SpritesMind.Net

Current working on building my own small computer (68k). I'm working on the video output first. Building/designing it in stages and adding features as I go.

The problem I'm running into, is finding certain 74 series chips to buy in single or low quantities. There are 7-8 hobbyist friendly sites to buy chips from, but none of them see to carry 8bit presettable counters. 74xx461(a) was the chip I was looking for. I've see '193s but really don't want to pair up a bunch of these. I'm using two 8bit presettable counters as programmable address counters and I have sectioned memory so I can read from multiple chips with multiple address counters in a single instance. My project is going to be complete done in TTL logic (or HCT), so no PICs or other MCUs. And I don't have a GAL programmer.

Any suggestions? Particularly for a replacement chip for the 74LS461 (it needs to be presettable not just clear, because I'm doing tilemap/tile logic). Something available in low 1-5 quantities amounts.

Here's the sites I have checked so far:

Mouser
Digikey
Jameco
www.futurlec.com
Avnet
http://consumertronics.net/TTL-Parts.htm <- they don't have prices listed. Haven't contacted them yet.
http://www.nuhorizons.com/

Edit: Looks like the 74xx469 is close enough to the 461. And http://www.memotronics.com/product?pid=291 carries it

Well, anyway. Anyone else have experience building SBCs or peripherals for them?

Here's a place in the UK that has the 74LS461.

http://www.littlediode.com/components/home.php

Have you tried ebay?

Graz wrote:Have you tried ebay?

That's where my link came from.

I actually had littlediode in my bookmarks, must have forgotten about them. Or maybe not... £10.05 per chip in quantities of 10-19. Whoa. A little too expense for my budget.

I found a few other 8bit presettable counters, like the 74ls590 - but I'd rather not use the shared input/output pin model if I can help it. But it's cheap @ $0.50 from Jameco. And the 74F579 which is more like the 461/469's, but not sure who carries it or if it's still available anymore.

I found that goldmine (ttp://www.goldmine-elec-products.com/prodinfo ... ber=G4968A ) carries the '461 at $1.05 a piece in quantities of 1. Woohoo!

Just thinking out loud, but for audio - I could build a Paula style setup pretty easily in all 74 series too. Just need programmable clock dividers, address counters, etc then to the DAC (either multiple DACs or a single with a logic adder). Not sure how do to volume. Maybe if I go the single DAC route, I could use a shift register for volume (0 shift = max volume). A 16bit address counter for each seems a bit small. Maybe a 24bit (3x8) range.

Or use PWM output instead of DAC. Sounds like an interesting project.

I'd realy recommend trying to jump up to programmable logic. Better if you learn VHDL (of verilog) and then creating heavy logic becomes a pure joy and art.

You can make your own programmer, consisting of only a logic buffer (i can send you a schematic) for LPT (printer) port and software is completely free, if you choose Xilinx parts, for example a CPLD: XC9572XL (that is now becoming obsolete, but still rocks) that costs a few bucks in a reasonable package (VQ64). FPGAs are also very very cheap. Development kits can be found starting from about 50bucks. Other vendors (lattice, actel and altera) could also be tried out, bet there are many crappy things I came across, such as written NDA agreement to be signed and negotiated with Actel, hidden tricks in even lpt programmer to limit software quality (hidden jumper that is missing in publicly available programmer schematics, when it is not present software behaves poorly - forcing you buy the official programmer), and so on...

I recommend xilinx just because I have used it for about 10 years in hardcore highspeed systems and it rocked. Actel also rocks.

You can download a lot of unofficial hardware available, including the PSG chip (!) and various peripherals (www.opencores.org) if you use programmable logic (of whatever vendor).

Here's a nice guidebook into programmable logic, just skip the xilinx marketing blahblah: www.dtech.lv/sega_dev/logic_handbook.pdf

However - if you choose the hellish 74-logic path:

Paula made in 74 logic will not be so simple It will be quite hardcore, realy. In old russian drummachine designs, where they did not have any similar asics (like roland, korg and yamaha did), dsps and signal processing knowledge, they used two current multiplying DACs in series. The first one generated volume as a current reference for the next one, which played samples from directly connected roms. I'd recommend perform shiftdowns for programmed amount, using downcounter with programmable load. Volume att. step would be 6dB, which is very coarse, but you need multiplicaion to make it more fine. Technically it would work by having 16 cycles for calculation and in the inital cycle load up 16bit shift register with sample and downcounter with the att. value. 16bit shift register shifting right occures while downcounter is not at zero. downcounter count down also executed while caounter is not at zero. result is read out of the shift register.

