DIY low-field continuous-wave NMR spectrometer

[pwlps]

As I see your project advances fast, I'm eager to see your first NMR signals 

What kind of sample are you going to try ?
Maybe this is too much anticipating your future results but remember in CW NMR knowing the sample T1 is important.  In standard CW NMR you need to avoid saturation to see the signal. In fact  you have two regimes.: 1)standard CW NMR with slow sweep through resonance where the condition to avoid saturation is: gamma*H1<<1/T1 ; 2) "fast adiabatic passage" (Abragam chapter III), where you deliberately saturate during the sweep through the NMR line, then the condition to avoid T1 effects during the sweep is H1/(dH0/dt)<<T1; but in this case the measured signal vs field does not give you the true NMR lineshape.  For (1) if you want to use water it might be good to add some paramagnetic ions to shorten T1.

[ChristofferB]

I am too! I'm honestly uncertain on that part of the theory, but I've read that a standard CW nmr test solution is concentrated copper sulphate (~1M) in water to shorten T1. I plan on attempting both fast and slow scans, as I use my oscillocope ramp gen. as timebase trying different things on an empiric basis should be doable.

The marginal oscillator as described earlier (not designed by me) is useless in sweeping the RF freq, since the amplitude isn't constant, so I'm building the sweep coils (first I'm trying 2 x 9 windings 3 cm diameter) connected directly to a lab psu.

I did however tune the detector to a point where inserting the CuSO4 sample would quench oscillations, at about 4.3 MHz, so perhaps I did hit magnetic resonance!
I'm not counting success before I see a peak, though.

The problem with using a ramp generator is that on the vertical down slope it'll rush back through the entire field, giving a graph on the screen that has a sorta good line, and then a return bad trace looping back in the background. Maybe it would be worth using the blanking input with a window comparator to blank the display on the return.

Thanks for the interest!

[Kleinstein]

For the sweep one could in theory use a triangle sweep, going up and down at the same rate. This would give two superimposed curves with possible hysteresis effects from the magnets and also possible difference due to a very high Q resonance and fast sweeping.
In this case averaging should be time based, not in XY mode as to get both parts separate.

I don't think blanking in run back is really needed. The up / down rates can be quite different if really needed. However the speed is limited by eddy currents - even if the current in the coil changes fast, the field will not change that fast.

The sample has two effects that can quench the oscillation: one is the NMR, the other is just eddy current due to the electrical conductivity of the liquid. So the liquid by itself would have an effect just by macroscopic conductivity.  To reduce the effect would be by using a non conductive sample, maybe some organic liquid / paraffin.

[pwlps]

The sample has two effects that can quench the oscillation: one is the NMR, the other is just eddy current due to the electrical conductivity of the liquid. So the liquid by itself would have an effect just by macroscopic conductivity.  To reduce the effect would be by using a non conductive sample, maybe some organic liquid / paraffin.

Eddy currents are independent of the field so it might not be a problem. On the other hand a non conducting liquid sample like parrafin will have a quite ridiculous spin relaxation rate (1/T1) making the detection much more difficult. I wouldn't care too much about eddy currents, after all there is for sure a lot of eddy currents in our body during an MRI scan 

[ChristofferB]

Many lab manuals for undergrad physics courses uses a solid plug of natural rubber as sample. That may be the reason. I just find that discouraging because that similar aliphatic chained molecules can't even be resolved on a 500 MHz 10T instrument, so I'd rather have a nice peak.

[pwlps]

Many lab manuals for undergrad physics courses uses a solid plug of natural rubber as sample. That may be the reason. I just find that discouraging because that similar aliphatic chained molecules can't even be resolved on a 500 MHz 10T instrument, so I'd rather have a nice peak.

Yes, rubber or another polymer might be a good sample because it is well insulating and at the same time has short relaxation because the motion timescale is much longer than in simple liquids making the dipolar relaxation mechanism very efficient.

[Mr polarbear]

First of all wow, your dedication and work is amazing. Before this thread I had never heard of NMR, after reading all the posts I started searching for any information I could find on NMRs, turns out there isn't a lot especially on the sweep/hermholtz coils.

In terms of things that actually are helpful and note worthy this is what I found;
(1) http://www.cco.caltech.edu/~derose/labs/exp5.html
(2) http://courses.washington.edu/phys431/cwnmr/cwnmr.pdf
(3) http://experimentationlab.berkeley.edu/sites/default/files/writeups/NMR.pdf
(4) http://www.conspiracyoflight.com/NMR/NMR.html
(5) http://www.jabolatorium.com/

Link 1 is a dumbified sweep coil circuit, might help with designing a circuit to drive the coils.
Link 2 uses 80hz triangle waveform instead of a ramp waveform thought that might solve your ramp waveform dilemma.
Link 3 has an interesting NMR-head at fig15 don't really understand the use for the paddle, I mean i get that it is used for changing the magnetic field shape, but why is beyond me. The rest of the article is not very helpful the equipment used is outdated.
Link 4 is the best and and can be a real building block in designing circuits for your NMR
Link 5 I know is an ESR, but it has some interesting circuitry.

