| Pages:
1
2
3
4
..
12 |
Twospoons
International Hazard
Posts: 1324
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
Instead of using such nasty shapes, you can build a halbach array using a ring of diametrically magnetized cylindrical magnets (available off the
shelf). I ran a quick sim on FEMM, and got a much flatter field than in you sim above. Shimming is achieved simply by rotating the magnets. Size
in sim is kind of random. magnets are N52.
More elements usually = flatter field.
Attachment: halbach.EMF (5kB)
This file has been downloaded 1317 times
[Edited on 13-9-2010 by Twospoons]
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
Be nice if that were so wouldn't it? But as your picture demonstrates, the magnetic field in the circular magnet designs is affected by there lack of
density (both magnetic and otherwise). Your design pulls a nice field, there is no escaping that, but with heavier duty magnets, but the maximum field
in the center is >0.5T (closer to 0.4 in fact). My "nasty shape" is pulling an homogeneous field with close to 0.7T (and I've designed others
pulling over 1T, but without the homogeneity). I've tried nearly every design I can think of with circular magnets and without resorting to sizes
& shapes that are outside what is easily available, I cannot break 0.5T (although with 8 Neo52 Discs, 1/2" dia, I can reach it with a "fairly"
homogeneous field through the center with a 4mm useable region averaging from ~0.47-0.5T).
With an even nastier shape, which I will share directly, I can get a homogeneous field around 0.95-1T. I want to install some software so I can draw
it in CAD & import it to FEMM first.
Here is one design - 10mm squares (1" high) & 10mm diameter discs (1" high), with the 4 circular magnets being used to tune the magnetic field.
Across the 1cm center the field fluctuates from 0.82125T to 0.82110T which is pretty bloody homogeneous (0.82T right across) with further tuning
available from turning the 4 circular magnets (all magnets are N40 - N52 would improve the field further).
[Edited on 14-9-2010 by un0me2]
[Edited on 14-9-2010 by un0me2]
quam temere in nosmet legem sancimus iniquam
|
|
|
Twospoons
International Hazard
Posts: 1324
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
I'm thinking of the tradeoff between the 'perfect' answer, and one that is easily obtainable (its what I do all day, as an engineer). I'm just
wondering where you will obtain odd shaped magnets with exactly the right magnetisation direction. If thats no problem for you then go for it!
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
watson.fawkes
International Hazard
Posts: 2793
Registered: 16-8-2008
Member Is Offline
Mood: No Mood
|
|
Quote: Originally posted by un0me2  | | I've tried nearly every design I can think of with circular magnets and without resorting to sizes & shapes that are outside what is easily
available, I cannot break 0.5T (although with 8 Neo52 Discs, 1/2" dia, I can reach it with a "fairly" homogeneous field through the center with a 4mm
useable region averaging from ~0.47-0.5T). |
Broadly speaking, you've got two kinds of noise to worry about.
I'm over-simplifying at this first pass to make a point about optimization. Feel free to expand as needed.
You've got underlying thermal and Doppler noise (and others). Increasing the strength of magnetic polarization increases the splitting of degenerate
energy states. The higher the energy difference, the more easily it's distinguishable from unavoidable noise in the sample itself. On an NMR graph,
this causes peaks to become shorter with respect to background.
The other kind of noise is noise in the magnetic polarization itself. This is both from the magnet of the device as well as from the Earth's shifting
field (also and others). This noise is observed as a frequency shifts in a resonance line. Mathematically, this kind of noise causes a convolution in
frequency space. On an NMR graph, this causes peaks to widen.
