Hello,
We had some additional responses since I submitted a summary for my question. I have
added them and a complete summary is included below.
Below are summaries of the responses to my question yesterday about spectral phasing
problems. I should have been more specific about when this problem occurs on our
instrument. We observe peaks which cannot be phased while running proton NMR on the
ethyl benzene test sample with standard sweep widths (12 PPM) and no receiver
overloading. This is completely correctable on the instrument using the JEOL process
of "machine phasing", but can sometimes give problems when the data is off-loaded to
be processed off-line.
>From the responses to my question, it appears that this phenomena does occur with some
but not all instruments. JEOL and Bruker have machine specific ways to address/correct
the issue. Since some instrument types reported the phasing problem while others of
the same type reported no phasing problems, the issue may be related to instrument
installation and calibration issues.
My guess is that it comes from a second (or higher) order phase response of the filters.
All systems (even very old systems) seem to have the second order phase correction need.
However, in the older systems, the size of the correction was small and we all said
"the effect near the ends of the spectrum are in the cutoff of the filter" and to be
expected. Reprocessing older data clearly shows the effect with peaks near the ends of
the spectral window AND it can be corrected with a second order phase correction parameter.
Our JEOL has a typical second order phase correction value near -150 degrees while the
older Nicolet/GE, Varian and Bruker systems typically have values near 30 degrees.
The observed result is that with older systems there was non-phaseable data near the ends
of the spectrum which we considered "normal" and ignored. With some newer systems the
second order phase correction can be much larger. This results in a higher percentage
of the spectral window having the phasing problem.
The phasing issue can be addressed in multiple ways:
* Increase spectral window to allow the peaks to be in the area correctable by linear phasing.
* Leave the spectral window the same, but increase the cutoff frequency of the filters.
* Have manufacturer supplied phasing corrections specific for each instrument design.
* Use phasing routines which can apply zero, first and second order phase corrections
woody_at_acornnmr.com
=========================================================
Original Question:
Our new JEOL NMR instrument produces data which when phase corrected with zero
order and first order phase corrections still has some peaks which are out of phase.
We have experimented and found that the addition of a second order phase correction
parameter allows all peaks to be phased. We have another example of data from a
Varian instrument which has the same phasing problems and solution.
Is this a general issue with newer NMR instruments or does it just show up in a
few selected cases?
If a general issue, how does one solve the "not all peaks can be phased at the
same time" issue?
===========================================================
Summary of responses
1111111111111111111111111111111111111111111111111111111111111111111111111
Is the problem only with peaks within about 10-20% of the filter
cut-off? Most (Butterworth) filters tend to show a non-linear phase
terms near the 3 db point.
2222222222222222222222222222222222222222222222222222222222222222222222222
I had a similar problem with our new MercuryVX console. The entire spectrum
would phase nicely, with the exception of TMS. I did not attempt a second order
correction, but it seemed logical to assume that such a correction would work.
Rather, I played around with the sw & fb settings (fb, of course, being
integral to sw unless subsequently modified) & found that the standard filter
settings roll off too close to the end of the spectrum, causing phase (&
presumably integration) errors. (I was able to reproduce this on our Unity 400,
as well.) I ended up modifying the _sw macro so that any spectrum in which
integration was really likely to matter (implemented if d1>10) had the fb
setting doubled, so the filters didn't mess things up. This entailed a loss in
s/n, but it was worth it. I'm going back now to figure out whether using dsp is
a better option. I believe it is, but I haven't had time to pursue it yet.
Otherwise, I can always phase my spectra nicely, barring instrumental problems
or rookie errors like wrap-around.
33333333333333333333333333333333333333333333333333333333333333333333333333333
No problem with all the Bruker spectrometers.
44444444444444444444444444444444444444444444444444444444444444444444444444444
We have a JEOL Eclipse + 600. I was not aware that you could use a second
order phase correction on this spectrometer in addition to P0 and P1. Can
you tell how you are doing this?
555555555555555555555555555555555555555555555555555555555555555555555555555555
I have recently seen this behavior on newer Varian Unity spectrometers. In
all cases, the problem is solved by extending the acquisition time. Using
line broadening > 1/at does not help with phasing. Only increasing "at" so
no signals are truncated works. The new spectrometers/magnet/probes can
produce H and C linewidths well under 0.5 Hz. I think having this type
resolution on a routine basis is what's new.
I do not know how this occurs. But I hypothesize the discrete FT can not
properly "model" truncated signals and phase distortion is a result of using
the DFT in a circumstance where it can not possibly succeed. I've increased
"at" (and the flip angle) and have never seen this problem again.
6666666666666666666666666666666666666666666666666666666666666666666666666666
I have Varian spectrometers (Inova and Mercury) and, basically, don't
have any phasing problems that can't be solved by properly setting
the instrument up (pulse programs, timing, receiver, dsp,
post-processing (lp)).
Indeed, we routinely do F19 nmr (among others) over fairly large
windows (>100kHz) with only small first order corrections following
careful adjustment of "dead-times" and filter delays. Remaining
*baseline* "anomalies" are removed with back predicting the first few
points. Proton, phosphorus, carbon, etc. are quite routine.
777777777777777777777777777777777777777777777777777777777777777777777777777777
I have seen this from time to time on various modern NMR systems. Almost
every case seems to involve receiver non-linearity -- after all the usual
suspects have been eliminated. That is, software people assume that,
because there is a 16 bit ADC in the system, the receiver has a similar
dynamic range in the analog world. I might add that it does not take a
clipped signal to produce this result. An FID that appears normal, but is,
in fact, at about the 0.1 to 1.0 dB compression level is quiet
sufficient. We found that on Varian 500 and 600 Mhz systems, it was
necessary to add a manually variable gain stage at audio frequencies and a
manual pre-amp attenuator in order to properly optimize the receiver
response. (I can supply further details if necessary.)
