6502.org

That's great, it's always good when things work in the end!

I'd suggest also running a full RAM test, to check you can write and read all RAM locations without corrupting others. You could also consider running the Dormann test site, which I only started using recently, but which has already caught a few failures in my systems.

Well done!

-Gordon

Strange issue: the RAM doesn't seem to be accepting write cycles. I've checked the connections and they're all correct, so I have no idea what could be going wrong.

Is it still the same circuit as posted here?

Yes.

The schematic seems to have no issues to me. Maybe double-check NAND wiring (U5A & U5D)? Do you have a logic analyzer? If so - you could check if RAM's /CS & /WE are properly asserted during RAM write cycles.

EDIT: Are you using decoupling capacitors? I don't see any on your schematic nor on the breadboard. To quote BDD: I even like to squeeze them right under the chips - it saves a lot of space and allows the traces to run under the capacitor (especially when doing PCB designs): I actually found its weak pull-up to be a very helpful indicator that I forgot to add capacitors near power-hungry components. In my case, ESP-01 was often triggering reset during WiFi reconnects, and this issue went away after I added an extra 10uF cap near it.

I'm single-cycling it using an arduino, and by poking the board with an LED I can see that /CE and /WE are behaving correctly. There are some decoupling caps on the board, but they might be hard to see because they're only little 100nf ceramic ones.

When you checked the wiring, did you look at VCC and GND too? If you happen to be using those little jumpers shaped like staples it's really easy to get them off by one row and accidentally leave (e.g.) your RAM's power pin floating. This sort of thing can be frustratingly pernicious, and can be pretty tough to spot (at least by my middle-aged vision) without a magnifier of some type.

There's 4.98V between Vcc and ground on the RAM chip, which is well within limits according to the datasheet.

Ok, I've tested the RAM chip by false-feeding it on a separate breadboard, and no matter what I do I can't make it accept writes properly. I also tested a few others I got in the same order, and all of those exhibited the same behaviour. I find it hard to believe that all of them were dead on arrival, but it seems to be the only explanation.

These things are hard to troubleshoot at a distance.  Where did you get them?  I will comment that I have never gotten a defective IC from a legitimate distributor; and I've been responsible for probably millions of ICs at work.  However, see the links in the second paragraph in our sticky topic at viewtopic.php?f=1&t=1953, about China's counterfeit IC market.  Also, how are you handling them?  Any MOS (CMOS, NMOS, PMOS, etc. silicon is vulnerable to damage from static discharge; so you should exercise adequate anti-static handling precautions when you're building.  I still suspect some simple forehead-slapper error that you'll eventually find.

I got them from Mouser. And yeah, it probably is a stupid error, but I can't think what. I've triple-checked all the pins and run it with false-fed signals and it still doesn't work.

Not sure what your flash programmer looks like but you might use that with minor hardware/software modifications to test your RAM chips...

Mouser is definitely a legitimate distributor, so no worries there.  I have no doubt that the parts were all good when they arrived at your house.  After some more head-banging, you'll be telling us what you figured out, and that you got it working.  In the mean time, maybe spending a few days away from it will result in your coming back to it with fresh eyes that see something you didn't catch earlier when you were only seeing what you were expecting to see, or you'll think of something you didn't earlier when you got tunnel-visioned.

Just a reminder that you won't know if a write was successful unless you're capable of doing a successful read. IOW, don't be too sure it's the write that's failing.

I agree with Garth. You might find some new insight after setting the problem aside for a while. (Also, do the reads and writes achieve the appropriate levels noted for reads and writes in the data sheet's Truth Table? Just checking!)

-- Jeff