A description of the hardware-level track format of the IBM media used on the BESM-6 I've found is as follows (translated by Google with minimal corrections):

             Track header structure:

At the beginning of each track, its header is recorded, which
is a summary of track information used for hardware and software control and allowing
to organize work with a package that has defective tracks.
  The header contains:
   - Address marker (AM) - 12 bits,
   - Identifier (I) - 2 bits,
   - Track address (A) 16 bits,
   - Address checksum (ACS) - 12 bits.
  This information is duplicated if the track is good. When
the track is defective, the replacement track address is indicated in the 2nd part of the header.

  Position of the header from the beginning of the track sweep:

     | 270  |          |  42   |          | 258    |
     | bit  | 42 bits  |  bit  | 42 bits  | bit    |
     |zeros |          | zeros |          | units  |
     |      |          |       |          |        |

   - Address marker code: 111111000001;

   - ID values:
       00 - working track,
       01 - the path is defective, the one adjacent to it is good
       10 - the track is defective, the one adjacent to it is defective
       10 - replacement track {Likely a typo, should be 11? - Leo B.}

   - Address structure:
       16÷9 РР - cylinder number,
       8÷5 PP - head number,
       4÷1 РР = 0.

   - Address checksum (includes ID)
     is formed taking into account the features of the hardware circuit

As reverse-engineered, ACS was a 12-bit "sum with carry to the right" of the lower 12 bits of A with the remaining bits.

The lower 5 bits of A are constant 0, presumably, because there was only one sector per track.

The questions are:

  • Does that resemble the track/sector header structure actually used by the IBM disk controller?
  • What was the actual structure of the whole sector? How was the beginning of the data part indicated, according to the IBM standard?
  • Hoping that the BESM-6 disks, kept in storage of the London Science Museum for 30 years, could finally be read.
    – Leo B.
    Sep 21 at 22:19
  • 1
    The packs don't matter, format is done by the host sending home address write command to controller (1311/2314). This format does not look like an IBM /360 DSAD as such would have a 9 byte home address, independent of being CKD or FBS (at that time most likely CKD, see here). Especially notable that those fields seem to fit 6 bit characters (6x7=42), so maybe some 14xx style recording? 2314 is highly unlikely as it was 8 bit byte orientated.
    – Raffzahn
    Sep 22 at 3:47
  • Should have read 2311/2314. 1311 is for 14xx - which ofc might be the example copied.
    – Raffzahn
    Sep 22 at 3:56
  • @Raffzahn The contents of the bit stream look very close to what you have described there, except byte alignment. You say that Flag Byte has only 2 meaningful bits, and here exactly two bits are used. It is only the track description record which is missing. I was wondering if the disk could be read by a regular controller, but most likely that is moot, if the data recovery facility has its drives connected to a rig which reads a raw bit stream to be interpreted in software.
    – Leo B.
    Sep 22 at 4:48
  • My knowledge about pre-360 formats is limited, so I only rely on manual reading for 6 bit era. Could it be that your translation of the track ID needs a second look? The way it is, even with correction for 11, does not make much sense to me. Also, already on 1301/02 drives teh home addrrss was written using 8 bit mode.
    – Raffzahn
    Sep 22 at 13:32


You must log in to answer this question.

Browse other questions tagged .