If you want to make dac, you can make your own PWM, but don't expect it to be any further than 10bits. Remember that you need samplerate*(2^bitwidth) cycles speed for such output system. Meaning that 32kHz output will require 32.768MHz clock for all this machinery. If you'd use sigma modulation techniques (synthesize DSD stream), than 2Mbit/s would be enough, but logic board for one channel will then take size of a snow shovel.

I'd remind that running it faster than a few megahertz is highly unlikely unless you can make a high quality PCB with short and length controlled traces instead of breadbording with wires.

I would recommend using a ready dac with one of the standard serial interfaces that has LRCK BICK DIN pins (as an example), and drive them serialy. You can even extract such DAC-s from some oldschool CDROMs and many old CD players. All you will need is clean +5V supply and dual opamp after it (can also be extracted from the same machines).

Using pure logic also makes burden due to specific bitwidths and gate counts per chip, so that you will frequently bump into bitwidths not fitting into logic, and if you add some logic, there's too many free, and if you borrow a gate from another chip and kind-of simulate the missing one, you find the timing got severly skewed.

Regarding VDP, creating anything more complicated than CGA or ZxSpectrum video sequencer, is quite impossible on discrete logic, too. Largest and the most hardcore video board that I've seen, that's built on logic was "olivetti" - kind of CGA with some extended video modes and video memory, that was full depth ISA card (kind of ~40cm by full height) filled with DIP chips.

I want to start looking into CPLDs and stuff sometime, but then again, load of 74 chips to make stuff look cool

TmEE, in your country xilinx representative is Silica (Avnet) most likely - can give you direct representatives in Poland (I'm from LV). Closest Actel's representatives are in LT, but they require signing NDA and safety stuff in person and wait for verification. Lattice are a bit clumsy CPLDs (resource problems - cells get wasted on routing and similar issues).

Logic can cover things visualy - I do it that way. Even light some vacuum tubes if necessary. Underneath the logic filled oldchool board there's just an FPGA that does stuff

I kinda had my hell with logic for too many years, now have some kind of strong repulsion to it Maybe because I had to deal mainly with russian (cloned) logic chips.

One thing I definitely would like to realy use in some freaky design someday could be ferromagnetic memory board (i've got one 1/2 meter by 1/4 meter in size for just 80bits!) Realy try to upset it with nearby sources of noise (DC brush motors, etc) and see how contents get glitched up.

Also code execution directly from perforated cards or tapes doesn't leave my mind for ages, but can't find a working punchcard recorder/reader/tractor anywere around.

I'd remind that running it faster than a few megahertz is highly unlikely unless you can make a high quality PCB with short and length controlled traces instead of breadbording with wires.

I've read on another forum how bad bread boards are for higher than 3-4mhz. Even some 65c02 SBC on a bread board @ 2mhz was supposedly pushing it (because of the extra edge case sensitivity in 65x projects verses some other processors).

I've also read that wire wrap boards can go quite a bit higher. Things start to get into the analog side of the world of electronics that I'm barely familiar with (probably barely enough to get by for my needs).

I personally like the idea of using 74 discrete logic chips. If the logic is laid out, it wouldn't stop someone from replicating that down to some GALs or in a CPLD. I have the logic worked out for most of what I want to do in video and the logic itself is pretty simple. But I'm still in the process of translating that into 74 series logic. And like you said, real world timing issues can come into play. Hopefully I can keep track of this allow timings to function a bit loose for any real world issues that I can't foresee.

I might be a bit of a masochists I guess, but I'm having fun I've seen people do crazier stuff (an emu author has built a working processor on nothing but relays). I would like to get into some DIY GAL stuff/programmers then eventually some higher level stuff like you've mentioned. Actually, I'd love to build my own processor all in 74 series some day So I guess I am a bit hopeless