It’s the only way to do it. This was never a “here is exactly how to build this thing I’ve already made” video series, it’s a project log. We can all just be happy I never hit a brick wall that needed a total rethink.

James does not make mistakes. He gains experience.

@@JanBruunAndersen Heh, something like that. Though I see no reason for it not to be a "yes and", instead. ;) (i.e. James makes mistakes, and from hem, gains experience!)

Glad you liked it! It was easier to explore possibilities once I'd confirmed it was the only serious issue.

He who never made a mistake is just better at hiding it! 😉

It’s very similar to program debugging once you get in the right frame of mind.

Glad you enjoy it!

Indeed. That software "bodge" [I think of bodge as a hardware term, somehow, so hence the quotes] was great! :)

The most painful thing about making mistakes in this project is having to edit the video watching yourself make the mistake in slow motion! :-)

Very kind words. I'm aware of Sam's stuff and Jeri Ellsworth's earlier work. Scary chemicals in those processes though so I'm happy build from discrete logic!

I’m thinking that the adapter might make a good proxy for it, but I also have half a mind to mount this backplane and then stick all the other mistakes into it. I wonder how many boards I’d. we’d to make to get a second working build.

Yes! I was overdue a mistake. Thanks!

I heat up the solder and blow with my mouth. Of course, neither of these solutions are "professional", but they work. Your note about the table surface applies...

Interesting. I'll have to give that a try.

Yeah, I could see a bunch of ways to solve a problem like and it was worth showing different ways to do it.

Very generous of you! Glad to hear you are enjoying it.

But the led's wouldn't have been right!

Good to hear you are finding them interesting! I'd find it disingenuous to hide mistakes, this is a project log after all!

Lol! In the same I get close to Mars when I climb the stairs! (At the right time of year)

Glad you enjoyed it! The honest truth is I wouldn't enjoy this kind of stuff as much if there were never any problems to figure out!

Thank you very much!

Thanks gorf, always appreciate your support!

Thanks! I get a kick out of solving problems so the highs and lows of "it's broken" & "I've fixed it" are often better than if all just works first time.

You're welcome!

I'll finish it eventually!

@@weirdboyjim This stuff always takes twice as long as you'd planned, and three times as long as you hope.

I agree! If you never make mistakes you are probably not pushing yourself!

Thanks! I’ve already done a bit of a tidy up of some signals that might have been a little vulnerable to cross talk.

Any chance that'll bump up the stable clockspeed a percent or so? Or are you strictly limited by the speed of your ICs?

Thanks!

Glad you like it! Try not to put any buttons on backwards!

Behind the scenes: I worked out the connector issues before I finished editing the previous video, I designed the adapter board and ordered it before the main troubleshooting ad a just in case. If there had been a longer wait I might have been more biased towards the bodge.

@@weirdboyjim I've always loved the dodge; it's pitting your mind against the problem at hand. It’s was one of the things that helped with my move from production test into R&D. way too many years ago. 😞 I used to take rejected production boards and rework them in to test jigs as an extracurricular activity to speed up testing. The technical director in the end said that as I was borrowing their kit so often to reprogram parts etc that I might as well move over and become a design engineer. So bodging has always been in the blood and something i love to do.

Thanks andymouse!

Thanks so much 😊

Thanks Amaury!

Ouch on that Board! I've been fortunate my mistakes haven't caused any extra damage.

Thanks! I'll keep going, it'll all be done eventually.

Yeah.... if only you could do a fix in easyEDA and as soon as you click "save" it's fixed.

Being able to turn all the led's off might be handy, but I really like having them there!

@@weirdboyjim I think the LEDs should definitely stay, just that they have a separate return path to the USB connector (like a star ground system).

Kind words! I'm sure the apparent humility is just a by product of trying to slow myself down, I can barely understand myself when I talk too fast!

@@weirdboyjim When I'm talking about the humility it's more like - You are the first one to admit making mistakes (and it makes you relatable, believe me). But then it becomes something positive because 1) You try to determine WHY you made the mistake, which can be just, oh well, wasn't focused on what you were doing (s..t happens) or something more profound that allows you to adjust your thought process (more like, not that you didn't do it right, but rather that it was the way you thought about it that needed some adjustment) 2) Gain something from that mistake, devise a plan to circumvent it with the less effort possible 3) come up with an even better or more clever solution than the original plan ! So any of those mistake may wind up as a game changer.. Or maybe a honest to god brain fart 😄

​@@ivanscottw Video editing gives you a new perspective on this, you get to watch yourself make mistakes in slow motion over and over again! On the plus side I have almost grown immune to cringing at my own voice! 🤪

Thanks! I've been fortunate with how few serous mistakes I've made so far.

I won't be entirely happy until I have a new backplane but at least this version lets me get on.

If it wasn't for the led's displaying the status on the pipeline then the fix would have been fine on it's own!

Thanks!

You really don’t have to! But I appreciate the sentiment more than you realise!

