DIY low-field continuous-wave NMR spectrometer

[ChristofferB]

Finally got my large ferrite magnets! Thought I'd share my progress so far!

I'm attempting to build an NMR spectrometer (it's like an MRI but for single molecules). Pole pieces are in the mail.

I'm going to make a central block of 30mm thick acrylic with a bore to accept the probe, then have pole pieces and magnets stacked on the outside of that, and outermost acrylic blocks with a pattern of tapped M6 holes to add steel bolts to to shim the field for homogeneity.

I've attached the schematics for the magnet.

NMR can be done 2 ways: continuous wave, where you have a continuous unchanging frequency and then a ramp generator driving a coil around the magnet or pole piece, sweeping the magnetic field, or pulsed, where you give a short RF burst and then do FFT on the resulting echo (modern approach).

If I ever want to sweep the magnetic field, I will add a coil to one of the pole pieces, if not I won't bother.

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.

I quicly made two crudely orthogonal coils and wrapped around a sample tube of water and tried it out with an RF gen on one coil, oscilloscope on the other and lo and behold! no signal on the scope until i reached around 4.5 MHz, then it grew and tapered off again. Unfortunately I think I've just found the frequency where the coils are in resonance because it makes no difference if i remove my sample tube. aw.

Either way, I've found a truly great review of all the electronic oscillators and detectors used in CW NMR, and it's freely accessible! give it a read!

My magnet is still very incomplete, I haven't ordered the acryllic yet, and i probably need to do a lot of grinding and hand lapping of the pole pieces to make them polished and parallel enough for this sort of work.

They have a small 5mm long 1mm wide cut in the edge towards the centre, I really hope that won't mess with my homogeneity too bad.

Either way hope you found this interesting! I'll share more when I've actually gotten something working!


Cool review:
Morris, Breed, Newton:
Advances in Electronics Prompt a Fresh Look at
Continuous Wave (CW) Nuclear Magnetic
Resonance (NMR)

Electronics 2017, 6, 89; doi:10.3390/electronics6040089

--Chris

[pwlps]

Cool review:
Morris, Breed, Newton:
Advances in Electronics Prompt a Fresh Look at
Continuous Wave (CW) Nuclear Magnetic
Resonance (NMR)

Electronics 2017, 6, 89; doi:10.3390/electronics6040089

--Chris

I don't know if it is of interest but I saw the Robinson oscillator detector circuit exactly as the one shown in Figure 10  (Faulkner and Holman, ref. 27) in a commercial NMR magnetometer from the 80's. The probe was using a frequency sweep (rather than field sweep).

[RoGeorge]

Very interesting! 

Isn't it supposed to be a closed magnetic circuit, in order to improve the uniformity of the magnetic field uniformity inside the gap?

My wild guess is that without coercing the magnetic lines through an external U shape magnetic circuit, the field lines will close through air for each magnet and thus the magnetic field inside the gap won't be very uniform.

If the magnetic field is not uniform over the entire probed volume, then different regions of the sample will resonate at different frequencies, thus making any resonance hard to see.

Again, I have zero hands on experience with NMR, but looking at the setup I'll say the field between magnets will vary a lot without a confined magnetic path.

I'm very curious to see how it will go, please keep posting the progress even when it doesn't work as expected, like you just did.  Thanks for sharing.

 

[ChristofferB]

Thanks for the interest! I found a modernized version of the marginal oscillator using a Jfet instead, and having some more fine control. The original author claims it can be swept easily by adding a varactor in paralel with the tank capacitor. I'm either gonna do that, or make a power amplifier to sweep the magnetic field, either way I'm going to rely on the ramp output on my oscilloscope, as it's already synced to the scope and I don't think I can do better anyway.

And you're right, having a flux return yoke would be good, but it's also a lot of additional metalworking, most of which I will have to pay others to do, so I'm gonna start without and see if i can shim it into homogeneity.

My probe so far consists of 90 windings 7 mm dia. coil 3 cm long. I have a bit of a cyclic problem, since I can't find the larmor frequency without the probe but I can't really design the probe without knowing which freq. it should be resonant at.

I'll keep making updates for sure.

--Chris

[Kleinstein]

For the spacer, holding the two magnets apart, I would not use acrylic or any other hydrogen containing compound. This would give a background from further out. The thin strips shown so far may be acceptable.  I would prefer something like ceramics, glass or aluminum.

An outer magnetic return path could replace the spacers and give a stronger field (e.g. something like 4-10 times). In addition it would reduce effects from magnetic material in the surrounding.  In the current form iron in some 50 cm around the magnet would effect the homogeneity of the field. I agree that it takes quite some effort and material.

p.s.: For finding the resonance it could help to measure the field strength with a separate sensor.  With a DSO or digitizing DMM one might get away with turning a small coil in the field.  One can tune the probe coil with an external trimmer cap to get a larger frequency range with a single coil.