You need to deal adequately with both kinds of noise in order to get meaningful results. Sharp peaks that aren't visible much above the noise floor
don't give you much. Neither do tall peaks that are so washed out that you can't see features adequately. Magnet optimization requires work in both
areas. It also requires a better understanding of the S/N ratio of the sensing electronics and a better model of error budget. The good engineering
design goal is a reasonable impedance match between marginal benefits along any axis of improvement. In other words, each element should be equally
difficult to improve upon.
|
|
|
densest
Hazard to Others
Posts: 359
Registered: 1-10-2005
Location: in the lehr
Member Is Offline
Mood: slowly warming to strain point
|
|
A note to the various good people doing field computations: the papers and commercial disclosures I've read use both magnets and ferromagnetic
unmagnetized pieces to achieve their best results. Some of them use only rectangular magnets, leaving all the exotic shapes to be machined from soft
iron or other high saturation material. Adding those pole pieces tailored to fit odd-shaped crevices might make it easier to achieve high strength
homogeneous fields.
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
@ Watson.Fawkes
There is no doubt about the fact that greater resolution, does at present, require 2-5T fields (or higher), that is one of the issues that is being
faced. That being accepted, there is also a major effort at present to make miniature and even micro (on chip) NMR suitable for portable sensors, such
as would be of benefit to schools, Universities, the Armed Forces and Homeland Security amongst many others. The remarkable progress in the
miniaturization of the electronic components makes this feasible if only small enough, useable magnetic fields (from 0.5-1T can be made using
permanent magnets). Consequently the massive number of papers on the subject.
@12AX7
Thanks for the input, I was hoping to utilize plastics simply because they are easier to manipulate/fabricate. That said aluminium is not all that
difficult to deal with (unless it has to be welded). Would you have any suggestions of easily available metals, or oxides, that could be utilized
instead?
As to the shim circuits, yes, they are going to be necessary, although I would hope to be able to build small servos (off 1 Integrated Chip) which
could automatically manipulate the tuning magnets so as to get the field as homogeneous field as possible, thereby reducing the need for shim coils
dramatically.
@Twospoons
I realise what you are saying, let me set your mind at rest in that regard, every magnet I have used, or attempted to demonstrate the use of, is
available online for very little cost. That is one of the major requirements of this project, commercial availability of each and every component
part.
@ Those & the rest, I am also trying to design the simplest possible electronic circuit to (a) convert Digital-to-Analog; (b) synthesize the
signals using a programmable synthesizer (IC) which can sweep from several KHz to several MHz (the IC containing all of the components necessary to do
(a-b are on one IC) so with minimal noise); (c) delivery of the synthesized signal to the radiating side of the double saddle coil; (d) collection of
the signal at the receiving side of the double saddle-coil; conversion of that signal to (i) isolate the peaks from the transmission frequency while
(ii) assigning the same to that frequency (all on one IC with the feedback loops, etc. all built in); (e) convert that information to from analog to
digital (part of the same IC) and a USB 2.0 connection to drive the whole device.
PS 12AX7, the homogeneity now would be what down to the 0.01% mark?
quam temere in nosmet legem sancimus iniquam
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by un0me2 | ...
As to the shim circuits, yes, they are going to be necessary, although I would hope to be able to build small servos (off 1 Integrated Chip) which
could automatically manipulate the tuning magnets so as to get the field as homogeneous field as possible, thereby reducing the need for shim coils
dramatically.
[rquote]
If those are PWM servo dirvers or similar, you'll need _really_ good filtering on their outputs.
| Quote: Originally posted by un0me2 | ...
@ Those & the rest, I am also trying to design the simplest possible electronic circuit to (a) convert Digital-to-Analog; (b) synthesize the
signals using a programmable synthesizer (IC) which can sweep from several KHz to several MHz (the IC containing all of the components necessary to do
(a-b are on one IC) so with minimal noise); (c) delivery of the synthesized signal to the radiating side of the double saddle coil; (d) collection of
the signal at the receiving side of the double saddle-coil; conversion of that signal to (i) isolate the peaks from the transmission frequency while
(ii) assigning the same to that frequency (all on one IC with the feedback loops, etc. all built in); (e) convert that information to from analog to
digital (part of the same IC) and a USB 2.0 connection to drive the whole device.
