1. list of detector types/data sources for support
   - hex-anode
   - quad-anode
   connected to
   - acqiris
   - TDC
2. sample(s) of data (experiment/run) which can be used for test/calibration purpose.  
   hex-anode : exp=xpptut15:run=280 
   quad-anode : ??? (Razib)
3. format of software (library of methods, a set of scripts, GUI wrapper, packaging, etc.)
  - detector-manufacturers software represents everythin as GUI
   - libraray of methods/ scripts/ GUI make sence
4. list of monitoring/calibration plots, types and format of constants for calibration purpose
   - Razib will make screen-shots of all interesting plots
   - calibration should provide for each delay line (t0, k) 
5. algorithms for implementation
   - peak-finding in waveform
   - plots stc.
6. list of analysis plots and output data formats
   - Razib will make screen-shots of all interesting plots

 Achim Czasch <czasch@atom.uni-frankfurt.de>
Wed 12/7/2016 5:57 AM
To:
Dubrovin, Mikhail;
Cc:
Osipov, Timur;
Action Items
Dear Mikhail and Timur,

I have compiled the lib on CentOS 7 (which is basically RedHat???) on gcc version

Here are 2 example programs that show how to use the lib:

a)
This program can read the LMF-data files that our software "CoboldPC"
produces. Timur Osipov may have some sample files from his beam times.

http://www.roentdek.com/download/LCLS_Linux/sort_LMF_1_detector.zip


b)
This is basically the same. But it does not read LMFs. It expects
a simple binary file format. So you will start with this one because you
have your own ADC data.

http://www.roentdek.com/download/LCLS_Linux/sort_non-LMF_from_1_detector.zip



Roadmap:

- measure the detector signals with your ADCs
- write a program that extracts the timing info from the signals.

Simple method:
look for the 2 samples which are just above and just below A.
A is the half amplitude of this signal.
The interpolate between the 2 points to find the timing value @ half ampl..

Advantage: Very robust. Works always.

better method:
Mimic the CFD-mechanism.
Superimpose the signal with an inverted and delayed copy of itself.
Look for the point where this new signal crosses zero.

Advantage: better resolution (about factor sqrt(2)).

But the simple method is already very precise.

- Now you have the timing values of all 7 detector signals.
Now you can use my example programs.



Some histograms that you will need:

Let's call the signals u1,u2,v1,v2,w1,w2 and mcp.
(values in nanoseconds)


- 1D:  u1+u2-2*mcp
- 1D:  v1+v2-2*mcp
- 1D:  w1+w2-2*mcp

- 1D:  u1-u2
- 1D:  v1-v2
- 1D:  w1-w2

- 2D   u1+u2-2*mcp  versus u1-u2
- 2D   v1+v2-2*mcp  versus v1-v2
- 2D   w1+w2-2*mcp  versus w1-w2

3 position images:

Xuv versus Yuv
Xuw versus Yuw
Xvw versus Yvw

with

u= u1-u2
v= v1-v2
w= w1-w2 - w_offset

and

Xuv = u * fu;
Yuv = (u * fu - 2.*v * fv)*0.5773502691896;
Xuw = u * fu;
Yuw = (2.*w * fw  -  u * fu)*0.5773502691896;
Xvw = (v * fv  +  w * fw);
Yvw = (w * fw - v * fv)*0.5773502691896;


fu, fv, fw are constants close to 0.7.
w_offset is a constant close to zero.
The exact values for these constants will be extracted from
real data later during the calibration. (Timur can tell you more about this.)


And it is also good to plot this data for each signals:

FWHM versus amplitude

You can calculate the FWHM (full width at half maximum) similar
to the "simple method" that I have described further above.
Just do the some on the trailing edge of the signal.

We will have to exchange some more emails until everything is working.

best,

Achim

> Could you remind me please, what is a right order
> to adjust parameters and calibrate new detector?

yes, there is a sequence that must be followed in the correct order:
a) First you must set the parameters in the config file so that the time sum peaks get shifted to zero.
b) Then you set the 'runtime'.
c) Then you set the xy calibration factors so the that image gets the size that you think is right.
(and set the xy offset parameters so that the image is well centered.)
d) Then you set 'radius'. This value should be 1 or 2 mm larger so that it really includes all hits.

Delay-line detector library use   
           
Achim Czasch <czasch@roentdek.com> 
2019-08-16, 12:00 AM Dubrovin, Mikhail             
     
Hi Mikhail,

> Scale factors converting time to coordinate are not important now. 
> because I am not aware about precise detector geometry.

For the DLD (not HEX) you must set the 2 calibration factors
manually later.
Example: If you have MCPs with 40mm active diameter then you must
the factors so that the image has a diameter of 40mm.
The factors for x and y will be slightly different to achieve a round image.

>  >> b) Then you set the 'runtime'. 
> Could you remind me please, how can I get this value(s)?

You plot x1-x2 (units: ns) in a 1D-plot.
Then logy-scale.

Now you can see a distribution that ranges from
-z to +z  (z is just a number here).
For 'runtime' you chose z + 2 ns.

> But, distributions of the time_sum for u and v look flat, even without calibration tables.

The errors that are corrected are within about +-4 ns.
So maybe you must zoom in a bit (in the y-scale).
But for a DLD this correction not so important.

> Still, it looks important to have an ability to calibrate potential differential non-linearity
> of the delay lines. I hope this is possible for QUAD-anode, right?

Not as easily as with the HEX. With a HEX we have a 3rd layer.
This layer provides the additional info that we need to make
an automatic linearity calibration of the 3 delay lines.
But with a normal DLD we do not have enough information.
So this algorithm can not be applied.
You can use a pin hole mask and measure the non-linearity manually.
Then you can apply another code from us that will spit out correction tables.
But this process is very very tedious and time consuming - and if you change something
(cable lengths, even voltages) then you may have to do it again.

best,
Achim