[ChristofferB]

Thanks for the interest, I agree that the spacer might not be ideal but the reason I'm sticking to this specific magnet setup is, aside from ease of construction, is that I've read a series of articles using one very similar, and actually getting structure resolution data out of it.

From the article:


source: https://www.sciencedirect.com/science/article/abs/pii/S1090780716302713

Maybe some stiff expanded polystyrene would be a good compromise, having a very low density and still being workable.

Another idea for the return path: getting a lenght of square structural steel tubing 10 cm wide on the inner diameter, with 5-6 cm inner height and put moth magnets and pole pieces on the inside. Tough to machine on the inside, but it could be filed reasonably flat and then shimmed with aluminum foil.

But that's in the future, I'll get to work on building the marginal oscillator, so I can actually see the resonance. The width of the resonant area for a single proton environment divided by frequency is a good measure for homogeneity.

See attached schematic. I plan on adding a BNC port somewhere to connect a frequency counter to.


Thanks for the interest!

[moffy]

Please keep us updated with your progress. The project is truly fascinating!

[ChristofferB]

So, I think I have designed a yoke that I can manufacture without machining or welding.

It's based on the (hopefully correct) assumption that the geometry and tolerances can be much sloppier than in the pole pieces and in the parallelity of the faces of the magnets.

I've cut an I beam in half (with a sawzall, it wasn't pleasant), and found some 60mm wide mild steel flat bar pieces.

My plan is to have a magnet on each face of the I beams, now T beams, i guess? and use the 60mm for distance.

The whole assembly should hold together just by the magnetic force, but I don't like the narrowness of the flat bar, in terms of it suddenly sandwiching in on itself, so I'll put a 6mm hole in each corner of the I beams and put some m6 threaded rod through for stability, to clamp it all together.

The faces of the I beams are definitely not flat, they're slightly concave, but I guess I can shim it with some thin foil of steel or aluminium without ruining the flux path too much.

I've modeled it, and the cut parts of the I beam doesn't seem to carry any flux, which is good. It also doubles as a magnet stand so that would be handy.

What do you think? Is this a yoke or a joke?

See attached.

Thanks for the interest!
 

[ale500]

As a chemist, I applaud  . as a fellow DIYer great work !. Old machines used to have an oscilloscope, too, the 60 and 90 MHz ones. The 90 MHz machines had a 1 T magnet, they also weighted 1 Ton, too .

[ChristofferB]

Thank you so much! Yeah, permanent use of oscilloscopes for scientific instruments such as XY displays, timebases and amplifiers totally justifies hoarding all those semi-vintage 20 mhz scopes!

So, my gaussmeter (hall effect sensor, SS49E) only really goes to 0.1T, according to the datasheet, but the probe reads 0.15T between the pole pieces.

I got the magnet assembled and painted! I'm pretty pleased! 8 kg of yoke should be decent! I'll keep off the shimming until I can get the oscillator done, and then I can tune after the narrowest peak. Parallelity will come with the pole pieces currently in the mail.

Either way I'm quite happy with it! The threaded rod isn't really needed, it clamps itself together just fine! It's just a "don't crush my fingers" measure, and also to ensure it doesn't shear sideways and fling out the steel blocks.

--Chris

[Kleinstein]

With the configuration of strong magnet and coil around the sample one can also detect mechanical resonances in metallic (non magnetic) samples. This may work for a first test as the frequency can be calculated from the dimensions and modulus. The frequencies for a few mm samples tend to be in the 0.2-2 MHz range - with quite a few resonances.
I have tested this in the pulsed mode of operation (high power (e.g. 50dBm) pulse + detection on decay), but it should also work CW. AFIAK the resonances should be stronger than from NMR.

The yoke looks good - maybe the side parts are a little close to the magnet, causing some inhomogeneity at the edge.
0.15 T is already quite a strong field and reducing the gap with the pole pieces will increase the field further.

[ChristofferB]

That's interesting! Kinda like what is seen in quartz crystals? I'll have to try that. It's also a good-to-know pitfall that could make me think i was seeing the proton resonance.

I would like the sides to be all the way out too, but unfortunately the field sucked them in and they're very tricky to move now. All the way out there's 10mm to the magnet edge, maybe I should make some acryllic spacers on the inside to keep them in check.

I hope i can shim it by just adding bolts or nuts to the back side of the yoke, that would be ideal. I'm unsure what effect the threaded rod will have there, maybe it'd have been better to add a few more.