... |
That's a fairly wide sweep range for good performance - precision, accuracy, and stability.
Also, if I'm understanding what you're saying, you intend to use a pair of coils facing each other with the sample between them. If so, that will
likely result in the driving signal swamping the sample signal; there's a reason that most NMR systems place the pick-up coils at right angles to the
driving coils.
Not only that, but the power levels needed for the drive side likely will induce excess noise in the receiver; it's better to pt power circuits on
separate PCBs with good isolation - local regulators for the receiver one and all that. You can do it on one, but good board layout is needed.
Note that FT NMR uses a short pulse to excite the sample, and it is possible to share the same coil for drive and pickup although you generally need
protection for the receiver. The frequency range needed is fairly small, a ordinary DDS chip will do the job - the frequency base is independent of
the sweep and generally covers a much wider range.
http://www.analog.com/en/rfif-components/direct-digital-synt...
|
|
|
|
Twospoons
International Hazard
Posts: 1324
Registered: 26-7-2004
Location: Middle Earth
Member Is Offline
Mood: A trace of hope...
|
|
The electronics you describe sounds remarkably like an RF spectrum analyser with an attached tracking generator - a fairly standard piece of radio
test kit (though not exactly cheap). I have a TEK492 analyser that operates from 50kHz to 22GHz, no tracking generator though. Cost me $2k on Ebay.
Maybe this will help you build your own : http://www.scottyspectrumanalyzer.com/ , or at least give you some ideas.
Helicopter: "helico" -> spiral, "pter" -> with wings
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
Yes Not_Important, that is precisely what I am saying, I was looking at using another integrated chip on the receiver side of the double-saddle coil,
to filter out anything on the transmission frequency. By careful timing of the sweep I should be able to do so, taking any remainder as resonance at a
particular frequency (ie. the necessity of the timing of the sweep, allows one to know what frequency the resonance belongs to). I'm busily looking
for the partner IC to the transmission IC, I'm betting there is one.
Anyway, that has to wait for an answer in the reference request thread, here is the latest go - 1/2" Square & 1/2" Diameter x 1" high magnets with
a 1/2" central area to give a 0.812 (+/-0.01)T field right across the 1/2" central area...
 
With further playing around with the tuning, I can get a dead flat line @0.84T, and I suspect I can get up to 0.85T with a +/-0.01 cross-section, now
that is a permanent magnet
[Edited on 15-9-2010 by un0me2]
quam temere in nosmet legem sancimus iniquam
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
Yes, pole pieces.
Yes
YES
YES!!!
Anything else is pure retardery!
Looking for fairly intense fields requires a lot of magnet, so there isn't really much room for pole pieces in this particular application. So maybe
pole pieces aren't as important here as they are for motive applications (motors, transformers..).
Still, you could fill in the crecent-shaped gaps between the round and square magnets with hunks of steel. You could also use larger magnets focused
into smaller cross sections through the use of pole pieces, so you get a smaller volume but a higher field strength. Just lining the chamber with
steel on the right sides may help.
Tim
|
|
|
arsphenamine
Hazard to Others
Posts: 236
Registered: 12-8-2010
Location: I smell horses, Maryland, USA
Member Is Offline
Mood: No Mood
|
|
Historical notes:
The first production NMR for chemists was the swept scan Varian A-60.
My university was so cheap, the profs were still using it when other schools had moved to 100MHz FT-NMR technology.
For NMR, 60MHz <-->1.4T field. In 1960's technology terms, that was near the limiting remanence of two 9"thick by~24" diameter Alnico II discs
in close proximity. Smaller magnets could produce the necessary field strength but nothing beats a big fucking magnet for field homogeneity. The
assembly weighed roughly 1500lbs, and was planted above the intersection of two walls.
Isotropic Alnico5-7 might have cut the necessary mass by half.
Using NdBFe magnets, you could probably achieve the same effect using between 1/7 and 1/3 that magnet mass, depending upon the desired gap.
