Lately I’ve been doing a lot of work on the rosco_m68k hardware, specifically on integrating the new Yamaha V9958 video board. As part of this, I wanted to extend the firmware to support basic “virtual consoles”, so that if a V9958 is detected it will use that for text output, or fall back to the UART if there isn’t a video card.
I’ve written up a longish post on Hackaday with the implementation details of the console itself (it’s worth a read if you’re interested in implementing a scrolling 80×24 text mode console on a thirty-year-old graphics chip), so I won’t go into too much detail here about how it works – I’ll just prove it does work with a picture:
One of the problems I faced while putting this together was that the ROMs I used for the rosco_m68k are only 16KB in size, and while the community has already come up with a workaround for this (thanks to RTS4E75!), I wanted the standard firmware to work on unmodified boards, so it had to fit in 16KB.
I also didn’t want to sacrifice any of the existing features supported by the firmware (because backward-compatibility matters, even in retro single-board computers!), so I was left with a bit of a dilemma – how to fit the video driver and associated text-mode font into the few bytes that I had left after the Kermit loader, Easy68k compatibility layer and basic machine initialisation code?
After playing around with a few different ideas, I eventually decided I would split the ROM in two stages – Stage 1 would contain the bootstrap code, and everything that needed to stay resident in ROM. Stage 2 would contain the program loader (so Kermit currently, though that can now easily be swapped out for something else such as the experimental SD card support we’ve been working on).
How does this split save space? Simple – Stage 2 would be compressed, and Stage 1 would be in charge of decompressing it from ROM into RAM and running it from there!
I’m not going to lie, I’m pretty pleased with this. I seriously doubt I’m the first to do it, but being able to take an almost-full ROM and fit in an entire graphics driver and text font, plus a simple Zip-like decompression algorithm (I didn’t write this, I used the excellent liblzg code – a LZ compress/decompress with a decompression written in M68k assembly) is, I think, something of a win. Making it all work together properly (especially tying up the parts in stage 2 that needed to call back to bits in stage 1) took a bit of work, but once I’d ironed out all the bugs and made sure I wasn’t doing anything silly (like overwriting the loader code with the bytes it was loading…) it all works really well.
Full Disclosure: For the purposes of this post, PCBWay supplied me with a few free prototype PCBs, so I guess you could say that this is a sponsored post. Nobody gets a free pass around here though – I’ll be keeping this review totally impartial!
By this point, I’ve had quite a lot of PCBs made out in China, with varying results, and I have a “go-to” manufacturer whose work I’m generally very happy with. I’ll refer to that manufacturer as Brand X in this post.
I’m always on the lookout for new and better things though, so when PCBWay got in touch recently to a) compliment me on my project, and b) offer me free PCBs in return for a review (this review, in fact) I was more than happy to accept. And not just because of the compliments (though flattery does go a long way) – I was genuinely interested to see how they stacked up against the producers I was more familiar with.
So here’s what I did – I sneakily had them make a board that I’d already (recently) had made by Brand X. I figured that way, I could do a side by side comparison and tell you about it. Sound good? Let’s get to it!
For the purposes of this, the board I ordered was the rosco_m68k bus board prototype (revision 0). This board is a local-bus backplane for my rosco_m68k Motorola 68010 project.
Technically, this isn’t a hugely challenging board from a manufacturing perspective. It’s a two-layer design, although the front layer is just one big copper pour, with all the routing happening on the bottom. The smallest trace width on here is 0.160mm (6.3mils) with clearances being the same. Everything is through-hole – there’s no SMT pads on here, no vias and no thermal reliefs or anything like that. In short, it’s a very simple board that should be well within the capabilities of any fab, and shouldn’t break the bank. Ideal for a side-by-side comparison.
The board itself is 160 x 115 mm, and for the purposes of the review, I ordered with the same options from both manufacturers – FR4 in blue, lead-free HASL, 1.6mm thickness and 1oz copper.
Comparison 1 – Order Process
Ordering PCBs is by this point second nature to me, and to be honest there wasn’t a lot to choose between PCBWay and Brand Xin this regard. Both websites just worked and ordering was a breeze – Upload a zip with gerbers and drills, get instant quote, set options and click Order.
The approval process did take slightly longer with PCBWay, but then this was my first order with them so I guess they check things a little more carefully. In any event both companies had passed my order to production within a day or so, so no big deal.
