Backing up Japanese 3.5" disks

Here's the R-Studio image, I think the disk's dead as a doornail. :sorrow:

http://www.segadatabase.net/etc/A.dsk

 

It's actually not as bad as you might think. Ok, there are a ton of bad sectors on the disk, but most of them are towards the end of the disk, in regions which were probably mostly blank anyway. The first sector also read as bad, but we know that the first sector can be readable, otherwise you wouldn't have gotten a directory list to appear at any point, so that tells us that a lot of these bad sectors can probably be "encouraged" to read successfully anyway. What you need to remember about floppy disks, is that they're actually an analog storage medium. What you have is a large surface area which can be magnetically charged and discharged in localized regions. This disk surface itself is divided up into tracks, which are a series of "rings" around the disk, which are linearly spaced from the innermost position to the outermost position of the used disk surface. Data is written to the disk in sectors, which are contained within tracks. Each sector has a header giving positional information, and some CRC data to verify the sector data has been read correctly. While reading a sector, your drive reads the average "charge" level in a particular localized region to determine the value of each bit, and at the end, calculates a CRC and compares it with the recorded value in the sector header to perform error checking.

The thing is this: when you're trying to read the data from that disk, your drive is only attempting to read in a very small, localized area of the disk surface, but in fact, the area that's been charged is larger than the area that's being sampled during the read. There's also the very real fact that the head alignment of your floppy drive is likely to be slightly different to the drive that originally wrote the data, so you may only be sampling on the edge of the encoded region rather than hitting it in the strongest region. I have found that physically shifting and deforming the disk surface, ever so slightly, while the floppy disk is actively being read and retrying bad sector reads, manages to shift the read head into a slightly different portion of the disk surface and successfully read the data. It's not as dangerous as it sounds either. You're bound to deform the disk surface too much at some point and cause it to contact with the read head, but this isn't as bad as you think, because 1.44Mb floppies have a fairly strong, smooth plastic cover over the magnetic surface, so you won't contaminate the read head or damage the data itself, it'll just make a bit of a noise as it slides over the head, and probably give the head a good clean in the process.

For floppy disk recovery, I suggest getting a USB floppy drive, and removing as much of the covering over the top of the drive as you can, while still keeping the drive operating. You can use R-Studio, set the bad sector retry count to something workable, say, 30, then try gently pressing and deforming the disk when the drive has difficulty reading a sector. Keep trying different ways of manipulating the disk surface until the sector is read successfully. You can combine your results from multiple read attempts to slowly "fill in the blanks" with your disk images, and then target any remaining really difficult sectors at the end, banging away against them as much as you need to until you manage to read them. I've managed to successfully recover every single sector from dozens of 1.44MB floppy disks with this method, even when there were hundreds of unreadable sectors on most of them. Once you get the technique right, you should find you can recover the data, with some effort. I can try and make a short video showing this technique if you want.


In terms of what you've managed to copy off this disk so far, there are a lot of files on there that have a "CSO" extension, and most of them seem reasonably intact. Thing is, I have no idea what these files are. That extension doesn't seem to match anything that makes sense for the 1995-1996 time period the files originate from, and I can't see anything identifiable from looking at the data itself. Any idea what they're for? If you're willing to try and rip disk images of the other disks too and post them, I can analyze them and try and work out what's on the disks, and which disks, and which sectors on those disks, are the most critical to recover.

 

Woah, that's quite a post. I think some of this info could be copied and sticked on this (Repair, Restoration and Conservation/Preservation) board. I appreciate all the information, it helps me a lot.

Here's another image from another disk from the same lot of disks.
http://www.segadatabase.net/etc/A2.dsk

The disks (Image A.dsk = A1, Image A2.dsk = A2)

I'm going to try and get some more (complete) images using the methods you described.

This is the drive I've been using btw:

 

I don't know why you think you can read these on a Windows PC....

It's using a custom filesystem and a platform specific tool should be developed to interpret the filesystem.


I was able to tell after downloading one of the sample images you uploaded and having a quick look on Winhex.

You can obviously image them on a Windows PC, but no the PC can't browse the contents of the discs. No wonder the PC want to format them.

Edit: By platform specific, I meant that the tool would need to open and interpret the image in the same way the original machine would deal with the floppy.
Edit2: I can see some form of FAT, which is 2 sectors long (2DD DOS discs have 3 sectors for each copy of the fat) and the directory entry table, here:

