A brief report on the noise characteristics of the IRS in the
first two anneal sequences for Campaign 18

9 Feb 05

Lei Hao and Greg Sloan

Abstract:

As measured by the SURN tests, SL, LL, and LH are still
noisier now than before Campaign 18, although the severity
of the problem has lessened from the startup anneal sequence
to the first mid-campaign anneal sequence.

Analysis:

    Noise characteristics of the modules (in DN)

           SUR data                SRS data
           noise   pre     post    noise   pre     post
   Module  limit   anneal  anneal  limit   anneal  anneal

   Campaign 18 startup

    SL      8.8     9.5    10.6    15.0    12.7    12.3
    SH      8.8     6.7     7.5    15.0    12.8    13.0
    LL     14.5    15.7    15.2    15.0    13.3    13.1
    LH     17.6    22.5    20.2    15.0    13.7    13.6

    Campaign 18 first mid-campaign anneal

    SL      8.8     8.1     9.0    15.0    12.1    12.2
    SH      8.8     7.0     6.9    15.0    12.8    12.7
    LL     14.5    14.7    13.0    15.0    12.7    12.9
    LH     17.6    19.3    19.2    15.0    13.1    13.2

  Both the SURN and SRSRN results have dropped from the
  start-up sequence to the first mid-campaign anneal sequence,
  although SL and LL still exceed the limit once (for a total
  so far of three failures out of four), and LH still failed
  both times.  The LH SURN levels are still high, but they
  are clearly dropping, so far in a monotonic manner.  It
  is very interesting that the SRSRN numbers, while never
  beyond the limit, are also dropping.

Comparing to the average SURN in all anneal sequences in 
Campaigns 15-17 to all those in Campaign 18 gives the 
following results:

          Average for     Max for   
  Module  Camp. 15-17   Camp. 15-17  Camp. 18
  SL       8.8 +/- 0.6     10.1       9.3 +/- 1.0
  SH       7.0 +/- 0.5      8.0       7.0 +/- 0.3
  LL      12.6 +/- 0.9     14.1      14.7 +/- 1.2
  LH      18.3 +/- 0.6     19.5      20.3 +/- 1.5

  The previous brief report only compared our Campaign 18
  data to the corresponding start-up sequences in Campaigns
  16 and 17.  We have expanded the comparison sample to all
  anneal sequences in the previous three campaigns, which
  has raised both the means and standard deviations.

  While SL, LL, and LH still appear to be degraded compared
  to data from previous campaigns, they are better now than
  before.  It remains to be seen if these modules return to
  the previous norm as Campaign 18 continues.