1. AWBIN: 941216_215200: Exact match; weights calculated from multiplying two identical vectors with triangular distribution. ---------------------------------------------------------------------- 2. IRBIN: 941220_214121: shift_code = 0x8008 -6 < diff < 4 941221_175601: shift_code = 0x800C -76 < diff < 55 3. ALIGN: 941221_004733 Exact match (XALIGN data only) 941221_020252 XALIGN data matches exactly YALIGN matches if treated the same as x-align data 21-Dec-94 Created sclsum.ips to test SCALESUM instruction; Upon running it, IP generated the "Invalid Opcode" (0x8020) error. The other error words are, 0x8020 0x01D6 0x3450 0x3F30 0x0123 Turns out that the code for SCALESUM is 0x1E0 not 0x01D6. Edited the .src file; compiled and perfed sclsum. No errors - but 30 non-compares; The pixels which are different occur on 8k boundries!! Talked to Mike. He pointed out that 8k was the size of IP working buffer. Also suggested trying irbin with a constant input value. Filled memory 0 with 700. Re-ran irbin test. No errors. File name: 941221_234546 Discovered that irbin.src extracted a 512x512 area from the raw image but downlinked only 64k words (256x256). This would mean that the raw data used for calculations would not be the same as used by the IRBIN instruction. Changed command to downlink the full 512x512 buffer. 941222_005036 file with new data type. chkirbin still generates lots of errors. Tried row order and skip take order - no change in error rate. Changed chkirbin to continually skip and take rather than calculate a new start address for each row. No significant change in error rate. ------------------------------------------------------------------------- 22-Dec-94 Wrote an IDL program (/usr/people/mathur/mk_irpat) to create a buffer containing checker board pattern. The progarm generates 512 bytes of checkerboard data and is then loaded into the IP by means of the generic data load command. (mbiptbl 0x0b...). This command did not appear to work. Wrote a control procedure to load a small checkerboard pattern directly into page 1. At first the irbin procedure still produced junk data. However it was discovered that the source page had been specified as 0x0001 which is actually page 0. When this was corrected IRBIN appearded to work ok. (data set 941222_184750) 941222_200741 contains a good data set. The checking program, /mdisw/idl/ucon/mathur/chkirbin, shows no errors on this data set. ALIGN: Found a problem (incorrect opcode) in align.ips. Changed repeat count to 512 (from 511). Raw and Xalign data looks ok. Yalign data has many vary large values (~64k). 941222_222343. ------------------------------------------------------------------------------ 3 Jan 95 1. 950103_230236: No data from the IP 2. 950103_230629: SCALESUM Data Mismatch every 8k words. 3. 950103_232410: ALIGN data No errors in XAlign 247 error in Yalign at 510 pixel boundries, starting at col. 255. ------------------------------------------------------------------------------ 5 Jan 95 1. 950105_142143: Contains a raw image, Xaligned and Yaligned images. Xaligned image does not match at all with the calculated data. Yaligned image shows 256 non-compares. This data set is useless as the Xaligned data was generated by the wrong IP macro. The macro was started at 0x158 rather than 0x151. 2. 950105_143806: No errors in XAligned data 767 errors in Yaligned data, at 255, 765, 1275... at 510 pixel intervals. In terms of 2d array, columns 255, 510 and 511 do not match. Column 511 contains garbage anyway (from Xalign). From row 256 onwards only columns 510 and 511 do not match! Is 255 a magic number? Error discovered in IDL routine - Col 510 was not computed! Fix reduces the number of mismatches to 256; Only col 255 shows mismatches. 3. 950105_145753: SCALESUM data Mismatches at 8k boundries, 8191, 16383, 24575, 32767... Total of 32. 4. 950105_162847: Vel, LineDepth and Cntnm; IP 1.6.01; DEP 0:26 4097 error in Velocity Image 1 error in Line Depth Image 1 error in Continuum Image 5. 