DFH - Defacto FreeEMS Hardware in KICAD

[Fred]

Thanks for the answers Fred, all much clearer now, especially the 4mA pin limit on the cpu. Thats not much room to play with.

You are most welcome :-)

They actually don't spec them as a pin, only as a bank. They spec an instantaneous max of 25 for a pin, but thats not a continuous limit, probably far from it. You could probably get 10mA out of them if the duty on a bank wasn't all that high etc. I have had my CPU lock up solid by over loading the pins on it... I thought I'd stuffed it, but it just latched up solid. All IO pins froze in their current states (= hydrolock and or burnt coils injectors ignitors etc on a real setup...)

My MS2 is setup with 330Ohm resistors straight out to external ignitors with no protection. I'm very unhappy about that and will be fixing it ASAP unless I'm ripping it out in favour of this instead :-) you should be using 1.6k with ms2 or this, and if that isn't enough to drive your load, you need a transistor or IC to buffer it. The vb921 at least requires a firm drive that the MS2 can't safely deliver. Other chips may vary...

[jharvey]

Here's freeEMS 1.0 A.14

https://sourceforge.net/project/showfil ... _id=621388

When you reply / review, can you note the A.XX number? Just to prevent confusion.

In A.13 RPM_input.sch, I think he's right, the LED won't light. It notes I12 saturation is 2ma. However different voltages may produce different saturation currents, it's still close to the limits. Perhaps more XOR's for buffering?

Oh another thought, perhaps it's already been addressed. If we want to prevent ROM flashing issues, maybe we should software limit the flash programming for when the switch is off. That way the XOR won't have power, and generally you can't have nasty things happen like injectors burning up ect. I mention it because I don't think we have a reliable switched path to the CPU yet. CPU might be able to get this from the battery sense R divider, but with out power pushing it down, where will it really be. Can we rely on that to be 0V when the switch is off. Perhaps we should have a direct path the the CPU to make sure?

Change log in included in the release. Look for it at SF.

[Fred]

Quite right, I hadn't explicitly said, but yes, their should be a key input on a digital line for our "go to sleep after shut down sequence" function. I don't know which pin, but yes, we do need that.

[jharvey]

I know Fred has mentioned CPU lockup, should we perhaps think about a watchdog timmer?

sample part for reference

http://delta.octopart.com/Maxim__MAX6373KA-T.pdf

[Fred]

Nah, that is only an issue if you severely overload the part. Given that we are ensuring that all pins that are used on this board are protected properly it won't be necessary at all. We can write some guidelines for those wishing to develop plug in add on boards in future so that they too make sure the designs are safe. In the event someone builds a shoddy circuit and plugs it in, that is their problem and they will suffer the failure, not us. Was it Jean or Brian that said "if you make it idiot proof they will build a better idiot" ? Either way, in most respects single protection of a device is sufficient.

There is also a COP watchdog built in (computer operating properly) however you have to set the time on it quite carefully and not trigger it too late AND not trigger it too soon, so it's trickyish to implement. I'm also not sure if it would get you out of that locked up state or not.

Fred.

[jharvey]

So I've stared blazing a trail for the PCB layout. The ERC produces 2 warnings that complain that VCC and GND aren't driven. I'm ignoring those complaints, as they really are driven, just not in the scope that KICAD sees.

I've also ironed out some bugs, typo's, ect and it now resolves to a net-list and creates the parts list for the PCB with out errors. It has one note, that I'm not using a "cmp" file so it's using the net-list instead. I guess I should learn about this cmp file thing.

Anyhow, just chomping away, and the bit is getting small.

[AbeFM]

Hey, my company makes some vary radiation hardened watchdog chips. Certainly a plus if you want to drive your car around inside the business end of a linear accelerator.

Probably's Fred's right, and we don't need to worry about it too much.

I got KiCAD running, going to take a look at things for the first time ever. Just like me, two minutes too late.

[AbeFM]

Heh, working in a real design house, it doesn't take long to get information.

I was showing my boss KiCAD, who's always interested in finding good programs, and pulled up the Freeems v14 schematic.

I showed him one block and he had a couple comments, in order of importance..
1) The 2200uf caps aren't shown as polarized, and they should be so it carries through at design time.
2) Generally, power in is on the left, power out on the right, it's against convention and a little confusing
3) The diode across the LDO neither of us got - it only dumps into the caps, but not to the +12 source. Is this on purpose? Seems there's a limit to how big a spike it could absorb, meaning more of it stays on the 5V rail.

Anyway, I'll keep looking as I get time. Definitely happy with things, though. It's exciting.

[Fred]

I originally drew the reg circuit that way left/right I think. Hence that's my fault, but I don't agree that it's confusing unless you are not looking at the parts and are basing on past stuff.

The caps I thought I used polarised, but maybe not.

The diode is present inside some/many regs anyway, it's to prevent the output being higher than the input at shut down time and damaging the regulator. We may not need it.

Fred.

[AbeFM]

Oh, ok, the diode makes sense, then. Just have to find out if it's needed, though I would probably still leave the hole in the board just in case someone buys a lamer LDO?