OrigamiSat-2 jpg creating on LINUX
process
I asked ChatGPT why it's not possible to convert this HEX data of the second
image from the bottom to a JPEG image. Then, the following response was received.
By performing the steps in the latter half on this my article, 100% JPEG images
could be successfully generated.
----------------------------------------------------------------
This HEX data has a structure that is very characteristic of a JPEG file.
In fact, it contains:
* Start marker: "FFD8"
* JFIF header: "FFE0...4A464946"
* DQT: "FFDB"
* SOF0: "FFC0"
* Huffman Table: "FFC4"
* Start of Scan: "FFDA"
* End marker: "FFD9"
So this is not “completely corrupted data.” However, there are
some problematic areas.
### Problem 1: Abnormal Huffman Table
This section: "FFC4001F00000105010101010117BB01..."
In a JPEG "FFC4" (Huffman Table),
the length-definition portion must contain properly structured values.
However: "...01010117BB01..."
The "17BB" part is extremely unnatural.
Normally, this area would look something like:
"0101010101010101" with only small values.
The sudden appearance of "BB" is typical of:
* a missing byte
* bit misalignment
* HEX contamination
* corrupted copying
### Problem 2: The Quantization Table Also Looks Suspicious
This section: "FFDB00430006040506050406060506070706080A100A..."
looks normal at first glance, but the latter part differs slightly
from standard JPEG tables.
This suggests that:
* part of the data was overwritten
* corruption occurred during transmission
* data may have been lost during log storage
### Problem 3: Unnatural Values in the Middle
This section: "...E6E70E05E8E9EA..."
is highly unusual for a JPEG Huffman definition.
Normally it would continue like:
"E7E8E9EA" in sequence.
This strongly suggests that "0E05" was accidentally inserted
into the stream.
### Conclusion
This JPEG:
* contains a valid header
* mostly preserves the JPEG structure
However, because the Huffman Table is corrupted, standard JPEG
decoders will likely report errors such as:
* “Invalid JPEG”
* “Corrupted file”
* “SOS marker error”
* “Huffman table error”
### What Most Likely Happened
There is a very high probability that your data was extracted
from something like:
* a serial log
* communication log
* UART dump
* memory dump
* partially copied data
and that a few bytes were either lost or inserted.
In particular:
* "17BB"
* "0E05"
* "69DD"
are especially suspicious. There is still a possibility that
the file can be repaired.
This JPEG appears to have:
* a mostly intact structure
* valid size information
* valid SOF/SOS markers
Therefore, recovery may be possible by:
* replacing the Huffman Table with a standard Huffman Table
* removing corrupted bytes
Because this is a Baseline DCT JPEG, repair using a standard Huffman Table
is especially effective.
In conclusion, the image data itself (everything after SOS) is very likely
still intact, while only the DHT section appears corrupted. Therefore, the
most reliable way to recover the image is to completely replace the entire
DHT with a standard Huffman Table.
----------------------------------------------------------------
Manually explaining all eight of the images above would be very time-consuming,
so I've semi-automated the process using a programming language based on Ruby.
The number of iterations (38c) in the rb file used to generate test1.hex should
be adjusted to match the number of rows in the CSV file downloaded from SatNOGS.
Note that after generating test2.hex, the last line containing FFD9 is generally
short, so you will need to manually edit that part. The final file, test3.hex,
has a JPEG structure with a single row of "FFD8...FFD9".
https://www.ne.jp/asahi/hamradio/je9pel/cas5atst.htm
Sample satnogs.csv: total_lines 908(d)=38c(h)
Image No.44(h) on 8 May 2026 by JH4XSY
test0.sh ... Shell script using Ruby to output ext044.rb
test1.hex ... Extract the right side including the frame number of each row.
test2.hex ... Extract the data body from each row, excluding the frame number.
Edit last line in it ending with FFD9 manually.
test3.hex ... Concatenate each line in test2.hex into one single line.
The result starts with FFD8 and ends with FFD9.
Edit it according to ChatGPT when JPEG image is not generated.
[ext044.rb]
> for i in 0..0x38c
> printf("grep 020003%04X satnogs.csv | head -1 | cut -c 77- >> test1.hex\n", i)
> end
[on Terminal]
$ ruby ext044.rb > test0.sh
$ sh test0.sh
$ cut -c 5-384 test1.hex > test2.hex
> Edit last line in test2.hex ending with FFD9.
$ tr -d '\n' < test2.hex > test3.hex
> Edit test3.hex according to ChatGPT.
$ cat test3.hex | xxd -r -p > test3.jpg
----------------------------------------------------------------
In summary, by modifying the HEX data as suggested by ChatGPT, I,JE9PEL
was able to convert only a small portion of the original image (top left)
into a JPEG image. (Still in the trial-and-error phase, 10 May 2026)
60509or0.csv
60509or1.hex
60509or2.hex
60509or3.hex
$ cat 60509or3.hex | xxd -r -p > 60509or3.jpg
----------------------------------------------------------------
2026-05-22 06:05 UTC in SatNOGS,
Even for the same sequence number, there were many different data sets,
so I displayed four rows of data for each sequence number using Ruby,
and then manually selected only one reliable row as the data for that
sequence number. At the end of the process, I used Google Gemini to modify
the Huffman Table in JPEG format. I was able to obtain the following image
after working on it all day since this morning, (Cf.JA3TDW)
[ext044.rb]
> for i in 0..0x50e
> printf("grep 020005%04X 60523ori.csv | head -4 | cut -c 77- >> 60523or1.hex\n", i)
> end
[on Terminal]
$ ruby ext044.rb > 60523or0.sh
$ sh 60523or0.sh
> Extract only one row reliable sequence number data.
$ cut -c 5-384 60523or1.hex > 60523or2.hex
> Edit last line in 60523or2.hex ending with FFD9.
$ tr -d '\n' < 60523or2.hex > 60523or3.hex
> Edit 60523or3.hex according to ChatGPT or Gemini.
$ cat 60523or3.hex | xxd -r -p > 60523or3.jpg
By using "cut -c 5-384" to remove the last two bytes (CRC Check) from
the end of each line of the HEX data, all the error messages that had
been occurring disappeared, and I was able to obtain 100% JPEG images.
[Summary]
> First, extract the necessary parts from Satnogs.csv using Geany.
> Save as 60730ori.csv.
[ext044.rb]
for i in 0..0x016
printf("grep 020005%04X 60730ori.csv | head -1 | cut -c 77- >> 60730or1.hex\n", i)
end
[on Terminal]
$ cd origami2
$ ruby ext044.rb > 60730or0.sh
$ sh 60730or0.sh
$ cut -c 5-384 60730or1.hex > 60730or2.hex
> Edit last line in 60523or2.hex ending with FFD9.
$ tr -d '\n' < 60730or2.hex > 60730or3.hex
$ cat 60730or3.hex | xxd -r -p > 60730or3.jpg
| | | |
Before deployment Membrane antenna
22 May 2026 | After deployment Membrane antenna
23 May 2026 | Earth and Membrane antenna
27 May 2026 | Membrane antenna
30 Jul 2026 |
Back to Top
Back to Home Page