Comparison 2 – Price
I knew Brand X would be hard to beat in this category as they’re widely considered to be the cheapest around. And, pound for pound (well, dollar for dollar, but whatever), I was right. Here’s how the pricing stacked up for 5-off:
PCBWay’s shipping was a fair bit cheaper than Brand X‘s, but the board price is considerably cheaper. Cheaper doesn’t always (or even usually) mean better of course, but this category is about raw price, and we have a clear winner.
Winner: Brand X
Comparison 3 – Lead Time
I ordered from Brand X on the 26th April, with the PCBWay order being made the next day, 27th April. I was interested to see how shipping time stacked up on this board as it’s a bit of an odd size, and blue (both of these can have impact on manufacturing time).
Despite the one-day lead Brand X had, both boards were actually handed to the carrier on the 29th April, so PCBWay are the clear winner here. Both were within the promised 2-3 day time, but PCBWay smashed it by almost an entire twenty-four hours.
Comparison 4 – Delivery Time
I didn’t think it was fair to compare delivery times for two reasons:
This isn’t a comparison of international logistics companies
International shipping is generally hosed at the moment due to Covid-19
For those reason, there’s no comparison in this category (suffice to say both took longer than they would have pre-Covid, with neither one markedly worse than the other).
Comparison 5 – First Impressions
First impressions of the boards were good – both boxes contained the ordered quantity, and the boards themselves looked good.
To me, the PCBWay boards beat Brand X boards though on purely aesthetic grounds – the blue was, well, just more blue. I was actually surprised by the difference, as I’d assumed that everyone would just use the same raw materials, but the deeper blue did look (to my eye) much nicer.
A cursory inspection also showed some silkscreen issues on the Brand X boards that didn’t show up on the PCBWay ones. For example, Brand X‘s board:
You might argue that these differences are just cosmetic, but how a thing looks matters. When you’ve spent many hours designing a thing and it comes back from manufacturing not looking its best, the disappointment is real.
Comparison 6 – Assembly and operation
Another category where there’s no clear winner, which is to be expected really as this isn’t a challenging board to manufacture by modern standards. In fact, if either board had been difficult to solder or failed to work I’d have been amazed.
As it was, both Brand X and PCBWay’s boards were great to work with, a dream to solder (there were no problems with even the fiddly bits like the Micro USB socket) and worked perfectly. As one would expect.
Overall Winner: PCBWay
In this totally subjective test, PCBWay take the victory, if only by a nose. Are they the cheapest? Well, no. But as I said, cheaper isn’t always better, and when it comes to quality PCBWay definitely deliver the goods.
Is that worth the extra money? Only you can decide that. Personally, I’m looking forward to redoing this with a more challenging board and really putting them through their paces.
Will I use them for every board from now on? In all honesty, probably not. I have a lot of prototypes made that inevitably end up as scrap, and for those cost is the driving factor (given I can be assured they’ll function correctly, of course).
But will I use PCBWay again? Absolutely. Because most of the time, quality matters.
Also, their blue FR4 is, well, bluer. And I like that.
If you enjoyed this post or even found it useful, and you have PCBs to make, why not give PCBWay a try?
None of the links in this post are affiliate links, and other than the free boards that were supplied for review purposes I don’t receive anything in return for your attention or clicks.
When you’re in the market for Motorola 68k CPUs (as I am), chances are at some point you’ll turn to eBay. And sure, there are plenty listed on there, with prices ranging from super-cheap to fairly-reasonable.
However, as always with eBay, caveat emptor is king! When you receive your chips, have a good look at them. They should look like this:
Notice that the Motorola logo looks good, the text is kinda small, and the date code makes sense (1992, week 51 in this case).
Now, check these out:
Notice here that the logo isn’t complete, the text is bigger than on the previous chip, and the date code suggests these were produced in the very first week of 2019 (or maybe 1919!?). I’m pretty sure Motorola weren’t producing these chips at that time…
You can’t see it too well in this picture, but the surface of the chip is sort-of rough, suggesting someone’s used some kind of tool to grind the part number off…
It turns out that these chips are either fakes, or re-badged original parts, and they’re not 68010s at all, but rather 68000s. If they’re fakes, they’re reasonably good ones – they work quite well as 68000s. Of course, the 68010 is pin-compatible with the 68000, so unless you have code that specifically uses the added features in the former, you might never know these aren’t the real deal.
