Reverse Engineering the Janome Memory Cards

Looking at the binary read out again, I notice a couple of issues:

1. Every 4 lines the text repeats itself - which is a 64 byte offset or that the A6 pin is not toggling

2. Every 8 bytes are repeating (half of the line), A3 pin isn't toggling

3. Sometimes the repeat parts aren't quite the same which indicates maybe a signal integrity issue (or power issue?)


Though I feel like I'm getting close. This memory card has the alphabet on there and this is close to that:


Found some incorrect pinouts in the firmware :(
Fixed that and still having issues but closer to an alphabet!

Every other letter is 0xF1.  Hmm .. pyembroidery makes is seem like I should be looking for 0x80 to indicate a control byte coming.

With the fixed pinout, I added A-1 back in and see saw the wrong deltas but the start of the file REALLY looks like a header: "832 DATA EXP L V1.0 R1.0"
I'll keep address A-1 in.

So sew files only give deltas of movements, not absolute values so if my starting point or one byte is off, it messes up the whole image. With this, I decided to just plot the deltas, expecting on my card there to be a lot of back-and-forth movements.

First I tested on the known good file (downloaded from the internet):

Yep - I expected it to be a circle around the origin.


And my readout:
that makes no sense. Something is wrong. Which sucks because the file opens with:
"832 DATA EXP L V" which looks like it could be a header. Then a little further along, there's this in a sea of non-ascii characters:
which looks very much like a version string. So I was hoping it would be right.

After some trial and error, I stopped using the A-1 address pin (which is suppose to be the LSB) and got this
That looks good - because this card contains text, it makes since that there's only side to side and up and down motions. But converting to positions yield:

Flywired the memory card to 0.1" headers then jumpered them to an Arduino Mega.

Here's the start of it:

I'm using pyembroidery to parse the results. Looking online, everywhere said these should be *.sew files and thankfully pyembroidery has a parser so I can see how it works. 

To test to make sure it works, Janome has some free design you can download and here's "Fairy and Ladybug": https://www.janome.com/inspire/Embroidery/fairy-and-ladybug/

And I get this on my readout:
Lovely. But it's far too few points. Looking at the source code of the parser, it seeks the file to a magic number 0x1D78. Looking at the file in a hex editor, it's all zeros until that address. On my readout, it looks like the data starts at 0x3000 so I set it to seek there. Interesting point, my filler bytes are all 0xFF not 0x00


More points but not right. I'm expecting something like this (Memory 101 column):

hmm ...