Soldered today a second Apple-1 board, the video part works well, I see the *@_" pattern on the screen and it's possible to clear the clear to one "@".
But after a Reset, I don't get the "\" prompt, nothing happen after a reset.
Did this tests:
- see the correct voltages (5V is 4.95V)
- cheched all chips, all ok
- changed the 6502, 68B21 and the PROMs with the ones from a working board
- I see the Reset goes from low to high when I press Reset
- I see the correct Clock Signals on pin 3 and 37 (see picutre)
- I see the correct Sync signal
- Mods are in place
I would be verry happy to get an advise where to look next.
Thomas
... means the CPU section does not work. Since you have checked the voltages and the clock and changed the 6502, the PIA, and the PROMs against known good ones, it only can be some open or short (solder blob ?) or one of the following common mistakes:
- not closing the "6502" and "no DMA" solder jumpers
- missing or wrong programming of the 74154 memory decoder patchboard
if all of this is OK, the most likely fault left is:
- DRAM does not work
The Wozmon won't make a prompt if the DRAM is not there or does not work correctly (at least for the locations the Wozmon needs). I had one case where it made three '_' and then died. But typically, it's just no prompt.
Now, if you have one of my A1, A2 PROMs with the hidden diagnostics page, you can use that to see what's wrong with the DRAM, as this diagnostics software was written such that it needs no RAM to give all the necessary DRAM error messages if the DRAM is bad.
You can also check if the 74123 oneshot produces a 480ns wide pulse, but you need at least some accesses to the DRAM to see that pulse on the oscilloscope. If the 6502 runs wild, there are chances that it may occasionally access the DRAM so you can see a pulse. But if the 6502 ever hits the infamous "KIL" instruction it will stop fetching instructions and that's it. No pulses to be seen until the next RESET.
Now you can see the conundrum. Without a diagnostics page in the A1, A2 PROMs you have a chicken and egg problem.
Thank you for your help Bernie!!!
Got it working, it was a very tricky problem. It seems I bought a weak bunch of 7432. I replaced chip by chip with the chips from the working board and ... surprise ... after a I replaced this 7432, I got at least a prompt after the reset. Second problem were three faulty 2504 chips. Tried to use AM2808 chips, one bunch of them works well, an other shows a "T" insted of the "\" as a prompt sign.
Will do more tests tomorrow!
In post #3, Thomas R wrote:
"It seems I bought a weak bunch of 7432."
"Second problem were three faulty 2504 chips."
Uncle Bernie comments:
These are the perils of buying Apple-1 parts here, there and everywhere ! You don't know if they will work. And according my experience while composing my famous 100% tested and burned-in Apple-1 IC kits, I can honestly estimate about 1/3 of the builds based on helterskelter ICs do fail. I have piles of bad ICs in the "graveyard" on the top of my monitor, all from making only 50 kits. And I ran into bizarre behaviour of certain TTLs, too. They don't work in the Apple-1 but work fine in some other vintage computers I use as "testers" in some cases. And with this I mean whole tubes of TTLs, from the same lot, all being "unfit for Apple-1" despite they work elsewhere. Weird !
The drama with those Signetics MOS ICs is even worse. I bought 8 x 2504 from the mythical horse and got 3 bad ones. Sent the bad ones back and got 3 good ones. User "auraksphere" reported 4 bad 2504 out of 7 from the same source. Seems they don't test them and there are too many duds ! But the worst PMOS shift registers I ran into were National Semiconductors MM1404 I bought from a broker, half of which were bad. Alas, my sole LSTTL based Apple-1 build needs MM1404. It refuses to work with any other type of 1k x 1 shift registers.
The 1970s Signetics MOS IC quality sucks in general. The largest stockpile of 2519N in the world sits in Jerusalem, ~5000 pcs, mostly date codes from 1977, and I found 10-20% of them are bad, and what is worse, not all of those are duds right out of the tube, but they die during the first days of burn-in. BTW, this is the same guy from which I have gotten all my 2519N in the past two years, but then he got greedy and quoted me $65 each at qty 100. Same price as he sells them qty 1 on Ebay. Bloody usurer ! No deal for you, Avi. I got the ones I needed for my next 50 kits elsewhere and I will never buy from you again.
So far not much trouble seen with the 2519N character generators. From the 100 or so I had, only one was bad, one output does not move, so one column of the characters is blank, which looks weird. This low failure rate is even better than with the 6502 and 6520 I had.
The bottom line is that if you use untested and unproven ICs to build an Apple-1, you must be able to find the bad one(s). The most painless option is to buy one of my famous IC kits (you can contact me via the Applefritter message system by clicking on my handle and then use "send PM"). By buying directly from me, the usurious Ebay fees can be saved and you get a much better deal.