Either way, I'm considering the probe design, I think the best would be to have a sheet of copper clad pcb riding on top of the magnet, with some nylon screws to adjust height and position of the sample, and then have a soldered copper box or pipe down the centre, with the coil in, so the sample tube is inserted through the pcb. That would also allow for coax connectors and maybe a preamp close to the coil.

-60/40 Sn/Pb solder isn't magnetic is it?

--Chris

[RoGeorge]

I don't think 60/40 solder is magnetic, at least I don't feel it attracted by magnets.

A way to test for very faint magnetic interaction is to put a small bit of the tested material to float on water then to keep a strong magnet nearby (almost any small blob of material can float on water because of the superficial tension, or by improvising a tiny boat/raft for bigger lumps).

Some apparently non magnetic metals are in fact diamagnetic (Landau diamagnetism), but I didn't tested for solder.

Last weekend I revisited an old NMR project that I put on hold many years ago (that is why I am so curios about your CW NMR).  I've seen solder blobs on some other DYI NMR machine (e.g. https://youtu.be/Bf3r9Db0TSs?t=269 at minute 4:29 is the reading coil for a FT NMR, with lots of solder around, but he is sweeping the magnetic field intensity, not the excitation frequency).

The FT NMR project I was trying back then was aiming to get 3D MRI (Magnetic Resonance Imaging) in the Earth's magnetic field.  The best found online at that time was the 3D image of a fruit and vague hints about its internal structure, with a complete 3D scan for an object the size of an apple being ready in "just" a day or two  .  Can not find that DIY project any more, but the results were similar with the 3D demo from this educational machine http://www.magritek.com/wp-content/uploads/2013/12/Magritek-MRI-2011-web.pdf (note that that 3D image took hours to scan, and the resolution is very low, so pretty unusable).

To read the NMR signal there was a sharp band-pass filter to reduce the out of band noise, and a few amplification stages.  All I've seen use to had filter + low noise pre-amplifier, thinking here it will be very unlikely to see the NMR signal with just an oscilloscope. 

Didn't know about the trick with the marginal oscillator, that was very interesting to learn, thank you.

Another thing that I found fascinating this weekend was to learn about how NMR can help to identify organic molecular structure (yep, never dive into that before).  Judging by the other related project of yours, the gas chromatograph https://www.eevblog.com/forum/projects/gas-chromatograph-diy-done!/ , I guess this is what you are after, organic molecule identification by NMR, is it?

These are the videos that I liked so much (linking them here so I could find them in the future, after I'll forget  ), from the 'Knowbee' channel:

[Kleinstein]

The marginal oscillator is kind of a very sensitive amplifier, a little like of old days regenerative receivers. So the signal can be quite large at the output of the oscillator. It also acts as a low BW filter.

The mechanical resonance would mainly be a problem (possibly confusion with NMR resonance) with a small (e.g. 3 mm ball) relatively isolated solid part.  The Q-factor and thus sharpness of the resonance depends on the material, but can be lowered with coupling. The glass test tube with a coil should have a relatively low Q, more like the LC resonance.

For tuning the LC resonance one should keep in mind that variable capacitance diodes have quite some loss and thus should not be used for too much. So the varicap should only be a smaller part of the total capacitance.  The coarse part should be better with switches capacitors or an air trimmer.

[langwadt]

As a chemist, I applaud  . as a fellow DIYer great work !. Old machines used to have an oscilloscope, too, the 60 and 90 MHz ones. The 90 MHz machines had a 1 T magnet, they also weighted 1 Ton, too .

why would it be called a 90MHz machine if it is only 1T? Hydrogen is ~42.6MHz at 1T

[ChristofferB]

Thanks for the continued interest!

I'm still waiting for my pole pieces, as well as parts for the marginal osc.

I found this design for a permanent magnet sweep coil driver, and honestly it seems much easier than trying to make a stable RF VCO, so I think I'll start there.

I'm currently trying to design the mechanical layout of the probe, since I need a coil form that's not conductive, but also doesn't show up on NMR.

I'm thinking, making a shielded 1 cm thick copper foil box that can be lowered down in the field, and then have a hole in the top for the sample tube.

Tube guide and coil form will be a - don't laugh - drinking straw epoxied in place.

It's very thin walled extruded polystyrene, and very low density. and has the right diameter. I think that's a winner. I think I will have both ends open, so I can make a variable height stop to position the tube with.

What do you think?

 

[ChristofferB]

Just did some more design of the probe. I think this is what I'll try.

