It’s best to think of slicing as just sampling the 3D reciprocal space. Then in the absence of errors fine slicing will improve signal/noise by reducing the excess background under the peaks. However [Mueller M, Wang M, Schulze-Briese C. Acta Cryst (2012) D68, 42-56] show that there are diminishing returns from slicing the spot width into more than 4 or 5 slices, and with finer slicing handling more data may be a nuisance. Shutterless data collection with modern pixel array detectors introduces very little noise / image, so this argument holds.
With older systems, there is a compromise with 1. shutter jitter, depending on the speed of data collection compared to the shutter speed 2. goniometer accuracy, to handle the stop/start synchronised with the shutter 3. read-out noise of the detector 4. reducing the background under the spots 1-3 favour thicker slices, 4 favours thinner Judging the balance with shuttered collection is complicated, but the advantage of shutterless data collection with fast detectors is clear Phil > On 1 Dec 2016, at 04:42, Edward A. Berry <ber...@upstate.edu> wrote: > > On 11/30/2016 10:16 PM, Keller, Jacob wrote: >>> If you fine slice and everything is then a partial, isn't that *more* >>> sensitive to lack of synchronization between the shutter and rotation axis >>> than the wide-frame method where there's a larger proportion of fulls that >>> don't approach the frame edges (in rotation space) ? Especially if you're >>> 3D profile fitting ? >> >> That is how the argument seems to go in Pflugrath 1999, but I would think >> that shutter jitter is a random error, so it would seem better to have >> several of these random errors for a given spot than just one. Perhaps >> measuring with high multiplicity would have the same averaging effect. >> >>> Is fine slicing more or less beneficial at high resolutions relative to >>> lower ones ? >> >> In terms of I/sigI, it seems to be the same proportional improvement across >> all resolutions. See Fig 4 of the Pflugrath 1999 paper. >> >> JPK > > I think the problem there is that, if the shutter jitter is random with a > constant sigma, it becomes a larger percent of the total exposure for that > frame. It would be like taking a 1ml pipetor with an error of 2% of full > scale, i.e. 20 ul. Because you want to average this out, you set it to 200 ul > and pipet 5 times. The sigma of that measurement would be sqrt(5) * 20 ul, I > think, so worse than doing it all in one shot. On the other hand if you take > a 200 ul pipet with sigma 2% of full scale or 4 ul, and take 5 times, the > error is sqrt(5) * 4 ul which is less than 20 ul. > Of course this would not apply to reflections that are fully recorded on one > frame since they are not reflecting while the shutter is open/closing. Then > it would be only variation in background. > >> >> Phil Jeffrey >> >> Princeton >> >> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ >> >> *From:*CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Keller, >> Jacob [kell...@janelia.hhmi.org] >> *Sent:* Wednesday, November 30, 2016 5:44 PM >> *To:* CCP4BB@JISCMAIL.AC.UK <mailto:CCP4BB@JISCMAIL.AC.UK> >> *Subject:* Re: [ccp4bb] Effects of Multiplicity and Fine Phi with Equivalent >> Count Numbers >> >> If the mosaicity is, say, 0.5 deg, and one is measuring 1 deg frames, about >> half the time is spent measuring non-spot background noise under spots in >> phi, which is all lumped into the intensity measurement. Fine slicing >> reduces this. But I am conjecturing that there is also fine-slicing-mediated >> improvement due to averaging out things like shutter jitter, which would >> also be averaged out through plain ol’ multiplicity. >> >> I guess a third equal-count dataset would be useful as well: one sweep with >> six-fold finer slicing. So it would be: >> >> One sweep, 0.6 deg, 60s >> >> Six sweeps, 0.6 deg, 10s >> >> One sweep, 0.1 deg, 10s >> >> Or something roughly similar. Who will arrange the bets? >> >> JPK >> >> *From:*Boaz Shaanan [mailto:bshaa...@bgu.ac.il] >> *Sent:* Wednesday, November 30, 2016 5:19 PM >> *To:* Keller, Jacob <kell...@janelia.hhmi.org >> <mailto:kell...@janelia.hhmi.org>>; CCP4BB@JISCMAIL.AC.UK >> <mailto:CCP4BB@JISCMAIL.AC.UK> >> *Subject:* RE: Effects of Multiplicity and Fine Phi with Equivalent Count >> Numbers >> >> Hi Jacob, >> >> I may have missed completely your point but as far as my memory goes, the >> main argument in favour of fine slicing has always been reduction of the >> noise arising from incoherent scattering, which in the old days arose from >> the capillary, solvent, air, you name it. The noise reduction in fine >> slicing is achieved by shortening the exposure time per frame. This argument >> still holds today although the sources of incoherent scattering could be >> different. Of course, there are other reasons to go for fine slicing such as >> long axes and others. In any case it's the recommended method these days, >> and for good reasons, isn't it? >> >> Best regards, >> >> Boaz >> >> /Boaz Shaanan, Ph.D. // >> /Dept. of Life Sciences / >> /Ben-Gurion University of the Negev / >> /Beer-Sheva 84105 / >> /Israel / >> // >> /E-mail: bshaa...@bgu.ac.il <mailto:bshaa...@bgu.ac.il>/ >> /Phone: 972-8-647-2220 Skype: boaz.shaanan / >> /Fax: 972-8-647-2992 or 972-8-646-1710 // >> >> // >> >> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ >> >> *From:*CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] on behalf of Keller, >> Jacob [kell...@janelia.hhmi.org] >> *Sent:* Wednesday, November 30, 2016 11:37 PM >> *To:* CCP4BB@JISCMAIL.AC.UK <mailto:CCP4BB@JISCMAIL.AC.UK> >> *Subject:* [ccp4bb] Effects of Multiplicity and Fine Phi with Equivalent >> Count Numbers >> >> Dear Crystallographers, >> >> I am curious whether the observed effects of fine phi slicing might in part >> or in toto be due to simply higher “pseudo-multiplicity.” In other words, >> under normal experimental conditions, does simply increasing the number of >> measurements increase the signal and improve precision, even with the same >> number of total counts in the dataset? >> >> As such, I am looking for a paper which, like Pflugrath’s 1999 paper, >> compares two data sets with equivalent total counts but, in this case, >> different multiplicities. For example, is a single sweep with 0.5 degree 60s >> exposures empirically, in real practice, equivalent statistically to six >> passes with 0.5 degree 10s frames? Better? Worse? Our home source has been >> donated away to Connecticut, so I can’t do this experiment myself anymore. >> >> All the best, >> >> Jacob Keller >> >> ******************************************* >> >> Jacob Pearson Keller, PhD >> >> Research Scientist >> >> HHMI Janelia Research Campus / Looger lab >> >> Phone: (571)209-4000 x3159 >> >> Email: kell...@janelia.hhmi.org <mailto:kell...@janelia.hhmi.org> >> >> ******************************************* >>
