I) Measure the crosstalk from magnet A and find the compensating magnet B
   A) Observe a BPM immediately upstream of magnet A
   B) Vary magnet A by current DeltaI_1
   C) Find a nearby magnet or magnets B in an earlier line which compensates the
      effect of magnet A with current change DeltaI_2

II) Re-initialize the knobber definitions
   A) Modify the file /home/erpopr/online/epics/pyiocs/iocs/knobber/initialize.py.
      For example:
       knob_defs = {
           "MR2DIP09":{"MR1CRV04":-0.045},
           "MS2CRV01":{"MS1CRV01":-0.25},
           "MS2CRV02":{"MS1CRV02":-0.02},
           "MS2DIP02":{"MS1CRV01":-0.25/7.0},
           "MS2DIP09":{"MS1DIP07":0.022/6.0},
           "MR2CRV04":{"MR1CRV04":0.08/2.0},
           "MR2CRV03":{"MR1CRV03":-0.03/3.0},
           "MR2DIP02":{"MR1CRV01":0.2/6.0},
           "MS2DIP08":{"MS1DIP07":0.01/1},
       }

      A 1% change (DeltaI_2/DeltaI_1 = 0.01/1.0) in MS1DIP07 compensates the change in MS2DIP08.

   B) Now run the knobber initialization:
      (cbeta) erpopr@erp39:knobber $ python initialize.py   

      The file knobber.init.json will be updated.

III) Restart the IOC as follows:
   A) Log in as epicsioc on erp129
      1) kinit -f [CLASSE username]
      2) ssh -Y epicsioc@erp129
   B) cd l0e/pyiocs
   C) Restart the IOC:
      (cbeta) epicsioc_erp129% ./restart_knobbers
-----------------------------------------------------------------------------------------------------------
Knobber diary - 26 January 2020. Francois Meot and Jim Crittenden
=================================================================
Method for making knobs in S3
-----------------------------
Observe BPMs in R1, B1BPC03-3 and D1BPC01-3.
Observe beam spot on view screen at entrance to S3: S3SCR01.

MS3DIP01
--------
Observe transmission in pass 4, which is presently about 50%, quite stable.
Vary S3DIP01 to the edges where transmission begins to suffer.
The range is 189.53 to 189.88 A. The nominal setting was at 189.86 A,
so it was on an edge. Set it to 189.66 A.
Put the S3 screen in.
S2DIP03 is initially at 115.137 A.
Take reference orbit and subtract it.
Change S3DIP01 +2 A, D1-3 moves -3 mm. Screen corroborates, but less accurate
Change S3DIP01 +3 A, D1-3 moves -6 mm. because of jitter.
D1-3 moves back to zero with 0.04 A in S2DIP3.
Define 
We find compensation using ms2dip3 rato 0.04/3. Test knob: it works.
Since acceptance is only +-0.25, range of +-3 a is sufficient.
Afer defining knob, find new optimal value of ms3dip01 to be 189.55.
Range is again +-.11
New saved value for ms2dip03 is 115.135 A.
NB: The transmission acceptance of MS3DIP01 and our knob works up to 3 A,
so its range is sufficient.

ms3dip02/3 No effect
----------

ms3dip04 range +-3 A, ratio with ms2dip08 is 0.04/2
--------

MS3crv04 affects D1BPC01 vertically by < 1 mm.
--------

Ms3crv03 +3-->-3 moves beam at d1bpc01 - 2 mm.
--------
0.54 amps in ms2crv03 corrects it, but with the wrong sign,
so it is using a quad off center somewhere.

ms3crv02 +2 A moves D1BPC01 -3 mm in Y and -4 mm in X.
--------
ms3crv01 +1 A moves D1BPC01 -3 mm in Y and -4 mm in X.
--------

Decide not define knobs for the S3CRV0x.
Need modeling study to understand how to shield.
---------------------------------------------------------------------------------------------------