1000 or so A-60's were made over a 15 year period. Most were retired for salvage but some were refurbished with modern process control hardware for
proton FT-NMR. NIST currently has an A-60 on display as a museum piece since it was such a game changer.
@un0me2
.81T puts the proton precessional frequency around 34.5 MHz, which is in the middle of a US Gov. exclusive radio band allocation. Remember that when
you go blasting 100 watts into the excitation coil.
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
That shouldn't make "too" much difference if the whole thing is carefully shielded should it? I mean, shielding will be absolutely essential given the
proximity of the magnetic field to the electronics, so running a shield around the whole shouldn't be too hard - on top of which, the range of the rf
excitation coil will be severely limited.
quam temere in nosmet legem sancimus iniquam
|
|
|
arsphenamine
Hazard to Others
Posts: 236
Registered: 12-8-2010
Location: I smell horses, Maryland, USA
Member Is Offline
Mood: No Mood
|
|
The magnetic field can induce eddy currents in adjacent shielding.
It may be more useful to shield the room instead of the magnet assembly.
Luckily, conductive screening (or simply heavy aluminum foil) is inexpensive.
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
Shielding 35MHz is no big deal, the skin depth (and therefore penetration into shielding) is quite small. It's a frequency which tends to stay in
wires if you're careful, so it's not freaky RF, easy enough to manage. A few well placed shields, bypass caps and ferrite beads is fine.
Tim
|
|
|
arsphenamine
Hazard to Others
Posts: 236
Registered: 12-8-2010
Location: I smell horses, Maryland, USA
Member Is Offline
Mood: No Mood
|
|
My point was that an incautiously designed EM shield distorts
the magnet field homogeneity because of eddy currents.
A coil connected to a power amplifier output is necessarily an EM transmitter
of some stripe. In the RF world, such a coil is often called an inductively loaded
antenna.
I would love to hear about homebrewers succeeding to produce a useful NMR device
but, much more than that, I would hate to hear of that same group facing legal action
because they'd encroached on a government perquisite.
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
| Quote: Originally posted by un0me2 | | ...I was looking at using another integrated chip on the receiver side of the double-saddle coil, to filter out anything on the transmission
frequency. By careful timing of the sweep I should be able to do so, taking any remainder as resonance at a particular frequency (ie. the necessity of
the timing of the sweep, allows one to know what frequency the resonance belongs to).... |
Note that receivers have a certain range of signals they will handle, both in absolute terms and in the amplitude difference between two simultaneous
signals. Nonlinearities in the amplifier can result in intermodulation between the signals, causing sidebands and ghosts. Keeping unwanted signals as
low as possible is better than just trying to filter them out after amplifying.
Why are you choosing to do a sweep NMR as opposed to a Ft one, given how cheap fast digital processing is these days?
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
Could you explain the difference? I'm quite happy to do so, I just require some information - presumably you mean transmitting a noise that
encompasses all the relevant frequencies, collecting the entire signal, filtering out the transmitted noise (like a digital notch filter?), then
perform a Fourier-Transform on the signal remnant left after removing the transmitted noise?
I'm just guessing, but I would prefer not to have to fabricate the "sweep" function, if a recorded sound (encompassing the frequencies) could be
simply pulsed into the sample, that would simplify (and reduce the size of) the instrument dramatically. I'm trying to work out how to get the digital
signal from the ADC to the laptop/pc where it can be dealt with using software (http://pdfserv.maxim-ic.com/en/an/AN4530.pdf is an example). I'd love it if anyone had dealings with this idea, the "firmware" doesn't appeal...
However, I would be extremely interested in the FT-NMR, how exactly is the excitation waveform determined? Presumably it is a function of the magnetic
field strength, but as a signal - surely it could be stored on the PC/Laptop and sent DIRECTLY to the rf coil from the USB port, as a nsec pulsed
signal, the response would be collected, converted to digital and then read by the software. That would greatly simplify the production of a "working
solution" for this project.
