ztx951). This only matters when you try to get close to the noise floor - you do have nearly enough gain yet.

Power with batteries, or linear power supplies. Switchmodes operate in the range your equipment detects.

You may need a Faraday shield around you pickup coil, balanced shielded cable to the amplifier.

You can also use a ferrite pot core microphone transformer to increase the input impedance and signal. That is also probably good for a 30x increase in signal. If you have an air-cored sense coil this will be the only way to get close to the noise floor, as no transistor amplifier has a low enough input noise resistance. (if you use a ferrite core resonant sense coil, you probably don't need it). You can resonate the sense coil with a small C on the secondary of the transformer.

\$\endgroup\$
  • \$\begingroup\$ I think a resonant circuit is the best idea, thanks, I need to use an air core pickup coil not ferrite. So I need to work out which type of circuit to use - by "capacitor across L2" do you mean in parallel with L2? The ferrite microphone transformer sounds a good idea, but how careful would one need to be with magnetic shielding? \$\endgroup\$ – MichaelT Jul 3 '18 at 23:41
  • \$\begingroup\$ The preamplifier is already battery powered. \$\endgroup\$ – MichaelT Jul 3 '18 at 23:42
  • \$\begingroup\$ Microphone transformers already have very good shielding systems designed for them, so this is a solved problem. You can wind your own using a ferrite pot core - these have very low leakage, and a matching shield can. You may be able to use one designed for a ribbon microphone - I don't know if they go high enough in frequency. You may be able to get one made for ultrasonic sonar microphones. \$\endgroup\$ – Henry Crun Jul 3 '18 at 23:46
  • \$\begingroup\$ Shielding matters more if there are in-band fields. If you make it resonant, then you won't pickup much 50Hz. Since most SMPS are about this frequency you could rewind a transfomer from a laptop supply. EI type transformers have poor stray field, but it would work, if not be ideal. \$\endgroup\$ – Henry Crun Jul 3 '18 at 23:49
  • 1
    \$\begingroup\$ Are you also generating the 25kHz excitation signal. i.e. is it available to you as a strong signal? Do you actually need to get to the noise floor? Is there interference? (magnets, power systems etc)? If you would like to talk this through, you could call me viber/whatsapp/phone, and I would be happy to discuss it. \$\endgroup\$ – Henry Crun Jul 3 '18 at 23:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.