Sciencemadness Discussion Board - do-it-yourself nuclear magnetic resonance spectroscopy

do-it-yourself nuclear magnetic resonance spectroscopy

Pages: 1 2 3

aliced25 - 22-10-2010 at 05:54

@ not_important

Thank you, that is pretty much what I was getting at. However I like the idea of the high multiplication/division rate on those PLL's, by using a simple SAW Filter at a high wavelength one could reduce the signal width enough that when it was fractionated on the next PLL, the bandwidth was smaller than is possible with DDS Chips. Shit, it may even be possible to utilize the pretty weak DAC's on many DSP's (using the onboard memory & perhaps even the system clock). That removes two parts and leaves us with a monotonic signal on a narrow bandwidth, highly accurate wavelength. I have seen numerous articles where the shape of the pulse produces different effects in MRI especially, but also with NMR. I do wonder how a simple sawtooth pattern would run, it wouldn't be a perfect SINE wave, but it would approximate one. By the time it is put through a phase filter, multiplied to the shithouse, filtered and then fractionated down to the target wavelength, I wonder what it would look like?

@ Twospoons

Yeah, but Lead Based solder & Bromine/Chlorine containing Dielectric "were" cheap because everyone did it too...

I strongly suspect the end is nigh for the etched PCB, bucketloads of waste polymer, more buckets of toxic waste from the etching process & a board that cannot be recycled as anything other than fill.

IC manufacturer's use wire almost exclusively to join pressed metal components, which are then sandwiched into a small plastic box.

The size of the traces is going to be the start of the end for them anyway, 0.25mm traces are just way too small for decent signal quality, but they are all that will fit with BGA's.

I strongly suspect Circuit Boards may go back to being wired, many components will have to be rejigged to enable this to be done in the most expeditious manner, but the integrating circuit between the IC's is getting to be an IC in it's own right, wiring reduces the expense involved in applying copper, then masking, exposing, cleaning, etching, cleaning then tinning, in terms of home users, they are going to be stuck between a rock and a hard place. The Green's will control the Senate here (from the next changeover date, used to remember it, now I cannot be fucked). There is little prospect that they will look favorably on any move to clean up the Electronics industry - changing to dielectric material that can be dissolved in solvents, leaving behind the SMD's, IC's and Copper as solids, ready for reuse/recycling, which does not generate massive amounts of toxic and unrecyclable waste during production, is a bigger step in that direction then Lead Free, Halogen Free Components.

As to the ease of mass production - mass producing circuit boards using the same technology used now to make smaller and smaller IC's, predominantly as a means of connecting those IC's to one another - is no great stretch to imagine. Robots are used to locate, weld, string, cut & weld the wire connectors between the pressed metal pads in IC's now. There is no solder (and little heat - they use Ultrasonic Welding/etc.). Why it should be any more difficult for major manufacturer's to tool up to make bigger circuits, with higher clearances, etc. is beyond me (in fact, I suspect it may be the next, necessary step in order to increase board density).

@ arsphenamine

There is a paper on the previous page from someone who tried to measure the field from 4 Neodymium Cubes (~0.6T IIRC) with a Hall Sensor. Their results were indifferent, so they tested the field by reference to the resolution they could achieve with a syringe full of water. In terms of testing techniques & sensors, that looks like it would be the most accurate by far (while it is rather easy to equate the testing procedure to the end-use).

watson.fawkes - 22-10-2010 at 06:47

Quote: Originally posted by aliced25

There is a paper on the previous page from someone who tried to measure the field from 4 Neodymium Cubes (~0.6T IIRC) with a Hall Sensor. Their results were indifferent

Of course they were indifferent. If you're trying to measure a spatial derivative of B, you don't go about measuring B in two different places and then subtracting. A Hall effect sensor measures B directly. If you had a measuring device of infinite resolution, this might work, but since you don't, it doesn't work. If you're looking at inhomogeneities in the range 10E-4 or so, it means you're knocking off 40 dB of dynamic range right off the bat, before any other noise sources. What you need is a differential sensor that senses spatial derivatives of B directly.

I've thought of three ways of doing this. A pair of Hall sensors and a differential amplifier is conceptually feasible, but you've got to worry about induced EMF in the other conductors, for example. The best way to deal with this is to make everything small and put everything on a single chip, but I don't have a fab available.

You can mechanically vibrate a loop or coil in the field and rely on electromagnetic induction. This directly measures delta-B because the induced EMF is proportional to the change in magnetic flux. Since we can assume the area of the coil doesn't change, the signal comes only from delta-B. This all assumes that the axial direction of the coil is parallel to the direction of motion.

You can also change the induced EMF in a coil by switching a diode in and out of conduction. This requires some fancy electronics, because you have to measure induced EMF on top of a changing forward bias current.

arsphenamine - 22-10-2010 at 12:33

Quote: Originally posted by aliced25

There is a paper on the previous page from someone who tried to measure the field from 4 Neodymium Cubes (~0.6T IIRC) with a Hall Sensor. Their results were indifferent,

It means they had not the will to try a little harder.

Even the humble Allegro linear Hall sensors can resolve to .1 G (1e-5 T) if you take care of external noise adequately.

The 150µV noise floor over a 4.5 Volt range is ~15 bit resolution. Unfortunately, they only have a +/-1400 Gauss range at most.

Quote:

so they tested the field by reference to the resolution they could achieve with a syringe full of water. In terms of testing techniques & sensors, that looks like it would be the most accurate by far (while it is rather easy to equate the testing procedure to the end-use).

The reported value was necessarily taken from the median of a broad peak.

There was no mention of the peak's Q which means there were no hard figures on the field homogeneity.

They told us where the ballpark was but couldn't find the bases.

aliced25 - 25-10-2010 at 13:37

That might be so, it is still the best route I can see to working out where the bases should be.

Anyone got any ideas on the shape & placement of Shim coils?

Twospoons - 25-10-2010 at 16:52

Quote: Originally posted by aliced25

As to the ease of mass production - mass producing circuit boards using the same technology used now to make smaller and smaller IC's, predominantly as a means of connecting those IC's to one another - is no great stretch to imagine. Robots are used to locate, weld, string, cut & weld the wire connectors between the pressed metal pads in IC's now. There is no solder (and little heat - they use Ultrasonic Welding/etc.). Why it should be any more difficult for major manufacturer's to tool up to make bigger circuits, with higher clearances, etc. is beyond me (in fact, I suspect it may be the next, necessary step in order to increase board density).

You've never been involved in electronics manufacturing, have you.
I mourn the loss of lead solder - its so nice to use, and the alternates are all more expensive. Its just a knee jerk reaction to the Lead Boogeyman. Bet you car has more lead in its battery than all the electronics in your house.

PCB and IC manufacture may seem similar at first glance but your really can't compare EUV lithography at 32nm, class 10 cleanrooms, CVD, etc with etching copper in a bucket of ferric chloride ! Wire bonding in IC's is done with gold - a flame formed gold ball is pressed onto the bondpad, the other end is pressed and smeared onto the lead frame (which is usually etched, not punched, for fine pitch parts). The bonds are weak and gold is soft, so the whole lot gets buried in epoxy. You cannot do this macro scale with copper! Ultrasonic welding is not trivial either - the welding head has to be designed to suit the job.
You still need some way to hold the parts down, and something to stick them to. Copper clad PCB with solder is still the most cost effective way to do this, and will remain so for a long time to come.
I think you are also missing the break between design and manufacture - manufacturing companies do only that, build stuff for other companies, hence their processes are generic rather than specialised.

12AX7 - 25-10-2010 at 18:39

Quote: Originally posted by Twospoons

The bonds are weak and gold is soft, so the whole lot gets buried in epoxy. You cannot do this macro scale with copper! Ultrasonic welding is not trivial either - the welding head has to be designed to suit the job.
You still need some way to hold the parts down, and something to stick them to. Copper clad PCB with solder is still the most cost effective way to do this, and will remain so for a long time to come.

Those black lumps you see on mass production stuff are direct mounted dies. I suppose they use wire bonds, which will be gold, but I doubt the traces are even plated. The whole thing is covered in a goober of epoxy. What better way to not only reduce expense, but also add a layer of copy protection, when you need a billion of something?

Tim

Twospoons - 25-10-2010 at 19:34

It can be done with wire bonds, or the chips can be attached upside down (flip-chip) using gold balls on the bond pads. Either way, a PCB is still required.
It only makes economic sense, as you say, when making a billion of something. Assuming you can get the chip maker to sell you naked die.

aliced25 - 29-10-2010 at 15:39

No, I wasn't thinking about working from the bare wafer. I was somewhere else entirely, I recall somebody saying that the density of Computers would double every x years, then he came back and stated that based upon what was happening, it would double in half that time, the number of Integrated Circuits on a single board is growing until it is veritably unsustainable. IIRC it was one of the old Microsoft people in an interview (or several interviews).

Anyway, that is waaaayyyy off-topic...

I'm wondering about the design for the shim coils, what shape, where should they be placed and is there any prospect of modeling them with FEMM or some other package to determine the interaction (good or bad) with the permanent magnetic field.

