I am getting closer to get my Apple I build to work. I had some trouble with the DRAM that was sorted out, to a certain point.
See that thread:
Now after replacing a few more evil DRAM chips found by Mike Willegal's RAM test program, I can run that RAM test for days and days, heat the machine with a heat gun, etc., and at least no more DRAM errors shown.
But when I run that BASIC checksum program from the above thread, it now runs "only" 120-150 hours before crashing with some garbled "Stopped" message.
Digging deeper for root cause would be tedious. So I wonder whether I just let it stay like it is. It's reasonable good for demos on Apple Fests etc.
This said, which reliability (or Mean Time Between Failures) do you have with your own Apple I, original or replica ?
Is it able to run for weeks and months without a hitch ?
Or do I just ask for too much perfection ?
Comments invited !
At VCF East and MakerFaire I have run both a original and replica Apple-1 (using original/period Mostek RAM) continually for 12 to 14 hours at a time. Always using a fan without failure. Last year at VCF West we had someone running their original all weekend without an issue also.
I don't think I have ever run an original Apple-1 more than 10 or 12 hours at a time since I would never leave something so valuable running un-attended, and I need to sleep at some point. As for a Mimeo, a few years ago, someone accidently left the one at the VCF Museum running for over a week (before we got remote power management) and was still running Apple the 30th the next week without a problem. That system is using early 80's Mostek RAM, it also has a fan and a onshot resistor changed to fix the RAM timing to spec for the chips.
Thanks for the info. My Apple I build certainly can run for 12-14h without a hitch.
And I don't need a cooling fan. The heat sink never gets hot enough to make the regulator go into self protection mode.
These regulators are virtually indestructible. Designed by Bob Widlar.
I am a bit curious about what timing is considered correct for the oneshot. I took scope pictures early in my build and saw that the 480ns shown in the schematic must be wrong. So I corrected it to have the MUX switch sooner after RAS, but still, I don't know what the builders consider to be the right value.
I have no issues with running my Apple I build around the clock, unattended. I have reformed and ripple tested the electrolytics and they stay cool. The glass diodes won't fail and the 1N4001 are modern and have flame retardents in their mold compound.
None of the ICs is a fire hazard. Even if the magic smoke escapes they just cause a stink. Never a flame. If you don't believe me, just take any plastic encapsulated vintage TTL IC and reverse bias its power supply voltage on a beefy lab supply. No flame. The magic smoke gets out, though, and unless you can't get that smoke back into the IC, it won't work again, ever ;-)
What I do as a precaution is to turn off the vintage B & W monitor when I am not nearby. This is a real fire hazard due to the aged components inside and due to the higher power in the circuits.
According to my experience with lots and lots of collectible computers / chess computers / etc. the most prevalent fire hazard is with the electrolytic capacitors in the power supply. They have a limited lifetime anyway, and the electrolyte dries out, making the ESR go up. Then, with ESR high enough, and the ripple current heating the thing up, it may explode and spill all its guts all over the board. The aluminum foil shreds may cause a short, and with the separator paper shreds still attached, there is nice kindling for a little fire. I always apply greatest scrutiny to electrolytics in vintage electronics. In most of my collectibles all electrolytics were replaced by newer ones. Alas, to keep the looks of an APPLE I correct, you need the blue Sprague capacitors, and the youngest ones I have ever seen were made in 1990.
The blue Sprague Capacitors are still available new. The 2400u (39D248G025JL6) is in stock at Mouser for $16.00. The 5300u (39D538G015JP6) can be ordered from Vishay for $9.914 each in a quantity of 83. I bought some 10 years ago (2010 date code) so I would have spares for my Apple 1 and had to buy minimums on both at that time.
I would expect it to run indefinitely.
Barring some unknown bug with the BASIC interpreter, there's no reason for your program to stop after 100 or 1000 hours. There aren't even any strings in that one to clog up string memory and require garbage collection.
This paper seems to indicate that actual experience with an array of 128 unknown type early 80's single board micros is that they would experience a failure on average, every three weeks. 128x3 = 384 weeks average MTBF for a single board or 384/52 = about 7 years. Your mileage will vary, but as someone that has used a number of personal computers for an extended number of years, that number "feels" about right to me. I've had many more personal computers go bad while in storage than while in use. :)