AMMRL Community,
A few weeks ago, I posted a query regarding suggested tests to run on a
new Varian cold probe upon installation. The response was great, and
here I present the compiled responses. I first summarize the consensus
tests, then post the six responses verbatim, and, at the end, restate
the question for reference.
Thank you all for your help. I you think of any further tests, please
let me know.
- Josh
***
SUMMARY
***
- The easiest test to run that’s not in the installation manual is the
battery of Autocalibration tests found within Varian’s BioPack suite of
programs. Though these take an hour or so and provide just about all
the information needed for running biological experiments, it seems the
engineer may not run this test unless asked. For this you need a
well-behaved 13C-, 15N-labeled sample; we use a 3 mM proteinG sample
for this purpose.
- PFG stability seems to be a problem. PFG stability is normally tested
with the Gradient Amplitude Stability test portion of Autotest. One
responder has updated the standard pulse sequence, and the modification
to be used with caution, as with any new pulse sequence is described below.
Gradient echo amplitude should be >98% of the original after 20 repetitions.
The installation engineer should reconfigure the settings on the gradient
amplifier to compensate for the extra inductance of the cold probe's gradient
coils. Settin gradientshaping should help, though I don't know if this requires
a waveform generator to shape the gradient pulses.
- 13C channel performance seems to be a problem, at least on the Varian
probes (see next point). One user encountered noise that was difficult
to detect because most users just did 1H-15N work. Another reported
sending the probe back twice to fix the 13C channel.
- Some cold probe owners have noticed something called "Cryo-noise".
This is manifested as occasional bad transients due to weak capacitors
(arcing?) – usually in the 13C channel. The test for this will be to
run real experiments that use high-power pulses with decoupling and
spin locking on 13C and 15N, examining FIDs directly with the command
dfsh. These experiments should last ~24-36 hours.
- Pulse stability. There is concern that 90-degree pulse widths will
change over the course of an experiment, responding to the temperature
increase of the coil. One suggested test for this is to compare S/N of
a spectrum, e.g. gNhsqc, taken with nt=4 and compare it to one with
nt=16. If the pw’s are stable, the longer experiment should give a
S/N twice that of the shorter one. One Bruker owner mentioned the
sensitivity of their probe head to heating by decoupling. One2 recommendation
is to use adiabatic decoupling. The gNhsqc experiment may be too forgiving
in the regard. Perhaps a gc_co_nh experiment, which uses a 13C spinlock
and 15N decoupling would be better.
- Water suppression seems to be difficult. This may arise from any of
several factors. The high Q’s of cold- and cryoprobes make them
especially susceptible to radiation damping. One respondent reported
poor suppression when gradient selection was used for suppression. On
reading the other responses, one wonders if poor gradient shapes may be
playing a role there. Perhaps the test for this should be comparison of
presat, Watergate, and WET suppression methods with the same sample on
a RT probe with those acquired with the cold probe.
- The effective VT range should be tested. Some people have found that
their probes cannot go beyond a 12C – 62C range. Personally, I don't
expect to go above 62C, but we occasionally go down to –3C with salty
aqueous samples and lower temps with hydrophobic peptides in methanol.
We may need to shunt these samples off to a spectrometer with a RT
probe, but we'll try to push the temp range of the cold probe.
Unfortunately, I can't find VT specs in the HCN Cold Probe installation
manual, though there's an entry for the VT range to be filled out by
the installation engineer.
- It's recommended that shimming specs be paid attention to for both
normal and Shigemi tubes. I am also going to seek that the installation
engineer shims a 3 mm sample for our 5 mm probe. Last year's ENC and
Varian user's meeting were replete with glowing reports of enhanced S/N
when samples were concentrated and placed in 3mm and 1.7 mm tubes. The
rationale is that, with cold probes, the dominant source of noise i
the sample, therefore less sample - less noise ("sample" here
being the whole solution, not just the compound of interest).
- Tuning may be difficult. The tuning knobs on the cold probe are
somewhat novel, in that they're not just independent knobs you turn for
adjustment. First, you need to pull down a selector knob to choose
which channel being tuned. The HCN cold probe installation manual
describes the next step as "push the knob up to engage the rod", then
adjust the capacitor. It's unclear to me what's going on here, as we
just pulled the knob down, then we're pushing it up again. Maybe thereare two knobs. I'll know more once the probe is out of its crate. I had
been concerned by reports that the tuning rods would break if they were
too close to the end of their range and the probe heated up. It turns
out that there's now a warning against this behavior in the HCN cold
probe installation manual.