The other real issue to solve is how to process the digital signal, which would presumably be on the DSP (or in Flash Memory) waiting to be called.

What would we need a basic GUI, hooked up internally to one or more of the free Spectral Database(s), which can compare points that have been assigned in order to find possible, probable and definate matches. As definate matches are found, their spectrum should be removed from the complex spectra, in order to allow the spectrometer & the user thereof, to try and eliminate other possibilities.

not_important - 30-10-2010 at 09:12

Quote: Originally posted by aliced25...

The other real issue to solve is how to process the digital signal, which would presumably be on the DSP (or in Flash Memory) waiting to be called.
...

If you're meaning that the FID data would be stored in Flash memory that might work. If you're intending to gather the FID data and filter+FFT using Flash as working memory, that won't work.

Flash memory is closer to a fast disk drive than to RAM - either SRAM or DRAM. While individual memory words may be read at random as if in RAM or ROM, writing is a different story. Common Flash memory requires updates to be done to blocks or memory, this usually means erasing the block so all bits have the same value, then writing data which will flip the values of some bits - you can only write zeros or write ones depending on the type of Flash, erasing sets all bits to the other value.

The erasure is moderately slow, so treating Flash like RAM makes for a very slow memory. If you just want to collect data for later analysis elsewhere then you use a pair (or more) of buffers in true RAM filling first one and then writing it out to Flash as you use the next buffer to capture the following sequence of data; this is similar to the way disks are used in data collection.

Another aspect of using Flash as RAM is wearout - writing/updating Flash causes it to slowly deteriorate past some point data errors start becoming important. Most current Flash runs between 105and 106 cycles exactly what this means depends on the type and brand/style. The point is that while using Flash as you would a disk generally creates no data integrity issues, using as RAM may well do so; look at how many accesses a 1K FFT processing would do.

As f or the automated analysis you'll need resolution to 1/10 ppm to make it worth while, spinning the sample controlling its temperature and so on. You'll also need a good reference, be it TMS in the sample or locking to the 2H of the deuterated solvent and adjusting the proton frequency data from that.

[Edited on 30-10-2010 by not_important]

[Edited on 3-3-2015 by Polverone]

aliced25 - 1-11-2010 at 22:25

Well, there ain't no deuterated solvent going to be used, that's for damn sure. I'm trying to look into the design of the coils and the placement thereof. Are you certain that the original concept of spinning the sample was down to 10ppm? I do remember reading about it, but I do seem to recall reading that when the concept was first suggested.

Anyway, we are slowly creeping down toward a couple of hundred ppm, without shims. I do remember reading that book/article you posted about shimming, why do they spin the coil as well as the sample? It seems to me that if they did not, then they would not have repeating spikes as the coil passes through different field strength zones.

Thanks for the heads up on the Flash, I'll have to look a little more closely. I may have to incorporate a memory chip on the board, but that would put the FFT/etc. on the PC/Laptop, which would allow for a less complex MCU/DSP (well we really aren't utilizing a tenth of what the one in question is capable of).

But memory can come later, I badly need some suggestions on shims. Anyone got any papers on the various approaches or ideas on what would be best suited?

not_important - 2-11-2010 at 12:28

Spinning is used in just about all lab-analytical devices, just about needed to get the needed uniformity. A NMR that can only resolve to 10 PPM is not much use save for measuring water content or the curing of glue/plastics that have active hydrogens that react with some other group. Most NMRs only spin the sample, read that again. You spin the sample so it sees all the field values, slightly broadening the peak but preventing a given type of proton from appearing as several peaks. remember the the FID is 100s of milliseconds long, each part ofg the sample sees much of the field a number of times during that time; and their field is rather uniform to start with.

Without deuterated solvents you are pretty much limited to compounds that are liquids, and you'll need to get TMS as a reference. Without some real trick fast sample swapping you can't depend on attempting to calibrate with a known then replacing it with an unknown, you need to calibrate while measuring the sample. You could use fluorine as the reference, something like perfluor-oneopentane but then you need a solvent for that, and fluorine's peaks aren't as clean as hydrogen's.

Shimming was what discouraged me from going ahead with building a FT-NMR ...

aliced25 - 4-11-2010 at 03:52

OK, so we need as much information about shims and their design as possible, this is where we come when we cannot figure shit out on our own is it not?

Alright, what do we know about shims?

They are small electromagnets, so PWM control should really allow for control of the same.

What do we know about the inhomogeneity of this design? The problem lies in the x,y coordinates approximately 1/4 of the x-axis from center and about the same on the z-axis (presuming y is vertical & x,z = N-S and E-W respectively).* In terms of spatial homogeneity this is not as bad as the small portable NMR-Mouse, etc. which do not use homogeneous fields at all.

But at least we are only shimming in the order of 50-100ppm (with some designs). It seems like we are on the verge of being able to come up with something at least. arsphenamine seems to have plenty of information on shims and their design, I've seen some articles on the subject, but they didn't spin as they were using MRI, in which the coils spin around the sample.

Can anyone work out how to model electromagnets in FEMM? That would have to be a start, even if we can only model them in 2D, it has to be better than nothing.

* Always presuming (and I hate to do so) that the magnet is at least moderately homogeneous vertically, which is a big call.

not_important - 4-11-2010 at 08:48

PWM with really good filtering, any ripple there is going to show up in the FID signal as shifting/jittering of the peaks.

MRI spins the coil because spinning a patent would add to the discomfort, and the coil isn't for field homogeneity but rather creating a gradient that provides a 'slice' through the patient that is the active imaging scan. Rreally quite different than you are attempting to do.

aliced25 - 4-11-2010 at 15:34

Spinning the patient, damn that would sure take away from the boring side of the discomfort though wouldn't it?

MRI came from NMR, it works on the same principles using electromagnetic coils to produce the magnetic field. There are several papers on using permanent magnets to produce a field so the researchers could image small rodent brains. Magnetic Resonance Imaging used to called Nuclear Magnetic Resonance Imaging, which kind of shows the relationship (NMR Imaging).

If small DSP's can be used to recover the signal from that and plot the position (ie.2D NMR Image) and control the gradient coils, I think shimming using smaller gradient coils should be possible. In fact looking at papers like this and this I suspect we can probably work it out (I don't expect it to be easy). What I find interesting with MRI is that they utilize the gradient coils to re-intensify the FID, as a spin-echo, why can't we use the same concept?

[Edited on 4-11-2010 by aliced25]

[Edited on 4-11-2010 by aliced25]

not_important - 4-11-2010 at 16:46

You can use spin-echo, just good timing and control of the RF is needed.

Do realise that imaging is a _less_ demanding task than structure determination; MRI is mostly looking at H2O or H2O vs 'hydrocarbons' such as fat - and how tightly it is hydrogen bonded to the tissues it is in while structure determination is looking at all the protons and their local-in-molecule environments. Please don't try to apply too much from MRI to your NMR, the homogeneity needed for imaging is less than for structure determination to start with and other factors are in play.

The 2nd ref - the PDF - gives most of what you need to design shims, now all you need is the tools.

[Edited on 5-11-2010 by not_important]

aliced25 - 6-11-2010 at 15:31

Yeah, that is why I added it

The plot with the Magnet Assembly shows that the magnetic field is inhomogeneous in a curve from North to South around the diameter of the circle (and through the center of the sample area). So I was thinking two coils or maybe even 10, (5 a side) to allow for the minute amount of additional magnetic assistance needed to get down to the ppm range.

Adjustment would be simplified by the fact that what you do to the first & fifth coil of each side is the same (so too the 2nd & 4th) - so 8 coils out of 10 would be adjusted as 2 coils (ie. 1,5,6 & 10 would be the same and 2,4,7 & 9 would be the same) the other two would be equivalent (or close enough, ie. 3 & 8), allowing for only 3 adjustments. Alternatively, 12 coils would present only 3 adjustment ranges, coil adjustment 1 would be 1, 6, 7 & 12; Coil 2 would be 2, 5, 8 & 11 & Coil 3 would be 3, 4, 9 & 10. If there were 16 coils, 4 PWM IO's could run the lot. That would mean that there were 4 coil adjustments and the miniature coils would be better (in my view) than 2 large ones, insofar as they allowed for finer adjustment.