I’m not sure why anyone is faking (or re-badging) these chips – given that these cost me a couple of dollars each I can’t really see where the margins are. I guess someone, somewhere has a bunch of 68000s, a facility for re-badging them, and the belief that they’ll sell much better if people think they’re buying a 68010. Who knows, maybe they’re right (I mean, I bought some, so… 🤷)
Now, I know that some people like to buy a 68010 as an “upgrade” for their old Amiga or ST (and maybe MegaDrive too, are people doing those?). It occurs to me that some of them might inadvertently purchased one of these imposters, and because the software on those machines isn’t generally written to take advantage of the 68010s features, they might not know it.
If you’ve upgraded your machine and your worried, and you have the facility to assemble some code, the following will let you know if you’re running on The Real Deal or one of these rebadged 68000s:
movec.l VBR,A0 ; Copy vector base register to A0...
movec.l A0,VBR ; ... and load it back into VBR
If you’re running on a bona-fide 68010 (or above), this shouldn’t do anything. On the other hand, on a 68000 (even a cleverly-disguised one) this will cause an instant illegal instruction exception (vector 0x04). How your computer tells you about this will vary, but I’m guessing it’ll let you know…
This weekend I’ve had a lot of fun playing with a proof-of-concept Arduino-based IO device for the rosco_m68k. Along the way I’ve learned quite a bit about the Arduino, identified a few future enhancements for the rosco_m68k design, and generally had a fun (if sometimes frustrating) time. It’s been a while since I’ve done any real hardware hacking on the rosco_m68k, so it’s been a nice change.
This post will probably run a bit longer than usual, so the TL;DR is that I wanted to prove that the expansion connector and bus layout of the rosco_m68k could be used to make new devices that would map into the computers IO space, and I (naively!) thought that an easy way to do this would be to use an Arduino hooked up to the computer’s bus. It (finally) works, but it wasn’t as easy as I’d hoped, and along the way I had to learn a bit more about the Arduino’s hardware while also becoming re-acquainted with the MC68010 and MC68901 user manuals and spend a lot of time looking at logic traces.
If you just want to skip to the end, the code is on GitHub.
An exciting day here – after several months of delays (firstly due to me procrastinating followed by production delays due to Coronavirus) the PCBs for my retro m68k computer have arrived!
I eventually decided to have these made by JLCPCB in China, and I have to say I’m very impressed with the quality and finish. On unpacking I immediately noticed a couple of things I didn’t spot before sending the gerbers off, but those are my mistakes – the fab is actually pretty much perfect. Drill is spot on and the silkscreen is lined up perfectly.
Obviously how satisfied I am may change once I get the components soldered on, but I’m pretty confident that the board will be absolutely fine. I’ll update here once I’ve done that.
Time for another update on the rosco_m68k, my homebrew retro computer project. As always, regular updates can be found on hackaday, so please do check there for the details (and like/follow if you want to!) – this will be more of a digest of what’s been going on over the past couple months.
So I guess the headline thing that’s been happening recently is that I’ve started working toward having a PCB manufactured for this thing. I realised I needed to press on with this after I took the breadboard in to work for a geek-out session with some of my awesome colleagues, and had a totally nerve-wracking commute with the breadboard (it survived, save for one current-limiting resistor which developed a loose connection).
So I started work on reducing the component count (more on that later) and actually designing a PCB. It’s now pretty-much ready to go off to OSHPark for manufacture (Yes, I could get it cheaper but I like purple and, y’know, yield) and looks like this:
So that’s the hot-of-the-press headline (I finished it today), and I’ll update more once the boards are made. In the meantime, if you want more details, follow the project on Hackaday for (usually) weekly updates.
Lowering the component count?
In the original design of this computer, I went for a full 7400-series TTL design for all of the glue logic, and that was cool. I learnt a lot about designing with discreet logic gates, was forced to really think about the way I wanted my circuit to work, came to understand propagation delay in a way I didn’t before, and bought a lot more breadboard. That last point is important – fifteen 7400-series chips, while cool, take up a lot of physical space. And PCB fabs charge by the square inch.
It had been pointed out to me before (thanks, Ken!) that I could do with a couple of programmable logic devices what I was previously doing with ~fifteen discreet logic ICs, so I decided (in the interests of staying sane and solvent) to investigate. After a good bit of reading and learning, I’ve now replaced most of the 7400-series on the board with two ATF16V8QL (link to PDF) CPLDs.
This is a win in more than one respect – not only has it saved me board space but it’s made routing the PCB much easier, cut down the current draw of the board, fixed up some propagation delay issues and allowed me to (finally) fix the bug with the expansion select line. These chips were available at the time the CPU I’m using was released, are still in production today, and cost me around £1 each (delivered, based on 5-off, from Farnell in the UK). They’ve taken over all address decoding and IO glue logic, and I highly recommend them.
