AMMRL members & interested parties,
I'd like to update you on our experiences with our recent cold probe
installation, especially the results of the tests we've run and tricks
we've learned. In December I submitted a query soliciting such tests,
and then submitted a summary of many excellent responses from AMMRL
members. Our cold probe has since been installed (a couple of times),
so this message describes the results of the tests and a few key
pointers we learned during the installation. The most important pointer
is item #12, and you may wish to skip to it to ensure you don’t fry
your probe.
I apologize for delay in this message (and its length). We had an
intermittent problem with the preamp that required some extra cycles of
return/testing/repair/receive. We're happy that we're now attaining
S/N ~ 5050:1 on our old 600 MHz Inova when we used to meet spec with a
reasonably recent RT probe getting 1080:1.
FYI, I'll soon be posting photos of our installation on one of our
unofficial websites:
http://homepage.mac.com/jkurutz/PhotoAlbum11.html
I think such photos will be useful to people who are about to install
cold probe systems but have never laid eyes on the real equipment. I've
also posted this document in Word format for easier reading on our
download page:
http://homepage.mac.com/jkurutz/PhotoAlbum13.html
First, let's cover the results of the tests.
1) The BioPack "Calibrate Full" test on our 3mM protein G standard
(13C/15N, 50 mM phosphoate, pH 6.0, no NaCl) has become our benchmark
measure of relevant performance. Even if most pulse widths seem to
calibrate within reason, the final comparison of first transients for
various 2D and 3D experiments can reveal whether performance is
generally messed up, though the cause may not be immediately obvious.
2) PFG stability. I continue to receive many comments regarding the
pulse sequence for testing PFG stability I described in the previous
summary. I have now used those pulse sequences safely, and I've posted
them on our unofficial website:
http://homepage.mac.com/jkurutz/FileSharing13.html
(Feel free to take anything from that website, especially the NMRPipe
and NMRView guides.) When I first tried the "gradientshaping "y" vs "n"
comparison, I discovered I had no such parameter. It turns out that
this parameter needs to be created deiberately, and our engineer had
not done so. See the back of the "HCN Cold Probe" manual for
instructions on creating this parameter and “qcomp”. Now that I've
created it, it seems things are OK. Our installation included putting asmall magic box on the back of the gradient amplifier that’s supposedto compensate for the particular issues associated with cold probe
gradient use. It's probably doing its job because it passed all the
AutoTest tests.
3) 13C. Our 13C channel does appear to be touchier than a RT probe, but
it's manageable. I've been keeping an eye on it, and it seems OK.
4) Cryonoise. We've encountered the “cryonoise” phenomenon (testing
with the BioPack HCCH-TOCSY and noesy-CN-hsqc), but it seems it can be
successfully managed by conditioning the probe – a procedure outlined
in the HCN Cold Probe manual. In my last summary, I mentioned it might
arise from arcing in weak capacitors. I have since been corrected, and
have learned it comes from material buildup on the probe inductor. The
accretion can be removed by conditioning, which involves spending about
10-20 minutes delivering 200ms high-power pulses to the probe; it's
important to have no sample in the magnet during the procedure. I wrote
a macro to set up all the parameters to set this up and get it going
nothing fancy but I won't share it because it sets power levels tha
are appropriate for my system, but which will fry the probes on most
other systems (see item #12).
5) Pulse stability. My experience with the BioPack Calibrate-Full
routine has made me lazy, so I haven’t done much comparison of
individually-calibrated pulses and automatically-calibrated pulses. I
figure that Calibrate tries to find the pulse widths, small angle phase
corrections, etc. that give the best real-world results regardless of
what the "true" 90's are, so I go with them. The routine involves
a lot of pulsing, so calibration is probably done while the probe coil is
moderately warm. In practice, I just place 128 or 256 ss scans before
any complicated 2D or 3D experiment, just to be sure the probe has
heated up and cooled back down by the time data acquisition starts. If
you really want to know if you’ve reached probehead thermal
equilibrium, you can watch the cryobay’s full display and monitor the
wattage required for keeping the temp down at 25K; I recall it's
usually about 5 watts, and this will rise if the inductor gets warm,
then go back down one it's reached steady-state.
