perfect (SPARCstation IPX)
Summary
TODO
Notes
TODO
Hardware
| Make | Sun |
| Year | 1991 |
| Model | SPARCstation IPX |
| Chassis | Lunchbox |
| Processor | Weitek SPARC POWER μP WTL 8601 |
| Memory | 16MB HB56A433SU-8A 72-pin SIMM |
| Ports | SCSI HD50 |
| AUI DB-15 | |
| 2x Serial Mini-Din 8 | |
| 13W3 | |
| Sun I/O Mini-DIN 8 | |
| Audio Mini-DIN 8 | |
| Graphics | Sun Turbo GX (cgsix) |
| Storage | ZuluSCSI Compact with 128GB MicroSD Card |
| Int. Peripherals | AMD Lance Ethernet |
| Turbo GX (cgsix) Framebuffer | |
| Sun Quad Ethernet (SQEC, 501-2062) SBus card | |
| Single-Ended Fast/Wide SCSI (SWIS, 370-1703) SBus card | |
| Dimensions | |
| Length/Depth | 26.5 cm |
| Width | 24.5 cm |
| Height/Thickness | 11.7 cm |
| Weight | 5.45 kg (12 lbs) |
Software
| Operating System | |
| Unique applications |
Links
- OpenBSD/sparc
- Lots of good info on supported hardware and such.
- Sbus 32MB Expansion Board
- Thread on getting the SS2 SBus memory expansion board fitted in an IPX. (I don't have one, yet.)
Log
A little lunchbox machine
Received a very well packed SPARCstation IPX shipped from Massachusetts. No markings to indicate its pedigree, but a decent condition unit overall. I bought the unit as "untested" from a lot of 2.
I opened the machine and it was pretty full of dust. The first surprise was a Weitek SPARC POWER μP CPU. The only thing cooler than that would have been an SBus memory expansion card, but I felt pretty awesome to have this nice upgrade.
After dismantling each half and blowing all the dust out I went to test it, but the machine wouldn't power on. The power LED briefly blinked when flipping the switch, and a quick chirp was heard.
This seems to indicate a power supply problem according to glitchworks so I took the supply back apart and inspected it more closely. Sure enough the secondary side capacitors had been leaking. There was some black corrosion on the traces.
Fortunately it appears the machine had been stored on its side so none of the goop could drip onto the motherboard. Some migrated towards the primary, but no corrosive effects there. So I'm going to need to recap the power supply.
In the mean time I have a few other things to gather for this machine. I need an AUI to 10BaseT transceiver, possibly a SCSI drive replacement (considering BlueSCSI), and some additional memory (it came with a single 16MB stick). I'll also need to mod a replacement battery onto the clock/NVRAM chip.
PSU Rebuild and testing
I replaced the power supply capacitors following the aforementioned guide. Even after cleaning, some traces were left with black corrosion. I tried testing the power supply unloaded, but that wasn't any good. I could see the 5V rail come up, but not reach 5V before relays clicked and it reset. The 12V rail only reached a few volts. So I went ahead and hooked the power supply up to the motherboard sans drives. The IPX booted. As expected the NVRAM battery has failed, and will need replacement.
Booting the complete machine yielded a SunOS 4.1.2 install called 'flower'. Mounting an external disk failed and dropped me into a superuser prompt. I executed /etc/rc.single to remount filesystems then edited /etc/fstab to comment out the external disk.
The next boot failed because the NIS domain was unreachable. From the ok prompt I did boot disk -s to get back into single user mode, and moved the /etc/defaultdomain file out of the way to disable NIS. I also reset the passwords for root, and guest.
After logging in and poking around I found a pretty bare system. There was an IBM MOTD indicating this was a IPT (Internet Pass-Thru???) Tester Workstation, and hostnames indicate it was from IBM's Fishkill, NY site. However it seems everything important was mounted remotely.
The machine ran fine for a couple hours, but it's pretty noisy. Cutting out the SCSI disk has become a priority. After that I can see about replacing the fan, but things might get too toasty once an SBUS card or two is added.
I also booted once with the AUI transceiver hooked up, but I didn't want to modify too much on the original install to see if I could pass traffic so I just know it sees a link right now.
New BootROM
I grabbed the latest (V2.9) IPX BootROM from https://archive.org/details/sunshack_bootroms. Prices weren't great on 27C020 EPROMs so I just opted to reprogram mine. First time using my cheap UV eraser and it worked (twice), but the bulb flashes a bit much. I wrote the ROM once with the TL-866II plugged into my front USB 3.0 port, but it hung on verification.
minipro -p "AM27C020@DIP32" -w IPX-29.ROM
Found TL866II+ 04.2.128 (0x280) VPP=13V, VDD=6.5V, VCC=5V, Pulse=100us Chip ID OK: 0x0197 Writing Code... 132.56Sec OK Reading Code... 8%^C
Switched to one of my USB 2.0 powered hub ports after another erase cycle and everything went well.
Found TL866II+ 04.2.128 (0x280) VPP=13V, VDD=6.5V, VCC=5V, Pulse=100us Chip ID OK: 0x0197 Writing Code... 132.30Sec OK Reading Code... 2.09Sec OK Verification OK
Machine boots. Says 2.9 now. Memory check (no NVRAM battery yet) seemed faster. SunOS 4.1.2 started with no issue.
More RAM
New SIMMs installed without issue. They don't look as loaded as the original 16 MB SIMM (both sides, and an oversize PCB to accommodate extra chips). I'll just have to live with single side loaded SIMMs that I'll rarely look at though.
Now the machine has maxed out RAM unless I can get my hands on one of those SBUS memory expansion boards.
