I wish to post a summary to the question that I posted below. I wish to
thank Bruce Fulton, Rich Shoemaker, Nagarajan Murali, and several others for
discussions. I would also like to thank Rolf Kyrburz and Mikhail Reibarkh
from Varian for providing definitive solutions for direct acquisition and
processing.
The basic problem with cold (or cryo) probe with C13-cold preamp is probe
ringing. A long delay, typically 350us in Varian's setup, is usually
required before acquiring data.
My post can really be divided into 3 questions:
1) What is the best way to acquire a C13 spectrum with a C13-cold cold (or
cryo) probe to avoid this delay?
2) With INOVA, how should we set up C13 acquisition to avoid probe ringing
as well as to have VNMRJ back calculate the spectrum automatically?
3) If data were already collected with a delayed acquisition, how can we
use Varian linear prediction to back-calculate the spectrum?
1) Several people pointed out that methods relying on linear prediction are
not perfect (there is often a small dip in the baseline around 20-40ppm) and
the best way to avoid the long delay is to use spin-echo. There are several
sequences in Biopack to accomplish this. The problem is that if one wants
to use Ernst Angle the 180 pulse inverts a large +Z component to -Z and
effectively increases the delay needed for the relaxation recovery. This
makes Ernst Angle optimization unusable. A better way is to invert the C13
spin in the beginning of the sequence as demonstrated in the "z-restored
spin echo" paper by Xia et al, MRC 2008; 46: 432-435. I attempted to code
the sequence, but have not made it to work. The paper described in detail
the use of chirp pulses for Bruker instruments but not very clear (to me)
for wurst pulses on Varian. It is likely I did not use them properly. I
probably should have use the simple composite pulse instead.
2) It was suggested to me that setting dsp='i' and qcomp='y' along with the
"coldprobe" macro will prepare the instrument for direct detection. It did
not work for me: "qcomp" does not exist in my system and it makes more
sense to use dsp='r' if there is a mezzanine board on the ADC board. The
following procedure is recommended by Varian:
the macro "coldprobe" just sets rof2 (post-pulse) delay to the values
appropriate for the probe.
If parameter alfa is not taken care of, it results in very large first-order
phase distortions.
The simplest way to deal with it is to phase it (lp can be easily in 300-400
range, this seems to be in error, it should be in the range around -4700),
then use macro "calfa" that calculates parameter alfa to have a correct
value - which gets rid of first-order phase distortion.
When alfa is set, you need to re-acquire the data; spectrum acquired with
proper alfa will have minimal lp.
I have not yet tested this. Also if this is done once, the parameters are
probably not automatically set up for other users who want to do the same.
The procedure is for INOVA. I think for VNMRS DirectDrive systems there is
a macro "setlp" to set "ddrtc", etc.
3) By multiplying 350us by sw, I know that I missed about 13 points
(13*360=4680 and My spectrum could be phased with rp=-54 and lp=-4700 with
extensive wriggles.) Thus it is necessary to right shift the fid by 13
points and backward-predict the beginning section of the fid. Rolf Kyburz
has demonstrated to me linear prediction with the following parameters works
quite well:
lsfid=-13 parlp lpfilt=32 lpnupts=np/4 strtlp=16 lpext=16 strtext=16
proc='lp' lpopt='b' wft
He made additional comments that I think are very interesting and I am
including his comments here:
it looks as though we need to predict a couple more than
the empty points that we just "shifted in" through "lsfid";
also, "lpfilt" should ideally be twice the number of peaks
that we use for the prediction - but as we are looking at the
start of the FID which is essentially defining the broad
baseline features, I think the minimum "lpfilt" should be
at least twice the number of complex points that we want to
predict: lpfilt=16 "sort of" works, but lpfilt=32 definitely
gives better results.
One thing I did not mention in my reply - partly because I did not
investigate that part thoroughly enough: I have indeed seen LP fail
completely on your dataset where at a first glance it should have worked,
e.g., with
strtlp=lpext=strtext=13
and a small lpnupts in lieu of np/4: the noise in the C13 FID is substantial
and may cause the fitting program (of which I haven't studied the details)
to "freak out", producing havoc; increasing lpnupts to use half the real /
acquired data for the prediction appears to avoid such failures (in this
example). So: while the original recipes (and the current "easy setup macros
for LP" may work flawlessly for typical proton / nD bio-NMR datasets, in the
case of a C13 FID one may need to play with the parameters a bit in order to
obtain usable results. Also, if the "source data" include distorted FID
points (such as the first few datapoints which I'm discarding with
strtlp=lpext=strtext=16), this adds to the noise effects and will not make
the job for LP any easier.
A last point: I again haven't checked this in detail, but as far as I know
it is still true that with LP you should set fpmult=0.5 in order to avoid DC
offset in your spectrum.
Hsin
--
Hsin Wang, Ph.D.
Department of Chemistry Phone: 212-650-5831
Marshak Hall - 1217 Fax: 212-650-8719
CUNY - City College Email: hsin.wang_at_sci.ccny.cuny.edu
160 Convent Ave.
New York, NY 10031
_____
From: hsin [mailto:hsin.wang_at_sci.ccny.cuny.edu]
Sent: Monday, August 04, 2008 11:23 AM
To: 'ammrl_at_ammrl.org'
Subject: AMMRL: C13 detect on a C13-cold cold probe on INOVA
We have spent some time getting C13 spectra on a cold probe in which both H1
and C13 are cold on an INOVA600. Knowing that probe ringing is substantial,
I have used the macro "coldprobe" to increase rof2 to 350us. The macro did
not seem to set up any processing procedures to deal with the data, however.
The phase seemed all over the place. I have previously done the same on a
DirectDrive system and there was no problem. I have posted this question to
Varian, but I have not gotten an answer. Does anyone know how to process
this kind of data?
Hsin
--
Hsin Wang, Ph.D.
Department of Chemistry Phone: 212-650-5831
Marshak Hall - 1217 Fax: 212-650-8719
CUNY - City College Email: hsin.wang_at_sci.ccny.cuny.edu
160 Convent Ave.
New York, NY 10031
Received on Tue Aug 19 2008 - 09:19:40 MST