6) Water suppression I. Gradient shimming, hence water suppression, is
very tricky with the Cold Probe. There's some parameter in the gmapz
pulse sequence that seems get misset every now and then, probably when
clicking on buttons to set up PFG-H1 or PFG-H2, and the maps look
lousy. This never happened to me using a RT probe. "Lousy" means
clearly misshapen maps when you "Display Maps"; they are not noisy, but
they're squigglier than normal, and you can sort of shim on them, but
the results are poor/fair, not good/excellent. Also, the large/small
profile pairs are invariably large/really small. When I encounter this,
I fix it by reading in the map & parameters that the installer used
(with a 4Hz D2O standard), then making a new map. I haven’t tracked
down the offending parameter yet (it's not d3, which would account for
the large/really small profiles).
7) Water suppression II. After my last summary, I received a discussion
from Jack Howarth (U. Cincinnati College of Med.) explaining that the
difficulties with water suppression arise from the cold probe's 8mm
inductor diameter. The probe thus sees a greater length of sample, so
the "tails" on the profile are more prominent, and water signal coming
from these ends of the sample are harder to suppress. He recommends
using Shigemi tubes to alleviate the "tail" problem. One of our Varian
engineers recommended manually setting your gzwin smaller than what
would be picked automatically so that you don't include the tails when
gradient shimming. Our other installation engineer recommended manually
setting gzwin wider than the profile so the tails and flanking noise
are explicitly included. I have done no quantitative comparison of the
two manual methods, but I'm getting OK results by letting the automatic
routine have its way. That said, most of the time, I'm performing
heteronuclear experiments, for which lineshape isn’t as critical to
water suppression as t is for homonuclear experiments. My homonuclear
water suppression hasn't been that good, but the homonuclear customers
seem to want to work between -5 and 10 C, which broadens lines and may
account for poor water suppression independent of cold probe
performance.
8) Water Suppression III – sample size. Our main PI has always
recommended 0.5 mL 5mm samples (and gotten good results), though Varian
has always said to use 0.7 mL samples. water suppression on the 0.5 mL
samples is pretty poor with the cold probe. Unless they’re using
Shigemi tubes, you should strongly recommend that people use 0.6+ mL
samples if they're using 5mm tubes.
9) VT range. Our spec VT range is 0-50 C, and it reaches 50 °C OK, but
it has trouble at the lower end. I tried running an experiment (on a
real sample) at 0 °C (by tempcal standard, the setting was really -3
°C) with the engineer-set VT air flow of 15 lpm and a (JK-set) bath
temperature of –19 °C. It took approximately 30 minutes to declare its
slow approach to 0 was not going to happen – the best it got was ~ +4
°C. After some intermediate trials, I succeeded in achieving 0 °C by
turning up the air flow to about 20 lpm and keeping the bath temp at
–19 °C. During the second day of acquisition, I noticed the tuning
apparatus was developing condensation from our room air (~ 55% rel.
humidity). I found that I could still maintain 0 °C sample temp with a
bath temp of –14 °C, but not 13 °C,
10) Shimming Shigemi & 3mm tubes. I haven’t shimmed a Shigemi sampleyet, but the one person I talked to who had worked with a gel sample
(plugged at both ends) said it was trouble. I advised him to be
patient, and he eventually got a decent shim, but it took a couple of
iterative rounds of mapping/shimming/re-mapping/re-shimming. The first
3mm sample I tried to shim was a real peptide w/ 10% D2O, and that was
a mistake. Although you get a sample S/N benefit for concentrating your
sample & using a 3mm tube, you have trouble maintaining a lock signal
with so little 2H in the coil. I gave up on the real sample, shimmed on
a straight D2O 3mm sample, then went back to the real sample, and
everything was ducky (“good” for non-MN/WI/IL/IA people).