@@weirdboyjim You're very welcome! I don't think you appreciate how much of a resource you're providing for others! People will be able to watch these for years to come! Just wish I had more free time! 😁

Best way to learn is to experiment yourself!

I was very relieved when I realized I was going to be able to fix it.

Always the way!

Your project could make for a fantastic learning tool btw, at a computer museum. (like a far more compact version of James Newman's megaprocessor at Cambridge.)

In some ways finding mistakes and fixing them is more rewarding than never making them!

@@weirdboyjim Yes, my projects tend to be very "rewarding" in that way, too. haha

I did think about putting this adapter board directly onto the backplane, I'd have needed to de solder the existing headers though.

Why thank you Bob! When I realized the connector was reversed I paused what I was doing and designed that adapter board immediately. I didn't go for ultra fast delivery so I actually got away with an order total of $5.34 (including the once a month $8 discount). I half expected to not use it. Much cheaper than rushing into a new back plane (which will cost much more than that in connectors to go on it).

@@weirdboyjim it wouldn't the same cost to rebuild the pipeline board?

@@weirdboyjim nevermind. Just read the "less satisfatory" answer below.

@@cussuol Getting the pipeline board remade would be a similar price, but I'd have to get new components AND have an annoying routing job if I'm going to keep the Led's in the intended places.

I was a little embarrassed by this one.

I actually spent time calculating the timing implications of trace length difference and it really isn’t much at these clock rates. Much more import to iterate the design away from long dependency chains.

@@weirdboyjim Absolutely. I know it doesn't matter. Just makes me twitch, what with my professional high speed PCB background. 😁

The monitor is slowly evolving towards an OS.

Wire-wrap has its own kind of beauty. A minicomputer with 18-inch boards spaced well apart with the area between filled with wire-wrap lets you envision how the circuits are woven like the pattern on a sweater.

I can see how this might have been easy to solve with wirewrap, but what if those connections were done early and lots of others crossed on top? I have no experience of how that scales.

@@weirdboyjim its not that difficult to make a change. essentially just unwind the 2 or 3 levels above (should it be on the bottom), then move them. The only issue I've come across is that wire wrap wire is solid and about 0.007 inch diameter (30awg) so it only tolerates being wound and unwound about 4 - 5 times before it breaks. of course you must have enough slack else you'll find yourself replacing the entire wire from point to point. the other issue is wire wrap IC sockets are rather expensive.

It would have been handy here in hindsight!

You are so welcome!

@@weirdboyjim I really love the way you operate. Good man!

Modern motherboards have 2-3 orders of magnitude tighter timings though. I doubt James has to worry about trace length just yet.

@@benjaminmiller3620 I agree with you. The signal will travel along the trace at the same speed regardless of the clock speed so it's a concern when overclocking or on the speeds of modern hardware. I still wonder about the signal timing. The motherboard has been overclocked before and the chips were operating successfully outside their documented timings. I'm just curious about the latency and if a new backplane revision is coming why not investigate latency and incorporate modern practices as a learning exercise?

I did run some maths on this stuff in response to another comment. At these clock rates it's not a big deal, one day I might need to worry about it. I really like the modular look of this build but the extra trace length probably wouldn't help if you push the clock a lot higher.

​@@weirdboyjim Seeing your processor already push the limits of it's chips was really impressive. Back when you had the processor on breadboards you had a stable build. It's great you can reach a high clock speed. You make building a processor look easy and that's a credit to your skill and knowledge.

I think it was onscreen on the previous video. It was $79.70

@@weirdboyjim Pricey, but for me it would be worth doing again. But that is probably my autism talking :)

I shall indeed!

If you watch my critical path I can account for the clock rate limit just by the latency of some of the circuits, so I'm not inclined to push it much harder.

I know you would have!

I’ve thought about it, I’ve checked and it’s actually about 50% more which is not as bad as I expected. The bus control board was something like 5x but It’s obviously not a linear scale.

@@weirdboyjim Ok cool, it seems to be going ok so far. I'm just worried you're going to have intermittant problems with noise with such a large board with no GND plane, as routing GND using tracks could result in the return path from some daughter-boards taking extremely circuitous routes back to source (can cause ground loops). In your case you don't have any really high speed signals, but a multi-layer setup can also help reducing EMI. You can then also use stitching vias next to the faster signals to maintain ground reference integrity when transitioning between layers. I'm going on a bit, as it's a subject I have to contend with daily (work on a lot of RF designs). Anyway, great work, looking forward to the next stages of the VGA implementation!

That was one of my early ideas, it would work but it wouldn't be very satisfactory.

Thanks for mentioning that. I was not previously aware of these projects. This is definitely where tech needs to go next!

A couple of people have been trying to get FPGA versions up and running but it's not one of my goals.

Well it will end up on it’s own board that plugs into the expansion port. You could do something like isa but it would be a lot of effort so I wouldn’t do it without an additional goal in mind.

