Puma board for FreeEMS

[jharvey]

I posted about the translucent on this thread

viewtopic.php?f=9&t=351&hilit=translucent+FR4&start=131

Here's the key links I noted.

http://www.mgchemicals.com/products/500 ... html#specs
http://www.currentcomposites.com/html/e ... nates.html

Currentcomposites natural is fairly translucent, and shows a picture of what it looks like. However you can't see through it, but a LED can light it up fairly well. I don't know of a picture of the mgchemicals FR4.

Also about the long term use of the case, think audio amp. Many cars are made for show, not the track. Ascetics can often be as important as functional. I also prefer functional first ascetics later.

I think you feel the need for a metallic case because of thermal resistance. Note that steel is a fairly good thermal insulator, so I believe you are really saying the case needs to be aluminium. If the drive silicone has a low impedance and is driving high impedance injectors, the heat dissipated by that silicone may not be that much, and a much higher thermal R may be allowed. I would need some empirical data to indicate how many BTU's are being dissipated, then I could use that data to predict the necessary thermal R, then I could predict what materials would be acceptable. I believe plastic has a lower thermal R, then you most folks might expect.

According to this chart, http://www.engineeringtoolbox.com/therm ... d_429.html
Aluminium ranges from 30 to 250
Steel is around 16 to 43
Glass is around 1
water .58
Ethylene glycol is .25
Acrylic is .2
fiberglass is .04

So water and antifreeze have nearly the same thermal R as Acrylic plastic.

Any how, that's really shed color at this point. For now the LED's should be purely for diag purposes.

[Fred]

I'm just setting up a MS box with 8 LEDs on the front, though only 6 will be used for now, 4 fuel, 2 ign. Purely for diagnosis, though.

Marcos, for once, I disagree with you about the protection. Here are my thoughts:

Any pin routed to an EXTERNAL connector MUST be protected sufficiently (we need to define acceptable methods for that). An internal header or connector of some sort designed for an expansion card or similar is fine, though.

Does that make sense and seem logical to you?

What I was suggesting and wanting was a dual inline header or similar that you could plug a custom board with various circuits on it, into. That gives expansion and customisation possibilities. Look at msextra, everyone customises everything. Not just to gain the basics that it should have had, either, but for other things that not everyone wants. Yes someone could put a shoddy circuit on that header and cause a failure, but that would clearly be their own fault. We could extend the freeems compat guide to expansion cards too, with simliar specs, circuit wise. The point is, that such a header takes up minimal board space and provides maximum future flexibility. If it's possible to do, then I would do it. Perhaps your board is too small and specific, though. I don't know. I think you'll regret it later if you can't plug in a board to add various circuits, though.

Jared, 0.58 vs 0.25, interesting info, I knew water was better, but I didn't know the discrepency was that large.

Fred.

[nitrousnrg]

wow, very cool indeed. I'll keep those laminates in mind.

About the case, I was thinking in metal because of RF protection... but a thermal analysis is quite good, specially with transistors working in the linear zone to regulate 1A or more (P&H drivers). What tool/method do you use to predict the necessary thermal R? A little link wold be awesome :-)

I never did it in that way (using BTU/cals), my thermal chain stops at the heatsink attached to the device, only because I see the case as an utterly complex piece to model. I'd like to go a little further in that aspect.

Any pin routed to an EXTERNAL connector MUST be protected sufficiently (we need to define acceptable methods for that). An internal header or connector of some sort designed for an expansion card or similar is fine, though.

Does that make sense and seem logical to you?

Yes, it makes sense, seems fair, and I agree :-)
That draws some limits about the whole board design too, which is nice.

What I was suggesting and wanting was a dual inline header or similar that you could plug a custom board with various circuits on it, into.

Look at page 12 of this thread (connector board template), then look at the first post of page 1 (cpu board).
I think your custom board is my connector board, which have plenty of space to put whatever you want on it. The board that would be installed in my car should have a db37 connector, and lots of pads for testing circuits, like a breadboard. As I said before, I want different connector boards (eg, a MS-like connector with prototyping area, or a plug&play board for chevys/fiat/vw, or a connector board for other projects).
So, the idea is a stable cpu board and a customizable connector board. In fact, including the stepper driver in the cpu board was a tough decision... not every car has a stepper IAC.

