although mainly the 74HC139 and the 1.8342Mhz oscillator orders are still at the 'airhub in china', the first batch of harris 65c51's did arrive, as did the mainboards, so the first board is now running..


after some initial illegible brabbelings due to the system clock being bridged to the rs232 clock and the baudrate being something absurd, (good enough for the scope, not good enough for a terminal some other workaround was found in just using a series crystal directly hooked to the pins in a double stacked socket while we wait for the proper parts



one day this should become a line of PLC's, as well as ATX sized 65C816 mainboards to replace our pcs with 'nsa proof' stuff that only does what WE want it to do.

so far no reply from that CS8900A company on the cost of a mil-spec cerdip 40 version of their chip so no ethernet net. (not gonna use any of that smd crap that is only held in place by the pcb traces. we're not -that- cheap, plus they'll have to get irq's to work in 8 bit mode or we'll just have our own chips made elsewhere.

Aha, PLCs.

So you are new to PLCs, EMC issues, ESD shielding of computers and such ?


Hm... why not just building your own CPUs, peripherals and Ethernet controllers from TTL chips ?
It isn't that difficult...

It might take years though!

But on-topic, I would like to know more about this project, or product, or whatever it is.

<-- so far for 'interference' besides the thing being one big groundplane (which, admittingly, is some sort of antenna by itself, not much we can do about that as the cases will also be full metal anyway so turning it into a 'raster' won't help much) also you'll notice foil capacitors of 1uf not just at the rs232 line drivers but ALL over the board (need to take one apart, see which side is the outside of the foil, and somehow find a way to make pick and place robots recognize that, as well as adjust the footprints to indicate the plate side for manual assembly)... on the first testboards i was like 'screw that just get one working for now' 4 more boards to do radio interference (both way) testing with and overvoltage testing etc

these boards are to replace the emulator with. (which only emulates 1 acia and no vias so far anyway). as for 'emc' there is this thing called 'metal cases' usually comes combined with something magical called 'IP65'. (you know, the type usually used for wiring boxes on ships and such

although the vias will probably drive solid state relays just fine (65kohm dc resistance, trigger at 3v, at least the 10 amp ones i tested with driving anything else directly with vias obviously requires a bigger board. also 24vdc seems to be the going standard in that field so the psu will be adapted to that as well. the current boards run of the 5v rail of an atx psu and that leaves no space for extra diodes and such or else it drops well under 5v

all caps are 1uf, and the whole thing is one big groundplane, even on the testboards. (could make some adjustments to kicad to drop vias every 5cm to connect both gnd layers on both sides or turn it into a ground raster (Which seems to workk better) but as the cases will always be metal anyway, and as for the plc's all interfacing will run through opto couplers, (which when combined with a zener diode and resistors limiting the voltage and amps also make for perfect universal inputs that can give a '1' on pretty much any voltage surrounding 1-40-ish, including ac with software debouncing) interference is not something i'm very worried about

looking at the 'competition' those toshiba PLC's (which btw market for several thousand euros -each- even second hand it's one big solid aluminium case.

as for pc replacements... pc's are one big source of radio interference anyway and always have been. the board is the -least- of the worries, try those atx psus and run them next to a spectrum analyser in a faraday cage. lol. priorities in 'fixing things' would include -that aspect- it's usually not the board causing the rf meters on your tranceivers to go up a few points constantly when placed in the same room as pc's... it's the psus.

as for building the ethernet out of 74hc logic, i looked into that, stumbled upon the original xerox alto ethernet board, and went like 'nah' anything up to the 'integration level' of an 8 bit ne2000 isa card from the 1980s would be acceptable tho, just with more modern magnetics and proper trace impedance. and just twisted pair but preferably a single chip solution. but a -solid- one. no 100 pin smt crap that does a lousy job at a whole lot of different things. preferably one designed to work perfectly as what it is, an 8 bit 10mbit ethernet tranceiver. (not some repurposed thing originally designed for 16 bit pc cards with useless eeprom interfaces and internal address decoder and just '5 volt tolerant inputs' (meaning it's actually a 3.3v device on which they removed the protection diodes) which besides being able to do all that stuff we don't need -does not- have working irq's in 8 bit mode. which we -do- need (as in that case you can just switch the whole thing into standby mode until a command arrives or something else triggers an irq or nmi)