[Edited on 18-9-2010 by un0me2]
quam temere in nosmet legem sancimus iniquam
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
Traditionally a simple pulse is applied (rectangular, or maybe rectangular windowed RF), during which time the reciever is blocked out (to prevent
overload), or watching from a precisely balanced coil (orthogonal to the excitation, so it doesn't pick up any of the transmitter's signal). The FID
(free induction decay) induces RF in the reciever coil, which can be sampled by the reciever. I suggest a traditional RF amplifier and converter,
pulling ~39-41MHz (center 40) down to an IF of 5MHz, then converting it again to 500kHz, and one or two more times, to low audio frequencies (~1kHz).
This can then be fed into a computer sound card and processed with standard audio techniques.
The purpose of downconversion is to increase the frequency spread in relative terms, making it easier to observe the differences. 1ppm at 40MHz is
4Hz, which is a whopping 0.4% at 1kHz.
Multiple samples can even be recorded and mixed together, averaging the signal and increasing SNR.
Tim
|
|
|
aonomus
Hazard to Others
Posts: 361
Registered: 18-10-2009
Location: Toronto, Canada
Member Is Offline
Mood: Refluxing
|
|
I'm not sure if any of you guys have the equipment to do so, but if you have any semi-mainstream DSOs, you can always get labview to work with your
DSO as the data capture device, and use a labview controlled signal generator to provide the IF for the mixer and RF amplifier for the excitation
coils.
Considering that DSOs are cheap now, no reason to try to reinvent the wheel if you have labview...
Just my $0.02 adjusted for inflation.
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
Ok, let's move on from the mid 30MHz range, let's imagine that the magnet design is capable of pulling a field of 1.215 (+/- 0.015) Tesla, what would
be the excitation range on the waveform? Also, if one were pulling 1.4-1.5T, what would be the dynamic range of interest?
Put simply, excitation at a single wavelength is far easier than trying to scan it, if I can be sure that it is going to work, then I'm more than
happy to build a programmable pulse generator and simply ignore any signal received while transmitting (saves fucking around). There is a paired-set of programmable transceiver-receiver @ sparkfun. That would make life a great deal fucking simpler - the integrated chips are a lot easier than trying to build the other.
[Edited on 18-9-2010 by un0me2]
quam temere in nosmet legem sancimus iniquam
|
|
|
not_important
International Hazard
Posts: 3873
Registered: 21-7-2006
Member Is Offline
Mood: No Mood
|
|
FT-NMR often has more complex receivers than old fashion sweep single frequency systems, with the intent of increasing performance of the unit.
Quadrature detection driving a pair of ADCs is typical, 12 to 18 bits of digitalisation is typical for units not aimed at monitoring fast processes.
The DDS chips I referenced generate quad output
The excitation signal depends on a number of factors, pulse shapes other than rectangular are often used.
A Web search for FT NMR will yield a lot of information, and you could read the attached doc.
Attachment: chem843-2.pdf (284kB)
This file has been downloaded 1383 times
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
Yes, a google search for FT NMR yields a SHITLOAD of results, trouble is the signal to noise ratio in terms of determining the resonance signal is
fairly low, which may explain why I asked. I am currently sitting on a design that is modeled as producing a 1.255 (+/- 0.002) Tesla (the field is
measured around the 5mm diameter circle that the sample would sit in. The graph resembles a sine wave, with the dip, 1.251 Tesla, from the Peaks,
1.259 Tesla, occurring twice, both times in the center of the design (halfway through the sample, East-West), whereas the peaks are at North &
South.
But from what I am reading, including the attachment you posted, a heavy duty irradiation of the sample at the Larmor Frequency, will do? That
followed by detection of the decay signal?