A little adjustment
I went to use the IPX to test a Sun Multipack, but the machine wouldn't turn on. Several tries, cursory inspection, but no go.
I found a this post in a thread on recapping power supplies that mentioned overvoltage protection. Same symptom of immediate power off.
I was able to reach in and adjust the +5V pot with a screwdriver without opening the supply. I gave it about a 1/8 turn counter-clockwise, plugged everything in, and it powered right up. 5V line was showing around 5.066V at the hard drive connector so I left it at that. For reference here are Sun's voltage limits from the Field Service Manual
| Rail | With Hard Drive | Without Hard Drive |
|---|---|---|
| +5V | 4.90 to 5.10 | - |
| +12V | 11.16 to 12.84 | - |
| -12V | -11.6 to -13.20 | -10.2 to -13.8 |
48T02 Repair, NVRAM programming, new fan, and drive solutions
SRAM Clock Chip 48T02
I built a handful of M48T02 battery/oscillator replacements with the glitchwrks PCB design. Instead of the recommended cell holder I went for these cheaper ones on amazon. These have a different footprint, but can still be soldered to the through-hole pads if you cut the positioning pips off the bottom.
Then I followed the glitchwrks guide for preparing the original 48T02. I cut a bit high, but eventually managed to crack the "CAPHAT" module as described. Then installed one of the replacements on the protruding pins.
New Fan
While I was in the machine I swapped the stock fan for a Noctua NF-A6x25 FLX. The fan frame had to be modified with a notch to position the wires correctly. I just used some diagonal cutters, an exacto knife, and a small round file to make a new notch.
I did not have to modify the fan wiring or connector. I just routed it around the power supply cabling so it was out of the way. Even though it's a 3-pin connector it's easy to install on the existing 2-pin header.
At stock speeds the fan is still noticeable, but a marked improvement over the original. I didn't want to slow it down anymore because I intend to run with two SBus cards. For comparison: the Noctua moves 17 Cubic Feet per Minute, while the stock MMF-06D12DL is supposed to push 21 CFM.
New plan for a drive replacement
I also found out the BlueSCSI isn't compatible with Sun workstations. I'd have known before if I bothered to look at the wiki note on supported systems or searched for issues. Anyway, I can hold on to the two I built for some future Mac II acquisitions. I did test the BlueSCSI anyway, but it did not enumerate.
My new plan (that I researched, but couldn't confirm will work) is to use one of my Seagate Savvio ST973401LC 73GB 2.5" drives with a SCA to 50-pin IDC adapter. I tested one inside the Sun MultiPack enclosure and was able to enumerate it with probe-scsi, but I didn't have the cables to also attach a SCSI cd-rom and do a test install. I tried to set up a jumpstart network install, but something went wrong with bootparams and I gave up troubleshooting.
Installing Solaris 7
With the 50-pin to SCA SCSI adapter in hand. I mounted the 73GB 2.5" Ultra-320 drive to a filler drive plate. Cabling the drive was a bit finicky. I had to resort to a lower profile molex connector. The stock power supply cable also had to be extended. The new ribbon cable position is offset about 5mm from a normal drive, but it doesn't cause any issue.
I'm installing Solaris 7 from CD and so far the biggest trouble has been CD read issues (aging AppleCD 600e). The installer had no issue dealing with 70GB and even large partitions. I made a 3GB root which should hopefully avoid any 4GB limitation (the 3GB choice was wrong, there is actually a 2GB limit on bootable root), but my home partition is 30GB so I expect to be able to utilize the large disk just fine.
The tiny drive gets surprisingly hot. I was doing the install with the case folded open so I opted to block the power supply's main vents so at least a little air was being pulled past the drive. Now it's just lukewarm.
SCSI drive solutions
After some runs with the full case assembled, I've found the 2.5" drive produces too much heat for the IPX's cooling system. It is not super noticeable in the fan output, but I did a finger check on the drive after powering down, and it's just not staying cool enough.
I'm not sure the best solution. A few ideas to consider in the future:
- Block part of the power supply's holes to force more air to be drawn from the drive area. (slightly compromises Sbus/CPU cooling).
- Get a different drive, but the options seem to be old and dying, or newer and not dying yet. I'm doubtful of finding anything reasonably quiet or cool.
- No internal drives, get a decent compact external enclosure.
- Go for one of the more expensive SD/Flash solutions like SCSI2SD.
ZuluSCSI, A quieter and cooler IPX
I got a ZuluSCSI Compact Kit while I was at VCFSW. I specifically asked for something that would work with the IPX, and the vendor (Rabbit Hole Computing) said he had one set up that way. It was $40 for the kit version ($50 pre-assembled, $60 for non-compact). I had to solder some connectors, jumper headers, and capacitors.
I decided to clone the 2.5" 73GB drive with initiator mode. I grabbed a molex splitter, connected the drive and ZuluSCSI to the same SCSI cable, and to the IPX's drive power cable. After the drive started ticking, and the LED flashing, I realized this is a long process, and I probably shouldn't have set it up on my desk. I carefully transferred the setup to a temporary platform.
With the jumper enabling initiator mode, the ZuluSCSI just goes to town imaging any drives on the bus. It provides a serial output with live status through the micro usb port. Speeds are not that impressive, around 2.88 MB/s. Hopefully I can get a little closer to SCSI-1's 5MB/s in target mode.
I left it to run overnight. One weird thing is the status output of ZuluSCSI only counted up to 30%, so it did 30% + 30% + 30% + 10%. I made a backup of the drive off the SD card.
The mounting plate I previously used for 2.5" drives had to be modified a bit, but still performs the original function. The ZuluSCSI seemed to run fine with just bus power, but I ended up attaching an adapter just so the power supply cable wouldn't be loose.
