Hello hackers, unfortunately the first version of this mail article didn't made its way onto the list, either because of crude formatting or because of using a news2mail-gateway or both. Now I removed the attachement, please get it from: http://xw15.de/cbmstuff/Universal-Floppy-Dumper-20110807.zip Read on for my first mail, Womo -=-=-=-=-=-=-=-=-=--=-=-=-=- Hello hackers, this article is about my successful repair attempts to resurrect at least the electronic parts of a batch of SFD-1001 floppies, a 8250LP and a 8296D. It will be followed by two other articles where I will call for your help regarding the mechanisms and specific ROM contents. In April this year Tommy Winkler released a first version of a firmware for Nate Lawson's ZoomFloppy (USB connected backend for OpenCBM) with support for Commodore IEEE disk drives. Because we always planned to have support for IEEE drives within the ZoomFloppy and already designed the hardware to support these, I collected some SFD-1001 drives and such alikes from eBay for two years or so. Most of them were not working or in bad condition otherwise. For example the 8250LP is missing its top cover. From a series of auctions where a german sold the bequest of his father I got some five or so additional SFD-1001 mainboards and an empty case in mint condition with power supply and metal cage. With this I started my repair attempts. Two drives very relatively easy repaired by only checking the blink codes and swapping some chips more or less systematically. Since one of these didn't produce sane results, when tested with the ZoomFloppy I changed the mainboard with one of the spares and voila, I had two working drives. I then started with checking all the socketed chips from all SFDs and the 8250LP and found one defective 6532 RIOT and two defective 6530 RRIOTs aka 901885-04. A 6530 from one of the drives was missing, missing 6532 chips could be replaced with chip that I pulled from the spare mainboards. A third drive could be assembled from the spare case with the known good power supply. One mechanism needed a thorough head cleaning procedure: 1) getting a tissue, folding it and then placing it between two R/W heads with the door closed, 2) saturating the tissue with isopropyl alcohol, repeating the saturation every five minutes and letting it solve all the dirt at the heads for 30 minutes or so, 3) mechanically cleaning and polishing the heads with a standard head cleaning diskette and the recommended procedure. For a fourth SFD-1001 drive one of the remaining power supplies could be repaired as well as the mainboard by getting a last working 6530 out of the 8296D. Unfortunately I ran out of working mechanisms. I did not make systemtic checks of all the remaining mechanisms from the SFD, 8250LP and 8296D since I decided to first work for a solution on another problem... Where to get new 6530 chips? ============================ Answer: Nearly nowhere. As I found out from different discussions it seems to be a problem for nearly a decade or so to get any replacement parts for the 6530, especially the mask programmed version for the SFD and similar drives. As Nicolas Welte pointed out in several such discussions, the 6530 not only has a mask programmed ROM, but also a mask programmed logic. This is something that could be compared with ASICs from manufacturing principles as of today. In the SFDs and 8250LPs Commodore did not make use of an own customised 6530 with a dedicated logic and ROM mask, but reused chip from the older drives like the 8050. They simply ignored the 6530 internal masked ROM and replaced it with an EPROM with new contents with the help of a little adapter board. So this means that the programmed logic gate masks for the 6530 were all the same or at least nearly identical to be reused in other floppy types with small external adaptions. In all the discussions from above Nicolas and perhaps some other people also often mentioned that it should be possible to replace a 6530 with a 6532, an additional ROM chip and some logic to recreate the internal programmed logic gate mask. One needs "only" to find out about the concrete implementation of that programmed logic. So the way to go for getting 6530 replacements was defined. How to create a 6530 replacement from a 6532? ============================================= Meanwhile, perhaps initiated by these discussions form the past some others already showed that such an adapter is infact possible to create. French pinball enthusiasts created an "adapteur" to replace defective 6530 chips in some Gottlieb Sys1 and Sys80 soundboards for pinball machines. A german company sounds a very similar board named "MIOT-Adapter". And finally the Micro-KIM replica of the MOS Technologies KIM-1 single board computer was created by Briel Computers and it came with an integrated replacement circuitry for the two 6530s of the KIM-1. Last but not least there is probably the mother of all these existing 6530 replacement circuits. Ruud Baltissen's "Build your own KIM-1" page where he described all the oddities about how the create a replacement for the 6530 by using a 6532, a ROM and some "glue" logic. This showed to me that such a replacement must be doable also for the Commodore floppies. I also was willing to find a way to get a 100% compatible solution for the "fourth" problem described by Ruud. I looked out for a way so that the additional 6532 IRQ generation method which depends on the behavior of PA7 cannot be enabled anymore. But at first the "fifth" difference, "the last and major one" needed to be solved, "the way the registers ..." of the 6530 "... are selected" in the Commodore floppy disk drives, namely the SFD-1001, 8250LP, 8296D, 8250, 8050, 4040, 3040, 2040. Analysing the CBM floppies' memory maps ======================================= From old Funet.FI files (now on zimmers.net) I saw that at least three people already did investigate the memory map of the floppy disk controller of different CBM floppies. And they wrote some programs for this purpose. Olaf 'Rhialto' Seibert described the memory maps both for the bus controller (BC) and the disk controller (DC) as well of a 8250 floppy disk drive. He also described a little routine executed as job code to copy one memory page from the DC onto the shared memory between the DC and BC. Afterwards standard M-R commands could be used to read out the contents from this buffer. William Levak used a similar routine and composed a memory map for the 4040 disk drive based on readout results. And then there is Andre