Hi everyone. I'm wondering if anyone has some insight into a weird problem I'm observing with an Apple 1 replica build.
I have recently completed an Apple 1 build using an open source board and managed to get it working well, passing memory tests etc. However I've only been able to get it running using a Synertec 370-6502 CPU. I've tried a MOS 6502, a MOS 6502A and a Rockwell R6502AP, and although the terminal section works as it should with any of them, hitting reset with all but the Synertec doesn't get the computer section running.
I'm wondering what could cause this build to be so picky about the CPU? Could it be that the specs of the Synertec are slightly different from the other processors and something marginal with the other components allows the board to work with it, but not the others?
Of course I can just keep the Synertec CPU in place and have a perfectly running board, but I'm curious as to why that one works and the others don't.
In post #1, "Silicon-Surfer" asked:
"I'm wondering what could cause this build to be so picky about the CPU ?"
Uncle Bernie answers:
A properly built und tuned-up Apple-1 works with any NMOS 6502 from any manufacturer. Faster speed grades (6502A, 6502B) allow for wider timing margins on the 480ns oneshot (which is essential for my famous IC kits, as the timing components can drift a little bit during soldering), but as long as you have an oscilloscope to adjust the 74123 oneshot to 480ns by change of the resistor, and use a stable (preferably: mica) 47pF capacitor, even the slowest 1975 data code, ROR-less, early 6502 will work.
As you say the RAM tests work, let's assume for now your "maverick" build is OK. (For Uncle Bernie, everybody not using his famous IC kits is a "maverick" ... not in the sense of "Top Gun" but in the sense of rancher speak, if you know what I mean, and no pun intended. I actually like these "maverick" builders - some give up but the few who post here and tell us about their desperation only show to the world that my IC kits have "hidden value" in them).
But back to your problem, Silicon Surfer. Take your time and read this thread:
and pay particular attention to tbe conterfeit 6502 from China you can see there. I never fell for those because I know that neither MOS Technology, nor Synertek, nor Rockwell ever used laser engraving on their 6502 (nor on the "family" members, i.e. 6520, 6522, 6532), but the Chinese counterfeiters do it, to turn CMOS 6502 of various origins (or even the Japanese knockoffs with no decimal mode) into "6502".
Which typically don't work in the Apple-1.
I could tell you stories about the entrepreneur who plans to build 100+ industrial wave soldered Apple-1 clones (a few have been sold already, testing the waters). He disregarded my warnings (I'm the technical advisor for this project) and bought hundreds of "6502" from China. But only a few did work in the Apple-1. These few may have been original NMOS 6502 which the crafty Chinese counterfeiters ran trough the same process - grinding off the original chip lettering, applying black coat, and then laser engraving with the nice "Rockwell 6502" text. And with a huuuuge Rockwell logo.
After this blunder he then bought the 6502 and 6520 he needed from me, at a slightly higher price (a few dimes more each maybe, I don't know what he paid for the crap / counterfeits).
I do have a reliable source, these are leftover stock which came out of Atari, date codes from the early 1980s, and never left the United States.
Draw your own conclusions. Try a Q-Tip with Acetone (nail polish remover) on your "6502" at a place on the upper side having no lettering.
- Uncle Bernie
As usual... Uncle Bernie is spot on... Fake 6502s from China... often relabeled 65C02s as those are more common these days.
A good place to get a good NMOS 6502 for a relatvely cheap price is from a VIC-20 or 1541 floppy drive off eBay. Lots of them out there cheap, and even if they are non-working, the 6502 is very rarely ever the culprit. And you can often re-sell much of the leftover parts and recoup most if not all of your money. Commodore fans probably will hate me for suggesting it, but there is really very little value to those and Commodore made literally millions of them.
UncleBernie - I don't mind being a maverick, I think part of the enjoyment of a project like building an Apple 1 replica is hunting for parts and troubleshooting and solving any problems that you have along the way. I know your IC sets are great value and guarantee a working build straight up, but for me that's too easy - "the journey is the reward".
I think you could be right about the counterfeit ICs because the Rockwell has an early 2000s date code and laser etching. I assumed that they still produced them at that point, but maybe not? It wasn't ordered from China, but that doesn't mean it didn't come from there originally. The MOS 6502 did come from China but the markings look original, and acetone doesn't affect it. It wasn't new so maybe it was just faulty. The MOS 6502A I had in my parts stash and was a pull from an old machine years ago. I assumed out of 4 that more than 1 would be good, but maybe I'm blaming the board for my bad luck with CPUs...
In post #4, "Silicon-Surfer" wrote:
"you could be right about the counterfeit ICs because the Rockwell has an early 2000s date code and laser etching"
Uncle Bernie's comment:
Here we are. As I said, none of the NMOS 6502s and family chips ever were laser marked. Rockwell made them in a slightly upgraded 1967 wafer fab, much the same generation as MOS Technology. I don't know exactly then they stopped making NMOS 6502, but I'm sure there are no "Rockwell" 6502 after 1998 because in this year Rockwell International spun off its semiconductor division under the new name "Conexant" - they had quite some success in the 1990s with modem chips etc, all made in CMOS.
The problem with CMOS 6502 in the Apple-1 is threefold - some have CMOS logic level thresholds which won't work in a TTL environment, unless you would reduce the CMOS part's supply voltage to 3.3V. Some CMOS 6502 have extra functions on the "NC" pins of the NMOS part, which need to be taken care of. I have a mod to run a WDC W65C02S6TPG-14 on the Apple-1, without the supply voltage trick. But the elephant in the room is the fierce current spikes produced by fast CMOS logic (that WDC can run at 25 MHz under certain conditions although it's only tested up to 14 Mhz) and speed means greed means *lots* of dI/dt spikes. And the Apple-1 PCB layout can't take that without adding extra bypass capacitors. In the 1990s, leading edge process technology CMOS got so fast that you could not use wire wrap or double sided PCBs anymore. Ironically, ECL100k can be wire wrapped (using impedance controlled twisted pair wires and special PCBs) and ran at least 10 x faster than these CMOS parts.
I have a rotten design of an IC prober in my lab which never worked alright because the CPLDs in it from the early 1990s could not toggle all flipflops in a segment at once without crashing the (internal) power supply within the CPLD PLCC package so badly that some other flipflops would get corrupted. Even soldering small SMD power supply bypass caps directly on the PLCC pins would not help. The evil bond wire inductor and the - most likely - botched internal ground network in the IC layout conspired against this rescue attempt.
Only after losing weeks of my precious RQLT on trying to debug this project I consulted the FAEs of the CPLD manufacturer and they said: "oops, sorry, you can't do that, in our new datasheets there is a disclaimer." Needless to say, I avoided this CPLD family afterwards like the plague.
Back in the day these fools designing those chips did not understand what they were doing. Slope controlled pad drivers would have avoided this blunder. But they were not invented yet ... plain digital designers would not understand them anyways.
- Uncle Bernie
Thanks Uncle Bernie. There's no doubt that you're a wealth of knowledge. I tried to search for an answer on when Rockwell stopped making 6502s and couldn't find anything useful. I should have just come to the source!