[Edited on 19-9-2010 by un0me2]
quam temere in nosmet legem sancimus iniquam
|
|
|
12AX7
Post Harlot
Posts: 4803
Registered: 8-3-2005
Location: oscillating
Member Is Offline
Mood: informative
|
|
| Quote: Originally posted by aonomus | I'm not sure if any of you guys have the equipment to do so, but if you have any semi-mainstream DSOs, you can always get labview to work with your
DSO as the data capture device, and use a labview controlled signal generator to provide the IF for the mixer and RF amplifier for the excitation
coils.
Considering that DSOs are cheap now, no reason to try to reinvent the wheel if you have labview...
Just my $0.02 adjusted for inflation. |
You need megs of RAM and a full width FFT window to observe ppm shifts. This is pretty well a DAQ thing (data aquisition) requiring a PC, unless you
have one of the incredibly massive mainframes that costs $50k and offers spectrum analysis this refined.
Tim
|
|
|
arsphenamine
Hazard to Others
Posts: 236
Registered: 12-8-2010
Location: I smell horses, Maryland, USA
Member Is Offline
Mood: No Mood
|
|
$9k USD
I wondered what acquiring a "cheep" NMR system would involve, before
and after purchase. In summary, it is expensive to keep one active.
Read on for details.
LabX.com is auctioning a functional Anasazi-modified Varian EM-360,
a permanent magnet 60MHz FT-NMR.
http://www.labx.com/v2/adsearch/detail3.cfm?adnumb=422363
On the full size images, note that it is connected to a water faucet; the
30 shim coils run hot. For nominal flow, the predecessor A-60 model
heated 60F water to 100F as I recall.
$9000 - auction price
$1500 - UPS shipping for 1/2 ton from Iowa to DC + fork lift at both ends
The original vendor blurb is at:
http://www.aiinmr.com/products/aii-instruments/
<s>Given its water+electricity consumption and the temperature controlled
environment requirements, the $20k/yr figure I've heard for the older
Varian A-60 seems rational, if disagreeable, but still cheaper than a
superconducting NMR.</s>
UPDATE:
Anasazi Instruments support states that the eFT-360 continuous power
requirements are 110VAC at 6 amps and no water required, more in line
with those of a middling refrigerator without an icemaker.
[Edited on 19-9-2010 by arsphenamine]
|
|
|
un0me2
aliced25 sock puppet
Posts: 205
Registered: 3-2-2010
Member Is Offline
Mood: No Mood
|
|
Yeah, that would be the point of what is under discussion wouldn't it? The fact that unless a way around the problems mentioned is found, there is
little hope for the use of NMR at home. There are units that are smaller, like the NMR-Mouse, which generates a non-homogeneous field above the magnets, but which are a lot cheaper to run (not to buy unfortunately). Most of which was discussed in quite some detail by the late Dr Dykstra in his PhD Thesis.
It may have escaped your attention, but there is a major change to the magnet configuration in this discussion, the generation of a virtually
homogeneous field above 1T (+/-1%) perpendicular to (East-West) to the poles (North-South), inside the magnet. As the NMR-Mouse doesn't use a
homogeneous field at all, I'm wondering how much shimming will be needed and whether that can be dealt with mechanically (additional magnets to shim
90' to the main magnetic field).
12AX7, the use of a quadrupole amplifier/detector, to comparator/DC converter, then a low-noise, high-speed DAC (to convert the teensy changes in the
DC output into columns of binary output), is the far end. It shouldn't be that hard, switching the receptor-coil off while it is ringing (here & here is a discussion) may be. That said, it has been dealt with before and should be susceptible to adaptation given the size of the dedicated integrated circuits are minute compared to those used in the past.
For mine, the AD9850 would seem like a good choice for the 1T range (this article) suggests that it wouldn't take a whole lot more than that to allow one side of the system to be set in stone with one chip. In fact,
replacing the chip on this board really ought to give the range we want (up to 50MHz), with programmable interface.
[Edited on 20-9-2010 by un0me2]
quam temere in nosmet legem sancimus iniquam
|
|
|
| Pages:
1
2
3
4
..
12 |
![[*]](images/xpblue/default_icon.gif){kind=icon}