Fachat whose program is not much different to the one from William, but also provides one which can be used to download the BC's ROM. Except for Williams memory map analysis for the 4040 drive, no information could be found in the net for all the other drives. Therefore I investigated memory maps on my own for the SFD-1001 and 8250LP. I also wrote my own tool which is able to a) read out the whole adressable 64KiB map of the DC (although actually only unmirrored 16KiB adress space are adressable by the CPU), b) read out the whole address 64KiB map of the BC and c) it does /not/ leave the floppy in stuck condition so that it needs to be reset afterwards. Please find a listing of the floppy side code as well as a little bash shell script which uses OpenCBM cbmctrl calls to control the uploaded floppy routine attached to this article. The shell script should be easily convertable to standard Basic commands since OpenCBM cbmctrl is nothing else than a simulation of standard IEC/IEEE commands. Creating a 6530 replacement circuit =================================== With additional read out address maps from all my working SFD drives and the 8250LP (only the electronics part is working yet) I recognised that the 6530 internal decoding depends on /CS2, RS0, A6 and A7 only. With the two existing Chip Select inputs from a 6532, the required glue logic could be reduced to two NOT gates. One is needed to enable the extra ROM chip whenever PHI2 is high (same as on the Commodore adapter boards), the other is needed to invert A6 is 6532-CS1 input while CS2 is connected with 6532-CS2. To have a more clean design PHI2 should be NANDed with R/W which is then used as ROM enable. RS0 is used as ROM output enable while A7 is connected to 6532-RS. Above I wrote that I wanted to have Ruud's "fourth" difference between the 6530 and 6532 be solved. This objective was reached by connecting 6532-A4 to a fixed high level. That way the described IRQ method cannot be enabled anymore (check the datasheets). A4 from the 6530 chip socket was connected with 6532-A6 instead. The latter effectively halved the 6532's RAM size from 128 Bytes to 64 Bytes. In the end this measure improved overall compatibility of the 6530 replacement in two points: supported IRQ methods and RAM size. Currently the chosen Flash ROM is 32KiB in size. Smaller ROMs are not available for reasonable prices. Therefore I decided to pack as much as possible different FDC drive ROMs into it and make all the banks selectable by DIP switches. With this it should be possible to support all drive DOS ROM versions of all Commodore dual CPU drives with only one final circuit and firmware. Currently I don't know how many such ROMs do exist, therefore one of the follow-up articles will follow. Supporting SpeeDOS 2K FDC ROMs ============================== Since nearly all TTL chips contain more gates than I needed for the circuit I tried out, if the 8250/8250LP EDOTRONIK SpeeDOS system could be supported by an alternative configuration option. And it was possible by using a 4-NAND gates chip. Switching from 1KiB FDC-ROM support to 2KiB support means that A10 DIP switch configuration option is lost and only 16 different ROM images can be supported with a 29C256 chip (or 28C256, 27C256, ...). When a 27C512 EPROM is used, maybe 32 different 2KiB ROM images could be used (untested). Since the 1KiB/2KiB option could be selected with a pair of jumpers even mixed 1KiB and 2KiB ROM images could be stored within the same ROM chip, so I hope that a '256 ROm chip is enough for all existing ROM image files. Current state of the development ================================ In the last months a hand wired prototype was developed, debugged -- Commodore mixed up PHI2 and R/W in the schematic of the 6530 adapter board which costed me at least eight weekends of debugging with a mixed signal o'scope and logic analyser -- then successively extended by the SpeeDOS option and tested again. Several other circuitry alternatives were constructed, optimised and tested until the current design was chose which only needs one TTL chip for the glue logic. No GAL, no difficult to find exotic components, just standard and easy to get parts. Currently only a SFD-1001 drive was tested thoroughly by the replacement adapter and only from an electronics point of view. I don't know of any good testing programs that could check, if there remain any compatibility problems from the replacement. I have plans for creating a profesionally manufactured PCB and already created a layout that fits into a 8250LP and a SFD-1001 as well. I don't know, if the same design would also fit into the 8250, 8050, 2040, 3040, 4040 and perhaps some other suitable drives. What else? ========== Before declaring that little project as being finished I would like to discuss some remain topics with you. The first one is about available ROM image files about all the different FDC-ROM versions that are documented by Commodore, but not available up to date (e.g. the FDC-ROM from the 2040's DOS-1). Please see my follow-up article on this topic which I will write up the next weekend or so. Next I would like to discuss options about repairing the mechanisms in the 8250LP and SFD-1001 disk drives or perhaps options about replacing the mechanism with a drive from a different manufacturer. To me it would not make any sense to have a 6530 replacement, when I'm still not able to repair the mechnisms. Maybe, if there is some interest in it and if we find solutions for all the other top-10 defects for all the old CBM disk drives, then a final 6530 replacement adapter could by available by the end of the years. But this would require also that someone else finds a source for 6532 chips. If the 6530 chips are impossible to source nowadys, the 6532 ones are at least very hard to find in new and factory packaged condition. Womo PS: If someone else also collected defective CBM drives and was successful in repairing some of them, maybe there is an option to exchange some parts. Currently I'm searching for at least five 100 TPI mechanisms used in the SFD-1001 and 8250LP drives. I would be willing to give away my original and working 6530 chips and perhaps some other electronic parts since now I know how to replace them and I already got some spare 6532 chips years ago. -- ------ to obtain more infos about me, look up the page ------ ------ http://www.wmsr.de | wm (at) wmsr (dot) de ------ Message was sent through the cbm-hackers mailing listReceived on 2011-08-09 06:36:17
Archive generated by hypermail 2.2.0.