VOGONS


First post, by athlon-power

User metadata
Rank Member
Rank
Member

So, a while back, I decided to try to force a motherboard from an IBM PC 350 Type 6581 with a Intel 486 DX2-66 into a Packard Bell PB400 case, and while it "worked," it had a few issues, which were pointed out to me in the original post. The case underwent extreme modification, and so did the motherboard (well, the riser card did at least), and it also happened to look ugly as sin. Skip forward a few months, and I've worked on it quite a bit and fixed some of the original glaring issues that were there, to try to attempt to make things a little better.

First major fix: the riser card.

In order to fit the riser card from the IBM system into the Packard Bell case, I had to cut 1/3 of it off. Yep, my instructor at my technical school took a diamond grinder wheel to it. Anyways, this left exposed traces and rough edges, which I immediately covered in electrical tape and left there. This, among an uneven cut, made it look genuinely awful, and I also had bent parts of the retaining arm down where it was cut just too short to hold it in place, which also looked awful.

I took the electrical tape off, and sanded down the board until it was as level as I could get it using sandpaper and subsequently inhaling what was likely an incredibly unsafe amount of PCB dust because I did it indoors and didn't wear a mask of any kind like an idiot, and glued over the exposed, albeit rounded and flat edges, with contact glue, and let it dry. It was the closest thing to PCB sealant I could get, so I had to work with what I had. I then ripped off the bent portions of the retaining arm, and took two heatsink clips that were lying around and manually bent them back and forth until they were the right length, and then bent them into 'v's, and put them through the holes in the retaining arm where they then gripped the riser card, preventing it from moving it back and forth entirely. I put masking tape on the clips themselves where they contacted the PCB to prevent shorts. This part took the longest, with me spending a week trying to find ways to prevent the riser from moving both left and right, using multiple ideas and methods, before I finally asked my mom and she suggested I put clips on it.

20200622_173839.jpg
Filename
20200622_173839.jpg
File size
476.51 KiB
Views
375 views
File license
GPL-2.0-or-later

Second fix: Making the ISA cards not bend when installed. Now, while I didn't actually fix the base problem, I was able to alleviate it by removing the SCSI controller and using an IDE CD-ROM drive instead. The idea here was that now that the SCSI controller was no longer there, I could put the sound card of my choice in the top slot and not worry about it bending. I feel like the sanding down of the riser card helped for some reason as well, but I'm not sure why.

20200622_173852.jpg
Filename
20200622_173852.jpg
File size
473.13 KiB
Views
375 views
File license
GPL-2.0-or-later

Third and fourth fixes:

When I was modifying the case to make things fit, I realized that the 3.5" HDD bay was right over where the processor socket was on the IBM motherboard, which wasn't a problem until I realized that the 486 I was using had a fairly tall glued-on heatsink attached to it, which caused it to not fit when it was installed. My instructor cut off all of the 3.5" bay, and now it fits perfectly, but of course, I now have no 3.5" HDD bay. What I ended up doing is using a garbage 3D printed 5.25" to 3.5" adapter which looked utterly awful, which I later replaced with a metal 5.25" to 3.5" adapter that originally went to a 1GB Seagate HDD that I got out of a 486 server that had been left outside for months, needless to say the HDD and most of the components inside were dead. Because of the way the adapter fit into the case, it actually still looked pretty bad, so I changed it out with a two-piece solution (one bracket on each side), which looks much better now.

The IDE CD drive I decided to use (or, the only free beige one I had at the time that worked) was a 48x drive- a massive upgrade from the 4x drive, but the problem was that it had "48x MAX," written on the drive's front, which made it obvious that the drive was far outside of that computer's time. A magic eraser later, and the problem no longer existed.

20200622_173914.jpg
Filename
20200622_173914.jpg
File size
432.63 KiB
Views
375 views
File license
GPL-2.0-or-later

While this isn't perfect, I think it's far better than it was before. This motherboard likely has bad capacitors as I completely cleaned it and it still:

A) Will refuse to work with cache over 256KB, saying the upper chips are bad, when they are known good chips. Does not happen with 128KB of L2.

B) Has an incredibly unstable PS/2 mouse port, which sometimes works, and sometimes doesn't. Have tested multiple mice, still only partial dice. Sometimes it will lock up the whole computer on POST if a mouse is plugged in, sometimes it will give a pointing device error in POST, sometimes it will work just fine at first but in games like DOOM glitch out all over the place after 5 minutes of play, and sometimes it works just fine.

20200622_173809.jpg
Filename
20200622_173809.jpg
File size
531.51 KiB
Views
375 views
File license
GPL-2.0-or-later

This photo of the outside of it I took a few weeks ago, but everything else other than some adjustments on the CD/floppy drives to make them more centered are the same:

20200421_193419.jpg
Filename
20200421_193419.jpg
File size
402.55 KiB
Views
375 views
File license
GPL-2.0-or-later

Reply 2 of 14, by mkarcher

User metadata
Rank Newbie
Rank
Newbie
pii_legacy wrote on 2020-06-22, 22:30:

I wonder maybe your mouse port needs diodes to prevent backfeed? (This might not be the issue at all.)