- It is suggested that performance be evaluated by comparing spectra
taken with a normal RT probe and the cold probe. It turns out that this
is problematic. The expected maximum S/N gain for a cold probe on a
nonconductive sample is ~4, and that this benefit is lower for salty
samples. The chief problem here is that different buffers are now known
to produce different S/N even when buffer pH and concentration are the
same. For instance, a phosphate-buffered sample is expected to not gain
too much from the new technology, but a Tris-buffered sample is.
Cambridge Isotope Labs now prints a list of buffers recommended for use
with cold probes: betaine, bis-tris, glycine, HEPES, imidazole, MES,
PIPES, tricine, and Tris. These are the ones they sell perdeuterated,
but you shouldn't need that for routine work if you’ve got a
isotopically labeled sample. The best real-world test of the cryogenic
sensitivity gain will probably be to measure S/N for BioPack
Autocalibrate spectra on the same Tris-buffered protein sample with the
RT and cold probes.
- The engineer should check the power output of the amplifiers before
hooking them up to the probe. The cold probes can easily be busted by
giving them too much power, and the best way to check that is to
measure the power output. One user is employing sub-maximal power
levels to extend probe life. As with all performance/preservatio
issues, though, the costs of reduced excitation bandwidth must be
weighed against the expected repair bills.
***
VERBATIM RESPONSES (edited for anonymity and relevance)
***
1) we have several Bruker cryoprobes and they occasionally misbehave.
I am not familiar with Varian probes but based on Bruker experience, I
could recommend several venues for checking.
1. Wobbing for salty samples (mostly 1H).
2. Shimming of all kinds of stuff, both shigemi and regular tubes.
3. Lock stability.
4. How much do probes warm during long decoupling pulses and do pulse
lengths change while probes warm up?
5. What happens during sumaltaneous decoupling on two channels?
6. Do you see any long term instability in pulses/powers (ideally over
a 1 week long expt).
If I think about any other sadistic expts, I'll let you know. Good luck!
2) with Bruker cry(o) probes, we were
sometimes disappointed about the limited
allowed VT range.
(It's something like 285k .. 335K)
Another limitation is the low amount of
heat / RF power tolerated by the probe head.
Peak power can be limited by attenuators but
the heating during 13C or 15N decoupling
(e.g. in HSQC experiments)
has to be lowered by short aquisition times
or adiabatic decoupling (e.g. CHIRP).
These restictions migth be similar with varian.
3) I’m sure you’ll get several opinions on this. I was
responsible for planning, installation, testing, and ongoing
troubleshooting and service for our 600 Cold-Probe. As part
of this, I've worked closely with Varian's Cold-Probe R&D group,
and have been involved with working out a variety of problems
with the cold probe (both mine, and others around the North
American continent).
I'm going to reduce my suggestions to just a couple (because I
wouldn't want you to have to reproduce the full year of headaches thatI've gone through).
A) PFG amplitude stability!! The inductance of the gradient coil
in the cold-probe is 4-5x that of a normal probe (and outside the
specified limits of the Highland gradient drivers). Depending upon how
well the installation engineer optimized the slew-rate of your gradient
amplifier, you might have oscillations in the rise/fall of your
gradient pulses which cause the voltage to "hit the rails" and thecurrent/voltage will oscillate (the leading/trailing edges of your
gradient pulses will look like a low-frequency FID). Setting
gradientshaping=’y’ helps, but won’t solve the problems if your
gradient driver is railing-out. Best test is to do the Gradient
Amplitude Stability test, which is just a
90-DephasingGradient-tau-RephasingGradient-tau-acquire sequence. A
form of this is in AutoTest, but the original was poorly written, and
I have my own that I re-wrote. Make sure your gradient-echo amplitude is
reproducible to > 98% over 20 (or more) repititions.
B)Real-World Performance!! To have a realistic idea of how well
your cold-probe is performing on real samples, I strongly suggest that
you use a typical (and stable) biological sample (with a realistic salt
concentration), and that you run a couple of simple 2D experiments
(i.e. gNhsqc, and a NOESY (or watever you want) )on your best Room-Temp
probe. Optimize your experiment as well as you can with your RT
probe. After the Cold-Probe is installed and calibrated, repeat the
same experiments on the same sample. Then you can compare the
real-world improvement in Signal:Noise vs. your RT probe. We did this
using identical parameters (of course, the Rf levels and pulse widths
were different), and we only had a 10-20% improvement at first, even
though our improvement for ETB was about 4.5 fold, and Sucrose was
4-fold better vs. the RT probe. Basically, this was the beginning of
how we discovered the problems with our gradient amplitude
stability/reproducibility.
