RAMAC Oscilloscope Observations

1
RAMAC Oscilloscope Observations

• Joseph Feng
• Magnetic Disk Heritage Center
• Computer History Museum
• June 22, 2007

2
Acknowledgements

• Thanks to Al Hoagland, Pat Connolly, and the
  students at Santa Clara University for bringing
  up the RAMAC
• Thanks to the IBM Corporation for loaning the
  RAMAC drive
• Thanks to Hitachi Global Storage Technologies,
  Steven Lambert, and Terry Whittier for providing
  computers and oscilloscopes
• Thanks to the Computer History Museum for
  providing the laboratory facilities

3
OD track (-01)complex waveform

• Top trace expanded view showing details
• Bottom trace one full record (20 of track)
  showing starting sync burst, start gap, 100
  characters of data, and end gap

• RAMAC readback waveform - bottom head at OD CE
  track (Tektronix)

4
ID track (100) complex waveform

• Same text data as on the OD waveform - vertical
  scale changed for top trace, same for bottom
  trace
• Note the wider pulse widths, lower signal
  amplitude, and poorer signal-to-noise ratio

• RAMAC readback waveform - bottom head at ID CE
  track (Tektronix)

5
Customer data track - blank data

• This is representative of the most commonly
  observed data pattern
• Repeating blocks of 10000001, which decodes into
  blank data (unpunched column in IBM card)

• RAMAC readback waveform - bottom head at a
  customer data track (Tek)

6
Decoding the RAMAC data1

• The RAMAC readback waveforms for two consecutive
  revolutions were captured with a LeCroy 9345AL
  digital oscilloscope
• 10 ms per waveform, no averaging
• 100ns/sample or about 100X oversampling
• 1,000,000 samples per waveform
• The apparently high noise levels in the waveforms
  is attributed to the increased bandwidths in the
  modern electronics and the oscilloscope
• scope bandwidth 500 MHz
• a 733-based preamplifier had an estimated
  bandwidth of about 10MHz and a net gain of about
  5X its two key functions were to provide the
  proper load impedance for the heads and to act as
  a differential-to-single-ended adapter from the
  heads to the oscilloscope

7
Decoding the RAMAC data2

• A program was written to decode the digitized
  waveforms
• After mathematically reducing the noise, 100 of
  the 10,054 sectors (out of 10,200 possible) were
  decoded without any obvious errors
• one sector had real errors, apparently due to bad
  overwrite
• two characters were decoded with proper parity,
  but not allowed
• 28 unformatted tracks were observed, most had
  low-amplitude transitions, variable envelopes,
  maybe due to low write current
• one track had been used for experiments by SCU
  students

8
Contents of the RAMAC data1

• Most of the CE tracks (-01 and 100) had test
  track identifying text
• Most tracks had blank data or tables with
  references to Canadian companies, including one
  to EXPO 67 CANADA
• Six sectors had non-boring readable text, with
  typographical errors
• One reference on three sectors to the final
  results of the 1965 World Series, with October
  14, 1965 date, the correct result, and a humorous
  (and obviously deliberate) misspelling of Sandy
  Koufax
• The 40-year old data appears to be
  well-preserved. There were no errors that could
  be attributed to data decay. In most cases, the
  tracks with low amplitude/bad envelopes were
  adjacent to good tracks with no signs of distress
  or readback errors

9
Waveform envelopes - 1well-aligned sectors

• Top trace 100 ms or two revolutions, showing
  data envelope
• Lower trace index pulses
• The spikes at the start of each record, seen on
  the ID side of the scan, is due to old data, with
  the new data written slightly OD and down-track
  compared to old data

• Head moved across one data track from OD (rim) to
  ID (hub) (LeCroy)

10
Waveform envelopes - 2showing cross track
misalignment

• Envelope modulation in 10ms blocks due to cross
  track offsets between sectors
• The modulation pattern on the ID (hub) side is
  the complement of the pattern seen on the OD
  (rim) side
• Decoded waveforms taken at the point of minimum
  amplitude variation

• Head moved across one data track from OD (rim) to
  ID (hub) (LeCroy)

11
Waveform envelopes - 3showing a bad sector

• The third sector after index shows a loss of
  amplitude compared to the other sectors
• Symmetric signal loss on both hub and rim sides
  means it was not due to cross track misalignment

• Head moved across one track data from OD (rim) to
  ID (hub) (LeCroy)

12
Waveform envelopes - 4showing localization of
bad sector

• This shows the signal envelopes for the tracks
  near the one in the previous slide
• The adjacent tracks do not show a similar loss of
  amplitude, so the signal reduction may not have
  been due to a scratch or other defect

• Head moved across five consecutive tracks from OD
  (rim) to ID (hub) (LeCroy)