11) Real samples. There’s now a button on my forehead you can press to
make me say “How much salt do you have in your sample?” Lots of protein
chemists/biochemists blithely prepare samples using 150-300 mM NaCl,
but find that they don’t get the magic S/N benefit they think they
should when using the cold probe. I’ve been reasonably successful in
getting them to take a little more time developing low-salt conditions
and employing larger-MW buffers such as Tris and MES instead of
phosphate. It makes a real difference. If you haven’t received your
standard cold probe yet and you work with protein samples, you should
strongly consider renegotiating your order so you get a new
“Salt-tolerant” cold probe instead.
12) Don’t fry the probe. In my previous summary, I recommended
measuring Vpp of the signal going to the probe to ensure it wouldn't
fry the probe. Our engineer didn’t want to do that. He had a good idea
for what the 90° pw’s should be – and they’re about the same as for a
RT probe, though the power levels are radically lower. THE FOLLOWING IS
REALLY IMPORTANT: Our engineer simply advised us to not use high power
levels (60-63 dB high power, 45-48 decoupling) because they could fry
the probe. When I asked about implementing safeguards against
inadvertently using high powers, we typed “config” on the VNMR command
line and set the max powers available to the different channels. It
turns out this only keeps you safe if you try typing in a power level,
e.g. “tpwr=63” may net you a tpwr of 54 if that’s your maximum in
config. BUT, if you load an exisiting file or use an experiment that
was set up before the config file was modified, or if you recall a
parameter file, theses config limits aren’t checked and you can easily
fry your probe. Egregiously, this is exactly what happens if you load a
set of standard parameters from /vnmr/tests to test your probe! ALSO:
in BioPack, you can set the parameter Bpcryogenic=1 to impose power
limits on BioPack experiments. But it turns out that only works when
going through a BioPack Calibrate routine and/or updating your probe
file. There’s nothing to stop you from setting up a BioPack experiment,
then manually changing power levels to dangerous levels, then typing
“go.” MY SOLUTION and, I’ve found out, Bruker’s, is to add more
attenuators at the amplifier inputs (NOT the outputs). Keeping the same
target high-power 90° pw’s, I added attenuators until all the safe pw
were achieved on the autotest standard with maximum power (63 dB). Now
our default power levels are high and users are not able to set easily
set them high enough to do damage. Our Varian engineers don’t like this
because they think it’ll train people to use high power settings on
cold probes, and when they move on to an institution with a
“normally-installed” cold probe, they’ll fry it. I'm sorry for the
managers to whom this might happen, but I’m keeping our attenuators in
place to protect our system, and I encourage you all to do the same.
In this second section, I cover extra issues that came up during
installation.
1) Probe swapping. Our renovation model assumed that we’d install the
cold probe on one 600 and give its nice RT probe to the other 600 to
replace its ageing model. During the repair cycles, we had to either
switch the probes around again or leave our premium instrument
probeless and idle. If you’re getting a cold probe, it would be
worthwhile to see if you could bundle in an equivalent new RT probe at
a discount.
2) Plugs/outlets. We were prepared to wire up a 3-phase plug that
matched our 220V/3-phase outlet, but we weren’t prepared to wire up a220V/1-phase plug for our outlet; the cryobay came with 220V wires
hanging out the back instead of a plug. This is not a major deal, but
it involved some hours of running around to find a matching plug before
we could get the power on so we could start establishing vacuum in the
system.
3) RF cables. Our last RT probe was shipped with a fat
“high-sensitivity” 1H RF cable, but our cold probe came with only
regular cables. I ordered a bunch of the fat cables (RG214/U) from
Pasternack Enterprises, and they’ve actually led to a measureable S/N
benefit. I think it was roughly ~5-10%, but that’s a great benefit for
a $40-65/cable investment! The big benefit came from replacing the
standard BNC/SMA M/M cable with the fat RG214/U cable, part no.
PE33519-60 (60” long); this connects the 1H preamp to the cryopreamp
driver.
