CTFFIND4 with --amplitude_spectrum_input --filtered_amplitude_spectrum_input

Hi all,

I am trying to use CTFFind 4.0.16 where I provide the input Fourier amplitudes (with background subtraction already performed). However there is some bug that renders the input into garbage and so the fitting fails.

The command line is as follows:

ctffind --amplitude_spectrum_input --filtered_amplitude_spectrum_input << STOP_PARSING
CTF4Input_cb580c2647bb8a66.mrc
CTF4Result_cb580c2647bb8a66.mrc
1.326
300.0
2.7
0.1
512
50.0
4.0
5000.0
35000.0
500.0
100.0
no
STOP_PARSING

Here is the input that I pass in as an MRC file (as float32):

https://drive.google.com/file/d/0B23Il6NoQDNZQnV0cmZON0hiczg/view?usp=s…

Looking at image_headers.f90 I am using the same mode=2 flag to write the file.

and just to illustrate, here is the rotational average of the input:

https://drive.google.com/file/d/0B23Il6NoQDNZTlYyNGpqMVBTd2s/view?usp=s…

Here is the diagnostic output MRC file:

https://drive.google.com/file/d/0B23Il6NoQDNZajFtYTNrcXI1VXM/view?usp=s…

As you can see, the underlying data is nothing like the input. Looking through ctffind.f90 I don't really see what might cause this. There is one code block starting at line 457:

if (amplitude_spectrum_input .or. filtered_amplitude_spectrum_input) then
call current_power_spectrum%ReadFromDisk(input_file,current_micrograph_number)
call current_power_spectrum%ForwardFFT()
call average_spectrum%Allocate(dims=[box_size%value,box_size%value,1],in_real_space=.false.)
call current_power_spectrum%ClipInto(average_spectrum)
call average_spectrum%BackwardFFT()
else

This is resizing the input image I think, although they are both 512x512 in this case, so the code should not have any effect aside from wasting a few CPU cycles.

Can I turn on debug mode without re-compiling the code?

Robert McLeod
robert.mcleod@unibas.ch