There is one alternative once you have a running Apple-1 build: you can use it to test and burn-in your own IC sets. I have described the "Step 3c" method in this thread:
https://www.applefritter.com/content/how-populate-empty-apple-1-ic-sockets
This is essentially the way I make my IC kits. The next step of course is to do the burn-in. I use the same diagnostics page which comes with my IC kits since mid 2021 or so, where I ran out of 256x4 PROMs, and had to use 512x4 ones. This diagnostics page really exercises everything in the Apple-1 and also the DRAM is stressed and thoroughly tested. Only a few DRAM die during the 4 weeks of burn-in but this happens. So if you do this procedure yourself, don't let the diagnostics just run for a few minutes. Most of the weaklings die within the first few days, but I had rare cases which died after 2-3 weeks.
So it is possible to succeed with building Apple-1 without my kits. But for anyone not yet having a functional Apple-1, I would recommend to get one of my kits. If you buy directly from me, outside of Ebay, you get a price you can never beat by buying the parts elsewhere. On Ebay, all bets are off, but you sure get less for your money. Ebay fees eat up the difference ! You can avoid to get less and make Ebay fatter and fatter ! (I wish there was another auction place which does not have usurious fees, but I did not find one where prospective Apple-1 builders would look).
Bernie, I know your perfect selected and working chip sets, bought already two of them ... ;-)
I needed a challenge and had the intension to build a system based on chips in my stock, and as you predicted, it's really not easy. Anyway, my system runs now since about 6h with your test, no errors till now.
Btw, my first system with your chip set shows now also sometimes the problem with no response to the "Reset" (short pin 1 to pin 9) ...
It stays interesting with Apple-1 replicas ...
In post #5, Thomas_R wrote:
"My first system with your chip set shows now also sometimes the problem with no response to the "Reset" (short pin 1 to pin 9) ..."
Uncle Bernie answers:
Try to run the diagnostics page in my A1, A2 PROMs on this machine. To avoid breaking off the pin #14 by repeated bending, just cut the short stem of the "T" shaped PCB trace below A1, on the PCB solder side. The short stem goes to the pin #14 of A1, and believe it or not, this time the layouter used his brain, and this is the only spot where both pins #14 of A1, A2 are connected to the rest of the circuitry. The one cut isolates them both but keeps them connected together. This is an easy mod for 512 bytes firmware space, for instance. This little mod can easily be undone by adding back a small jumper wire from pin #14 to pin #8 (ground, logic L). The chip select signal which goes to these pins #14 before the mod is superflous, it's the PHI-zero clock input of the 6502 inverted by the NAND gate at location B-1 as seen in area B-7 on the Apple-1 processor section schematic. Oh the irony ! One of the lesser known design flaws in the Apple-1: you can write to the PROM area, and yes, there will be a bus clash. But the PROMs always will win (bipolar is stronger than NMOS) and the 6502 will not get damaged as its output drivers are too weak to blow themselves up.
If the diagnostics run OK, it's certainly not an IC problem. The 6502 has its own quirks, and one of them is the RESET actually being no real reset which clears things, but it's yet another interrupt (like the NMI and the IRQ). So if you have a jittery signal, the RESET may fail under rare circumstances of being not there all the time until the internal sequence is completed, which takes many cycles and pushes stuff on the stack like IRQ and NMI do. The whole topic is a bit elusive as you need the transistor level schematic of the 6502 to understand what is going on, and the datasheets avoid the topic. I have succeeded in getting occasional botched RESETs, when using a jumper wire on the keyboard connector, or the "gimmick switch", but so far it did not happen with a real switch, but it could happen. To really defuse this situation, you would need to use a switch debounce circuit which not only debounces the RESET switch, but also synchronizes the RESET signal to the CPU clock.
However, if you do have some IC problem (some IC started to limp) then the diagnostics page will unveil it by either freezing the screen output, or producing garbage, or spitting out DRAM error messages. If garbage is produced, change the PIA first. If it freezes, change the 6502 first (the diagnostics do not need any RAM). If a DRAM message comes along, move the offending DRAM to see if the bit error location moves with it. If so, replace it with a good one. Other than that, all the checks mentioned in my post #2 are helpful to zero in on the culprit. In case of DRAM errors that can't be nailed down to one offending DRAM IC, a nasty unstable carbon composition resistor not properly making the 480ns timing maybe ? Easy to diagnose with an oscilloscope while the diagnostics are running. All other builders who see mysterious DRAM errors which do not change bit position when DRAM ICs are swapped, but having no scope, use the metal film resistor which came with the kit, and put it into the "27K" place at the 74123.
This procedure should help to fix most of the misbehaving Apple-1. But if you don't have the diagnostics page in PROM, it's hard.