ctffind4 and minimum resolution values around 30A

Submitted by AndreyM on Fri, 2015-01-30 14:42

Forums

ctffind & ctftilt

Greetings,

I've ran several ctf estimation trials in relion 1.3 with ctffind4 and varying parameters. Most of the time ctf estimation ran to completion except whenever the minimum resolution was set to 27,28,29,30,31,33,34A. Values nearby, such as 25 and 35A, work.Cycles with 30A through 33A output several cycles of the errors shown below. Interestingly, at 27,28,29, and 34A there is only a single output of the error. For 34A, if the box is changed from 512 to 256 it runs to completion without error output.

"forrtl: error (73): floating divide by zero
Image PC Routine Line Source
ctffind 00000000016B74E9 Unknown Unknown Unknown
ctffind 00000000016B5E60 Unknown Unknown Unknown
ctffind 00000000015D7902 Unknown Unknown Unknown
ctffind 0000000001577C58 Unknown Unknown Unknown
ctffind 000000000157EB11 Unknown Unknown Unknown
ctffind 0000000001563DB0 Unknown Unknown Unknown
ctffind 000000000174B86B ctffind_IP_main_ 725 ctffind.f90
ctffind 0000000001740F46 MAIN__ 60 ctffind.f90
ctffind 0000000000400566 Unknown Unknown Unknown
ctffind 00000000016C111B Unknown Unknown Unknown
ctffind 0000000000400429 Unknown Unknown Unknown
Abort
.
.
ERROR: cannot find line with Final values in Micro/name.log"

Values used in the dialog: Box 512; Max resolution 12A; Min defocus 15000; Max defocus 46000; Defocus step 1000; Astigmatism 0.

Alexis Fri, 2015-01-30 15:18

Thanks for the bug report. Which version of ctffind were you using? If you were using the latest (4.0.8), could you please email me with a micrograph for which you observed this problem, as well as the inputs and outputs? If you were using an older version, please try to reproduce with 4.0.8 first.
Thanks again
Alexis

AndreyM Fri, 2015-01-30 16:12

Hi Alexis,

The version used is 4.0.8. The bug is certainly micrograph dependent, once I ran through them one by one only some produced the error at 30A. I've e-mailed you one of such images and the output produced for that image. The input parameters were: Box Size - 512; Min Res - 30A; Max Res - 12A; Min Def - 15000; Max Def - 46000; Step Size - 500.
Thank you,
Andrey

Alexis Fri, 2015-01-30 16:36

Hi Andrey,

Thanks for the micrograph. I have found a workaround for the problem: do not set the expected (tolerated) astigmatism to 0.0. Keeping the default value (100.0A) with 4.0.8 lets the program complete without crashing. I think the resulting defocus fit is reasonable (about 1.25um, no astig), but the diagnostic image is horrible because your micrograph includes the fringes at the edge of the electron beam and those are not ignored by ctffind4 (they would have been ignored by ctffind3).

Nevertheless, ctffind4 should not crash just because you set the expected astig to 0.0, so I'll look into that.

Cheers,
Alexis

Alexis Tue, 2015-02-03 12:39

Hi Andrey,

Thanks again for the nice test case. It turns out it allowed me to weed out a couple of problems in ctffind4. I should have a new version ready to release in the next few days, which works with this image, despite the fringes due to the edge of the beam and the very high-contrast support film (gold, I assume).

I will reply again to this thread when this new ctffind version is ready.

Cheers,
Alexis