950106_013810: SCALESUM; IP 1.6.02; DEP 0:26 No Errors 6. 950106_014603: VLCTY, LDEPTH and CNTNM; IP 1.6.02; DEP 0:26 1 error in Velocity Image 1 error in Line Depth Image 1 error in Continuum Image 7. 950106_025005: ALIGN; IP 1.6.02; DEP 0:26 Changed repeat count for YALIGN from 64 to 255 to cause C256 to be the first column processed on a call to YALIGN. Called YALIGN 3 times (processing a total of 765 columns). Then downlinked 512x512 section of the image. No Errors !! 8. 950106_033920: ROWBIN, LRBIN; IP 1.6.02; DEP 0:26 IP error; 1D02 01D0 3450 AAAA 0169 Error in LRBIN - Invalid List; parameter AAAA at PC 0169; Next Instr = 01D0 ---------------------------------------------------------------------------- 6 Jan 95 1. 950106_141738: ALIGN IP 1.6.02; DEP 0:26 Set the repeat counts as follows - 64, 64, 64, 63, 65, 64, 64, 63, 1. The number of columns totals 512. However, processing for column 256 begins on a new call to YALIGN No Errors!! 2. 950106_144329: ALIGN IP 1.6.02; DEP 0:26 Created a new version of align to process the whole image. The repeat counts are set as 255, 1, 255, 1, 255, 1, 255, 1. First Attempt at checking reports 255101 errors in X-Align and 254901 errors in Y-align. Panic Over - Discovered that d-factor for x-align had been specified as 8196 not 8192 used for cheking. Using 8196 resulted in no errors. 3. PURGELST Tests ckpurge control file generates single and multi-bit errors in list memory and dumps data. data set 950107_011521 contains examples of single bit errors, single bit errors in different 64 bit blocks and multiple bit errors. It contains 7 data sets; i) Initial list, ii) list containing a single bit error, iii) purged version of the list (ii) iv) list containing two single bit errors in different 64 bit blocks, v) purged version of the list (iv) vi) list containing multiple bit error and vii) the purged version of (vi). rdtfr,fn,img,hdr,tblk for i=0,6 do begin print,img2(14:24,i) and 'ffff'x,format='(11z5)' & endfor for i=0,6 do begin print,hdr2(391:398,i) and 'ffff'x,format='(8z5)' & endfor ----------------------------------------------------------------------------- 9-Jan-95 Loaded v 1.6.12 of WCS code Modified align.ips to test to combined XYALIGN instruction. The new version is called align3.ips. Made corresponding changes to chkalign, creating chkalign3. Chkalign3 tests only 1 pair of images rather than three images as chkalign. 950109_223732: contains data from XYALIGN instruction. 130900 mismatches starting at 130305 or (254, 257). Looks like the same old Yalign problem. Instead of attempting to align the whole 512x512 array, split it up into 4 calls processing 255, 1, 255 and 1 columns respectively. 950109_231747: contains data from XYALIGN instr. No errors. ----------------------------------------------------------------------------- 11 Jan 95 Loaded V 1.7.00 of WCS code Wrote 2 control files, copycb and docopycb, to load an IP program to test the COPYCB (Circular Buffer) instruction and to execute the test respectively. After debugging a good data set was obtained; 950111_235159. It contains 14 data sets. Each set contains the data from the circular buffer. 4K data is written to the circular buffer on each call to the COPYCB instruction. Output data therefore contains the following pattern; 0x1111 junk junk 0x1111 0x2222 junk 0x1111 0x2222 0x3333 0x4444 0x2222 0x3333 0x4444 0x5555 0x3333 0x4444 0x5555 0x6666 0x7777 0x5555 0x6666 etc. Loaded Lev9t27; Re-ran copycb test; Looks OK; 950112_003452 execute veriptbl; Data looks strange - all 0xAAAA's !! re-run ipstrtf -- same result reloaded t26 + ipstrtf - good tables reloaded t27 + ipstrtf - still bad - table data is not being written to IP. reloaded t26 dofftst: 950112_011506 - 1 error each in flatfield and badpix, no errors in scale. awbin: 950112_015318 No errors dorblr: 950112_020051 IP Error 1D04 - invalid repeat