Anyways best of luck in comleteing your NMR (I really hope you do)

[langwadt]

why continuous wave?

[ChristofferB]

Man, that you guys dig up this many sources and references is making this endeavor a whole lot more achievable - and enjoyable.

Thank you so much! I'll sift through it.

quick thoughts: 5) The ESR meter looks cool. I did attempt to collect enough waveguide hardware, couplers and gunn oscilators to build an x-band ESR/EPR spectrometer, but honestly, the amount of samples that are EPR active compared to NMR is very limited. Also I don't have experience with the real instruments, but I do with NMR.

4) I've wondered too, and I'm unsure what the copper paddle is. It might be flux steering - see fig. 4 in this article:
https://www.mdpi.com/2079-9292/6/4/89/pdf


Looking further it looks as if the illustration is borrowed from that article.



--CW nmr was chosen because it's circuitwise simpler. This entire project stands and falls on whether I can construct a magnet homogenously enough.
  I won't know until I see the resonance, so until then, simplest, safest, most time-proven is the way to go.

Reusing the system, along with a pulse generator, capturing the FID (maybe with a soundcard?) and doing FFT and such on multiple scans requires much more work. Maybe some day but it's definitely an option.

 

[pwlps]

--CW nmr was chosen because it's circuitwise simpler. This entire project stands and falls on whether I can construct a magnet homogenously enough.
  I won't know until I see the resonance, so until then, simplest, safest, most time-proven is the way to go.

Reusing the system, along with a pulse generator, capturing the FID (maybe with a soundcard?) and doing FFT and such on multiple scans requires much more work. Maybe some day but it's definitely an option.

I'm tempted to think a pulsed NMR electronics could actually be simpler to assemble (tweaking a marginal oscillator to get a good signal isn't that simple either) if you can afford buying critical components like a good LNA (short deadtime) and a power amp with a gain quenching capability (to limit noise injection from the power amp).  On the other hand it might be difficult to achieve a very short deadtime at such low freqs so that good field homogeneity would still be required.

[langwadt]

--CW nmr was chosen because it's circuitwise simpler. This entire project stands and falls on whether I can construct a magnet homogenously enough.
  I won't know until I see the resonance, so until then, simplest, safest, most time-proven is the way to go.

Reusing the system, along with a pulse generator, capturing the FID (maybe with a soundcard?) and doing FFT and such on multiple scans requires much more work. Maybe some day but it's definitely an option.

I'm tempted to think a pulsed NMR electronics could actually be simpler to assemble (tweaking a marginal oscillator to get a good signal isn't that simple either) if you can afford buying critical components like a good LNA (short deadtime) and a power amp with a gain quenching capability (to limit noise injection from the power amp).  On the other hand it might be difficult to achieve a very short deadtime at such low freqs so that good field homogeneity would still be required.

I'd try https://en.wikipedia.org/wiki/Spin_echo

[RoGeorge]

I just realized I already have magnets big enough to try NMR.   



Bought them for approx. $10 about a year ago, from LIDL, as helping-hands to keep in place ferrous materials while welding or assembling.  If I didn't messed up the calculations, the field measured between the 2 middle spacers is about 66.5 mT.

I measured the field strenght, |B| of the magnet in the center of the probe holder and it measures about 1200G or 0.12T. This gives me a hydrogen Larmor frequency of about 5 MHz.

How did you measured \$B\$, and did the value changed much after mounting the magnets inside the metal yoke?

[ChristofferB]

Cool! The more the merrier!

They're a little on the narrow side for homogeneity I think, maybe unscrew them and put them the other way? On the other hand then there's a hole in the middle..

I measured my field strenght with an ss49e Hall effect sensor I'd epoxied to a non magnetic handle. Then just reading out the voltage on a multimeter and using datasheet value of 1.4 mv/mG (uncalibrated) to calculate field strenght.

[Kleinstein]

The large read magnet may still be on the small size. The difficulty is usually to get a very homogeneous field. Just to get it running the filed does not even have to be that large - for the extreme case there is "zero" field NMR, using just the earths field.

The usual way to get the field homogeneous, it to use only a relatively small part in the center of the magnet. So it's a good idea to have the field over at least some 5 cm across, even if the actual probe volume in more like  5 mm across.
It may still be possible to add larger pole pieces to get a larger field volume. Chances are the steel plates on both sides are a little on the thin side.

[shadow.dark]

I am very interested in NMR. I have a VNA. Can I implement the NMR function by adding simple attachments?

[ChristofferB]

Thanks for the interest!

I think you'd have more success with a spectrum analyzer. Modern NMR is done by short RF pulses and then the decay is fourier transformed. Problem is the differences in frequency are very small so you'd need a VERY narrow-band high-res spectrum analyzer.

If your VNA can do some kind of FFT it might be feasable though.

The project is on a bit of a hiatus at the moment, since im developing HPLC detectors at the moment. I'll revisit it at some point though.

