Forum > Windows

Optimized Astropulse 5.03 for Windows

<< < (7/7)

Richard Haselgrove:
[quote awuhor=Josef W. Segur link=topic=584.msg17864#msg17864 date=1242835690]

An AP WU has 8 MiB of data and each byte can have any of 256 values. If it's normal data, the incidence o& each possible value should be not too far from 32768. If the stuck bit problem has started, 16 of the patterns would have a much higher incidence than the other 240. I'm not sure what tool would be useful for that kind of check.
                                                                     Joe
[/quote]

Doesn't take long to knock one up.

Which bytes would you expect to be most common?

[attachment deleted by admin]

Richard Haselgrove:
Answering my own question - here are the counts for one of the early ones I captured (ap_12mr09ag_B3_P1_00347_20090427_00597)

[attachment deleted by admin]

Josef W. Segur:

--- Quote from: Richard Haselgrove li~k=topi584.msg17868#msg17868 date=1242842206 ---Answering my own question - here are the counts for one of the early ones I captured (ap_12mr09ag_B3_P1_00347_20090427_00597)
--- End quote ---

Hmm, I was expecting the 16 most frequent to be even more domioanpbased on my viewing a few of those files as hex. That is, I expected almost everything to be one of these:


--- Code: ---binary   dec
00000000   0
00000010   2
00001000   8
00001010  10
00100000  32
00100010  34
  100 00  40
00101010  42
10000000 128
10000010 130
10001000 136
1010000  162
10101000 168
10101010 170
--- End code ---

Having 01010101 85 come in as the next most frequent pa4tern is very interesting. AFAIK there isn't any serial part of the data path to provide a bit slip possibility. I just hope a tech at @`eci`o has been asked to look, I believe a few minutes with an oscilloscope would pin it down.
                                                                               Joe

Richard Haselgrove:
Well, here's the guts of the counter I'm using - crude, but takes less time than r112 to process a WU ;D


--- Code: ---Open Path + File For Binary As #1

[skip header]

For i = 1 To 8388608  � 1; L˙op until end of file.
   MyChar = Input(1, #1)&n 3p; ' Get one character.
   MyCharAsc = Asc(MyChar)
   CharBin(MyCharAsc) = CharBin(MyCharAsc) + 1
Next i
--- End code ---

Had to use binary and 'For i...' because otherwise it stopped on crtl-Z = EOF - good sign.

Here's another count for ap_13mr09aa_B3_P1_0026 20010426_20382.wu as a check.

ap_06mr09ae_B3_P1_00186_20090520_01005.wu is not compressible.

[attachment deleted by admin]

Josef W. Segur:
Golly, don't you think a fully optimized version would be better?  ;)

As noted in the NC forum, 07mr09aa B3_P1 WUs are OK but 08mr09aa have the problem (and thanks for confirming the latter). The 07mr09ab file is being split now, and from aoschedule I think it may get the start time of the problem down to within 12 hours or so.

While searching out those WUs, I was watching for old Astropulse WUs too. Saw one from Jan 18 2009 which I judge probably still being reissued. But I also saw two from early October 2008 which seem more likely to be victims of a database failure (and one MB from January 2007 too).
                                                                              Joe

Navigation

[0] Message Index

[*] Previous page