The Nicolet spectrometers, with their built in pre-amp attenuators and gain
control of both receiver IF and audio gain were particularly well suited to
this type of optimization. The more modern spectrometers seem to use
scaling algorithms that work for many "typical" samples, but not all. I am
not familiar with the modern JOEL system software, but you may want to
start by getting explicit control of pre-amp, IF and audio gain and explore
their effects.
If the problem is strictly associated with the filters, then your solution
is general. I seem to remember either a vendor application note or
publication on this subject a few years ago. BTW, do you have filters
implemented via a DSP chip set? If so, there are known second order and
round-off effects, depending on what filter function is used.
88888888888888888888888888888888888888888888888888888888888888888888888888888
My experience has been that the warped peaks are usually solvent or other
long T2 resonances. I teach our technologist not to phase using solvent
peaks.
99999999999999999999999999999999999999999999999999999999999999999999999999999
First of all this is a general problem. This is an artifact of the audio filters
in all kind of NMR spectrometers. Using digital filters with high bandwidth audio
filters using acoustic surface waves the effect is comparative small. The effect
is very high using classical LC butterworth filters. Nevertheless it is easy to
calculate this nonlinear phase error. I don't know the situation using
Jeol spectrometers, but in the case of Bruker devices there is a function
PKNL (phase correction non linear) available. You only have to switch it on
and then the higher order phase correction are done automatically using a built
in table. Works very well.
You see this problem within Bruker spectra. The effect really is not too large but
with high quality proton spectra you have a phase error of approximately 5 degree at the
edges of the spectrum without using the nonlinear phase correction. But I agree: you have
to measure rather carefully.
10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10_10
We have a newer JEOL 500 that this does occur in occassionally. However, I was
just using the blip function to reconstruct a few data points then processing to
eliminate the phase problems. Please send me (not all of ammrl) directions on
how I can add a 2nd order phase correction, etc. to the delta software.
11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11_11
We have ~ two year old Varian Mercury 300 and 400 NMRs that
have this problem. I figured it might be non-linear phase
shift in whatever they are using for an audio filter, but
that's just a guess, and I haven't looked into it any further;
it's not a serious problem for most spectra.
12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12_12
We had a similar problem with a new Bruker but it was due to vibrations from steam
pipes that run through a corner of our NMR room. These vibrations didn't show up on
the initial engineers report. It was "fixed" by vibration dampers. I assume you
have ruled that out?
13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13_13
News to me. We always assume peaks that cannot phase by 0 and 1st
order corrections have been folded in the spectral window. Be
interested in a summary of the responses you get.
Just a bit more thought brings to mind the possibility of resonance
offset phase problems. The longer the pulse width (lower the power),
the more resonance offset problems one will have as go away from
on-resonance. I have a hard time seeing how 2nd order phase shifts
get introduced, although I suppose analog filters possibly could
present some small problems. But with large sweep widths, and low
power rf, we have seen resonance offset phase shifts which increase
as one goes to the edges (high frequency) sides of the window.
14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14_14
My first reaction is that the selected audio filter is too
narrow, so that you are getting into the first part of phase roll off
farther from the carrier, which is only approximately second order.
Test: select a wider audio filter.
A second possibility that comes to mind is a non-uniform ADC
sampling rate, which also would give larger phasing errors at higher
frequencies away from carrier.
15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15_15
Interesting.
My one limited datapoint on that is I don't think it's general -- I've seen
no problems of that sort on a Varian inova installed 1 year ago. (Except
of course folded peaks and other artifacts.) I did have an application
recently that reminded me about the pre-acquisition delays (Varian's "pad"
parameter, and I think also rof1 and rof2) to adjust the time between the
read pulse and acquisition. These have a strong effect on phasing. Do
you have equivalent delays you can adjust on your JEOL?
16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16_16
we have 3 JEOL NMR-spectrometers; two of them are of the Eclipse series. In
some cases we also had similar problems. Particularly when you are observing
a large frequency range and have signals over the whole range it is sometimes
not possible to phase them all exactly, even applying second order phase
correction parameters. For which nuclei do you observe this phenomenon? For
some nuclei, e.g. 15N, 77Se or 31P we observe spikes, which often are like
signals and can be phased, but not exactly. We have both problems with the
old GSX/EX series as well as with the new Eclipse. I am not sure, what the
reason might be.
17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17_17
Are you sure that the acquisition window wasn't to small? In this case the
peaks situated outside the spectral width will be folded inside the window
with non coherent phase. Try to increase the acquisition window (spectral
width)! This is the case of a single 90° rf pulse. If not, many other
reasons may produce this effect. For example if the rf pulse in not well
calibrated to 90 or 270 ° and are around 180°; or when between the
excitation and acquisition there is a delay long enough for allowing one
some coherence to evolve under the scalar coupling into a superposition of
in- and anti-phase terms (product operators approach!)
18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18_18
Following up on the filter theme, another possible source may be from the
digital/frequency shifted quad. detection (fsq in Varian speak). It can be
used with dsp='r' on the Inova. It moves the observe frequency and the
filter offset outside your sw so the quads are removed as the are outside
the filter window. Some phasing problems may arise from that, especially
from high intensity peaks. I'm not sure if the JEOL has this feature, but
it may be something worth looking into, epecially if the second order phase
correction is a pain to do.
19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19_19
I have experience this myself usually phasing most of the peaks and having
TMS reference peak out of phase. Please send me any answers you get.
20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20_20
Received on Mon Aug 07 2000 - 10:36:37 MST