What About Software?
Things haven’t stood still on the software front – since my last blog I’ve fully integrated the MC68901, gotten timers and interrupts working, and made the UART work such that the computer can now talk over serial (via a CP2102 Serial-to-USB converter I had lying around). Work on an OS for this thing continues apace, and it already has some basic memory management, a fully async serial driver, and has been demonstrated with some (very) basic multitasking. My immediate plan is to implement a way to load software via serial (Zmodem) in the immediate term to make it so that, as I continue to develop the kernel, I don’t have to physically pull the ROMs to reprogram them.
It’s Going Well, Then?
I like to think it is, yeah. I’m having tons of fun, and this thing is starting to look like a product. Once the PCBs are made and I’ve done some more on the software, I think this thing will be at a point where other people might want one. Obviously it’s a 16/32-bit computer in a 64-bit era, but I kinda feel like it might make a great starting point for people wanting to really get to grips with bare-metal, without all the lies abstractions that modern architectures force on them.
The PCB design gets me to a point where the core computer is done, and the expansion connector I put in means I can continue to develop new bits without worrying about disturbing the ever-so-fragile connections on the existing breadboard. This will mean graphics, networking, and who knows what else.
And even if it never goes anywhere beyond the three prototypes I’m going to be building soon, I’ve learnt a hell of a lot from this project, and continue to do so every time I work on it – and that’s really all that matters, right?
I’m still having a ton of fun with the rosco_m68k single board computer, and regularly writing project logs on Hackaday. If you’re interested, be sure to follow the project there as that’s where the regular updates happen, but here’s an in-brief update on what’s been happening.
Firstly, the board now has a lot more wires:
The ROM and RAM are both now hooked up to the buses, along with a MC68901 Multi-Function Peripheral that handles IO.
The system is successfully running code from the ROMs, and the RAM is working fine. I have code that does a full write/read test of the on-board RAM, and other code that just uses it and both are fine.
There’s some new glue logic to support the MC68901 (the top board in the picture) which has upped the chip-count a bit, and I’ve completely redone power distribution which was starting to get a bit shaky due to the way the project has grown.
Everything is connected together via the buses in the center of the picture, and the control bus has grown to accomodate all the CPU control lines and the address decoder outputs.
I’ve had a bit of a nightmare this weekend trying to get the MFP working properly (See this and this on Hackaday if you fancy a laugh) but it’s finally working, and I’m almost at the point of sending “Hello, World” to a serial terminal. After that, I’ll be off to the races with this project – debugging is so much easier when you can actually output stuff rather than just run logic traces.
That’s all for now – as I say if you want to know more be sure to like and follow on Hackaday!
I’m writing regular logs on hackaday.io about the m68k-based retro computer I mentioned I was building a while back, but I thought I’d post a quick update here too. If you’re interested in the project, be sure to follow it on hackaday as that’s where the regular updates are going.
The main progress so far has been:
Get the CPU free-running
Decide on a memory layout, and design an address decoder to support that layout
Build the address decoder
Decide how I want to lay the buses out on the board for easier interconnection
Actually wire up the buses accordingly
I actually designed the address decoder a while ago, but gathering the parts and then wiring it up has taken some time – I finished it yesterday. Since then, I’ve been breaking out the CPU signals onto control, data and address buses, which I’m doing on an otherwise-bare breadboard in the middle of the main boards.
Right now, the rosco_m68k looks like this:
It’s still free-running at the moment (I’ve not wired the memory to the bus yet as you can see) but all the control pins I care about (basically everything except the legacy 6800 peripheral pins and the function code) are now on the control bus. The jumpers you can see there are configuring those control pins for free-run mode and will be replaced by connections to the RAM and ROM.
I did come across a couple of design errors while building the address decoder, some of which I fixed and others I’ve decided I can live with for now (e.g. an issue with the expansion select line I only discovered once I’d completely built the decoder and hooked up the LEDs at bottom left).
Anyway, that’s where it’s at right now. I expect that after another day’s wiring, it’ll be ready to run real code, and I can’t wait.
Somewhere around the time I got bored with Advent of Code 2018 (well, the challenges grew to become total time-sinks), I decided my project for the new year was going to be a homebrew retro computer.
I decided that, rather than something 8-bit based on the Z80 or 6502, I wanted to go up a generation and build something 16-bit.