C) Another good test is to (again) use a real-world sample in typical
salt, and do a simple gNhsqc, using nt=4. Then repeat the same
experiment using nt=16. Compare the S:N of the two, and be sure that
the second is 2x better than the first! If not, you have a stability
problem somewhere.
D) As soon as possible, run an experiment that does reasonably
high-power pulses and decoupling/spin-locks (within the specified
limits of the cold-probe, of course) on both 13C and 15N (i.e. a
simple HCN experiment). Make sure you run it for at least 24-36 hours,
and examin the FIDs directly (i.e. dfsh(x,y) ). Look for spurious
noise popping intermittently that looks like (but may or may not be)
what Varian calls 'cryo-noise". You may have bad FIDs, but won't
notice big problems after 2D/3D FFT; however, the bad FIDs are very
obvious if you examine the time domain data directly. We and others
have had problems with breakdown of capacitors (or other tuning
components) in the 13C channel. We didn’t see it for a long time
because our users were mostly doing H-N experiments. The 13C channel
is definitely the weak-link in these probes, so you should test this as
soon as possible.
These are, in my opinion, a minimum set of experiments that one should
do before believing that their Cold Probe is ready for prime time.
Good Luck with it!!
(cont'd) The only change I made was to convert the gradient pulses
from the old "rgpuls" syntax used in AutoTest, to "zgradpulse"
statements. The new parameter "gradientshaping" only works with
gradient pulses generated by a zgradpulse statement.So, the test was
inconsistent with the normal parameters used with the cold-probes.
I'm attaching two different sequences. The g2pul_rks.c just replaces
the g2pul.c sequence, and the ATgecho_rks.c is the sequence used by
AutoTest to do gradient echo stability. I found the most rigorous test
is to compare the gradient-echo stability doing a single gradient echo
using relatively long gradient pulses, and then do multiple echos with
shorter gradient pulses such that the multiple echos add-up to the same
total gradient time. If you are having problems with the leading/trailing
edge of your gradient pulses the long gradient echo will perform better
than the multiple, short-gradient echo experiment.
Have him set up some kind of 3D experiment on a sample in water. The best
would be one that uses gradient selection. See if he can get good enough
water suppression that the spectrum doesn't have artifacts.
We've had cryoprobes for almost two years (Bruker), and we still very
often can't get adequate water suppression in some experiments. The
experiments that perform the worst are usually experiments that use
gradient selection. In those experiments, the gradient selection is
usually the only water suppression method so there's nothing to optimize
(like water flip back pulses). So if the residual water signal is too big,
there's not much you can do about it.
5) Our Inova600's cryo probe is mainly used for macro molecules, so we
use Varian's BioPack auto calibration tests to check the performance of
the probe on a protein sample. The installation engineer doesn't
perform this unless you ask. And, of course, the BioPack should have
been installed on your system. It can be downloaded from Varian
website. The auto calibration can be done on a protein sample with a
good concentration. We use a 3 mM 13C and 15N labeled protein
containing a little over 100 amino acid residues in mostly H2O and a
small amount D2O for the lock. The auto calibration takes about 1 hour
and determines 90 pulse width and amplifier performance (e.g., qcom
parameter) for all the channels, and runs the first FID for all the
important multi-dimensional experiments and prints out results.
Probe safety:
We have several users using this cryo probe on different protein or
peptide samples. The 1H channel is the only one they need to
tune/match, the 13C and 15N tuning don't change from sample to sample,
so they are left untouched to prevent damaging. To further protect the
probe, BioPack allows parameters BPcryogenic, BPtpwrcryo, BPdpwrcryo,
BPdpwr2cryo, and BPdpwr3cryo to be set up according to the maximum
power levels for each channel.
You need to make separate probe directories for cryo probe and other
probes. This way all the cryo probe calibration can be stored under
the cryo probe directory (in /vnmr/probes). For example, the cryo
probe gives short 1H pulses so the power must be set so that the pulse
width is high than 6us. Other probes normally require higher 1H power
to reach the same pulse width. Each user must set the probe parameter
to the name of the cryo probe before calling BioPack experiments which
have been calibrated specifically for this probe.
