[Diy_efi] RE More ECU progress

[James Holland]

I get the digest version of this so I apologise if others have already
covered this.
> 
> I have so far done the PIC based parts with only input protection
> modeled on a successful project - Megasquirt - so any failings my simple
> design likely is the same as that design which is in use all over :-)
> That is not a disregard in any way of your input - being an engineer I
> prefer massive overkill on my designs and I will take every point you've
> raised and see what can be done to either eliminate it or make the
> decision that it enough of a boundry condition as to not likely apply
> for our intended usage.


I have seen the Megasquirt project although I'm not very up to date on it.It
is as you say well proven. It also makes a good reference design. If you
look at the evolution from MS1 to MS2 then you will see that a lot of what
they have done involves additional protection and grounding improvements.


> 
> And I don't know about what is currently present on the web - but the
> power supply has a 1 farad cap about .5 inches from the chips on the CPU
> - if that isn't enough I don't know what is :-)


I actually thought the 1F was a typo. Logic devices require bursts of
current as they switch, the decoupling capacitors are there to provide this
current. My earlier response really didn't represent what I meant to say. I
did mean 100nF not 100uF! And the reference to 'at least' was to the
frequency response and not the actual capacitance. If you consider the PIC
itself then you have switching currents at 5MHz (the PC) and other switching
at lower frequencies as other functions operate. The caps need to provide
current fast enough to meet demand or the voltage will be pulled down
(usually not by much). I haven't had occasion to use caps of that size myslf
but I would guess that the frequency response is just a few khz. A 100nF has
a self resonant frequency at around 10MHz with a fairly low impedance above
that all the way to around 100Mhz. However being a small cap it doesn't
carry much charge so it is usual practice to use one per power input. Lead
length and track length contribute to the inductance of the cap and reduce
the bandwidth so they do need to be close to the pin

I don't think the 1F will have any real effect on decoupling. It may give
you further problems with meeting the rise time specs on Vdd and by holding
up the power rails after you've turned off the supply. Something around 68
or 100uf would be more effective.
> 
> The 'daughter card' has full clamping on all inputs and all of it's
> inputs to the CPU card are via SPI - all inputs on the daughter card
> have 5.1V zener, and reverse voltage input protection - although in any
> normal auto usage the reverse stuff is overkill.

Clamping is good, the reverse protection isn't overkill. In almost all
modern cars now there are inductive loadings being switched all the time,
motors, solenoids etc. These can create big current spikes on turn on and
big reverse voltages on turn off. Car manufactures tend to run all their
wires together in one relatively long loom and there is a lot of cross
coupling. The 5v1 zener may be a little high

> 
> The latest rendition of the power supply has MOV's and 200V bridge
> followed by MOV's to make sure any rational DC input is dealt with - and
> lots of totally irrational ones too I expect.

MOVs are good, doubt if you need the bridge. Make sure that the MOV is of a
suitable voltage. They don't handle lots of current for very long so the
rated voltage shouldn't be too low. Remember that the source impedance is
very low so they can pull a lot of current when they conduct. A polyfuse
might help to protect the MOV but they tend to be slow and the trip point
varies a lot with temperature

> 
Snip
> 
> Going from a 2 side board to a 4 side board to have the various
> signal/ground/etc separated will make the board itself fairly pricey -
> that will have to be driven by some level of actual board/system failure
> -vs- 2 layer - here you need to convince me :-)
>

That's pretty much the eternal dilemma. Double sided boards I can make at
home for peanuts, Four layer are very expensive. You can do a lot to improve
a double sided board. Copper fill is your friend. The copper is already on
the board, why etch it away when it can be used to create nice big low
impedance grounds and power supplies

> 
> Not sure if I touched on this - but the daughter board which provides
> all I/O is .20 in away and sandwiched via your usual header things - so
> I don't think any ringing will occur even at max frequency likely to be
> present.
>

I got that. It does keep the track lengths to a minimum but those tracks can
act like aerials and propagate noise from board to board. I can't see it
being too much of an issue in this case but if you start switching injectors
and solenoids then it could start causing problems. Not sure if they'll be
room but a grounded plate between the switching board and the rest could
help.

> 
> And PLEASE keep up the feedback!!!!!!!
> 
> Oh - funny - the PS I designed for the actual ECU has MOV in front, 200
> PIV bridge, MOV in back and then the rest of my design including a zener
> clamp in front of and behind the regulator - and regulator is able to
> handle 2X 'normal' input voltage :-) There is no more enjoyable overkill
> than massive overkill :-)
>

Two times voltage is more than enough but don't forget that 2x voltage is 4
times the power dissipation

>
> 
> Given the whole Interceptor fits in a cast aluminum box - hell it may
> even survive some level of EMP - the rest of the vehicle sure won't...
> 
> 

Note point above about power dissipation. The metal case can work for you if
you get heat transfer to it, otherwise it may hold the heat in. I think
you'll be stuffed on EMP, the box is kind of ruined once you cut holes in it
and bring all the wires through :-)


There were a couple of other points that you mentioned in your email, I'll
cover them here. You stated that you aren't using the A/D on the PIC. I
hadn't realized that so a lot of my other comments aren't relevant.

The 'fast' A/D I'm curious about. Is this being fed directly to the ARM? I'm
not familiar with them. IIRC You mentioned 400ksps before, that's a pretty
fast data rate. Way above what you need for most automotive sensors and
quite a heavy loading for most microcontrollers. As I understand it the high
data rate is intended for the ION sensing stuff. Unless that's implemented I
would slow the A/D rate down a fair bit to save power and reduce the
overheads.

I haven't looked at the ION sensing stuff. I will have a read through it at
some point.

Last point. I get called upon to review other people's designs before they
go out to manufacture. There are a lot of professional electronics designers
who get very upset if their designs don't come up to scratch. Actually I'm
probably one of them too. I know how it feels to have someone pick through
months of work pointing out what may seem apparently trivial details and
sometimes making wrong assumptions. It sucks but I learn from it.

Cheers
James