I love that case / setup.

Both the clock and the data line on the PS/2 mouse port are bidirectional. I don't see how to put diodes into these lines without disturbing the communication. I diode on the +5V line might help in theory, but I don't see where any energy should come from that could be backfed through this line, especially as the +5V line has a lot of capacitance, so you need a lot of energy to cause noticable voltage changes on +5V. Furthermore, any sensible mainboards has an interference suppression inductor on the power line, and L/C lowpass filters on the clock and data line for the PS/2 mouse. If these suppression filters are replaced by 0-henry-indictors (straight wires) and 0-picofarad-caps (just no cap placed on the board) to save costs, adding a proper filter might indeed help.

Reply 3 of 14, by pii_legacy

User metadata
Rank Newbie
Rank
Newbie
mkarcher wrote on 2020-06-22, 22:43:

...any sensible mainboards has an interference suppression inductor on the power line, and L/C lowpass filters on the clock and data line for the PS/2 mouse. If these suppression filters are replaced by 0-henry-indictors (straight wires) and 0-picofarad-caps (just no cap placed on the board) to save costs, adding a proper filter might indeed help.

Yeah, this is what i meant, i apologize. I think a lot of older serial-mouse only boards might be confused by applying ground on one of the RS232 lines if the board doesn't have the necessary filtering, because I wonder if the occasional crashes are coming from the PC reading noise on the line as serial data? I dunno tbh. I apologize for taking the thread off subject.

Reply 4 of 14, by mkarcher

User metadata
Rank Newbie
Rank
Newbie
pii_legacy wrote on 2020-06-22, 22:59:

Yeah, this is what i meant, i apologize. I think a lot of older serial-mouse only boards might be confused by applying ground on one of the RS232 lines if the board doesn't have the necessary filtering, because I wonder if the occasional crashes are coming from the PC reading noise on the line as serial data? I dunno tbh. I apologize for taking the thread off subject.

No need to apologize. Finding out why the mouse crashes the system is helping the OP and other people encountering similar problems. I don't consider this as taking the thread off-topic.

Still, you should be aware that PS/2 mice and serial mice are connected to different ports with different characteristics. The RS232 ports used for serial mice are extremely robust and well-isolated from the processor (any voltage applied to them goes through a RS232-to-TTL level converter, then the UART, and touch processors only after these two stages), whereas the PS/2 mouse is in direct connection with the keyboard controller. The keyboard controller (as introduced with the AT) does not only handle the keyboard interface (and the PS/2 mouse interface), but it also does system management tasks like assisting the 286 processor to show its extended capabilities of accessing 16MB in a (mostly) 8088 compatible system. It needs to control the memory access pattern (using the infamous A20 gate) and the CPU reset line to achieve that.

If you manage to confuse the keyboard controller, this can easily crash the whole system. A PS/2 mouse gets ground and +5V (hopefully with suppression inductor) from the mainboard. So the only voltage levels it is capable of applying to its clock and data lines is between 0 and 5 volts. The keyboard controller is supposed to handle any level between 0 and 5V just fine. Still, the PS/2 mouse (cable) is able to pick up radio interference, which adds up to the voltage levels sent by the mouse, so the sum can (for a short period) exceed +5V or go below 0V, which could put the keyboard controller into an indeterminate state.

To prevent this from happening, IBM placed a filter consisting of an capacitor (shorting high-frequency noise to ground) and an inductor (blocking high-frequency noise on the way to the keyboard controller) between the keyboard / mouse ports and the keyboard controller chip. These filter components are not only found on the original PS/2 boards, but on any quality board that implements PS/2 like mouse and keyboard interface.

The crashes observed in the original post might come from different reasons:

  • The PS/2 mouse picks up noise on the data/clock line that crash the keyboard controller (should be prevented by filtering)
  • The PS/2 mouse interferes with the +5V supply line on the mainboard (should be prevented by filtering of the +5V supply line)
  • The PS/2 mouse protocol implementation in the keyboard controller is broken, and the keyboard controller crashes on normal mouse behavious
  • The keyboard controller chip is defective
  • The PS/2 mouse handler in the BIOS is broken and can not handle mouse data arriving at certain times, causing the POST operation to crash

My recommendation is to find out whether other people using the same computer have similar problems using PS/2 mice (which might indicate a design error in the system), or the OP is the only person to experience the problem (which seems to indicate broken hardware).

Reply 5 of 14, by athlon-power

User metadata
Rank Member
Rank
Member

Thank you for the ideas! I know that serial mice will work on this system, but from what I've seen, PS/2 mice connected to a serial adapter will not work. It also was more common with a Keytronic keyboard I have than the Mitsumi keyboard I'm using on it right now (shame, the Keytronic keyboard is much better overall). I can't think of anything else at the moment.

The one serial mouse I have is awful, but I have a plentiful amount of quite high quality PS/2 mice, which is the ironic part.

Reply 6 of 14, by mkarcher

User metadata
Rank Newbie
Rank
Newbie
athlon-power wrote on 2020-06-23, 00:49:

Thank you for the ideas! I know that serial mice will work on this system, but from what I've seen, PS/2 mice connected to a serial adapter will not work.

