Welcome to Ben Eater's Builders!

[enso]

Here is my wire-wrapped version of Garth's design, using 3d-printed sockets.

Only the big sockets are homemade. The pic was taken before I added the crystal (I am taking 10MHz from Daryl Ricter's SBC-4, upper left).

[plasmo]

Single-sided copper clad pc board is an inexpensive medium for prototyping. The copper foil is a great ground plane, and you can score and peel off copper foil with a X-acto knife to form isolated area. I built this prototype 68302 SBC on single-sided copper clad pcb using salvaged 68302 quad flat pack and piggy-backed RAM/EPROM. The ground pins of 68302 were bent down and soldered to the copper foil while VCC pins were bent up to form a ring in the center of the package. The rest were point-to-point wiring with 30 gauge wirewrap wires. The clock was 16MHz but this type of construction can run 30MHz+ without problems.
Bill

[drogon]

Over the years I have looked at various board build "solutions" - I have a laser cutter, but it won't cut copper (it may etch photo or other etch resist though). There is also the CNC router route - I've seen some good results with those, back in the late 70's/early 80's I did it by hand with rub-down transfers or, when we got super sophisticated, we'd design using some CAD (On an Apple II IIRC) then print it, photocopy to acetate then use that on photo resist board.

Modern chemistry and tanks have improved a lot too, but I find it's just too easy to either use stripboard or just upload a gerber to JLCPCB...

Maybe I'm more lazy than I thought!

Cheers,

-Gordon

[Agumander]

drogon wrote:

I have a laser cutter, but it won't cut copper (it may etch photo or other etch resist though).

I've prototyped a couple of small boards using my laser cutter to engrave away a coating of black spraypaint. It probably takes about as much work as a method using laser printer or a photoresist. I think the main benefit it gives is consistency.

CNC routing is nice since you can skip the acid step entirely, but you do need a way to deal with the dust it generates. It's also critical to fully probe and level the bed with every new board or tool change. A better use of the CNC machine IMO is to use it only for holes, and use another method to etch the traces. Especially if you only put one hole size in your design, so you don't have to change the bit!

[enso]

Likewise, I've used my lasercutter to vaporize a coating of resist. I never found a really clean material - some residue wound up baked into the foil, and the results were so-so. I was considering making a wax/ink/solvent solution but never followed through.

I've used the lasercutter to reflow a board, by focusing the laser onto the pads, but it was slow and kind of pointless -- a hotplate is a lot more useful.

At one point I was thinking of coating the board with a solder-paste solution and melting the traces - tinning them in the process (and wiping off the rest). My knowledge of chemistry is insufficient for finding a solvent that dissolves copper but not lead/tin resist.

I use the lasercutter to make stencils for SMT solder-paste. I use coated sticky labels, which makes it really great.

[BigDumbDinosaur]

If there is any one thing in favor of using a PCB it is that of component density. A dense layout means short paths, which means transmission line effects are less likely to cause problems. Also, multiple layers are possible with no more effort than laying out a standard double-sided PCB. Multilayer boards can be more dense than double-sided, and more convenient to lay out.

I'm patient. I can wait for the time it takes to receive an order from a board house.

[J64C]

BigDumbDinosaur wrote:

I'm patient. I can wait for the time it takes to receive an order from a board house.

They are extremely quick to deliver too, I find. I've had a few orders now and they have always been on my doorstep within a week.

With respect to signal lines, what sort of distance would you recommend they be separated by? So far I have packed them fairly close, which I figured might have been a bit of a 'no no' at the time. But, I've never really been able to find an answer to recommended separation distances.

[GARTHWILSON]

BigDumbDinosaur wrote:

If there is any one thing in favor of using a PCB it is that of component density. A dense layout means short paths, which means transmission line effects are less likely to cause problems.

For thru-hole parts, wire wrap allows placing the sockets with no space between them, meaning density is as good as you can get with thru-hole. You can also put skinny DIPs under wide ones, as shown here:

Connection lengths can be "as the crow flies," as short as possible, without difficulty in routing. There will be extra length from the wire to the ICs' pins (which is not the best for ground and power connections), but the maximum connection length across the board is reduced.

The PCB has the density advantage though if you use SMT, and even more, if you put parts on both sides of the board. (Some parts these days are not available in thru-hole anyway.)

Solderless breadboards definitely don't give anything resembling a transmission line, only antennas and inductors with tons of crosstalk. The only way you could avoid that problem is by twisting every signal wire with a ground return wire connected nearby at each end, something that's not practical.

[GARTHWILSON]

J64C wrote:

BigDumbDinosaur wrote:

With respect to signal lines, what sort of distance would you recommend they be separated by? So far I have packed them fairly close, which I figured might have been a bit of a 'no no' at the time. But, I've never really been able to find an answer to recommended separation distances.

Ideally they would be much closer to the ground plane than to each other; but putting them close will make it easier to keep the layout small and hold the maximum line length down. So I wouldn't worry about it. Most board manufacturers won't start charging extra until you get below .006"/.006" trace & space. I'm thinking of a web page illustrating the effects. I'll see if I can find it again.

[J64C]

Nice! That would be awesome!

[BigDumbDinosaur]

J64C wrote:

With respect to signal lines, what sort of distance would you recommend they be separated by?

I've mostly used 0.025" center-to-center, with 0.006" wide signal traces. That produces a spacing of 0.019" between adjacent traces. When running those traces at a 45 degree angle, the spacing is reduced to 0.013". I could go tighter, which most board houses support, but that spacing allows me to place a 0.026" via (with 0.008" hole) in a trace running parallel to others without having to dogleg the adjacent traces to maintain clearance.