as for 'it might take years'. this board took about 2 days (each production round of test boards takes a week to ship from the factory tho - we do have a cnc pcb mill but it's 500km away and also isn't exactly 'real time' . after the initial versions for trying out in which order to place/rotate the chips so the busses would be somewhat in the same spot without a million vias. (notice how nintendo solved that by just making the cpu and the ppu mirrored images of each other in the pinouts, something we can't quite do yet for the 6502 and it's periphial chips ourselves kinda odd that with most of the the busses in the same order the 6522's are 'upside down' from the 65c51s but so be it

for the atx version it would probably suffice to just print out the top layer on paper, insert the connectors and try wether it fits the average atx case and existing blindplates. but as the atx version is supposed to get the full 16MB of memory (preferably not in 32KB each sram chips but somewhat larger sram chips and would not be much use without the definitive ethernet version (although 8 bit isa bus slots will be included so just like it can hook up to 8 bit isa video adapters of any kind it can also hook up to any 8 bit ethernet adapter on isa) and such that is not much use yet.

other than that, commodore IEC bus on everything. need to work out how to get it to properly implement on a 65C22 as it seems the bugs that originally affected it's functionality on 6522's have been fixed a few decades ago. also need to figure out a way to emp-proof that a bit. as commodores own crap isn't exactly resistant to 'hotplugging' let alone deliberate tampering.

it seems like umc is selling cloned nintendo nes/arcade (actually the same chip with the color bits swapped around in some versions) ppu video chips to anyone for clone nes'es so including multiple of those could be considered (low resolution is acceptable, as long as it has sufficient outputs for individual displays). but they can only do 30x8 bit wide char wide displays (besides the useless game related stuff). (seems to be the only directly interfaceable video chip on the market atm tho. unfortunately good old 'AT' din 5 keyboard interface (whomever came up with that crappy minidin thing always falling at ibm out should be hanged and good to go onto the desktop.

minicom on linux doesn't seem to be playing along with the CTS/RTS game regardless of the 'hardware handshake' yes or no setting nessesitating a little loopback cable. other than that it's the normal pc mainboard header pinout not the directly wired IDC crimp pinout (not sure but i guess i prefer the direct IDC pinout for later versions, the pc connectors are kinda flimsy on the inside with the soldered contacts all over the place just to keep them in db9 pin order the 1:1 IDC with crimp connectors on both ends seems much more sturdy). probably the next version will have that max232 version with the correct number of tranceivers to also cover dtr/rts -and- some way to make either or both of cts/rts dtr/dsr do a local loopback as it is a rather annoying aspect of the 65C51 that it can't transmit in any way if CTS isn't asserted. (probably using some pins on the first 'system control' via that'll also pull reset and control some lcd display - controlling 1602A lcd displays directy seems to 'work' but their datasheets (mostly in chinese) say something about 1350 nanoseconds delays on some actions so that's definately not a 'guaranteed to work' setup. plenty of via pins for that.

hmm either that or the single nullmodem cable i've tested it with so far has seen a bit too many racks and tie-rips lemme check that thing. lol. all of the ones we've bought over the past 2 years are red. this one is still 1990s 'beige' lolol. could be just a screwed nullmodem cable lol.

0 volts from the pc side of things. on all handshake lines. not -10 not +10 just 0 on either of the 2 onboard rs232 ports or a usb one. well done minicom (or linux) for screwing up something simple people already got right in the 1960s.