Probe box will be 2 layers of double sided copper clad pcb, sandwiched together with brass standoffs. On the periferie of the brass standoffs a copper foil band will be soldered all the way around. Perhaps one of the PCB layers could be foil too. Either way, straw glued into the center, and a coil (or multiple) wound onto that. I'll probably start with around 50 windings, depending on the straw diameter.

You can see how I'll shield the coil signal all the way on the attached schematic.

Only problem is getting it straight in the center of the field. I think I'll make some small clamps with nylon screws that can be adjusted and clamp onto my yoke.

Maybe I should get some small thick O-rings to slide over the NMR tubes so they'll just hang at a set height.

Oh, and I just butchered a transformer for some laminated electrical steel shims. When I get the oscillator running I'll try using small (3-5x10mm) shims on the magnet poles to homogenize the field.

[Kleinstein]

Adding an electromagnetic field to the magnet could work. However it could make it a little more difficult to make sure the field is homogen.
The circuit shown is good for a small current up to about 1 A. This may be enough, if the coil has enough turns. A high electric power is problematic anyway, as the coil would than heat up the yoke or magnet and this way may change the field. It may need some protection (e.g. trans-sorbs  parallel to the coil to prevent a possible damage from voltage spikes (e.g. during turn of or the trace return phase).  I would expect the ramps to be rather slow anyway, so that the coil inductance in more like small.

At some 0.25 T the frequency would be in the 10 MHz range - 50 turns may very well already be too much than. The oscillators usually have an impedance range where they work best. The coil is essentially like an air coil an there are programs to calculate the inductance. The shielding / case would reduce the inductance a little.

I would consider have the oscillator really close to the probe, possible even inside the magnet., inside the box. This would reduce losses and possible other effects of cables. At least the main tank circuit should be there.

If the field is good, the exact position of the sample should not matter.

[ChristofferB]

I honestly can't imagine I need that much current. I have to sweep the field something like 60 miligauss to get a 50 ppm frequency shift.

I am unsure how to design the sweep coils, so after finding the larmor frequency I may just try out two identical coils in series of an arbitrary number of turns, and then have a lab supply to drive them, and then sweep over current ranges to see which range would be useful. maybe 5-10 windings of some thin enameled wire.

I've added the pole pieces, after filing and grinding them as flat as I can. If they need to be better I'll get a machine shop to do a gentle mill pass on them.

The field strenght is constant within the accuracy of my hall effect probe over almost the entire area at 0.115 T, or 1150 gauss.


I think I've found a cheaper alternate to the probe: making it from perspex/plexiglass coated in adhesive copper foil. There's apparantly a company who sells them made like that:




Thanks for the hints!

[ChristofferB]

So I've been productive! Almost ready to see magnetic resonance!

I've completed the marginal oscillator. It took me forever to troubleshoot the simple AF amp (see schematic in prev.post). It turns out the V- pin of the metal can LF356 is connected to the can, and of course I'd just dead-bug soldered the can to my ground plane... It doesn't even mention it in the datasheet!

Now the oscillator works as expected, any change in RF amplitude gives a quick AF spike.

either way, I've designed a probe. Haven't settled completely on the coil yet, as it's slightly too high (5-7 MHz).

Anyway, I feel the mechanical design turned out great! Hope you agree!

--Chris

[Kleinstein]

The mechanical setup look good.  However I would prefer to have much of the capacitance for the probe resonance down at the coil to reduce the current going through the BNCs and possible connector effects.

[ChristofferB]

That sounds like a good idea. I'm just worried about the component leads being magnetic and skewing the field but I guess it's a very small mass of magnetic material. I'll add 100pf across the coil. Maybe I'll even get it ton desired freq range

Thanks!

[Kleinstein]

Magnetic material in the cap is a good point. Some ceramic caps may contain a nickle layer. I would expect film caps (e.g. PS) to be non magnetic, with only aluminum, copper and maybe zinc used as metals.

[ChristofferB]

That may be right. I'll see what I have on hand.

I'm also considering implementing the gradient coils (at first manually operated with a lab psu) as just being smaller (3 cm dia maybe) coils inserted onto each side of the probe. I havent yet solved how best to put coils around the pole pieces, as there arent much space, and thin wire coils that large would be very floppy. Perhaps The best would be to get some coil forms that fit over the pole pieces 3D printed, and then just slipped over.

or maybe I should just wind them on a mandrel slightly larger, using thin wire and maybe not care about neat windings, and when they're in place, epoxy them into a bundle and slip them over, adding little plastic wedges to keep them centered.

[Kleinstein]

It should not matter much where the windings for the field sweep are. They should not be too close to the sample though, to to couple too much AC wise - though at a different angle, but nobody is perfect. So the extra turn could even be around the spacers of the yoke.