Weird Compaq Portable III CMOS Glitch

[T-Squared]

I bought one of Glitchworks' DS1287 battery modules a while back, and I've been having some oddities with it:

1. I've lost a Modem/Memory Card connector module because of what seemed like bad memory. (technically it still works, the pins of the smaller male header connector were bent so far out of alignment that I just took a heat gun and desoldered the header without much trouble, but now I can't find the proper 100-pin connector that fits with it.)

2. I get a 102 - System Board Failure message at startup when the chip is powered with the battery.

3. The internal clock seems to be counting up Minutes : Seconds : Milliseconds (with the last two zeroes chopped off), instead of Hours : Minutes : Seconds, because when I started the diag disk, and got to the clock setting, the internal clock started acting more like a stopwatch than an actual clock.

 

[modem7]

I bought one of Glitchworks' DS1287 battery modules a while back, and I've been having some oddities with it:

Which is the situation:
1. It has been working fine in the Portable III since 'a while back', but is now showing oddities; or
2. It was fitted to the Portable III 'a while back', and the oddities have been present ever since then; or
3. Although bought 'a while back', I am only trying it out now because of DS1287 failure in my Portable III.

1. I've lost a Modem/Memory Card connector module because of ...

I cannot relate this to "Weird Compaq Portable III CMOS Glitch". Was it intended for another thread?

... when the chip is powered with the battery.

I notice your use of "when". Are we infer from that that the 102 error is not present when there is no battery in the module ?

Or are you informing us that there is a battery fitted to the module?

 

[T-Squared]

I bought the module last year, around September, as a replacement for a similar idea I used to have with the battery. (It was a wider board with a CR2032 battery holder that would connect to the original four power solder points on the system board.)

However, I started to have problems almost immediately when the first original clock chip was powered with my original wider board idea. I had a "Disk Controller Failure" at first, and I thought it was a real problem, because those controller chips are next-to-impossible to source. I figured I was going to have to trash the motherboard.

Later, I found that the disk controller was just fine, and my original idea was not sturdy enough, so I decided to get a replaceable battery module from Glitchworks, and It's only recently that I identified the RTC/CMOS settings clock chip. I carefully desoldered that with a heat gun, and replaced it with a socket to make sure I could easily switch out clock chips that wouldn't work. Soldering the Battery Holder module to the first chip seemed easy enough, until I started to have crashes and memory detection problems. (It would not detect the 4MB of memory I added to the system properly)

I also kept having problems starting the system, and wouldn't go away unless I got rid of the memory expansion card.

Eventually, taking the Modem/Memory connector board off so many times had bent the pins that connect into to the motherboard. So I desoldered the connector. The board still works, but it's missing the necessary connector pins now, and I'm using another Modem/Memory connector from a system I bought for replacement parts.

Recently, I've gotten back onto trying to make it work with a new RTC/CMOS chip with the same Glitchworks module, but it still doesn't want to work correctly.

The 102 error only comes up when the battery is in the Glitchworks module, and the clock runs fast with it as well.

 

[modem7]

I read at [here] that there were two possible CMOS/RTC options for the system board: MC146818 or DS1287. And I see photos of both variations on the internet.

I have found a thread ([here]) on this subject that you created late last year. So your system board had the MC146818 option.

Per [here], the MC146818 option is what the IBM AT uses. And from what I read, on the Portable III, the battery supporting the MC146818 option is the very large square block, a 3.4V Lithium battery - your reference to "four power solder points on the system board".

So now you are swapping to the DS1287 option, not using an actual DS1287, but a certain Glitch Works product in combination with something else.

I bought a power module from GlitchWorks to solder to my chip, so I can replace the battery more easily whenever it goes low.

... so I decided to get a replaceable battery module from Glitchworks

Soldering the Battery Holder module to the first chip seemed easy enough, until I started to have crashes and memory detection problems.

Recently, I've gotten back onto trying to make it work with a new RTC/CMOS chip with the same Glitchworks module, but it still doesn't want to work correctly.

What exactly is the Glitch Works product you have? A photo or model number.

You write of soldering the product to a chip. What chip is that?

One Glitch Works product I see is the GW-1287-1, quote: "These repair boards can be installed on a NOS DS1285 or can be used to repair an existing DS1287 module with a dead battery." I see that the GW-1287-1 includes a crystal, understandable because the DS1285 (but not the DS1287) requires an external crystal. Perhaps you are using a GW-1287-1 on an old DS1287.

 

[T-Squared]

Yes, it was originally a lithium block battery. (two positive terminals, two negative) Back in 2013, to replace the battery, I had also originally just wrapped two LR44 cells in some electrical tape, pressed against soldered battery terminals with wires connecting to where the terminals were on the board, but that was not a clean-looking solution. In 2017, I tried to make a prototyped battery board that I could pull out and replace the coin cell on it, but the board was too heavy for the terminals. And yes, the GW-1287-1 board was the one I decided to buy and use.

I do have a Motorola MC146818 chip for my system, which was powered by that lithium battery.

I actually did have that idea in mind, that the two clock crystals were conflicting with each other. I'm thinking of removing the one on the motherboard.
EDIT: Or is that not something that needs to be done?

 

[modem7]

I actually did have that idea in mind, that the two clock crystals were conflicting with each other.

I infer from that that it is the GW-1287-1 module that you have.

I do have a Motorola MC146818 chip for my system, which was powered by that lithium battery.

I actually did have that idea in mind, that the two clock crystals were conflicting with each other.