65c02 board on the rs232 tester: txd green (+10V), rts red (-10V) DTR nothing (logical as it's not wired but looped back to the DTE)
WTI CMS-16 ethernet console switch on the rs232 tester: TXD green, RTS and DTR red
WTI remote powerboot+: TXD green, RTS and DTR red

PC running minicom regardless of hardware handshake setting or port used on the rs232 tester: just TXD green (that being said it doesn't 'ignore' CTS, it just doesn't seem to assert DTR in any way). whomever fucked up minicom, unfuck it. now. lol.


ok... time to configure something the pc/linux/minicom cannot screw up. now if only i remembered the dip switch settings to those wti things or the ip address of that zyplex rs232 server. lol (probably still something in our old blocks, not in the new ones as as far as i see anything i plug into anything nicely

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

Diodes Inc. ZXMS6005 works nice as a relay driver, but it's too slow for PWM.
ST L6374 line driver even could drive small relays, but the chip is a bit... exotic.
ONsemi LA6500 is a nice OpAmp for driving small DC motors and such.
Analog Devices AD7608 8 channel 18 Bit A\D is worth a look, parallel interface and SPI.
Analog Devices AD5360 16 channel 16 Bit DAC, outputs need to be buffered. SPI only.
TI OPA2209 is good for buffering DAC outputs if you don't need much output current.
Maxim MAX6250A is a good voltage reference, might be needed when using the AD5360.

Nowadays modern A\D and D\A converters usually are not available in DIP packages, sorry.
Looking forward to see, what cb3rob uses as a watchdog timer.

I'll just point out that a solid state relay is not an electro-mechanical relay. The driving side is just an LED and a resistor. You only need a few mA, and there's no inductive kick-back. There's no need for external drivers.

I have to admit, that I'm having some problems with digesting big blocks of text containing only lower_case letters...
but sooner or later cb3rob is going to need drivers for something.

;------

Routing a PCB of that sort within two days is OK.

I just feel a need to add my two cents:

;...



For better EMI, try to avoid 90deg. turns in PCB traces if possible. Especially in power supply traces.

;...

A trick for better shielding a PCB is to have a polygon at the edge of the PCB, connected to GND by a 100Ohm resistor or such.
Polygon shouldn't go around the whole PCB, but needs to have a small gap somewhere (else, you would be building antennas).
With this, an ESD discharge to the edge of the PCB (or to the casing of the computer) would be less "confusing" to the computer.



BTW: picture shows "customized" octagon vias with a 3.5mm drill (for M3 screws) as mounting holes for the PCB.

Also, it's good to have thicker PCB traces going to connector pins to improve long term reliability of the design,
and maybe to remove some of the solder mask on the botton to give a bit more "flesh" for soldering the connector pins.
Hmm... since we are at it: marking pin 1 of every connector (and maybe every IC) with something like '<' makes a design more "service friendly".

If one via won't do for the current, just place a group of vias close together connected to the same trace.

the pin 1 are marked by having a square solderpad, in as far as they are ic's they also have the outline including the orientation hole on the soldermask, the component number is printed at the pin 1 side as well.

all dip packages use the large pad footprint, traces are as thick as i could get them while still being able to get one in between 2 pads and maintaining the distance to the pad.
vcc traces are slightly thicker than an entire pad

for the xtal oscs we probably should get our own sockets made. currently they're in dip 14 in which 'short formfactor' (dip ones can be moved down away from pin 1 (not to have a dead trace on clock lines, just have a dead trace on vcc under long ones any dedicated xtal osc sockets i could find only supported one of the 2 formfactors. (funnily enough most of those came from some 'arcade repairs parts' company, not from actual mass production retailers)

actual xtal osc sockets seem to have higher legs than dip14 so just getting a goldplated round hole dip14 and removing the middle pins won't do for production quality.

slowly the rest of the parts are dripping in, now all that's still missing is the 74hc139n in the address decoder (the big board sticking into it's socket is 2x74hc254 doing the same job, at least, as long as you don't wiggle it - good way to make sure the memory test works - which it actually does. and some proper power connectors (it actually is designed for the 'green' phoenix 5.08mm plugin things but i just ordered a bag of blue chinese screw connectors with the same legs for the test boards. (now if only we could get wima to make the caps in blue as well as so far the caps are the only thing that are not blue or black i think siemens foil caps are blue tho lol. the test boards also have the 'easier to insert' cheap sockets, not the goldplated round hole ones, as getting chips in and out of there on a regular basis is a pain in the butt. (they do stick a lot better once they're in tho


foil caps seem to work fine for everything including the rs232 linedrivers btw. no more elcos. elcos are ****.
production boards will obviously have screwholes with ground around them.