[Fred]

I'm glad you agree :-)

I understand, but the connector board is much less useful/customisable if you can't get many spare pins attached to it. If you get a whole bunch out, then the possibilities are endless. My thoughts are something like:

30 pins come out of the plug, 60 pins come out of headers to attach other boards.
40 pins come out of the plug, 50 pins come out of headers to attach other boards.
50 pins come out of the plug, 40 pins come out of headers to attach other boards.
60 pins come out of the plug, 30 pins come out of headers to attach other boards.

etc.

I don't understand how having all the pins exposed internally like that can use so much board space. It seems very much like a good use of space anyway, it will ensure future upgrades of any sort are possible.

If there simply isn't enough room, sure, but the more you can get onto such a header, the better.

Do you have a reset button? (no need as both firmware and serial monitor can do this with a serial connection, but can be nice)

Do you have a load/run switch? you 100% need this.

Fred.

[jharvey]

wow, very cool indeed. I'll keep those laminates in mind.

Also keep in mind that copper is not translucent, so what you can illuminate will vary based on the copper routes. Also also, keep in mind, flooded areas often do not need to be fully flooded, they can often be a patter, such that the copper can allow the PCB to shine through.

What tool/method do you use to predict the necessary thermal R?

CAELinux is good for thermal analysis/prediction, but empirical data is best. Figuring out the thermal R is really based on datasheets, ect. Predicting the temperature of a heat sink or component can be done with fairly simple math. If it starts to get complicated, then CAELinux seems to work reasonably well.

http://www.caelinux.com/CMS/

On my project page, I posted about a thermal design I did for my work.

http://jaredharvey.com/Files/projects/thermal.htm

[nitrousnrg]

I understand, but the connector board is much less useful/customisable if you can't get many spare pins attached to it.

33+ I/O seems quite a lot to me. I have 6 spare digital I/O, and 3 analog spares. I aim to fill the 40-pin signal connector with a few more I/O, maybe 2 more digital and 2 more analog. I'm not counting the SPI/CAN/UART header.

30 pins come out of the plug, 60 pins come out of headers to attach other boards.

what?
I think I don't know the meaning of a plug, or something else in that sentences.

don't understand how having all the pins exposed internally like that can use so much board space.

Its not the space that the header uses, its the space that the tracks use to route things. With a couple of extra layers that wouldn't be a problem, and the component density could be really high without much hassle.
For example, I wanted to keep power tracks to a minimum (those are big, fat tracks). Thats why they go directly up from the transistor to the connector board. Signal tracks were treated different. Since I want free space in the connector board, I've put the 40-pin header against the border, in such a way that allows me to have a big, continuous, free space in the center of the connector board. Otherwise,the headers would be around the cpu, which is in the middle, and I don't want things in the center of the upper board.

Do you have a load/run switch?

ooops
Do you know why the adapt schematic points it as optional? <- EDIT: here is when the fan stopped and I had to disassemble the whole pc to clean it

CAELinux is good for thermal analysis/prediction

Whoa! I'll try to learn something about it. I played with FreeCAD and seemed nice to me, it doesn't has many features yet, but if it can import/export data it would be useful for thermal anaysis (I don't want to use a live cd for a single application)

On my project page, I posted about a thermal design I did for my work.

Interesting project man, its good to know what tools we have available in a penguin box.

[Fred]

I understand, but the connector board is much less useful/customisable if you can't get many spare pins attached to it.

33+ I/O seems quite a lot to me. I have 6 spare digital I/O, and 3 analog spares. I aim to fill the 40-pin signal connector with a few more I/O, maybe 2 more digital and 2 more analog. I'm not counting the SPI/CAN/UART header.

You should flag (skip/not do) the uart and make it SPI/CAN/I2C instead.

As for 33 seeming like a lot, 32 is what ms2 has and it's NEVER enough for any install with some extra functionality. People are always crying out for more. Also, 33 is 1/3 of the cpu io pin count, and that's for the lower pin count processor as found on the TA card. The bigger package has even more. It just seems like a waste to me to trap them in there with no way out. Does that make sense?

30 pins come out of the plug, 60 pins come out of headers to attach other boards.

what?
I think I don't know the meaning of a plug, or something else in that sentences.

by plug i meant the external connector to the loom. if you have 30 coming out into the loom, then 60 should be available on headers or something inside the box. That's my opinion anyway. It's your design at the end of the day.