Anyway, just for the hell of it, have a look at these (It'll take a little while for me to even get a slight grip on the maths):

Louis-S. Bouchard, 'RF Shimming for Ex-Situ NMR Spectroscopy and Imaging Using B1 Inhomogeneities' (2007)

Shapira, et al, 'Spatially encoded pulse sequences for the acquisition of high resolution NMR spectra in inhomogeneous fields' J.Mag.Res. Vol.182 2006 pp.12-21

Sakellariou, Meriles & Moule 'Variable rotation composite pulses for high resolution nuclear magnetic resonance using inhomogeneous magnetic and radiofrequency fields' Chem. Phys. Let. Vol.363(1) 2002 pp.25-33

Pryor & Khaneja, 'Fourier decompositions and pulse sequence design algorithms for nuclear magnetic resonance in inhomogeneous fields' J. Chem. Phys.125 2006 pp.194111/1-94111/6

Heise, Sakellariou, Meriles, Moule & Pines 'Two-Dimensional High-Resolution NMR Spectra in Matched B0 and B1 Field Gradients' J.Mag.Res. Vol.156 2002 pp.146–151

Chang, Chen & Hwang 'Single-sided mobile NMR with a Halbach magnet' Mag. Res. Img. Vol.24 2006 pp.1095–1102

Frydman, Scherf, & Lupulescu 'The acquisition of multidimensional NMR spectra within a single scan' Proc. Natl. Acad. Sci. Vol.99(25) 2002 pp.15858–15862

Chen, Cai, Chen & Zhong, 'Fast acquisition of high-resolution NMR spectra in inhomogeneous fields via intermolecular double-quantum coherences' J Chem Phys. 2009 February 28; 130(8) 2009 pp.084504/1-08504/11

Blanton, 'BlochLib: a fast NMR C++ tool kit' J. Mag. Res. Vol.162 2003 pp.269–283

Many, many more... Seems if the maths can be sussed out, the need for greater homogeneity can be ruled out...

[Edited on 7-11-2010 by aliced25]

not_important - 7-11-2010 at 10:42

Do remember that often papers that you reference are dealing with specialised applications, or are bleeding edge. For example in Chen, Cai, Chen & Zhong, 'Fast acquisition of high-resolution NMR spectra in inhomogeneous fields via intermolecular double-quantum coherences'

Quote:

Our preliminary results on a 3 T whole-body MRI scanner indicate its feasibility.
...
Further work is needed to quantify the limit of the size and concentration of solute molecules to which the method can be applied. Further development of the IDEAL methods is also necessary for better water suppression for practical applications.

not to mention the increase in computing power, storage, and excitation signal control those methods require.

Keep in mind that old Project Triangle statement - Fast, Cheap, Good - pick any two. Cheap and good takes a lot of work to achieve.

arsphenamine - 7-11-2010 at 14:01

Quote: Originally posted by not_important

Keep in mind that old Project Triangle statement - Fast, Cheap, Good - pick any two. Cheap and good takes a lot of work to achieve.

The Prototype Triangle is: fast enough, cheap enough, good enough.

I like to finish things before the Heat Death of the Universe.

Another point: the paper on rapid spherical field calcs for shim coils applies to superconducting solenoid bores and doesn't translate well to generic rectilinear gaps.

The old patents from the 1950's and 60's are more applicable to the Halbach array previously discussed.

not_important - 7-11-2010 at 20:33

However they refuse to state what "good enough" is, I've ased several times. I assume that the intent is to develop a NMR that can be used to determine molecular structure, which is one of the more demanding tass and requires a fairly high lovel of goodness.

I wored too long at a company that went by the prototype model, almost everything they shipped was little past the prototype stage. I'd say that at least 3/4 of engineering manhours were spent on fixing problems on products in the field, which is usually more difficult and time consuming than doing so on the worbench, and repeatedly fixing the same problem as it cropped up at other sites. As another engineer said on his way out the door, "I want a job that's more than a ongoing 'quick hack'"

arsphenamine - 7-11-2010 at 21:43

Quote: Originally posted by not_important

However they refuse to state what "good enough" is, I've ased several times. I assume that the intent is to develop a NMR that can be used to determine molecular structure, which is one of the more demanding tass and requires a fairly high lovel of goodness.

Who are "they"?

I am not responsible for the short-sightedness of your previous employers
but sympathize with you since it sounds as if your managers would need
facial reconstruction were a clue stick a genuine physical entity.

A prototype is not a product, and ought to look like junk so no one will think of selling it.

"Good enough" means that the prototype proves the concept's validity and
informs you of production pitfalls that weren't obvious from ab initio reasoning.
It allows you to chart corrections before going to production.
It means you can fuck up a little without suffering a lot.

One more time:

It's easier to make a small 'X' and then a large 'X', than to make a large 'X'.

Regarding NMR magnets, I am deeply apprehensive about wrestling with
1"3 magnets that exert a 75lb hold, out of fear for my fingers.
Watson.fawkes assures me that 75lb isn't so bad but I remain incompletely convinced.

When aliced25 gets to the magnet array construction stage, will his posts
decrease in number and length as a result of severe hand injury?

aonomus - 7-11-2010 at 22:52

In regards to spinning samples, it increases field homonogeneity, increasing the FID and decreasing peak widths.

At my workplace we run a 400MHz shielded bruker, and the spin function on our probe is broken (and has been for years), yet we can acquire 1H/13C NMR spectra good enough for structure determination. It isn't an absolute requirement, we just shim it *very* well with X/Y coils (shimming with Z, Z^2, Z^3 is the major requirement when spinning is required, but without, X and Y require extra work if the tube spinner isn't working).

If you guys increase the field strength to increase the FID, and shim it very well, you could probably get away with it....

Other tricks could be used to decrease SNR like cooling coils, using better preamps, and concentrated samples and give the instrument better resolution.

aliced25 - 8-11-2010 at 07:54

Hmmm, I'm not saying that I intend to try and use a completely inhomogeneous field, but with shimming we still would probably be fucking lucky to get down past the ppm range. Spinning in such a small field is going to be hellishly difficult, but the use of a range of pulses and then a mathematical model to derive the structure from the constants therein sounds like it might be part of the answer. The NMR-Mouse is actually in production and as stated earlier, it does not use anything even vaguely recognisable as an homogeneous field, the field homogeneity in this model (particularly with shimming) and the strength of the magnetic field (about twice that of the NMR-Mouse) suggest to me that while the answer is not the best answer possible, it is one of the better solutions we'll come across (if it can be made to work). You'll note, I am not trying to utilize a field that is completely inhomogeneous, I'm just trying to avoid the spinning once we are down to the sub-ppm range. Various solutions that were never before plausible are becoming plausible due to the sheer size and speed of the processors in the humble PC (and people wonder why I roll my eyes when I see it being used for idiotic games).

I have no intention of building something that is going to be a partially working prototype, followed by an equally partially completed prototype, etc. I want to design something that will work. It doesn't need to be able to perform to the level of the 10-20T instruments, but it does need to be able to perform. If raw processor power can be used to avoid having to come up with a motor design that will function in the presence of a fairly solid magnetic field, so be it (even better if that means we don't have to open up the bore in order to accommodate magic-angle spinning. The fact that the spectra can be mathematically derived from several broadened spectra developed by a variety of means, suggests that there is a degree of feasibility to the concept, mathematics is too precise to admit a coincidence of that nature. Remember, people were burned at the stake for suggesting less heretical ideas that turned out to be true, while I don't yet suggest it is going to be the answer to the problem, the articles on the subject make interesting reading (and it isn't as outlandish as the fucking thermite in the WTC crap).

The previous suggestion that there be anything up to 8 or coils per side (4 pairs, or 4 sets of 4 around the outside of the bore) seems to have been ignored. Of course, I have not been able to model the field in 3D yet, but the obvious inhomogeneity around the outside of the bore (which is matched if one maps the field N-S) which is lowest near the poles and highest at the E-W centerline, suggests to me that if the mini-coils were set up so that each set of 4 were adjusted equally, then we might be able to alter the field enough to get to the ppm level. We really aren't far from the PPM level as it is.

aonomus - 8-11-2010 at 17:03

Re: motor that functions inside an intense magnetic field: air turbine.

not_important - 9-11-2010 at 11:47

(sigh)

Indeed, it has been said here several times before that the sample tube spinning is done with a small air turbine, commonly of plastic.

As for the NMR Mouse

Quote:

A hand-held NMR sensor, the MObile Universal Surface Explorer or the NMR-MOUSE® allows measurement of NMR relaxation and diffusion parameters in surface-near volume elements of arbitrarily large objects. Considerable effort has been spent on development of NMR methods which are applicable in inhomogeneous magnetic fields. Applications are in process and quality control of polymer and elastomer products as well as in Medicine

Not structural determination, but things lie determining the crystallinity of plastics.

Also note that several of the papers you have referenced have stated that while the model of the field may show a few tens of ppm inhomogeneity, real world physical magnet arrays are an order or two magnitude worse in their performance - 100 to 1000 ppm. Both permanent magnet and electromagnetic shimming are needed to reduce that to acceptable values. Plus you still need a way to determine field strength on the fly - deuterium or TMS locking.

Shimming, an example - how many coils do they seem to be using:

nmr_shim.gif - 15kB

aliced25 - 10-11-2010 at 14:22

Air turbines require compressed air - that requires a compressor & a tank, which kind of excludes portability. The bore of the magnets discussed by the researchers who've papers have been attached to here don't have the space for magic angle spinning & neither does the one I put up. Nor do other researchers, who've authored papers that I have also linked/added to this this thread. There has to be a way around it, I'll keep looking, but sending a broadish, inhomogeneous rf burst(s) at the sample, while alternating the current to the electromagnetic shims (so the sample "thinks" it is spinning), while reducing the level of inhomogeneity in the magnetic field to ppm figures may not be as silly as it sounds. It has the potential to avoid spinning the sample, by "spinning" the shim-current around the sample.