Your favorite board house can provide you with their design rules so you can verify that your design doesn't jam things too close together.

[J64C]

Cool. I've mostly gone with fairly wide traces at around 0.0085". I'd have to check the center to center side of things.

[enso]

GARTHWILSON wrote:

...The only way you could avoid that problem is by twisting every signal wire with a ground return wire connected nearby at each end, something that's not practical...

I've been toying with an "experimenter's" '02 board which is mostly a well-decoupled 4-layer PCB (busses, power, clock, some control lines) and a wire-wrapping area for decoding. I broke out high address lines and control lines with a ground pad next to each one, to allow just that - a Cray-like twisted-pair ground/signal wire-wrap connection.

Since there is only a handful of these, it may be practical after all.

[unclouded]

Agumander wrote:

CNC routing is nice since you can skip the acid step entirely,

I love it for this reason. Even with HCl and H2O2 that can supposedly be rejuvinated, storing chemicals is not my favourite.

Agumander wrote:

but you do need a way to deal with the dust it generates.

I like FR1 rather than FR4 because the dust is supposed to less dangerous, but yeah, a respirator is definitely required.

Agumander wrote:

It's also critical to fully probe and level the bed with every new board or tool change.

I wrote G-code for a full Z probe map and it makes for a fun visualization when Z is scaled up but I'm too lazy to do that every time. Z depth is still critical however, so I just cut deep enough to avoid problems even though this approach does limit the minimum trace width and spacing.

Agumander wrote:

A better use of the CNC machine IMO is to use it only for holes, and use another method to etch the traces. Especially if you only put one hole size in your design, so you don't have to change the bit!

Totally, one hole size (0.8 mm). I bore holes to 0.8 mm at the surface, isolate and then cut the board out with the same 10° cutter because it avoids tool changes. If there are few holes then I can drill through manually. For many holes then it's worth changing to a 0.8 mm drill or endmill but then I have to change again to cut out the board. Sometimes it's nice to pause the job prior to cutout to manually buff with a cup brush on a rotary tool and do an e-test while the axes are still homed if the board needs further work. Doing three boards at once saves on time per board of course.

Lately I've been spraying the component side white then putting the boards back on the bed for some laser etched silkscreen, which ends up looking quite nice.

[Individual_Solid]

floobydust wrote:

Beyond being a copy from Garth's primer (the internet bible for 65xx newbies), Ben tends to explain circuitry reasonably well, but is completely absent on critical areas. i.e., how you take a design and implement it into a properly functioning circuit that can be implemented by anyone.

Perhaps a follow-on to this topic would be areas that should be sorted out, like the reset circuit... proper power supply and bypassing. Also, a simple warning to those who buy the kit... having it work reasonably well doesn't make it the perfect candidate for a PCB layout... there's more to it.

There's definitely something to be said about the accessibility of Ben's videos. The primer has a whole list of pre-requisite knowledge that can be hard to get a grasp of (in context of the project or just in theory). I think it's important to note that Ben's uploaded a lot more videos than just the 6502 series and quite a few of his earlier videos do add detail on how transistors work, how gates work, how logic is composed, how to read datasheets, etc... A favorite is the one where he shows off why building these circuits on a breadboard is probably a bad idea and introduces concepts like capacitance and inductance in a visual and also concrete way.

When you're creating videos where the level of prerequisite knowledge is a lot lower than what's needed for the 6502 primer, you're gonna go light on some details. I definitely don't know what I don't know, so there may be some really obvious topics he's skipped over the years. On the other hand, if you don't know anything, it's a very good starting place! It's an especially good starting place paired with something like Garth's primer - where you know you'll be doing a lot more googling.

I'm still trying to make my PCB work, so I'm certainly not authoritative on it (or any topic), but I have kicked around drafting a list of tips for success if you're trying to do the Eater build. As floobydust notes, a lot of the questions I've seen on here and on Reddit tend to rely around power distribution, clock, reset. A lot of folks on the reddit independently come around to using an Arduino to provide a cleaner single step clock than the 555-timer module (which Ben admittedly did not design for use in this circuit!) Decoupling caps are mentioned offhandedly in at least one video, but definitely deserve a better discussion. The most egregious problems often involve floating pins in an early step of the build. That's one thing I wish Ben would be a lot more vocal about doing correctly. For reset, I'd like to actually experiment with a DS1233 (and the required caps/resistors) - not quite as convenient as the DS1813 but it's accessible from Jameco, so an easy recommendation to cart for anyone who is planning to buy the Ben kit.

All in all, I think Ben's videos do a fairly good job of teaching you to read the datasheets and start thinking outside the box when troubleshooting. Some show great examples of how to dig your heels in when something goes wrong. But what's been drilled into my head while lurking here on the forums is that there are simply too many variables to play fast and loose with everything. Yes, a circuit may work fine under some conditions and not under others. You can't possibly control for everything, so good engineering practices are vital. Others have said it in wiser ways than I have.

As for the attribution of the memory decode scheme, I'll admit I do find it a bit baffling. Ben was very up front that his 8-bit 74series computer design came from Malvino's Digital Computer Electronics. I find it strange he never cites any source or inspiration for the BE6502. Then again, we're all baffled how he did the entire 74series build without any current limiting resistors on his LEDs (an issue not seen in his later videos, but has truly cost so many beginners headaches and lost time).

Ultimately, I'm incredibly thankful for everyone out there who is trying to make the 6502, retrocomputing, and electronics engineering more accessible to self-taught folks like me. Especially the folks on this forum answering questions and helping newbies from frying their gear or wasting their cash.