Look at what ms3 has:

http://stuff.fredcooke.com/ms3/expansion/

I realise that they have no conditioning circuitry on the card, but they do bring out many pins in a header for expansion. 40 go through the dip connector to the main board, and it looks like another 40 through a ribbon and anotyer 10 through another ribbon.

The basics should be covered without such an expansion card (whereas with ms3 the expansion card provides the basics... fail...).

don't understand how having all the pins exposed internally like that can use so much board space.

Its not the space that the header uses, its the space that the tracks use to route things. With a couple of extra layers that wouldn't be a problem, and the component density could be really high without much hassle.

Is your design 2 layer? Or what? If so, I understand. Did you read the docs on the CPU, though? It specifies 4 layer minimum. I realise the TA card isn't 4 layer, but that is still freescale's recommendation.

Do you have a load/run switch?

ooops
Do you know why the adapt schematic points it as optional?

You need it because we will use the SM for firmware updates. I would just put it on the card as a 2 pin header and run a cable to a switch on the outside of the box. No need for an actual switch inside the case, though that is a valid option, I prefer to not have to open it up for new firmware...

Fred.

[jharvey]

Did you read the docs on the CPU, though? It specifies 4 layer minimum. I realise the TA card isn't 4 layer, but that is still freescale's recommendation.

I'm a fan of the 2 layer approach. The connector card gets us a pseudo 4 layer effect, so the ability to route these things isn't really that much of an issue. I'm sure Freescale makes this recommendation for some mild performance enhancements relative to the power and ground planes. They expect routing to be done on the top or bottom layer, just like it's routed now.

The TA card had it easy, they just had to route most traces to header pins. This layout has several circuits on a board that is nearly the same size as the TA card. I would agree that this layout probably can't have all MCU pins available. The goal with this version isn't an all inclusive MCU. It's a very flexible MCU, but for a specific set of features. Allowing some extra GPIO would be good for diag, and can allow some prototype growth where this one isn't planned to grow. If someone did use the GPIO as a feature, I would recommend they make it a real feature such that it's on the proper ground planes, ect. I would hope nitrousnrg puts in what ever vias can easily fit, and moves on. If someone want's features that this doesn't offer, they can spin their own board based on this one to add such features.

To me it looks like several pins can be fairly easily routed, external to the MCU footprint. MCU pins 49-56 could have perhaps 3 pins (might only be 2) routed to a via, MCU pin 96, MCU 91 or 92, MCU 80 or 82, MCU pins and 100-105 could have 1 pin perhaps 2 pins.

Fred do you have a recommendation about what 2, or perhaps 3 pins in the MCU range of 49-56 would be nice to have as a thru hole?

What 1, or perhaps 2 pins in the MCU range of 100-105 would be nice?

Should 91 or 92 be routed?

Should 80 or 82 be routed?

nitrousnrg, if Fred has a recommendation, would you be willing to add those pins? If Fred doesn't have a recommendation, can you add what ever fits and has a short trace? I see 5 perhaps 7 general purpose pins with the ability to be routed for general purpose use with out compromising the board layout.

Oh also one RFI note we had made in the past, was to place an isolated non conductive ring around the perimeter of the board. The theory being that it's a form of antenna, and it can capture some RF and short it to ground. This helps prevent RF emissions, also it prevents RF from inducing voltages on the boards traces. To me it looks like this layout has room for that kind of a conductor. Currently that real-estate is a ground plane, if you separate a 10 mill ring around the perimeter, that will likely help lower the chances of RF issues.

[jharvey]

CAELinux is good for thermal analysis/prediction

Whoa! I'll try to learn something about it.

Ah man, don't do that. If we need something, I can wipe it out. Both FreeCAD and CAELinux can get you tied up for long periods of time. If we need something, I can wipe stuff up much quicker. I'm hoping you don't get side tracked by these two software tools.

[Fred]

As for which pins, Marcos can decide that just as well as I can and depending upon what he would like his users to have option wise. Provided the core pins are correct, I'm not too worried. I still don't know what the core pins are, yet, though.... which could be a problem, but so far the only one I haven't thought much about is the fuel pump. Marcos has probably gone by what is in the pin out doc, and hopefully when I chose that I researched it well. Hopefully if I hadn't he would have cried fowl on that too :-) Maybe I'll check it now to make sure it makes sense, though.

Fred.