Economical yes, but far less satisfactory!

I’ve been thinking this pcb once I’ve replaced it can become the host for all my other bodged boards. How many will I need to remake in order to have it working I wonder.

Hmm, power button. A physical switch would be easy but a digital toggle using fets would be more interesting.

​@@weirdboyjim (Digital Button && relay switch) || (button + powerful mosfet transistor) 😉

Good outlook!

No plans for that, it's very much designed to be sat on a desk with it's gut's exposed. I may do some designs in the future that would be more suited for conversion to an fpga, that would make it easier to play with concepts such as multi-core without needing to double the circuit up.

Good question, it isn't making a difference here. The CPU is clocked at 4Mhz and signals will go through a wire at about 90% of light speed. At most my bus traces are a couple of centimeters different so if you multiply it all out that track variance is about 0.03% of a clock cycle. Not something we need to worry about but at higher clock rates it can become significant.

On the backplane? I suppose I could but it's not going to be seen very much.

Power switch isn't the worst idea, had to plug/unplug this a few times during the troubleshooting.

Yes, but would be difficult with that header, as long as things are just outputting a standard signal I don’t think there is any way to trigger anything bad.

LS are actually a little faster than HC/HCT. I've not had any difficulty getting either from the larger suppliers, mostly mouser.

I'm buying from china. It's easier with tax and customs. I'll check your references. Why don't you use LS if they are faster than HCT? Nice videos BTW. I've seen almost all of them ;0)

Thanks Peter.

I did mention that as a possibility in the video. I'd either have to do some very tricky routing or accept the led's being in a different order than I'd like.

I've had various offers from multiple pcb vendors (including them) but nothing has quite felt right. I'm very hesitant to take on commitments.

I have a simulator for testing code on but I don't think there are any tools that would have protected me from this mistake.

@@weirdboyjim nor do I. Ideally you'd want code to verify that the pcb layout matches your circuit diagram by firing test signals into the pcb and associated circuit diagram and making sure you get the same output. Then presumably if you integrated all your circuit diagrams into one logic simulator you could test code on it. I'm guessing its the PCB to circuit diagram testing that's not available?

_We're going to need a bigger frame_

I think this one was late evening

This series needs a financial recount like the gold mining and crab boat reality shows have at the end.

Did notice "transparent pcb" was an option at one PCB printer the other day. Back plane? Blinken front plane!

Fortunately that connector was just signal outputs. One with power on would be trouble!

Those wouldn't be too tough to add as peripherals. I did actually get myself an RTC chip for some testing ukposts.info/have/v-deo/qJ6lg494b499mas.html

@@weirdboyjim Can a Timer/Counter and RTC be made and implemented from such logic components? - Similar to the other CPU Components.

If you want the fastest possible bit reverse I'd do it with a divide and conquer strategy swapping neighboring bits, then pairs, then fours and finally swapping the high and low byte. It would out perform the code you posted by quite a margin. The code you posted is better than what I wrote on the fly but that was on the spot code while focusing on a circuit problem just to test something. Your implementation looks ok at first glance but the compiler will never turn it into an optimal implementation.

If you ever get access to a time machine please sent that to me a few weeks ago!

Surely that would apply better to the breadboard version with messy wires?

Do you mean the ROM's in the pipeline or the program code roms? For the former I wrote some custom code in c/c++, it's a bit rough around the edges but gives me a simple way to generate the data. The program code I generate with my own assembler.

Your REVERSEW() uses what I call a divide and conquer strategy. It’s fun to play around with clever bit operations like that but you have to be careful. I’ve written a lot of very tightly optimised functions of this sort over the years (I have a reputation for it) but they take time to get right and tend to involve lots of care and debugging to be sure all the cases are right. Your implementation for example would not have worked, it would have set all the control outputs to zero (the word length was wrong). The function I wrote was not great from a performance perspective, but it was just a bit of quick test code written on the fly while recording a video to a prove the issue. I was balancing both confidence, speed of writing but also clarity of what I was doing to the user. I ran it once as a test and then threw it away so I think not writing optimal code was justified . If you are interested, here is the d&c version of the function I wrote: unsigned int BitReverse16(unsigned int rhs) { unsigned int out; out=((rhs&0xaaaa)>>1)|((rhs&0x5555)2)|((out&0x3333)4)|((out&0x0f0f)8)|((out&0xFF)

@@weirdboyjim that is the 16bit version of the 32bit version but there is other bitreverse() is actually part of the FFT(long double[],long double[],int,int); and has to be very fast, I have since rewritten the whole FF() routing in machine code and is again order of magnitude faster.

@@DAVIDGREGORYKERR the one you posted was actually 64 bit. One of the reason's you have to be very careful with this stuff is 64bit "clever" operations like this on a 32bit cpu often create terrible code depending on the compiler and architecture.

@@weirdboyjim i can use the statement short unsigned int A.