Converting float or negative integer to hexadecimal in Borland Delphi

Question

Ive written a program that communicate with some hardware using a serial connection. It sends a lot of hexadecimal values my way (sensor readings) and every once in a while it s

Answer 1

If you want to separate the exponent and the significand, you can use a variant record:

  TExtRec = packed record
    case Boolean of
      false:
        (AValue: Extended);
      true:
        (ASignificand: uint64; ASignExp: uint16;)
  end;

I think this helps to understand the structure of the floating point number.

Example usage:

var
  r: TExtRec;
begin
  r.AValue := 123.567;
  ShowMessage(IntToHex(r.ASignExp) + IntToHex(r.ASignificand));
end;

Output:

 4005F7224DD2F1A9FBE7

You can calculate it back:

v = (-1)s * 2(e-16383) * (i.f)

With

• e = $4005 = 16389 and
• i.f = $F7224DD2F1A9FBE7
• i.f = 1.930734374999999999954029827886614611998084001243114471435546875
• v=123.566999999999999997057908984743335167877376079559326171875

To convert i.f, i've used a binary converter.