After long period of not touching 65c816 i've tryed to add some "linear" memory to the system
system runs at clock rate of 1.77MHz, and some of You would say one can run with a murder with such slow clock rate
so i've did what datasheet says - added inverter to phi2in signal, and used this as my LE signal driving AHCT574 - system won't even boot up

okay, time for debug with logic analyser
LA says, that there is less than 10ns (even at 1.77MHz) to latch address bits, but also shows that propagation delay for my inverter gate was too large to enable
simple way out to use '123 chip and shorten LE signal, so it could be back in time with address bus/data bus multiplexer

and so i did

system boots up, but shows that RAM chip is bad

if i pull latch out of the board, and use jumper wire tieing A16 (i'm using 628128 SRAM chip) to any constant potential, then it behaves like it should

i'm pretty clueless at the moment, as it seems that bus addresses are latched properly (i'm having constant 0 on A16 when latch is latched) - maybe my data buffer is too small to see anything wrong with the circuit

Anyways, any help would be greatly apreciated

Candle

You need a latch, not a D-type, to capture the high address bits. The data bus carries the high byte of the address during phi1, your latch should be transparent during phi1 and then close for phi1. I'm not sure, but are you capturing the databus right at the beginning of the cycle? That would be too early - you'd be likely to capture the data bits from the previous cycle.

I think you need a '573, not a '574

I agree with BigEd, use a 74'573.

My 65816 uses a CPLD with latch logic vs. register logic and it works great.

Daryl

sorry - my bad
i got 573

there was bus conflict with RAM chip as my equations for RD was just !RW rather than !(!RW & PHI2)
after fixing this i got it up and running
Thanks

next step - to make it run faster

Excellent!

I guess you're using GALs for the glue logic, so quite easy to make changes? (We considered GALs for a few paper iterations, but in the end we skipped straight from 74-series glue to CPLD. Then we could put all the glue in one device.)

I recommend a 74ABT573. You want the absolute minimum delay on the gates.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Wow, that is fast! CMOS inputs too. I must have discounted 74ABT years ago because my National data book on 'ABT starts with the 74ABT125 and '126, and then goes to '240 and up, mostly being bus-oriented logic. I see Digi-Key does have a few (very few) of the low-level ones, and the prop delays are super fast. Now I'd like to find the '521 or '688, maybe a '30,... but I'm not having much luck.

Mouser seems to have a pretty good 'ABT inventory, quite a few different parts, in fact. However, I couldn't find any 'ABT521s or 'ABT688s. Best I could find was 'HC.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Very nice. Thx BDD.

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

You're welcome!

All of the glue logic on my POC 1.0 is 74AC because I had it laying around from other projects. POC 1.0 uses one each of 74AC00, 74AC04, 74AC08, 74AC20, 74AC32, 74AC74 and 74AC138. All except the 'AC138 have 74ABT equivalents. The only negative is the available 74ABT parts are SMT, mostly SOIC-14, SOIC-16, etc. Those packages have gull-wing leads, so they're a bit easier to work with than SOJ types. My SRAM is SOJ32 and I got that slobbered in place on the first try. So I'm game to try working with SOIC.

SMT parts obviously have their advantages when trying to pack a board as tightly as possible. For most of the prototyping we do around here, ExpressPCB's ProtoPro service works out well. The maximum allowable board size is 21 sqaure inches and the maximum allowable hole count is 650 (which count includes via). SMT would allow me to stay with the ProtoPro service by reducing component footprints and eliminating all those holes needed for DIP parts.

When time allows, I'm going to see if I can rig up a makeshift reflow oven to do the hard work. The temperature profile needed to solder most SMT stuff seems as though it can be implemented with a decent sized toaster oven. The principle problem will be in maintaining proper air circulation within to evenly heat everything. I'm sure there's a workable solution to be had.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

I was checking out the ABT family on NXP's website. Just FYI the 'ABT821 is avail from digikey, I think it's a functional equivalent to '521. No '688, I've used them in the past too...

Good point about the drill holes for a PCB board, I've never considered that while stocking up IC's. I did find most of ABT IC's you mentioned in PDIP from digikey. All except the dual 4-input NAND.

_________________
65Org16:https://github.com/ElEctric-EyE/verilog-6502

The '821 is a flip-flop. The '521 and '688 are identity comparators, which I had in mind using for address decoding.