Apart from gaming and a bit of BASIC programming on 8-bit machines like the Spectrum and C64, my first “real” computing was done on the Amiga, a machine which still has a special place in my heart today. For that reason, I decided I wanted to build something based on the Motorola 68000 series processor.
I chose the 68010 which (as far as I’m aware) was never used in the Amiga line, but which is slightly more capable than the 68000 (e.g. interrupt vectors can be moved around and it can support virtual memory when used with a 68451 MMU, though I’m not likely to do the latter in this build). Mine is rated for 10MHz, but I’m only running it at 4MHz at the moment).
The key differences between this and an 8-bit Z80 or 6502 build are:
Dynamic registers, so the clock cannot be stopped or single-stepped (this can be emulated by manipulating /DTACK though).
Asynchronous data bus and DMA support, with /BGR, /BGA, /BERR and /DTACK to take care of.
16-bit data bus, so more wires needed.
24-bit (with some caveats) address bus, so lots more wires needed, but 16MB of address space available!
Internally, the m68k family is 32-bit.
/UDS and /LDS lines for selecting upper and lower data bytes, so a more complicated address decoder and ROM/RAM in pairs.
No separate IO space, further complicating the address decoder.
More complicated requirements around /RESET and /HALT handling, especially at power-on.
I started gathering parts in January, and so far have the crystal clock circuit, the reset/halt circuit, and have gotten the CPU freerunning with /DTACK grounded and the various other bus management lines tied high or low as appropriate.
The address decoder is designed, and I already have the RAM and ROM chips ready to go. I’ve also built an EEPROM programmer because I’m too impatient/cheap to order one and wait for it to arrive.
All this means that, pretty soon, I should be able to get it running some actual code!
Fun times are ahead! If you’re interested, you can follow the project over on Hackaday.
Today is day 10 of Advent of Code 2018, the fun daily coding challenge I mentioned I was taking part in this year, and I thought it a good time to take a quick look back at my story so far.
I’ve still not managed to rank globally for any of the challenges (the closest I came was on day 5 when I completed part two 696th) but things are going reasonably well on our private BJSS leaderboard, and the challenges have mostly been fun. So far I’ve stuck with Ruby as my language of choice, although I almost broke out Unity 2D for today’s graphical challenge (Early-morning me obviously had grand ideas which were quickly set aside when I realised I could zoom out my terminal and stick with Ruby).
Here’s a quick rundown (with possible mild spoilers) of the challenges to date. The links take you to my solutions on GitLab, which obviously do contain spoilers, so don’t click them if you haven’t already done the puzzle for that day.
Day 1 was a very simple problem involving summing a list of numbers, and finding the first duplicated number. Nothing much to say here, nice and quick.
Day 2 involved finding characters duplicated two- and three-times in a string, with part two focusing on finding strings that differed by one character. I went with hash-based solutions for this one.
Day 3 was about finding overlapping rectangles. I made a simulation for this as the maths escaped me. This was the first solution that was properly object-oriented Ruby, and I got stuck for a while because I accidentally transposed my X and Y coordinates in one of the methods!
Day 4 involved finding the guard who slept the most, and (in part two) the minute in which they were most often asleep. Another simulation (I quite enjoyed this one).
Day 5 was about reducing a string by removing pairs until none remained. Regex and recursion to the rescue here – a really quick solution.
Day 6 was, frankly, a nightmare. I couldn’t really visualise the problem at first, and it took me a lot longer than it should have to really get to grips with it. Solved in the end, but after many hours on-and-off. I did learn something new though 🙂
Day 7 involved a dependency graph and another simulation. I had a lot of fun with this, and even made an ANSI-based visual mode so you could watch the simulation unfold in the terminal. Great fun.
Day 8 was a simple tree problem involving both breadth-first and depth-first traversal. The depth-first part had a little twist that meant it took a little thinking about.
Day 9 caused me a few problems – partly due to some vague wording in the puzzle, and partly due to my not understanding the problem properly for a while. It was made worse by the fact that I initially decided to write it with the wrong data-structure because I felt it would be quicker. After a few false starts I gave up on that and finally made it work using circular linked lists.
Day 10 was another fun one involving points and velocities to work out when the points (stars) align to display a message. Initially I planned to switch to something with proper graphical support (in a fit of complete overkill, I fired up Unity) but settled in the end on a Ruby solution after I realised I could make the output fit in a terminal.
And that’s it so far! The most fun I’ve had so far was probably on day 7, while my least favourite has been day 6.
Taking part? Let me know in the comments what your favourite puzzle has been thus far.