PCB layout - JHarvey

[AbeFM]

Yeah - that's for the web-based router. The program is awesome/frustrating. It's very easy to use - you cna push componants around, and it'll reroute in real time as you do. Vias and tracks and componants will "squish up" taking all the traces with them as they go. Then, when pushed "too far" things will jump and cross and reroute. It's effortless, like I'd picture a 3D sculpting program would be.

When it routes, it runs for a while (20 minutes?) all the while shortening overall trace length and number of vias.

It took the default trace length from KiCAD, but then has it's own table.

The bad news: Each time you change the board, as best I can tell, you have to load it in. You can't alter what's connected to what. Same with KiCAD. I wanted to add some pads to a set of unused jumpers on the block (the test points you see in the picture) and I couldn't. If I'd wanted to do it right, I'd have had to edit it in the schematic (which I did), THEN update the board, then reupload it to web program, (maybe it would take the current information?), let it reroute, which means recreating that table of trace widths (I ended up doing that table like 6 times - not that it's hard, but it doesn't mean I wanted to keep doing it), routing, reoptimzing by pushing the traces around, and then bringing it back into KiCAD.

Now I'm thinking maybe I could load the almost done board in and it would respect the trace widths on the nets. If so, it's a lot easier as I could just go back and forth. I haven't tried it.

But I really like the autorouter. if i'd had the layout about right and the schematic 100%, the whole routing process would have been under 30 minutes, including set up. It's not as complex a board as you've been doing, but its not bad, either.

Of course, I just thought "gee, what if someone wants launch control" so I may come back with info on how to upload a done board, or I may route those two traces by hand.

[jharvey]

So what's this web auto-router? Do you upload a KICAD file to a web page, or does KICAD come with a web page that can be used as the auto-router locally?

[AbeFM]

right most button at the top brings it up. Export a file, go to the webpage (the autolaunch doens't work), hit "start here" or whatever.

Load the file you just made, route, save it as a .ses (you need to put the extension on yourself for some dumb reason) then back in kicad you import it from the same little tool.

That's all there is. Pretty easy. I was afraid of it forever, but I'm glad I tried it. Saved me, oh, 45 vias.

[AbeFM]

Well, there was a short break. Ha!

I'm looking at getting my board actually made, and added some more free circuits Jason on www.miataturbo.net posted, and in relaying out my board, I realized... I'd lost all my tracewidths! So now while trying to figure out how thick to make them.... I find ridiculous amounts of current, I need planes, not traces.

Injectors:
Resistance: 2.5 ohms
Voltage: 8-15 Volts, nominal 13.5 Volts
Amperage: Peak 4.0 Amps/ Hold 1.5 Amps
4 amps x 2 = 8 amps.
8000 rpm x2inj/cycle = 7.5 ms/fire
Opening time ~1ms, 20% "peak" current, ~ 8 * .2 + 3 * .8 = 4amps effective (10.4*.2+3*.8=4.5amp) -> 20*c raise: 71 mils

If you imagine an engine running flat out, call it 8000 rpm, with a 100% duty cycle (still you'd have the 1ms/cycle with the full current), that's effectively 4.6 amps - times two for two injector pairs, and you're at ~10 amps (effective, peak is higher).

Now factor in 800 mA for the fans, another 800 each for a boost solenoid, variable intake, idle solenoid and VVT... And you're looking at over 14 amps. On a once oz pour board, that's, um, 300 mils! My board rapidly gets ridiculously out of hand there.

So, ideas:
1) Come ON, be realistic, the MS has no traces over 100 mils wide, just leave it
2) Perhaps 20 degrees C rise can still be pushed to 30?
3) Honestly, injectors shouldn't be over 70% duty cycle, idle 35%, boost control 80%.... I could probably knock 40% off my current figured right there
4) Perhaps I can find a 2 or 2.5 oz board
4a) Thicker pours mean large minimum trace widths, so I lose space again.

Or, do I just have to make my board twice the size or 4 layer? Curious if any of you have any ideas.

Also, as a total aside - how common on ECU controlled alternators? Perhaps we should roll this functionality into the FreeEMS?

[AbeFM]

One more idea... What if I made two nets - a power ground net, and a ground reference net. The bulk of the board doesn't need much current (the biggest sources being the transistors to run relays), so I could just make a second net, basically one fat trace, to go from the OEM connector to the DB-37 from the MS.

Sound good?

[gearhead]

1: The peak is only for a very short time. 2: you will be running sequential, not ganged, so cut the numbers in half. Also at 100%, you no longer have the peak, just hold...

The high power stuff is the reason that the high power ground needs to be separate from the rest of the unit, electrically. Short fat traces. They only sink, so it is the ground that needs the beef and a ground plane of decent size is required.

Gearhead

[AbeFM]

playing the whiny loser:
1 ms (.8-1.5ms) opening time, 8000 RPM, 2 squirts per cycle gives you a 20% duty cycle. The MS isn't smart enough to shorten the opening time I don't think, is it?

I'm running two banks of two injectors, so effectively hold current is 3 amps with peaks of 15 (15 amps -> (15V / 2.5 ohms ) x2inj + hold current for other pair)...

Of course, I imagine peak current is more like 4-5 amps/injector since the pintle moving takes resists current.. My earlier numbers were a weighted average current assuming you just turned off the current a split second each revolution.

Really I just need cheap place to get 4 oz pours. Jean, what do you do about large currents?

[jbelanger]

I haven't' had any problem using 66 mils on my p&h board with 1oz copper. Average current is well below 2A even at high load/RPM and this is for short durations in almost all installations. The actual current average is less than 1A for most of the time. And I have 1 ground wire per injector all connected to the common analog ground plane.

On my 4-channel driver board, I have 100 mil traces for the ignition drivers and I'm not worried. If the dwell is set correctly, the average current will be about half of the maximum current in the worst situation (again for a short duration) and if dwell is not correct then something else will give before the traces (the driver or the coil).

Jean

[Fred]

and if dwell is not correct then something else will give before the traces (the driver or the coil).

Or the fuse!! ?? :-)

[jharvey]

Here's a way to cheat about trace widths.

http://www.geocities.com/capecanaveral/ ... eWidth.htm

It's backing math is based on industry standards, and has a fair bit of field testing to confirm it's accuracy. I seem to recall it's a java script that the fellow is willing to let others use if they want to copy it. I thought I read something about geocities going away. Perhaps we want to archive that script just in case.