As there are, apparently, mathematical correlations between the responses, broad and narrow and the narrow, structurally determinant response can be arrived at by algorithms, broadening the pulse so that it excites the whole sample, then narrowing it until it only excites part also makes sense, kind of. There is a relationship between the width of the response and the width of the rf excitation, so obviously they must be extrapolating back from that in order to determine the "theoretical" response of the sample if it were in a homogeneous field and excited with a narrow-pulse. I visualize it as being like spraypainting a moving object (that moves within certain boundaries). The bigger the boundaries, the wider the spray, the smaller the boundaries, the smaller the spray. Extrapolating back from that, you could work out what the pattern would be if the boundaries were such that untoward movement was impossible and the spray was a marker pen instead. ie. it should be possible to determine based upon the multiple data points (fan shaped) what the narrowest range would look like (provided there is at least a slight correlation between the broadest, incomprehensibly wide response(s) and the narrower, semi-accurate responses. As that is what they appear to be saying in those papers (that such a relationship exists and that there is a numerical construct that can be used to solve for x,y & z @ infinity), it kind of makes sense. That said, there seems to be multiple formulae, which suggests the solution has yet to be found, as Einstein said, 'if you can't explain it, you don't understand it well enough'.

As to shims, I have mentioned a design several times, do you think current-output DAC's would be more useful than PWM controllers? That would give precise control over the current sent to the coils (or sets thereof). I'm just trying to work out a decent way of getting the job done electronically, such as would allow for someone using the control program to check the spectra when the shims are moved, then using the updated spectra, move one or more shims then check again.

As to determination of field strength, water (so too Tequila/alcohol) has a known spectra and has been used by various Author's cited in this thread to determine field strength (Or did I imagine that?). The known spectra would also allow for the correction of the shims, until the broad response(s) narrowed to an acceptable level.

not_important - 10-11-2010 at 19:15

Calculate the field strength the shim is going to be generating, then consider the effect of a 10%, 1%, 0,1% ripple on that field, and finally how much ripple is the PWM going to impose on the shim field. DACs + power amps will give smoother currents, but with much higher power dissipation. You need to determine what is allowable, ripple and heat generated, for your system in order to decide DAC vs PWM

Quote:

To go into the details of all this would be too lengthy and boring, but you sure see the logic behind it. Marcel Golay ended up with 13 sets of thin coils (4 of which zonal) which were glued into grooves machined in the Golay plates. The result was a shim system capable of producing field homogeneity of 1 part in 10^8 (without rotation; sample rotation improves the specification by another order of magnitude). That was quite sufficient for the 30 to 100 MHz iron magnets of the 50's to 70's and permitted, for example, the measurements of long range couplings smaller than 0.1 Hz. Due to the pretty good orthogonality it was also possible to shim the field manually by 20-turn current-source trimmers (though some empirical skill was needed and not everybody was up to the task).

If you wonder about that name : http://en.wikipedia.org/wiki/Marcel_Golay

note the 13 sets of coils, more or less pancake-like. High field superconductor systems take 30 to 40 shim sets as their fields are messy.

The pulse shimming is a nice trick, but both fairly new and more demanding in comprehension and equipment. Given that you've repeatedly confused fairly basic concepts, including that bit with MRI vs NMR spectroscopy, I'd not want to bet on pulse shimming being an easy solution for your project. Note that pulse shimmy has mostly been used with external sample systems, with a very large volume of sample being excited - in effect they are doing MRI to isolate a small section of the sample that lies in a fairly uniform portion of field, then other tricks to further compensate.

The size of compressor needed to drive the spinner is fairly small, this isn't shop tool stuff.

As for a reference for field strength, it needs to be in the same location as the sample during the time the same is being examined, although a guestimate could be done by doing a quick series of scans Ref,sample,Ref If the reference is present at the same time as the sample it can't contribute proton signals that could overlap with those of the sample, thus alcohols are likely out of the running.

aliced25 - 11-11-2010 at 07:16

Yeah, Golay is famous for a lot of things, sensors, algorithms (including the smoothing function) - a modified version of which is involved in that inhomogeneous NMR routine by the looks. The moving average would smooth the results, then the inhomogeneous field 1 is compared to field 2 (slightly more homogeneous) and the constant determined (unlikely to be linear). I've got to finish reading the articles, but it appears that by acquiring a statistically useful set of samples, they can work back from the known inhomogeneity and the FID from the same and compare it to the FID of more homogeneous fields (same sample). I wonder how far the peaks move? Do you reckon the peaks move as far as the bottoms, or would it be more like the FWHM diagrams?

Why bother with 13 coils? Set up 8/9 per side, then 2 more above N & S (no field until the pulse-spin is put through them) they will be separated (given the equality of the field at the equivalent positions in the NW,SW,SE & NE Quadrants) - so 1,9,11&19 would be equal (as would 0 & 10 {N-S} and also 5 & 15 {W-E}). 2,8,12 & 18 would be equal to each other as would 3,7,13 & 17 & 4,6,14 & 16. Quite simply the result would be better than Golay's as there are 7 additional coils (with the poles covered). The only reason for this is to allow for the pulse-spin. 0 & 10 would get a small pulse (increasing the field slightly), then the pulse would move on through 1, 11 then 2, 12, 3,13, 4,14, 5,15, 6,16, 7,17, 8,18, 9,19, 0,10 and so on... It is effectively spinning a small part of the magnet around the sample. In MRI they use that to regenerate the FID, or something along those lines...

Personally I think DAC's would be the way to go, but I keep running into articles talking about using PWM instead...

not_important - 11-11-2010 at 10:58

You're making big assumptions about the actual field shape; you're dealing with a 3D shape with small deviations not in the simulation as a result of actual physical artifacts in the magnets and supporting structure, plus from the electronics nearby such as tuning capacitors.

BTW Golay's setup was CW not FT and no pulses. I would like to see a rigorous analysis showing that your configuration is better than Golay's. He was a pretty smart dude, somehow I doubt he picked the arrangement of coils because he liked the number 13.

DACs vs PWM is in part going to depend on how much current/power is needed for the electronic shims. Do a good job with PM shimming and the electronic shims can be lower power.

aliced25 - 12-11-2010 at 06:10

Golay was a genius, I doubt He did anything much for fun, or from lack of thought. That said, he didn't have the extent of the control that is available now insofar as controlling the impulse to the magnet coils with anywhere near the accuracy we can achieve now. The Golay Cell for IR-Measurement is another one of his, it too suffers from the fact that processing speeds were so fucking slow a child with a new calculator could probably have kept pace with his mainframes. As the whole of MRI was developed upon sending pulses to gradient coils so that the magnetic field spun about the patient, and there are articles about spinning permanent magnetic fields about the sample (instead of spinning the sample), the pulse-spin routine would work, how well remains to be seen.

As far as assumptions about field shapes, yes, I am making assumptions. I'm basing these upon several articles where the author's have built "similar" shaped magnetic fields and found the problems lay not along the z-axis but on the N-S part of the X,Y. They've got the experimental data to show it, they also used solutions to measure the fields as Hall Effect Sensors & Reed switches, just cannot cut it. By comparing their results with the known spectra of water, the inhomogeneity of the field they produced was made clear. But note, they did get a spectrum. If there is an issue with the Z-axis, it can be reduced with Hemholtz pairs inside/outside the shims.

But there has to be some experimental data, I'm engaged in the search for ways to get light into and out of optical fiber at present, can we return to this argument in a little while? We are not going to find a way to make shit work without trying out a couple of ideas and seeing if they work out.

PS What do you mean FT in terms of Golay's set up? I don't recall suggesting that it was anything but a constant wave, but like all things, if improvements are ever to be made, they are made on the basis of the work of others (and applying new ideas & minute changes at great concepts). Golay was into a lot of things with spectroscopy, his least squares/moving average data smoothing is the basis of statistical data manipulation. That said, there are a lot of improvements, amongst those are the ones coming from sheer horsepower in processor output. Crunch enough numbers from the same sample, with various fields and you'll start to see patterns emerging, you know the "true" outcome, the difference between it & your results (statistically) must sooner or later show a range of responses that are either curved or that correlate in some way.

"10,000 monkeys at 10,000 typewriters" type of stuff, throw enough results at an anomaly, with meticulous records of field inhomogeneity/pulse inhomogeneity corresponding to those results and all of a sudden a pattern MUST emerge. Irrespective of how broad some of the results would appear, narrowing each of the two variables (field & pulse homogeneity) in a linear manner and we know that we get a nice narrow result-set. So the results from narrowing one at a time were finally measured (or something) and someone found a relationship... It was improbable that they wouldn't, the results broaden differently (presumptively) if only one of the variables is changed, compared to what they do if the other is changed. So there is a relationship, tentative, but a relationship nonetheless. So what happens when we broaden one and keep the other narrow? and vice versa? is there a possible complex relationship within the two matrices? Work out the relationship between keeping a narrow pulse and a broadly inhomogeneous field and all of a sudden shimming becomes immaterial.

aliced25 - 15-11-2010 at 03:04

C'mon, I'm sure everyone here has at least basic physics (well, at least who is keeping up in this thread) and advanced mathematics...