The crystal on the GW-1287-1, if connected to the piggybacked chip, connects to pins 2 and 3 of the chip.
On your motherboard, configured for an MC146818, pins 2 and 3 of the IC socket are connected to a crystal circuit (crystal plus some other components) on the motherboard.

If you have connected the GW-1287-1's crystal to pins 2 an 3 of the piggybacked MC146818, and pins 2 an 3 of the MC146818 are going into the IC socket, then there is a problem. That, in effect, adds an additional crystal to the motherboard's 32.768 kHz crystal circuit.

If instead, you have connected the GW-1287-1's crystal to pins 2 an 3 of the piggybacked MC146818 by bending up pins 2 and 3 of the MC146818 (i.e. pins 2 and 3 of the MC146818 are not going into the IC socket) then you end up with an MC146818 with only a crystal attached to the MC146818's pins 2 an 3. The MC146818 data sheet shows that for a crystal type connection, more that simply a crystal is required.

I'm thinking of removing the one on the motherboard.

It should be the one on the GW-1287-1 that gets disconnected, and if applicable, reconnecting pins 2 and 3 of the MC146818 to the IC socket. You end up with the MC146818 back to using only the crystal circuit on the motherboard.

Other possible issue:

I expect that like the IBM AT, and at least some AT clones, there is circuitry on the motherboard that feeds +5V to the MC146818 when the motherboard is powered on. See the '+5V switch' at [here]. If the Portable III's system board has that circuitry, then you need a diode between the GW-1287-1's battery socket and where (pin 24 ?) it connects to on the MC146818.

Note that such a diode is going to reduce the 3V going to the MC146818 when the Portable III is powered off, and so use a diode that has a low forward voltage drop (e.g. someone suggested an FR107).

 

[T-Squared]

This is going to sound weird, also, but I actually had this same problem even before I got the battery/crystal module. That's what started this whole thing, basically. The startup gave me a "Disk Controller Failure" warning when I used my hacked-together prototype battery board back in 2017. Like I said, I figured the system was basically trashed, because I don't think you can even get those chips anymore.

 

[modem7]

This is going to sound weird, also, but I actually had this same problem even before I got the battery/crystal module. That's what started this whole thing, basically. The startup gave me a "Disk Controller Failure" warning when I used my hacked-together prototype battery board back in 2017. Like I said, I figured the system was basically trashed, because I don't think you can even get those chips anymore.

When you headed down the 'prototype battery board' path, did you open up the Compaq-supplied battery module to see what was inside it? Some modules, in addition to an actual battery, contain a diode. Some also contain a resistor, which I think is there to limit current if the module gets externally overloaded in some way.

-----------------------------------------------------------------------------------------------------

If we suppose that:
* The Compaq-supplied battery module contains a diode; and
* The MC146818-configured system board (excluding battery module) has no diode (because the engineer knows its in the battery module); and
* The MC146818-configured system board has the '+5V switch' shown at [here]; and
* You substituted a 3V battery in place of the Compaq-supplied battery module.

Looking at the diagram at [here], the above would be equivalent to not having the two diodes (CR1 and CR2) in the circuit, and having a 3V coin battery connected to J21. When the system board is powered on, the '+5V switch' would be trying to put +5V onto the terminals of the 3V battery. In my opinion, that won't necessarily overload the PSU, but it may cause an unstable +5V running through the system board. I certainly won't be testing that hypothesis on my IBM 5170 motherboard. Maybe an engineer here may comment on that.

(Not to mention the 'badness' happening to the 3V battery.) (Damn, I mentioned it.)

 

[modem7]

Adding to my previous post:
Turns out that there is an earlier thread [here] on these forums in which it was established that the Compaq-supplied battery module contains a diode and resistor.

 

[T-Squared]

That's right. Now I remember; I did at least see a resistor inside of the original block battery.

Before I found this, though, I tried powering the chip normally with the Compaq diag program running (i.e. turn on the computer, start the diag/config, and then insert the battery), but it gave me "The configuration device failed the read/write test. Please contact Compaq for assistance."

I think we've established the problem. My module doesn't use a resistor and diode. I'll see if I can get something cobbled together, if it can be wired up in some weird way. I think Glitchworks might want to make a new verison of the power module board.

 

[glitch]

I think we've established the problem. My module doesn't use a resistor and diode. I'll see if I can get something cobbled together, if it can be wired up in some weird way. I think Glitchworks might want to make a new verison of the power module board.

Can you provide a picture of what you have installed in your system? I'm not sure if you have a GW-12887-1 (what you would've gotten if you bought a pre-assembled module from us) or if you've tried to stick one of our repair boards on a MC146818.

 

[T-Squared]

Here you go!

 

[glitch]

OK, if that's a prebuilt GW-12887-1 it will almost certainly not work in a socket intended for a MC146818. You *might* be able to get away with a GW-1287-1 made up using an original DS1285 chip, but even that's not guaranteed. The best course of action on systems that use a MC146818 is to replace the original system battery pack. On a Compaq Portable III or Portable III/386, the usual thing to do is add a 3x AA or 3x AAA holder, or a 1/2AA and a 3.6V lithium thionyl chloride cell.

There's nothing you can really do to make a GW-12887-1 work in this situation. We could possibly design a piggyback module to add a battery to the top of a MC146818, but 99% of the time you'd just replace the system battery pack. Here's an example with a 1/2AA holder and a 3.6V cell:

http://www.glitchwrks.com/2018/07/01/siig-s286

As you can see, I removed the trickle charge resistor to keep from damaging the new non-rechargeable lithium cell. You can also just add a diode in the battery pack, as Compaq apparently originally spec'ed.