4) Magnet type. We have an older round-bottom magnet, and the hardware
we were shipped was for a newer flat-bottomed magnet. (The round/flat
question was not included in the site readiness checklists, but I had
sent them photos of the magnet-to-floor region to demonstrate
clearance, and these clearly showed its round bottom.) This meant the
probe could not be positioned properly with the new bracket that
enables the probe to hang from the magnet rather than the shims. The
real solution is to have a couple of new stainless components of the
bracket made shorter, but there were no such parts in existence on
installation. So our installation engineer, ordered a standard bracket
FedExed to us for the next morning so we could proceed. Until the next
time we warm up our probe and Varian has parts ready for us, our probewill be hanging from the shims. If you have a round-bottomed magnet,
make sure Varian ships you the components appropriate for it.
5) Bubbles in water line. We encountered a brief drop in water
pressure during the (first) installation, which was probably due to
bubbles in the line. I wasn’t aware the water lines had to be
bubble-free, but this caused the He pump to shut itself down in an
early stage of the installation. Our engineer quickly restarted the
system, cooling it back down. We’ve since ignored the problem and
nothing else has happened.
6) Electrical power. Our initial installation began the day after a
big snowstorm, which means the Chicago electrical grid undergoes
occasional millisecond-scale brownouts. We had purchased a 3-phase 40
KVA Powerware 9330 UPS to let us ride out things like that and it
worked great! It supplies power to the cold probe hardware, including
the turbo pump and He compressor, our three four-channel Inova’s with
FTC chiller VT control, and our spectrometer host and server computers.
We’re currently running at 49% capacity, but turning on the 8 kW He
pump from a dead stop overloads the system, causing a one- or
two-second bypass to the line feed, followed by normal UPS protection.
It’s no big deal.
7) Water supply shutdown. Our last (and most successful!) installation
came during a heat wave that overloaded our massive water chiller
(apparently shared with an air conditioning system), shutting down our
chilled water supply. All safeguards worked as advertised: the He
compressor shut off, the sample ejected, etc. Service was restored
within ~ 20 min, and our engineer got things running again by opening
up a side panel on the cryobay and blowing air over some temperature
sensor within. It turns out that such an abrupt shutdown gives you an
hour and a half to get the pump back to work before the probe starts
really warming up to RT. Even if this happens, though, it’s supposed to
be no big deal, risk-wise. It’s a pain in the butt, to be sure, but the
probe shouldn’t be harmed by a slow equilibration to RT.
8) Carousel compatibility. On this same spectrometer, we also have a
9-sample carousel, and over the last year some of the Varian people
expressed doubts about how compatible it would be with the cold probe.
One concern was over the length of the upper barrel, onto which the
carousel clamps; this was not a problem for us. There was also concern
that the automation software might not be compatible with the cryogenic
parameters. This has not been a problem, though I recently had to
increase the eject air pressure to raise the samples higher upon
ejection.
9) Gauges. Don Frank, our old Varian engineer, had recommended that weattach temperature and pressure gauges to our chilled water line so we
could monitor the pressure and temperature changes across the He
compressor. We did this fairly cheaply with standard brass pipe
fittings pipe fitting and gauges from McMaster-Carr and Cole-Parmer. Itis reassuring to see our temperature and pressure as real numbers.
10) Use brass fittings. Though my latest edition of the site planning
guide had said our water lines should end in 1/2” male NPT fittings,
our engineer expected quick-connect, and had to go to Home Depot for
adapters. One of the pieces he got was steel, not brass. After several
months, the outside was rusty from condensation. During our latest
installation, we replaced it w/ a brass fitting; the inside of the
steel fitting was OK, but it was too disconcerting to live with the
rusty steel piece.
Whew! I congratulate you if you’ve read through all of this. I hope you
find it useful! Thanks for your attention.
- Josh
Josh Kurutz, Ph.D.
Technical Director, Biomolecular NMR Facility
University of Chicago
Center of Integrative Science, room W123B
929 E. 57th 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 Fri Jul 15 2005 - 18:08:16 MST