1. If a result-set is altered by the changing of a single variable, the result-set has a relationship with that variable (oversimplification perhaps, but also I believe a technically accurate description).

2. If the alteration of the result-set increases/decreases in line with the increase or decrease in the variable, then we have a linear/semi-linear relationship.

3. If we have a liner/semi-linear relationship between a result-set and a variable, then it should be feasible to reduce that relationship to a mathematical formula, to solve the problem spectra(A*

=Spectra(B

* or / functionx.

* The theoretical spectra solved for would be the one from a theoretically "perfectly" homogeneous sample.

Yes, the instrument is used in chemistry, but it belongs to physics, therefore the mathematical constructs and formulae that come from physics can and should be applied to it, where-ever a linear/semi-linear relationship between a variable and a result occur. This isn't rocket science, it is complex algebra, matrices and the effect of vectors on matrices. As results from various machines at each level of inhomogeneity are reproducible on other machines at the same levels, using samples, the spectra of which are able to be reproduced on the same machines without inhomogeneity, then inhomogeneity is not the problem. Ignorance is. Solving the riddle would remove the need for complex shimming, which in turn would make portable NMR a realistic choice in instrumentation.

As there is published (and peer reviewed) data on the problem and the numerous groups looking to solve the riddle (NB Many of these groups seem to struggle to present the problem in plain English, which suggests they aren't looking at the problem from the direction it may best be solved), if homogeneity is described as each axis being at 0, then a homogeneous field would be Field (X=0, Y=0 & Z=0

and the inhomogeneous field is represented as Field(X=1, Y=1, Z=0

, then the difference in results should be measurable. If the effect on the results is measurable, then it should be theoretically possible to ascertain a relationship between the magnitude of the variables that were changed and the magnitude of the change in the result. If one were to systematically pursue such a line of investigation, the several possible/potential relationships would be narrowed down by the number of results (and the fact that many theories wouldn't fit most results). The result would be an electronic/digital shimming algorithm, which could narrow the result set based upon known algorithms, with the magnitude of the inhomogeneity (and which axes) fed in.

That would make mechanical/electromechanical shimming a thing of the past, a hangover if you will. The study groups are on the problem, so it is likely to be solved (provided they don't get off-track). But I repeat, there is an obvious relationship between the homogeneity of the system and the accuracy of the result-set, an obvious relationship is nothing more than a problem awaiting a solution.

[Edited on 15-11-2010 by aliced25]

arsphenamine - 15-11-2010 at 13:46

Quote: Originally posted by aliced25

That would make mechanical/electromechanical shimming a thing of the past, a hangover if you will. The study groups are on the problem, so it is likely to be solved (provided they don't get off-track). But I repeat, there is an obvious relationship between the homogeneity of the system and the accuracy of the result-set, an obvious relationship is nothing more than a problem awaiting a solution.

In principle, yes, but only in principle.

Proving the possibility of a solution
is not a blueprint for a solution, and
a blueprint is not an implementation.

Similarly, while quantum chemical computations have elucidated a great deal of electron cloud structure, most NMR shift prediction software is a reasoned mixture of ab initio calculations tempered by empirical data.

While it is possible to strain shit out of the Ganges and declare it potable, the prospect of doing so can only be expensive and a last resort in the absence of more practical methods.

It is not necessary to replicate 60 years of field shimming technology.
Given the initially high homogeneity of a Halbach or Stelter array, it should be possible to do first order correction using a smaller array of magnets.

State-of-the-craft superconducting NMR devices may have close to 40 different shim coils. You could probably get by with much less since the Halbach/Stelter is so good to begin with. In this case, I would look at the field homogeneity patents up to but not including Golay's original patent.

The perfect is the enemy of the good
when you want adequately useful results in a timely fashion.

aliced25 - 15-11-2010 at 16:38

I'd agree with all of that arsphenamine, the difficulty is that the studies are underway and good results are starting to come from them. We can ignore those results, which are based upon the existence of more processing power on the average desktop/laptop (with which to implement the line-narrowing electronic shimming of imperfect results) or we can try and play with the big boys in terms of trying to fit enough shims into a tiny package to do enough to get us a decent spectral result. When you look at it, in order to get a decent result, we are going to have to implement Christ only knows how many shims, plus spinning.

I honestly think that this concept, the whole idea of portable NMR based upon permanent magnets and a laptop/pc would be better suited to the electronic shimming system. I mean, it would make the whole system a hell of a lot easier to implement, it would use nothing that wasn't already in use (well additional algorithms in the software) and would ensure good results from minimal equipment (notice I didn't say perfect - perfect requires ever greater superconductive magnets & coils and a mainframe).

I suspect watching & waiting might be the answer, as unpalatable as it may appear. Once there are known algorithmic approaches to effective spectral determination via the digital/electronic shimming routine, this project becomes feasible in a big way. At the moment it is at best borderline.

But that is not to say that the papers and the complexity of the field they are dealing with are overly complex (the mathematics looks it, but it always does). Take a good look at this paper & especially page 4/9. The design of the pulsed spin is based upon the NMR Mouse, which is unique in that the field is above the magnet. Here, where we are dealing with inhomogeneity of relevance only in the X,Y region of the area we are (1) going to irradiate and (2) going to get our response signal from, the pulse signal would have to travel around the circumference of the sample bore. They rotate left, then right and then essentially subtract anything that is one and not the other to be left with an essentially useful NMR Spectra.

Obviously for this to work, the pulsed spin induces the broadening to the left or right of the "actual" signal depending upon the direction of the spin. I'm not suggesting for a moment it is the "WHOLE" answer, but it would be something to have a bloody good look at. It even makes sense in a way, the central signal (the one we are after) would have to be the same in each regardless of the direction of the spin. The spin is a rotating magnetic field which allows for the excitation of different areas by bringing them from just below the field strength to just above it as it rotates. The RF Pulse is not rotating, it is static, so the broadening of the pulse would have to be determined by the rotating nature of the field, not the RF pulse. The broadened pulse comes from the FID of molecules in slightly lower/higher Magnetic Fields vibrating with the slightly higher/lower ends of the RF Pulse. I suspect that what the paper is discussing is the fact that when the rotating pulse is applied, it narrows the line, but also introduces broadening specific to the direction of the pulsed-rotation, comparing the clockwise/anti-clockwise spin pulse FID's would therefore allow for the manipulation of the resultant data. It would appear to be similar to the other variation of NMR where they remove the magnetic field completely (rapidly) inducing FID.

[Edited on 16-11-2010 by aliced25]

arsphenamine - 15-11-2010 at 22:41

In software, you can remove a lot of undesirable artifacts or
emulate some hardware functionality but
that doesn't make it a good idea.

If you do more work at the front of the tranduction chain,
exponentially less work is required downstream.

Polverone - 17-11-2010 at 16:59

By chance I happened upon a company called picoSpin that claims to have developed a compact 45 MHz proton NMR spectroscopy system with 80 ppb resolution based on permanent magnets. Units start shipping Q1 2011 at $20,000. This suggests that if you are very patient, future cheaper iterations, imitators, or second-hand units may eventually be available to the reasonably serious hobbyist. If the company or any of its team members have filed patents relating to their work, they may well be worth reading.

[Edited on 11-18-2010 by Polverone]

arsphenamine - 18-11-2010 at 06:20

Quote: Originally posted by Polverone

By chance I happened upon a company called picoSpin that claims to have developed a compact 45 MHz proton NMR spectroscopy system with 80 ppb resolution based on permanent magnets.

Spectacular disruptive technology.
Benchtop spectroscopy is back!

Other features are ~7lb. weight in a shoebox, solenoid magnet shape, network output using JCAMP-DX format
(~version 5.x, I assume), and a 20 μl sample volume in a 300 μ diameter capillary.

The last bit tells us that they have a ~2mm long sweet spot for the field
homogeneity, suggesting that the magnet is probably quite small, perhaps
10mm height.

Elsewhere, they mention using an internally doped water sample for field strength
calibration -- something that has recurred from the very beginning starting with
Varian patents from the 50's.

I wonder if it runs embedded Linux.

not_important - 18-11-2010 at 13:57

"temperature controlled permanent magnet" also argues for a smallish magnet, that is one way to keep field drift down. It does appear to be good for qual and quant both.

At a selling price of US$ 20K, that suggests a production cost of 1500 to 4000 dollars.

[Edited on 18-11-2010 by not_important]

aliced25 - 21-11-2010 at 20:22

The permanent magnet is not the problem, why on earth it would need to be temperature controlled is beyond my comprehension. It is getting the spectra out of the inhomogeneous (well only mildly inhomogeneous in comparison to some set ups) sample core that lies at the heart of the problem. As inhomogeneity in the RF pulse is something we'll have to live with (and moreover, something that the researchers of a shitload of papers of note have artificially contrived in order to hit the inhomogeneous field with a pulse that had matched inhomogeneity - the spectral width being based upon the MHz correspondent to the Upper & Lower levels of the magnetic field (the Larmour Frequency of the upper & lower modeled field), gives a decent spectral return - all of the protons are excited by the pulse (which hits all of the Larmour Frequencies) and by the looks the average of the spectral returns narrows the spectral linewidth appreciably. It is nice to actually read the information on the subject, obviously there is quite a lot of study going into it.

arsphenamine - 22-11-2010 at 11:52

Quote: Originally posted by aliced25

The permanent magnet is not the problem, why on earth it would need to be temperature controlled is beyond my comprehension.