The simple small PS/2-to-serial plug adapters do not convert any PS/2 mouse into a serial mouse. For these adapters to work, you need a special kind of mouse that supports both the old serial protocol as well as the new PS/2 protocol. The mouse automatically detects whether it is plugged directly to a PS/2 port , or it is plugged to a serial port using this adapter, and switches its communication protocol.

If you connect a PS/2-only mouse to a serial port using one of these adapters, you should not be able to crash a correctly working computer, but the mouse should just not work.

Reply 7 of 14, by chinny22

User metadata
Rank l33t
Rank
l33t

Not sure how the high and mighty IBM parts feel about being in a PB case but to the causal observer it looks like a standard PC who wont realise what a Frankenstein it really is.
Or in other words it looks like a nice tidy job 😀

Reply 8 of 14, by appiah4

User metadata
Rank l33t++
Rank
l33t++

Butchered IBM parts in a Packard Bell case sounds like the ultimate insult, but it works so who am I to judge 😁

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 9 of 14, by athlon-power

User metadata
Rank Member
Rank
Member
chinny22 wrote on 2020-06-23, 09:06:

Not sure how the high and mighty IBM parts feel about being in a PB case but to the causal observer it looks like a standard PC who wont realise what a Frankenstein it really is.
Or in other words it looks like a nice tidy job 😀

appiah4 wrote on 2020-06-23, 10:19:

Butchered IBM parts in a Packard Bell case sounds like the ultimate insult, but it works so who am I to judge 😁

I never realized the irony in this until just now. I just wanted a better case because the old one looked like this:

20190822_150720.jpg
Filename
20190822_150720.jpg
File size
474.36 KiB
Views
255 views
File license
GPL-2.0-or-later
20190822_150647.jpg
Filename
20190822_150647.jpg
File size
252 KiB
Views
255 views
File license
GPL-2.0-or-later

These pictures are from my old house, so excuse the mess. The new case is also more compact and is better in general than the old one, which is an even greater insult to IBM because I did this due to what I felt was poor case design on their part, so I instead forced the internals of it into a cheap knock off case because I felt the cheap knock off case was better. O w.

I'm also going to make a pretty aggressive jab at IBM here, because these parts weren't all high and mighty by the time this machine was built. IBM was relegated to making computers that were practically the same as their clones at that point, they no longer had control over the market, period, and were just another PC manufacturer. By the time they had created the PS/2 line, it was over. It just took time. So I don't think these parts really care because at this point, they were no better or worse than the next PC clone.

That's what IBM gets for trying to make a monopoly. I say this as I type this on a computer running Windows 10, which might make me a hypocrite(?)

Also, thanks, I was hoping I managed to make it look presentable. It looked god awful before this, and with how OCD I tend to be with these computers, I couldn't let it stand for long and it not drive me insane.

Reply 10 of 14, by athlon-power

User metadata
Rank Member
Rank
Member

For those who have the morbid curiosity to see what this once looked like:

20191116_192503.jpg
Filename
20191116_192503.jpg
File size
389.53 KiB
Views
253 views
File license
GPL-2.0-or-later
20191116_192535.jpg
Filename
20191116_192535.jpg
File size
361.45 KiB
Views
253 views
File license
GPL-2.0-or-later
20191116_192835.jpg
Filename
20191116_192835.jpg
File size
438.2 KiB
Views
253 views
File license
GPL-2.0-or-later

Help.

Reply 12 of 14, by chinny22

User metadata
Rank l33t
Rank
l33t

That IBM case isn't pretty but has a certain business no nonsense charm. The fact they have the P/N and S/N on the front shows they were thinking of the IT staff having to support 100's of these things in an office.
Your right of course though, even by that time IBM were nothing special, (excluding laptops) but the brand name was still big blue and I bet even some of their own staff thought they were something special.

Don't think Packard Bell ever thought they were anything special but some of their machines were nice.

Not that any of this matters for your build. That chunk missing from the bottom right of the IBM case is reason alone for the transplant.

Reply 13 of 14, by appiah4

User metadata
Rank l33t++
Rank
l33t++

Personally, I would have kept the IBM case.. To each their own though 😀

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 14 of 14, by athlon-power

User metadata
Rank Member
Rank
Member

I actually quite liked the aesthetics of the IBM case, but it had:

The aforementioned chunk missing.

The entirety of the case was coated in a thin, borderline non-removable layer of automotive residue vile chemical pollutant disgust that caused it to smell of chemical vapor toxic evaporation mixed with heavy metal air pollutant byproduct. Inside and out, it was like it had a fling with a car and they both got a little too excited.

Stupidity aside, it was used in an automotive shop since the Triassic period, so it had built up over time. I know this because before I formatted the HDD, there was automotive training software all over it and 1x1 pixel resolution .BMPs of cars that were made before man had lit the first fire.

Also, the IBM case was needlessly tall and heavy in my opinion and the PB case is much more compact while keeping almost all of the functionality.