Offset 0 1 2 3 4 5 6 7 8 9 A B C D E F

00000A00 46 49 4C 45 4E 41 4D 45 43 53 4F 20 00 00 00 00 FILENAMECSO
00000A10 00 00 00 00 00 00 81 59 52 21 02 00 00 0C 00 00 Â￾YR!
00000A20 52 4F 4F 54 44 49 52 30 43 53 4F 10 00 00 00 00 ROOTDIR0CSO
00000A30 00 00 00 00 00 00 55 5B E5 1E 05 00 00 00 00 00 U[Ã¥
00000A40 54 45 58 54 30 30 30 30 43 53 4F 20 00 00 00 00 TEXT0000CSO
00000A50 00 00 00 00 00 00 25 59 F2 20 0D 00 9E 01 00 00 %Yò ž
00000A60 54 45 58 54 30 30 30 31 43 53 4F 20 00 00 00 00 TEXT0001CSO
00000A70 00 00 00 00 00 00 76 90 5E 1F 0E 00 A0 12 00 00 vÂ￾^
00000A80 54 45 58 54 30 30 30 32 43 53 4F 20 00 00 00 00 TEXT0002CSO
00000A90 00 00 00 00 00 00 10 A1 CD 20 14 00 B0 0B 00 00 ¡Ã￾ °
00000AA0 54 45 58 54 30 30 30 33 43 53 4F 20 00 00 00 00 TEXT0003CSO
00000AB0 00 00 00 00 00 00 79 89 ED 1E 2D 00 58 0E 00 00 y‰Ã­ - X
00000AC0 54 45 58 54 30 30 30 34 43 53 4F 20 00 00 00 00 TEXT0004CSO
00000AD0 00 00 00 00 00 00 B6 8E 82 21 1B 00 68 19 00 00 ¶Ž‚! h
00000AE0 54 45 58 54 30 30 30 35 43 53 4F 20 00 00 00 00 TEXT0005CSO
00000AF0 00 00 00 00 00 00 48 55 FB 1E 3D 00 9E 37 00 00 HUû = ž7
00000B00 54 45 58 54 30 30 30 36 43 53 4F 20 00 00 00 00 TEXT0006CSO
00000B10 00 00 00 00 00 00 07 6D 5B 1F 81 00 9C 16 00 00 m[ Â￾ œ
00000B20 54 45 58 54 30 30 30 37 43 53 4F 20 00 00 00 00 TEXT0007CSO
00000B30 00 00 00 00 00 00 9A AC 8B 21 61 00 42 3F 00 00 šÂ¬‹!a B?
00000B40 54 45 58 54 30 30 30 38 43 53 4F 20 00 00 00 00 TEXT0008CSO
00000B50 00 00 00 00 00 00 EC 99 45 1F DE 00 A0 0E 00 00 ì™E Þ
00000B60 54 45 58 54 30 30 30 39 43 53 4F 20 00 00 00 00 TEXT0009CSO
00000B70 00 00 00 00 00 00 9D 99 68 20 3E 00 64 1D 00 00 Â￾™h > d
00000B80 54 45 58 54 30 30 31 30 43 53 4F 20 00 00 00 00 TEXT0010CSO
00000B90 00 00 00 00 00 00 08 52 97 20 2C 00 2C 16 00 00 R— , ,
00000BA0 54 45 58 54 30 30 31 31 43 53 4F 20 00 00 00 00 TEXT0011CSO
00000BB0 00 00 00 00 00 00 9C 65 8B 1F 1D 01 84 18 00 00 œe‹ „
00000BC0 54 45 58 54 30 30 31 32 43 53 4F 20 00 00 00 00 TEXT0012CSO
00000BD0 00 00 00 00 00 00 E3 50 C4 20 4F 01 1C 1C 00 00 ãPÄ O
00000BE0 54 45 58 54 30 30 31 33 43 53 4F 20 00 00 00 00 TEXT0013CSO
00000BF0 00 00 00 00 00 00 6F 7B 26 20 51 01 16 1D 00 00 o{& Q
00000C00 54 45 58 54 30 30 31 34 43 53 4F 20 00 00 00 00 TEXT0014CSO
00000C10 00 00 00 00 00 00 51 AC 1D 21 14 01 8C 16 00 00 Q¬ ! Œ
00000C20 54 45 58 54 30 30 31 35 43 53 4F 20 00 00 00 00 TEXT0015CSO
00000C30 00 00 00 00 00 00 28 53 25 21 27 01 FC 17 00 00 (S%!' ü
00000C40 54 45 58 54 30 30 31 36 43 53 4F 20 00 00 00 00 TEXT0016CSO
00000C50 00 00 00 00 00 00 82 59 52 21 81 01 06 08 00 00 ‚YR!Â￾


Edit 3: Is that a SEGA SATURN disk with SRAM backup saves ?

 

Well hey don't shoot me, I didn't know what type of filesystem these disks used, that's why I asked here

I don't think it's a Saturn floppy disk drive unit backup, if that's what you mean with your Edit 3 question.

 

x

 

One thing that's standing out really strongly to me after looking at this second disk image you've posted, is that most of the bad sectors in your read attempts seem to occur in exactly the same sectors on each disk. This strongly indicates most of these read errors may be due to different head alignment between your drive and the drive that wrote the data. Do you have access to any other 1.44MB floppy drives, even an internal one on a different machine? It would probably be a good idea to try reading the disks using another drive, you may find a lot of those currently unreadable sectors manage to read correctly on a different drive with slightly different head alignment.

I'm still not sure what the files themselves contain. The filenames and folder structure is very unusual. It does strike me as something that's been automatically generated by something, or is designed to be read on a very specific device. The generic nature of the file and folder names themselves kind of suggests to me it might be a form of test disk, IE, something you'd put into another system to verify it's able to correctly read and work with the filesystem. The file contents themselves aren't making a lot of sense to me at this point. The data definitely has a form of pattern and structure, but I'm not sure how to interpret it. The filenames suggest it's text, which some of the files look like they definitey are, but I'm not sure what character set to use to interpret it, and I can't read japanese either, so I wouldn't have a clue if what I'm looking at is correct.

 

I have just found valid english text which is being encoded into some kind of unicode format, where a leading byte of 0x23 appears before each ASCII byte. See the data starting around 0x59B00 in the first image A.dsk. The file seems like it's probably a string table of sorts, mapping lookup names to actual text. In this case, the strings seem to be listing moves, possibly for a fighting game (IE, "far spunch", "near bpunch", "jump skick", etc). Is someone with more knowledge on Japanese text encoding than me able to identify what character encoding they're using?

 

I've found some more text of interest. On A2.dsk, there's references to the "Sega Saturn". There's also some technical documents, including what appears to be a JAMMA pinout. Could this deveopment be arcade-related by any chance? There were also references to the PC-98 on this disk, but I'm not convinced that indicates that a PC-98 system wrote these disks, although it is a possibility.

 

x

 

They came in a lot with other Maajan Doukyuusei Special stuff, so it's likely they contain related stuff.