NdFeB magnet fields have a 0.1%/degree C temperature coefficient.

That's 1000 ppm/degree field variation, or 43.5 kHz Larmor Frequency drift in a 1 Tesla field.

When you want to resolve something at 0.4 ppm, it's an important consideration.

not_important - 22-11-2010 at 14:13

Continuing on that riff, several of the journal papers you yourself referenced discuss the temperature sensitivity and the need for temperature stabilisation. In some configurations and applications the temperature drift can create inhomogeneity in the magnetic field.

12AX7 - 22-11-2010 at 15:36

Good way to handle that would be a Peltier controlled box of heavy aluminum. Two of them, one inside the other, with lots of insulation between (aside from the Peltier). With a suitable thermistor and controller, that should keep it within a few milikelvin of nominal (give or take a few Hz shift). Shim coils will have to go outside, since they'll add heating from odd sides. The effect of the work coil and test sample may require an aluminum or copper housing inside the chamber to shield the temperature difference, which will affect the field noticably (dia/paramagnetism).

Tim

not_important - 22-11-2010 at 15:57

Or run at slightly elevated temperatures, say 35 C, with some but not really heavy insulation, and just use a heater outside the magnet region - put the magnet in a "warm box". Shim coils need to be fairly close to the sample area, else they need to be strong which will introduce another set of problems. If the basic magnet is fairly good, or shimmed to be so with permanent magnets, then the shims use little power.

aliced25 - 22-11-2010 at 23:21

I was thinking that the shims would use minimal power, they aren't being used to cover massive amounts of inhomogeneity, I still don't think that temperature would be the concern it is being rated as, but be that as it might.

The low-field or zero-field articles are of interest in terms of the spin-pulse design, the area of homogeneity in such small magnets is going to be minute, say four or five shims high, wrapped on a core and some seriously intricate spin designs could be tested. I'd be very interested to compare such tests, particularly with something that has a known spectra (in a homogeneous field) and keep trying out options and then comparing them to the known spectra. If the pulses are running while the RF Pulse is run, then stopped/allowed to keep going while the FID is occurring, what happens? There are a lot of seriously gray areas here, left by researchers with access to multi-million dollar equipment, who had no need to find out. Necessity is the mother of invention, so I'd be interested to see what comes of the interest in the designs.

As to cooling, why go with a Peltier, generate heat at the outset and use a modified "Icy Ball" design to keep the magnets at -33C/240K for the duration of the spin-echo experiments. If temperature is going to be that big a concern, then it would shit all over peltier cooling. Actually, would be just the thing for a lot of electronic cooling, lasers & the like, think about it, build the heat sink (the ball that needs heating) into your heatsink (on your PC) and let that continually regenerate the NH3 (or use the Einstein variant, which I'm suspecting uses lower pressures... It would alter thinking on desktop cooling wouldn't it? Of course at that temperature we start running into semiconductors, which would be an interesting idea...

watson.fawkes - 23-11-2010 at 05:16

Quote: Originally posted by aliced25

I still don't think that temperature would be the concern it is being rated as, but be that as it might.

That's because you haven't done the calculation yourself. Go do an estimate about what happens if one side of a permanent magnet array is 0.1 °C cooler than the other because of, say, a draft in the room.

arsphenamine - 23-11-2010 at 12:08

Quote: Originally posted by aliced25

I still don't think that temperature would be the concern it is being rated as.

It's importance was covered in this thread around 17/10/2010.

aliced25 - 24-11-2010 at 15:55

I don't expect that a draft in a room would affect a magnetic array in a Copper/Aluminium Enclosure, the sheer bulk of the housing and the insulating effect of the same would normally be expected to ensure that the various parts thereof don't have major differences in temperature. I don't see how you would alter the entire volume of the magnets temperature with a peltier cooler on the base anyway, the cooling would have to be through the center, through the core, through the horizontal axis, etc. If it were such a difficulty, I'd imagine there'd be more mention of it in the journals where they are using permanent magnets to make magnetic fields. The only sure-fire way to ensure proper cooling distribution would be to run a liquid coolant over the surface (depending upon the substrate too, if it is too thick, surface cooling will only exacerbate the issue) and circulate that through a cooling pump. I've yet to see anyone bother.

not_important - 24-11-2010 at 16:34

The journals are research devices, not manufacturing; they may be in a well controlled environment. As such much detail is left out. However several of the papers you have listed do explicitly state that various enhancements would be needed for practical devices, including temperature control. Others are for devices that have much less demanding quality requirements than analytical NMR, MRI can get away with nearly an order of magnitude greater inhomogeneity than structural analysis, pore-filling and elastomer curing measurements are yet another order of magnitude less demanding; several of the papers you've listed are refering to such applications.

For example, from 2 papers that I'm fairly sure you were the source of:

Quote:

A simple, small and low cost permanent magnet design to produce homogeneous magnetic ﬁelds
B. Manz *, M. Benecke, F. Volke

doi:10.1016/j.jmr.2008.02.011

As spectral resolution is enhanced through the measures
described above, the system will ﬁnally be susceptible to
very ﬁne temperature gradients and ﬂuctuations. An eﬃ-
cient thermal insulation as well as a frequency lock will
be mandatory to commercial high resolution portable
NMR sensors. Finally, remaining ﬁeld inhomogeneities
will have to be compensated by a set of shim coils, as it
is common practise for other high resolution NMR mag-
nets. However, for many NMR applications, even proton
linewidths of 50 ppm are suﬃcient for a quick characterisa-
tion of samples via relaxation and/or diﬀusion contrast,
like gels, polymers or high viscosity liquids.

Small Magnets for Portable NMR Spectrometers
Ernesto Danieli, Juan Perlo, Bernhard Bl.mich, and Federico Casanova

DOI : 10.1002/anie.201000221

Further improvements in the strength, size, homogeneity,
and temperature stability of the magnet are a matter of
technological refinement. A higher field can be achieved by
increasing the outer diameter of the magnet array. For
magnets built from SmCo, 1.5 T (corresponding to a proton
frequency of 60 MHz) can be obtained with an outer diameter
of 7.5 cm. This material has better thermal stability than
NdFeB, but remaining field drifts arising from fluctuations of
the magnet temperature must still be eliminated, for example,
by combining standard temperature control schemes with a
field-frequency lock.

There are a number of ways of obtaining the required uniformity and stability, you picked one of the more complicated ones.

You might also consider why such functionality would be designed in, if it was not at least useful if not required. Every zloty of parts boosts the sale price by roughly 10 zloty, so there's price pressure to keep unneeded hardware to a minimum.

[Edited on 25-11-2010 by not_important]

aliced25 - 11-1-2011 at 22:06

Anyone on here with experience getting the bloody 3D models to work, Comsol, etc. would be greatly appreciated.

I've decided to go with the AD9850, it runs at 125MHz (so it is useful to what, 62.5MHz theoretically?), with a 1T field, with a median inhomogeneity of +/-0.001T, it should be possible to work out some shim pulses to take it up to what we need. The TI ECG Front-end-type ADC, the 4-channel ADS1294 should be ample (with a small modulating frequency). Using the DDS Chip, the ability to generate some funky-as-fuck pulses should come easy enough (according to the person I asked who is doing the research with the bloody thing).

Quite simply, the shim-pulse, spin-echo technique(s) are probably the only way this is ever going to be suited to "home" or "portable" use, building in the controllers for the X,Y,Z shims would suck (although I suppose a PID controller could be utilized) but that would require feedback and apart from using a known spectra, I suppose distilled water would do, feedback would be difficult to say the least.

With regard to PCB's there is a move toward Cu/Al2O3, especially for high-temperature circuits, there are a lot less toxic requirements, particularly if the board is machined rather than etched. Given the existence of the same and the toxic output of the alternative (plus the fucking greens being in charge of this Cunt-tree), I'm half-expecting to be seeing the same in more widespread use.

aliced25 - 21-1-2011 at 15:15

Strangely enough, there is a 3D version of the Halbach/Stelter array which uses exceptionally simple magnets, which if a 0.2" gap is left in the top and bottom pole piece, gives a fucking incredible homogeneous field, especially through a 0.2x0.2" (~5.4x5.4mm) square in the center of the unit and allows for the sample to be put in and for the electrical connections to come in from the other end.

I've decided to purchase a small lathe & mill - with a view to setting up a CNC Conversion on both (X,Y,Z via stepper motors) and a design for a small-ish X,Y router table, with the Z-axis being restricted to on or off. The table will be used to route copper coated alumina or plastic to make a shitload of boards at once. With the small router I can also design the paste-overlay for the solder paste and with almost all of this using SMD's I'll try some of the ideas suggested by sparkfun to heat the paste/units properly.

I've also decided to stop fucking around with what others have designed, if what they have designed works, the only changes will be to reduce the parts count and simplify the setup.

aliced25 - 3-9-2011 at 05:12

Quote: Originally posted by un0me2

...