--Chris

[mrflibble]

I've been following this with interest. Cool project!

Did you have a chance to measure the homogeneity of the magnetic field? I wonder if maybe a Halbach array can create a stronger field with good homogeneity. The main tradeoff being smaller volume. And the other tradeoff being having to take more care to not get your fingers crushed during assembly of the array.

[ChristofferB]

Thanks for the interest!

I haven't gotten further with homogeneity, my plan for testing is to make circles of 5x5 mm graph paper and attach to the pole pieces and then measure the field directly on top in every square and see how things look.

I'm still at the problem of making magnetic sweep coils, and it's not looking good. I don't think haphazardly wrapping enamel wire around my yoke is going to give anything usable, so i think ill re-visit using some capacity diodes to sweep the frequency slightly. maybe just slowly manually at first or maybe connected to the scope ramp out. I think that's by far the easiest solution.

In reality i'll just parallel up 5-10 1N40XX diodes, since all diodes exhibit some capacity. I can always tune myself away from the uncertainty by attenuating the ramp signal and adjusting the freq.

[ChristofferB]

[ Specified attachment is not available ]Ok so the pressure of people following the project inspired me 

I made a varicon control with 2 1N4006 diodes in series, parallel over the tuning cap and coupled the center point out to a BNC through a 28K resistor, and was able to tune 80 KHz from 0 to 12V ! that's good enough for me! quite probably far too big a sweep! Anyway I hooked it up to my scope and adjusted the feedback to obtain this. (attached). This clearly shows a sweep is going on, as the oscillator is quenched in half the sweep! i imagine NMR resonance would look like a very sharp quench/unquench in this setup but I need to play around more with that.

Either way diode tuning is a go!

[Kleinstein]

The diode tuning is definitely an option to tune the frequency. However to be sure to get a magnet field dependent effect, one might still want a small coil to change the field.  The sweeps may still use the diodes and the coil only for slow shifts to a few different field setting.

The extra coil could have problems with the joke part of the magnet not instantly reacting to changes in the magnetic field. A missive piece of iron kind of slow down field changes and may also cause jumps / nonlinear effects. So a coil around the joke could cause trouble.
On the other side a coil close to the sample could dampen the detector coil.

[ChristofferB]

My main problem right now is I don't know what field strenght i'm working at, so i'm changing the tune range more or less at random.

My hall effect sensor don't go up to the strenghts i expect so I don't trust its reading. I think the ideal next step would be to ask my university's NMR responsible guy to borrow a research grade teslameter and get the magnetic strenght absolutely nailed down.

Being able to vary the magnetic field slightly might be a really nice idea. especially in the event that there's some strong RF noise source that's right on my resonance freq. Maybe just 10 turns of some 2.5 cm2 stranded installation wire on either side of the yoke would be sufficient at low voltage/high amp.

[ChristofferB]

Ok so sorry for updating this long-dead thread but I thought it better to keep all info here than to make a new one..

I did the terrible thing I didn't want: I made my magnet worse to get it into a range where i can measure the field strenght. I made the gap bigger (almost 10 cm now) and now I have a field strenght in the gap at around 43 mT. this gives a Larmor freq of around 1.8 MHz.

I also re-designed the probe to have a larger inductance coil, but for whatever reason, I cannot get the frequency below 1.85 MHz. Any larger inductance/capacity and the oscillator collapses. I'm clearly missing something about oscillator design. I have the varactor diodes and main tuning capacitor in the electronics box, then a 50 cm RG58 cable to the probe, which has a 40 pf trimmer and the coil.

Can anyone propose an improvement to the oscillator design that would help with this? Do i need to move more of the tank reactance into the probe? Or can i have a smaller coil in the probe as pickup and have the rest of the L in series in the oscillator enclosure, maybe wound on a toroid?

I'm at a loss.

Thanks in advance!

--Chris
 

[Kleinstein]

I don't think it's a good idea to have another inductor at the oscillator, as this would makea weaker coupling and more power going through the coax - so possibly more loss.  It is more like some other changes to the oscillator (e.g. larger caps at the supply, higher or lower current).

[RoGeorge]

Great to hear news about this project!   

From the history of the thread, I remember you tried a few different methods.

- Which one is now, oscillator frequency sweeping while looking for an amplitude peak, or magnetic sweep?
- A schematic to the oscillator you are testing now, or a link+page to the exact schematic would help, maybe a draft hand-draw of the setup, too.
- Does the oscillator work at low frequencies if you solder the LC tank on it, without using the 50cm transmission line?

[ChristofferB]

You're right! This should come with an updated schematic, see attached!

I am counting on a frequency sweep in a static magnetic field for now, When i get a low-impedance magnet driver made I might just use that as a fine adjust.

I was assuming with a big tuning cap, I couldjust make a range of coils for different frequencies, as is often seen in dipmeters and old ham receivers.

Trying to eliminate the larger R from having a coax between my L and C is a good idea, I'll update you on how that goes shortly.

Thanks again for the interest!