count (0) 01D0 - Op Code 3430 - IP state 0000 - Parameter 0169 - Address of next instr dotwavg: 950112_023130 No errors dovldc: 950112_023955 Error in List header, list was apparently not loaded. (based on the data dumped). doalign: 950114_004628 No Errors. (doalign loads and run the align program). ---------------------------------------------------------------------------- 16 Jan 95 950117_002427 - genvldc 950117_012437 - rblr ---------------------------------------------------------------------------- 17 Jan 95 950117_184822 - doiptst; Error in RWBIN; RWBIN list apparently did not load Checked 950117_181151 which contains data from veriptbl control file. The med-l (RWBIN) and Line depth (VWBIN) lists were not loaded correctly. ---------------------------------------------------------------------------- 18 Jan 95 5k test - works 950119_013006 - iptst (CROP/RWBIN/VWBIN); VWBIN errors 950119_013810: rblr test; No errors ---------------------------------------------------------------------------- 19 Jan 95 950119_013006 - is OK after all - The problem was caused by addition of large numbers overflowing. 950119_193908 - VLDC data; Many errors - Possible DEP SW problem ---------------------------------------------------------------------------- 20 Jan 95 Roger fixed the DEP code to put the Miscellaneous control bits at the correct address. They had moved to one address before the correct address. This caused the DEP not to process the DMA and IP control words properly leading to incorrect IP processing (The Camera header was not updated). The Miscellaneous control bits should normally read 0x1F6F. 950120_183313 - Vel, Line Depth and Continum data; Line Depth and Continuum show 1 error each (The last pixel?) Vel shows 8087 errors. 950120_191554 - Vel has 8098 errors; Ld and Cntnm have 1 error each Modified genvldc program to call VLCTY instruction only once to process the whole image. 950120_221922 - Vel data; No line depth or continuum data due to IP error. IP error was caused by dovldc calling IP programs at the old addresses. IP wcs load modified to use the old "uncompressed" velocity routine. The new version is 1.7.02. A new version of DEP lev9t27 is also loaded. It allows for longer exposures in the shutter timeout routine. 950121_001638 - Vel still contains a large number of errors Russ found an obscure error in Velocity routine. 950121_021557 - Vel works! ---------------------------------------------------------------------------- 23 Jan 95 doaw: 950123_183906 - No errors dorblr: 950123_184401 - No errors in ROWBIN, 262137 errors in LRBIN 950124_004832 - No Errors in ROWBIN; No errors in LRBIN doirbin: 950123_185617 - No errors in irbin dotwavg: 950123_190019 - No Errors doalign: 950123_190400 - No errors in XALIGN, No Errors in YALIGN docopycb: 950123_190705 - dofftst: 950123_191121 - No errors in FLATFLD, BADPIX or SCALE doiptst: 950123_192716 - No Errors in CROP, LF, RWBIN. 42 errors in VWBIN 1 IP error (8071, 0164, 3450, A188, 0168) Loaded dfc1000 (DEP Flight Code, V1000). 950124_002715 - No Errors in CROP, LF, RWBIN 53 errors in VWBIN No errors in IP dovldc: 950123_230431 - No errors in VLCTY, LDEPTH or CNTNM SCLSUM: 950124_010411 - No Errors in SCLSUM purgetab: 950124_012202 - purgelst: 950124_012802 - ----------------------------------------------------------------------------- 24 Jan 95 dorblr: 950124_223957 - 0 errors in ROWBIN; 262137 errors in LRBIN No IP errors; No TLM VCFC errors 950124_224919 - 0 errors in ROWBIN; 262137 errors in LRBIN T No IP errors; No TLM VCFC errors doiptst: 950124_230608 - No Errors in CROP, LF, RWBIN; 53 errors in VWBIN dorblr: 950124_232315 - 0 Errors in ROWBIN AND LRBIN!! Russ pointed out that LRBIN uses offset registers; Modified rblr.ips to add offset registers to LRBIN Call. Reset IP by ipstrtf. 950125_000828 - No errors in ROWBIN; No errors in LRBIN! Power cycled IP; ipstrtf; 950125_001916 - No errors in ROWBIN or LRBIN. :wq