A.Prof Takeda is currently working with a 1T magnet (with poor homogeneity reportedly, requiring X,Y,Z coils, or so I've been told) in order to try and work out a portable, home built system.

...
[Edited on 26-9-2010 by un0me2]

Ok, I may (or may not) have spoken to someone in a Japanese lab who is not only interested in the latest magnet design (simple as fuck, N40 & N52 Magnets with a 1" central void (which the "person" suggests would suffice for coils, shims, etc.) which is modeled at 1.184 (± 0.001) T or a N52 only design which is modeled at 1.231 (±0.003) T (inhomogeneity of around >0.5%) through the central 0.2" sample holder. Currently working on the design of a modified Cylinder/Cube based upon the design, if that works (based upon the strength of standard Halbach Cylinders v arrays), multiple Tesla field strengths are possible.

The person I may or may not be dealing with is a published author, with a working spectrometer based upon open-source software and hardware configurations. The magnets would cost less than $50 and the electronics would be less than $200 per unit (less with volume). If it works it would be envisaged that pre-assembled kits/units could be sold online.

PS Anyone here any good with modeling 3D magnetic fields using Matlab/Mathematica/Comsol/etc.? It is starting to become vital to the cause.

[Edited on 3-9-2011 by aliced25]

aliced25 - 9-9-2011 at 18:17

Quote: Originally posted by arsphenamine

Regarding NMR magnets, I am deeply apprehensive about wrestling with
1"3 magnets that exert a 75lb hold, out of fear for my fingers.
Watson.fawkes assures me that 75lb isn't so bad but I remain incompletely convinced.

When aliced25 gets to the magnet array construction stage, will his posts
decrease in number and length as a result of severe hand injury?

Sitting here laughing about the number of blackened fingernails and blood-blisters (aside from cuts from shattered magnets) I've acquired. That isn't the reason for the lack of recent activity, the necessary designing & machining, plus the need to complete an insane number of papers (yeah, back at Uni) while getting fitness back to a minimum standard all take time

That plus kids now

arsphenamine - 21-9-2011 at 10:51

Quote: Originally posted by aliced25

Sitting here laughing about the number of blackened fingernails and blood-blisters (aside from cuts from shattered magnets) I've acquired.

Urgh. Sorry, mate. Wish it was otherwise.

Quote:

That plus kids now

Spare time and uninterrupted sleep will become distant memories.

aliced25 - 23-9-2011 at 19:21

here is a bloody interesting paper - it really goes into the nitty gritty of how the researcher 'did' build a working NMR Spectrometer and then goes off into conjecture into how he'd improve it.

Given the accessibility of the TMS320 series and the McBSP channels integrated into most of them, it comes up with some interesting ideas. Reducing the number of DDS chips from 2 to 1 also makes a great deal of sense. Why this needs to work with a smart phone when any real research work would presumably have at least a laptop/etc. is beyond me. If the complex processing/mathematics were handled on the PC/Laptop/Tablet/whatever then the complexity (parts count, layers, IC's, etc) on the board (plus temperature problems) become less of an issue.

The current magnet is 1.132T with inhomogeneity in the >1,000 ppm range which would appear to be significantly better than that used in that article (and in others), with the complex & phase-shifted signals used to cancel out inhomogeneity on a much larger scale.

aliced25 - 7-10-2011 at 22:05

Ok, playing with Mathematica and FEMM, got two designs on the boil, we'll be playing in the ~60 & ~90MHz range using small Permanent magnets.

Spin Echo pulses were first reported by Hahn (1950) (in the ref request thread), discussed by Solomon (1955) and also by Carr & Purcell (1954).

It would appear that the industry (and research) has gone off on a tangent from that point, the calculations involved in working back from T1-T2 to get a narrow linewidth spectrum are not trivial (particularly in 1950's terms), whereas Golay's invention of shim coils made work on the concept unnecessary. It was easier and probably a great deal cheaper to design and build bigger and bigger magnets than to even try and contemplate quantum computing (which of course would allow the complex algorithms to be run in next-to-real-time).

Mildronate - 8-10-2011 at 01:41

Anbody had schematic on modern components, not from vacum tubes

aliced25 - 10-11-2011 at 04:09

Ok, looking very seriously at this now, as stated in several articles, we aren't going to be transmitting while we are receiving, so if the output can be digitized effectively (24-Bit Sigma-Delta ADC), a bandpass filter could* "potentially" be used to block the unwanted pulse-signal (thereby significantly reducing the complexity of the signal chain - no carrier wave, no modulation/demodulation, etc.) which is significantly higher than the ppm FID signals (which would be, given that 1MHz = 1,000,000 Hz, so the 0-15ppm shifts below 1KHz). The idea being that there is no data collected until the signal coming out of the receiver coil drops below the bandpass (could we use 1KHz?)

* Anyone got any reason why it wouldn't work?

aliced25 - 2-12-2011 at 20:13

According to FEMM if one were to utilize a certain design (using 25mm cubes as well as other components), the output in a central bore section is 125MHz. The homogeneity is 2.93794-2.93798T through that bore, which is around 13-14ppm prior to shimming.

totalsilence - 13-3-2012 at 08:23

Quote: Originally posted by aliced25

Ok, looking very seriously at this now, as stated in several articles, we aren't going to be transmitting while we are receiving, so if the output can be digitized effectively (24-Bit Sigma-Delta ADC), a bandpass filter could* "potentially" be used to block the unwanted pulse-signal (thereby significantly reducing the complexity of the signal chain - no carrier wave, no modulation/demodulation, etc.) which is significantly higher than the ppm FID signals (which would be, given that 1MHz = 1,000,000 Hz, so the 0-15ppm shifts below 1KHz). The idea being that there is no data collected until the signal coming out of the receiver coil drops below the bandpass (could we use 1KHz?)

* Anyone got any reason why it wouldn't work?

you are going to need to build a tuned LC circuit. The resonant frequency of this circuit will need to be match the lamor frequency as dictated by your sample / static field strength. The LC circuit is resonant whilst the pulse(s) is applied and then damped afterwards to stop the oscillation in the circuit. In the read phase the tank is resonant in order to "amplify" very small signals and tune out noise.

I am building a set-up for my doctorate. It is not a simple project!

DavidJR - 16-12-2018 at 04:23

I'm planning on doing some DIY NMR experiments. Still awaiting a bunch of deliveries I'll need, but I have assembled a Halbach array:

wg48 - 16-12-2018 at 08:40

Quote: Originally posted by DavidJR

I'm planning on doing some DIY NMR experiments. Still awaiting a bunch of deliveries I'll need, but I have assembled a Halbach array:

That was probably tricky to assemble. It looks like you used a jig to hold them then cast the plastic cylinder round them ?

How will you adjust the field strength or do you sweep the generator and receiver frequency using modern electronics?

DraconicAcid - 16-12-2018 at 09:10

I will be in awe of anyone who can build a functional NMR at home.

That being said, commercial NMRs have come down in price, enough that a benchtop NMR can be had for about $30K.

Texium - 16-12-2018 at 09:16

Quote: Originally posted by DraconicAcid

I will be in awe of anyone who can build a functional NMR at home.

I second that... and if they could also make their detailed design publicly available here so that others could replicate, that would be even more amazing. I'll certainly be following this thread!

Ubya - 16-12-2018 at 10:26

Quote: Originally posted by DavidJR

I'm planning on doing some DIY NMR experiments. Still awaiting a bunch of deliveries I'll need, but I have assembled a Halbach array:

shouldn't the field be as uniform as possible?

DavidJR - 16-12-2018 at 13:54

Quote: Originally posted by wg48

It’s not cast, I 3D printed the blue thing and then inserted the magnets. It was pretty tricky to get them all in though...

No varying the field for scanning, i’ll vary the frequency as needed instead. I’d rather use a Fourier transform approach and not do old school scanning though.

Quote: Originally posted by Ubya

shouldn't the field be as uniform as possible?

Yes, it should be, at least over the sample area.

I’m not certain this magnet design is good enough but it’s a start. I may well need to modify the design. I may also need to add some small shim coils but we’ll see how it goes.

The magnets I bought are 15x15x50mm, N48.

[Edited on 16-12-2018 by DavidJR]

[Edited on 16-12-2018 by DavidJR]

wg48 - 16-12-2018 at 16:55

Quote: Originally posted by DavidJR

No varying the field for scanning, i’ll vary the frequency as needed instead. I’d rather use a Fourier transform approach and not do old school scanning though.

Yes that's the way to do it these days. Look how a modern cell phone works or a the front end of a satellite receiver or a stick TV receiver for a PC. Though i suspect for the best signal to noise it is still narrow band tuned receiver but there is probably not a significant difference unless your trying to pick up signals from one of the Voyager spacecraft.

I read a few pages of the thread. The analysis of the shimming coils is similar to the way multi axis optical pointing systems trim out pointing errors due to mechanical inaccuracies between the axes, alignments of the sensors and geometric distortion of the sensors. Its complicated maths (big equations) but in the final analysis it just a polynomial correction for each axis. I guess the trick in shimming is working out where to put the coils to generate the polynomial correction. The first colour TVs had complicated trimming to align the three electron beams and correct image distortions.

