The hp8510c calibration kit file format.

A cal kit file is 1426 bytes long.

string definition
   1 byte   field size N, maximum string length
   1 byte   actual length
   N bytes  string data, ascii text

float definition  
   4 bytes  data
            Signed big endian 24 bit fraction followed by an unbiased signed 
            exponent.  Note: fa,fb,fc are unsigned characters, exp is a
            signed character.
            number   = fa.fb.fc.exp 
            sign     = (fa & 0x80) ? -1 : 0;
            fraction = sign+(fa & 0x7F)/128+fb/(128*256)+fc/(128*256**2)
            value    = fraction * 2^exp
            zero     = {0,0,0,-127}

class definition
   1 byte   number of elements in this class (0..7)
   7 bytes  standards in this class (1..21)

0000  7 b magic constants {0,12, 0,3,0,3, 0}
0007  1 b version {1|2}
0008  8 b magic constants {0,1, 0,0, 5,128,5,128}

cal kit label
0010  \n (field size 10)  
0011  \t (run length 9) 
0012  "PROBE LRM "

SOLT classes: 13
001C  1 b # in S11A class
      7 b opens

      1 b # in S11B class
      7 b shorts

      1 b # in S11C class
      7 b loads

      1 b # in isol class
      7 b isol stds

      1 b # in forward transmission class
      7 b fwd stds (thru)

      1 b # in forward match class
      7 b fwd stds (thru)

      6*(1+7) b for reverse measurements

      1 b # in response class
      7 b response stds

0084  13 class text definitions
       1 b \n = 0x0A = 10 (field size)
       1 b length
      10 b text

      open -20fF  short  load 7.5pH  isol'n std  thru  thru 
      open -20fF  short  load 7.5pH  isol'n std  thru  thru  
      response

Calibration standards: 21
0120  21 standard definitions  50 byte/record
      36 b data  (9*4 byte numbers)
           start freq
           stop freq
           Z0
           delay
           Kloss  
           C0 C1 C2 C3
       1 b \n field size 10
       1 b length
      10 b text
       1 b type 0=open 1=short 2=load 3=thru 
       1 b flags
           bit 0 : waveguide delay  tau_{eff}=tau_{lin}/sqrt(1-(fco/f)^2)
           bit 1 : sliding load

      damping in Np = Kloss*delay/(2*Z0)*sqrt(f/1 GHz) 

TLR classes: 4
053A  4 class definitions
      1 b number of elements in thru class 
      7 b thrus

      1 b number of elements in line class
      7 b lines

      1 b number of elements in reflect class
      7 b reflects

      1 b number of elements in adapter class
      7 b adapters

055A  4 class text definitions
      1 b field size 10
      1 b strlen
     10 b text

     line 0ps  match  undefined  undefined 

058A  6 b zeros
0590  2 b 0xff 0xff

Servaas Vandenberghe