As with many things in NMR gradient shimming is something that is
inherently fairly simple, but on first attempts seems impossibly
complex. But then once you are in the swing of things you wonder why on
earth you had any problems previously!
I'm no expert but this is what I would tell anyone else that asks - but
if there are better suggestions.....
> The revised GUI is pretty much as bad as the old one, with a few new
> things put in to make it worse. Almost all the text is now in a font
> that's so small that it's illegible. The graph that's supposed to give
> feedback about shim quality has both axes unlabelled, and the x-axis
> scale changes arbitrarily so that you can't measure or compare anything.
> The scale cannot be changed by the user.
The Y axis corresponds to the spectral width of the measured profiles,
the numbers are the data points that are used in the iteration values.
The X axis is a measure of success, generally if your final wiggly line
is within +- 0.1 units and your shim is still bad then you need to look
somewhere else (off axis shims, sample problems etc.). It is a shame
you cannot zoom in, but there you go. Always remember that the last
curve displayed is the profile _before_ the shims are adjusted - if you
want to see the curve of the result then you should click the 'show
current field profile' button.
When you have things setup and are using it in automation then you don't
need to bother with all these windows anyway!
> The actual procedure for improving the shims is a complete mystery. All I
> know is you have to improve the straightness of the displayed graph.
> Allegedly you do this by choosing different groups of shims to adjust,
> and altering their range of adjustment by typing a number in a box.
> Unfortunately, I have not discovered what combination of shim groups and
> numbers you have to use to generate improved resolution. The feedback
> graph certainly changes, sometimes straighter, sometimes more curved.
> Usually it's pretty much invisible because each iteration writes another
> thin line in a pale pastel colour onto a pale cream background.
Ideally you would need 1 shot adjusting all the shims, and this is the
case if the starting shim is not bad (typical of recalling a previous
shim set). If the starting shim is very bad then you might need more
iterations - the first attempt should correct gross errors giving better
data for the next attempt. This assumes that your shim maps are
reasonable as if you do not have good maps then things will be more
difficult!
> There are recognised good sequences for manual or simplex shimming when
> you are maximising lock level or FID area, but what do you do with
> gradient shimming? Nobody has been able to tell me. The results I've got
> so far have been dreadful. I might just as well have spent the time
> twiddling knobs by hand. After several iteration cycles, using gradually
> more Z-shims, over various ranges, the result was a water line that had
> visible large errors from all the shims I was adjusting. I quickly got a
> better result from shimming by hand.
>
> Clearly, I'm doing something wrong here. So, how _do_ you get good
> resolution from gradient shimming? Does anyone know?
Things to check:
1) Do your profiles look ok?
recall the gradshim parameter set you are using, then zg and fmc
The profile should look something like a rectanglar block of tar
that
is starting to melt. If it is very noisy then you should check the
p0/pl1 and rg, or that any selective pulse is in the correct place! If
the wings are not completely flat then probably rg is too high (it is v.
important that you do not clip the echo). If it is not centered or the
profile is very small then adjust o1 (especially if you are using a
different sample, such as chloroform) and sw (the profile ideally will
fill say 90% of the sw). If you do use something like chloroform you
should increase d1. If you have a large ripple then increase de. For
compatibility with future software check that aunm is 'au_zgonly', and
aunmp is proc_no (to do nothing: GETCURDATA QUIT). If you made any
changes then write out the parameter set.
2) Do your shimmaps look ok?
Either remeasure, or select 'recalculate shim maps from data'.
You should see a series of curves for y=x, y=x2, y=x3 etc, again
with
little noise, pretty well centered and of broadly similar intensities.
Note that at the ends the intensity will drop back to zero, which
corresponds to the baseline region of the profiles, and just before this
the intensities may wander a bit as the s/n in the profile is no longer
enough to get a good value. It is worth noting the data point number at
which this happens (usually around datapoint 22-28, depending upon the
profile width and s/n) If things look completely bad then go back to
step 1!
Remember to write the best shims you have to GSHIM before
starting the
shim map.
On old gradient systems check dc offset before starting!
3) Is your iteration control file sensible?
The exact steps and numbers depend upon your setup and what you
need to
do.
Each iteration step is defined by the number of datapoints in
the
profile that you want the system to use in the calculations (+-) 0-32,
and by the shim group that you want it to change. You should never use
more datapoints than the profile provides! (the number you noted in step
2). If you want to correct high order shims (z5) then you need to use
as much of the profile as you can (say a point or two less than your
absolute maximum) as these shims make more effect at the edges.
Adjusting z6 is most likely not too productive as this is a matrix shim.
Assuming my profiles fade out at point 26: I would look to use
for a
single step something like z-z5 with 24 points, for two steps something
like z-z3 with 18 points, then z-z5 with 24 points. I would think that
if you do not get a good result in at most 3 steps then you never will -
either check what you are doing, or take a new shimmap and try again.
Remember
these numbers are not absolutes, they depend upon what _your_ profiles
are like - if in doubt I would aim to use a smaller number of points.
> Before you ask, I have Z-gradients only, so I was adjusting Z-shims only,
> using 1H; the probe was a 5mm 1H{13C}, and was being shimmed from
> scratch; the X and Y shims had already been adjusted; the shim map
> included all the available Z-shims (Z1 to Z6), and was remeasured at any
> time there seemd to be an improvement to the shims; the sample was 50:50
> H2O:D2O, 5cm deep, and I wasn't spinning it.
The times I've had to shim something from scratch (not one of may
favourite jobs!) I used a chloroform lineshape sample, roughly got z,
z2, x and y by hand, measured a shimmap, got z-z5 adjusted (which also
corrected my attempt at z and z2!), twiddled a few more off axis by hand
and was most of the way to spec in far less than an hour including
coffee. But maybe I was lucky.
As a last comment, for those with the BSMS 'real time control board'
then 3D shimming is
possible even with only a z-gradient system.
Best wishes,
Andrew
-- Coventry telephone numbers will change in April 2000. The new number (which can be used now) is shown below. The old number will continue to work until the change over. /**************************************************************** Dr Andrew Gibbs Phone (+44) 02476 855 200 Applications Scientist Fax (+44) 02476 465 317 Bruker UK Limited email: andrew.gibbs@bruker.co.uk Coventry CV4 9GH England *****************************************************************/