An interesting project do keep us informed of your progress.

mattharbowy - 8-2-2020 at 09:06

FYI: just stepping in to revive this 2012/2018 thread on DIY NMR. I'm just going through what's been published before on the subject, and I'm planning out some of my first initial experiments and gearing up my toolset. In the next couple of months I'll try to post more on what I find out. At the end of this it looks like I'm going to have to learn a lot of AC analog electronics I never bothered to learn just to get to first base with this project, but it looks interesting and worthwhile, and if anything, the resurgence of benchtop NMRs doing 2d pulse sequences gives me hope that I can get to a spectrum of ethanol by end of year.

Ubya - 9-2-2020 at 04:41

Quote: Originally posted by mattharbowy

FYI: just stepping in to revive this 2012/2018 thread on DIY NMR. I'm just going through what's been published before on the subject, and I'm planning out some of my first initial experiments and gearing up my toolset. In the next couple of months I'll try to post more on what I find out. At the end of this it looks like I'm going to have to learn a lot of AC analog electronics I never bothered to learn just to get to first base with this project, but it looks interesting and worthwhile, and if anything, the resurgence of benchtop NMRs doing 2d pulse sequences gives me hope that I can get to a spectrum of ethanol by end of year.

i'm looking forward to it, it is a really interesting project, it is one of the many things i'd like to do. as amateurs we don't have many modern ways of checking if the compound we made is what we meant

wg48temp9 - 9-2-2020 at 05:47

I was curious as to how difficult DIY NMR detection or spectroscopy would be. Apparently detecting NMR in cobalt nuclei can be done with a grid dip meter but that is a special case. A quick google reveals several DIY projects. Here is one one with a description the instrument https://www.youtube.com/watch?v=b2n1-nvo7d4. Below is a pic of the instrument,

Below is some general info on NMR

mmkhttps://www2.chemistry.msu.edu/faculty/reusch/VirtTxtJml/Spectrpy/nmr/nmr1.htm

Attachment: NMRBasics_2016.pdf (4.6MB)
This file has been downloaded 1351 times

Ubya - 9-2-2020 at 12:41

yup i saw that when i was researching the subject. radar magnets are not something that is usually sold in italy lol, i'll have to make my own electromagnet from pure iron and a helmholtz coil. the RF part of the project is what bugs me, and i can't find an easy enough "tutorial" or at least explanation, it is all or very theoric, or too much technical, so i can't follow (i don't have an electrical engeneering background)

wg48temp9 - 9-2-2020 at 14:25

Quote: Originally posted by Ubya

yup i saw that when i was researching the subject. radar magnets are not something that is usually sold in italy lol, i'll have to make my own electromagnet from pure iron and a helmholtz coil. the RF part of the project is what bugs me, and i can't find an easy enough "tutorial" or at least explanation, it is all or very theoric, or too much technical, so i can't follow (i don't have an electrical engeneering background)

I was hoping to get a feel for the magnitude of the task in terms of the strength and homogeneity of the required magnetic field, the sample size and the signal strength and noise. But so far I have no data. From the construction of that DIY system it does not look like it has any thing exotic. i am guess that the RF part could be achieved with a digital radio receiver and perhaps a low noise pre-amp.

[Edited on 2/9/2020 by wg48temp9]

Ubya - 10-2-2020 at 02:39

i have no clue, but the receiver isn't the only part, you also need a transmitter that can produce a high power, very short, broadband signal with a precise and tunable timing

[Edited on 10-2-2020 by Ubya]

Ubya - 10-2-2020 at 04:02

i think i found a solution, probably the best as of right now
http://kuchem.kyoto-u.ac.jp/bun/indiv/takezo/opencorenmr2/in...

it is open source, he/them described the board and the software
https://opencorenmr.github.io/opencorenmr-docs/

in theory one could send the eagle files to a PCB manifacturer, ans the populate the board with the right components by hand

wg48temp9 - 10-2-2020 at 05:57

Quote: Originally posted by Ubya

i have no clue, but the receiver isn't the only part, you also need a transmitter that can produce a high power, very short, broadband signal with a precise and tunable timing

[Edited on 10-2-2020 by Ubya]

I don't know how precise or broadband the drive signal has to be but you can get an idea from the diy unit I linked to. That unit used 555 timer chips for the envelope of the drive signal which are simple RC timers probably about 5% timing accuracy. So the envelope timing is not critical.

You can buy a 50MSa/s sample rate, 14bits vertical resolution Arbitrary Waveform Generator for about £100 that will generate that signal.
https://www.ebay.co.uk/itm/FY6800-60M-DDS-Signal-Generator-2...
There are potentially problems with fine control of the rf signal.
Though a few 555 and an LC oscillator could used with at least one stage of amplification to drive and isolate power amp.

Ebay sells 70W linear power amps 3.5 to 30MHz for £24 including postage

I have made some progress on the task size for the magnet and receiver. The DIY unit I linked to states the magnet is 4,500 Gauss and that in can be varied by a few hundred Guass with the 10A coils. That implies it could be replaced with an electromagnet with about 20 times the ampere turns. Ok that's a big coil at least x20 of size of the small adjustment coils and much bigger to keep the power down. However that set up does not use all the surface area of pole faces. With more focused pole face the total flux could be reduces and hence the coil size could be reduced. I think the magnet construct is big task but doable probably using transformer cores as the main material with sold soft iron poles faces or the neodymium version shown in this thread or if your lucky and find an other big magnatron magnets.

Again from the linked system the receiver uses a amp with voltage gain of 10,000 apparently driving a scope displaying signal at about 100mV implying the input signal is about 10uV. Typical short wave radios are in the 2.-3uV range so suspect a digital radio stick will be similar for example :

cost about £20 direct digitisation up to 28MHz but only 8bit It has a built in FFT. I don't know how easy it will be interface to it. The hardware and some of the PC software to use it is open source.

This is just the initial look at the feasibility and practicality.

Dr.Bob - 24-2-2022 at 18:30

For any of you electronics people, I have a pile of old electronics books, guides, and such, some may be obsolete, but not sure. If you have an interest in NMR, I have a few books on that, more on the electronic side, as well as Techtronix 2465 Ocilliscope Servcice manual if anyone wants it. If anyone here likes NMR, I can give them to you, as well as some tubes, caps, and other stuff. I have many tubes (most of which are dirty or etched), as well as hundreds of caps that were used once. But if you want tubes and can clean them, I can provide a bunch. Good luck, if anyone here gets close to building one, like DavidJR, and is still active, I also know a guy who fixes them, and he has a garage full of parts and pieces. He is the only person I know who took apart a 300 mHz superconducting magnet in his garage (it was not energized.) I mean with a sawzall. Great photos. But know I understand the mechanics of them much better now.

Rainwater - 24-1-2023 at 17:06

Getting closer. The youtuber, Applied Science released this yesterday
https://youtu.be/JO_EHceV9sk

Twospoons - 24-1-2023 at 20:07

I saw that too. Interesting, relatively simple, but a PITA to retune for each thing you want to look for.

Rainwater - 25-1-2023 at 01:37

True.

Rf switches have come a long way.
SKY12212-478LF for example
$10usd
100w power
.6db tx/rx loss
46db rx/tx isolation

Would be better than tuning for each frequency.
But present its own problems.
Limited stopping power.
Can be in series with each other if isolation is not enough.

Signal generation will no longer be a problem with many ic operating over 100mhz clocks
You can generate a 80mhz signal with an esp32 and control it with nanosecond timing,

Recreating oscilloscopes level resolution and memory depth still remain a challenge but the ic chips are here now if not expensive
I/q modulation might be well suited and has come along way as well

I love living in the future

Rainwater - 28-1-2023 at 16:28

Read this thread 3 times and there seems to be a big controversy on
Well. Every aspect of a design and what is required.

I would like to try to nail down exactly what we need to make the hardware do

Will the following parameters provide a sufficient starting point for.

Receiving and recording the signals generated
Generating the transmission signal. (Not powering it)

0.2 seconds of recording.
200msps data capture at 12bit (4096 increments)
Good for detailed analysis less than 50mhz
Poor spectrum analysis at 100mhz

An 860mhz 50% duty cycle PWM generator or
360Mhz iq modulator, basically generates any waveform you can dream up.
Higher frequency, less control

All with already written and vested software from the open-source community
DAC, Modulators, ADC readers. It's all been written and licensed under open source
It just has to be cut and pasted together

a setup with a custom board a
Adc - TI ADS4128 200msps (Link)
And a nano 9k FPGA (Link)
can do all this. And it's cheap like $50usd

[Edited on 29-1-2023 by Rainwater]

Rainwater - 31-5-2024 at 19:15

15 months ago, i started down this road. By far the most difficult part is the digital RF receiver. Been having so much trouble figuring this out im taking my bsee and getting it upgraded to a MSEE in rtl design and digital signal processing. It will defendanty be cheaper to buy one, but there is no fun in that!

Pages: 1 2 3