Side note:
Once the cryo probe is in use, you most likely don't want to change to
other probes routinely. Our cryo probe's 13C coil had failed twice
shortly after the installation and the probe had to be sent to Varian;
our helium compressor had failed once and had been replaced by Varian's
engineer. Let’s hope these are not going to happen, but all of these
incidents need the cryo system be shut down, so that will be a good
time to switch to another probe.
6) I have a few suggestuions regarding cold probe. We have one on our
INOVA 600 since last June (operational).
a) You would make sure that an installer (enginner) carefully check out
the spectrometer console, particularly AMT's (not assuming they must be
that and that ---- somebody had done this sometime ago somewhere and
broke the probe beacuse it was too much power (for 13C)).
b) Water suppression with 2mM sucrose sample is good one. Make sure
that you will get an excellent s/n and line shape.
c) Do not push too hard to achieve shorter pulse lenght(s). I have
pw90 6.5 us at tpwr=55 (proton), pwC90 13.9us at 58dB (I use 15.1us at
57dB for my work), pwN90 34.4 at 56dB (I use 37.9us at 55dB for my
works). RF power(s) were determined by using power meter so we know
the exact RF powers (in watts on the 3 channels). The installer
wouldn't carry power meter normally so they would use pulse lengths as
a guidance. You might feel the pulse widths I mentioned above seems
too conservative. However I have decided to keep the life time of my
probe longer rather than pushing pulse length a few microseconds
shorter..
d) On any probe snapping the capacitor(s) is unwelcome but it happened
sometime particularly with multiple users situation. The tuning on a
cold probe is different from those of warm probes. For proton tuning
there are "tune" and "match", while C13 and N15 have only "tune".
Normally you tune H1 by "tune" first and followed by "match". If probe
is properly tuned you may need to check only "tune" from sample to
another, provide they are similar - solvents, temperature etc). In the
beginning tuning meter may not come down to "zero" but about ~4 but you
don't need to worry too match (eventually you may be able to come down
to 2 or 1 or even zero after some months of experience).
Carbon13 tuning may settle around a little less than 80, again you
wouldn't worry too match.
Nitrogen 15 tuning may end up around 17.
Important point of probe tuning is that you must make sure the tuning
rod is properly engaged -- first you will select the position (H1 tune,
for example) by rotating the selector and the push up the tuning rod.
When the rod is properly connected then there will be no space between
the rod (upper part - Varian calls this rod as tuning selector knob, I
think) and and bottom of the function selector. If the rod is not
fully up then it is not properly connected to the capacitor. If they
are misaligned and the rod is forthfully pushed up it may cause damage
(I haven't done it so this is my pure speculation). If the rod is not
engaged to the capacitor then tuning doesn't work.
You would need excerse to become familiar with this tuning operation
and then watch out how students understand and handle this opeartion.
Since the installation of our cold probe (ours is 5mm-pfg HCN probe) we
have been using this probe both tube mode and flowcell mode. Sofar the
probe has behaved nicely. We needed to replace seals on a scroll pump
after 6 months of constant operation and the compressor stoped once (we
suspect electric power surge was a cause of failure).
The probe itself shows an excellent sensitivity and lineshape (both
tube mode and flowdcell mode) and we are quite happy with the
performance.
***
THE ORIGINAL QUESTION:
***
I am seeking help with designing a battery of real-world tests for our
new Varian cold probe before we accept it and begin our warranty
period.
We have received our crates of equipment, and the unit will be
installed in a few weeks. To officially "accept" the unit, I am sure
the installation engineer will run a number of standard tests (S/N,
lineshape, etc.) to assess its performance. We would like to run
additional tests to ensure the reliability of the device under normal
facility conditions, i.e., sub-perfect operation by non-NMR-intensive
graduate students. For example, we have heard that the tuning rods may
break if they are tuned to one extreme or the other and the probe warms
up; we would like to test this hazard while we have the engineer there
and before we accept the probe. While we are not aiming to destroy the
probe, we want to push its limits under foreseeable normal
circumstances.
I would appreciate it if you were to tell me of problems you have
encountered with Varian cold probes (or Bruker cryoprobes, if the
problem is sufficiently general) that we could test during our
installation process.
Thank you.
- Josh
Josh Kurutz, Ph.D.
Technical Director, BSD/BMB NMR Facility
University of Chicago
Cummings room 201C, Mailbox #2
920 E. 58th St.
Chicago, IL 60637
Office: (773) 834-9805
Spectrometer room: (773) 702-4052
Cell (773) 315-5732
Fax: (208) 978-2599
nmr.bsd.uchicago.edu
Received on Wed Dec 29 2004 - 10:50:09 MST