PCB layout - JHarvey

[AbeFM]

Good news!

1) Installing the newer version fixed (so far) all my issues going from Windows to Linux, and it seems (so far) I can go back and forth without issues!
2) Putting a "power flag" on each power/ground line (don't forget to add a junction!) got my error count down to.... wait for it..... ZERO!

Hope this all helps.

I suggest (after proper backing up, etc) we all adopt the 20080715 version of KiCAD, as there are numerous bug fixes. I'm also implying we stay away from the RC's until they are officially "stable".

[Fred]

I'm also implying we stay away from the RC's until they are officially "stable".

Agreed!

[Fred]

So you don't want the CPU pin outs to move from their current positions because it might interfere with a non developed feature that pin could be used for. I'm sure some are forced to pins because of timers, ect, but I'm also sure the AN's can go on what ever AN pin is available. I'm also sure other pins can be moved around as well. My understanding is that you don't won't to limit future growth by using those pins now.

My feel is that you should have a table that notes something like "PE1 = injector 1 output", ect. That way if we decide to change the pin out, you can simply change PE2 = injector 1 and your all set. However if that doesn't work for you, perhaps we can simply run these pins to the connector board, then back to 1.0, and we can physically change them when ever we feel like it.

It does work for me, but only for optional features. All core features will live a fixed pin life for all sorts of reasons. Yes, some pins can be shuffled, but some can't and some won't be even though maybe they could. Time limitations are major at my end and if anything costs me time it means less code further away. That is not really an option as I'm it when it comes to code at least until it's fairly functional and not a bare shell. The features that have pins assigned are mostly coded to some extent or other. The ignition pins for example are on the data bus pins which have no special functions. Putting them there ensures we don't shoot ourselves in the foot later. Putting them anywhere else ensures that we DO shoot ourselves in the foot later. Timer pins are fixed. I guess you can hook up your current sense/thermistor stuff on any pins provided that they are floating traces without the components installed. ADC wise, the two banks are different with respect how they can be used. Mainly though, I've spent significant time wiring up my JimStim to them and it would be a fairly major pain to change it. Especially when you see my schedule... I'll post on that now.

I'm basically at the point where I need to connect the injectors to the CPU. I've been putting that off as long as I can, but I don't see much more I can do until I know where I can put the pins.

The injectors and RPM inputs you can hook up as they are 110% fixed. Ignition and fuel pump I need to research more before making any calls. ADC pins are as described in the pin out document. Even if you convince me to blow away one of my weekend days changing the code around and rewiring my board I would strongly prefer that the core sensors were all on one adc block and the other was all for optional stuff. It just keeps the documentation and conceptual stuff clearer. It's also neater and more clean overall than randomly hooking them up.

I hope that helps put it into perspective for you.

Fred.

[baldur]

This circuit is what is used by an OEM. It filters and adds some hysteresis. The hysteresis could easily be done with some form of a Schmitt trigger, but then you'd also have to filter out some noise junk. The pro's of this circuit are that you can adjust the filter and hysteresis, where other circuits are harder to control. I'm sure it could be done several different ways, this one has some field testing, so it got included.

OEM in what decade? Was it with a crank trigger or a trigger inside a distributor? 1 tooth per event or many?

[Fred]

latish model mazda mx5/miata, low tooth count, crank shaft vr sensors but with the vr conditioning built into the sensor.

[jharvey]

I've been stewing this one over a bit. Some pro's of routing to the connector board include the chance to measure delays and inductive and capacitive components of the trace, then add components to trim the delays. However, my gut feel tells me the cap and inductive components are in-significant, and can be safely ignored. So perhaps not much of a pro. Con's include it makes it a pain to sandwich the board together, and increases cost.

Pro's of just routing it include allowing for a single board, and simpler design. Con's include if I mess it up, it's gonna be hard to change, and perhaps a little costly. Hmmm feels like there has to be a better way. Perhaps a beer(s) would help enlighten me.

[AbeFM]

I like a board with traces for routing. Worse case, add a via by each one. Then you can "reroute" by cutting the offending trace and running a wire.

It's a prototype board, and while you should do what you can to avoid mistakes, I'd rather have something bigger, slower,etc etc that has experimenting in mind.



Back to general PCB stuff:
How are you doing the 3D CAD stuff? Are you using Wings? Is it compatible with ANYTHING else? Or are you just finding done components and adding them in? I would make this stuff in 5 minutes in any other program....

Also, I'm starting to try to route my adapter board... It's taking a while, but esp since I don't really have a footprint for the little wire blocks it's pretty annoying, and ultimately useless.

[jharvey]

I guess the driving engine for KICAD is wings, but I don't know much about wings. I did the 3D in KICAD, I edited the module in the module editor. Under the module properties I found I could add 3D components. I simply picked some pre-existing shapes and rotated them into place.

I'd agree the 3D modeler does seem to lack, but 3D at all is quite impressive really. Who knows maybe some day KICAD will work with Blender or AI.

Back to the PCB. So far I've added 6 or 7 via's that were not required for thru hole components. To get the last three injector signals, I'm going to need to use several vias. Not really all that bad, but I've found PCB failures for me have been bad adhesive and bad VIA's. I know via's are typically very reliable, but still tend to try to avoid them if there is a way to do it with out them. I guess in the end, you can always use a jumper if you suspect a via is acting up on you. Perhaps I'm just being picky.

[AbeFM]

I guess I'll have to check out the editor built into KiCAD. Or see if there's a way to import something like an IGES.

I think via's aren't too bad! The only problems I've had with them are rework on FPGA's, etc, where you heat the board differentially and over a large number of them.

But for every day work, ESP in a 2 layer board? I wouldn't have any serious concerns about it.

[Fred]

If it doesn't cost and there is space, you could always use two vias at each pass through to double the chances of getting through. Toyota do this for current capability with 5 or so in some places.