russian239

For the last two years I am playing with my own from scratch ECU. A week ago I've got a new test mule - surprise-surprise it's a honda, so hopefully it would be fine to start a thread here

russian239

Just got a new board - so far the power supply works and I am now trying to figure out VR trigger input. Please tell me it is VR in this 1995 Accord distributor?



Couple of links Re: this project: http://hackaday.com/2014/03/29/frank...-control-unit/ http://www.autoweek.com/article/2014...WS02/140429923

Hopefully it's appropriate to post this here. This is not a commercial product - this is an open source work-in-progress projects, so hopefully I am not breaking the forum rules.

russian239

I am using max9926 chip for VR trigger signal input, here's what I get for TDC and CKP signals:



Does anyone has a description of the signal shape, the angles at which these rises and falls are happening?

DC_Legacy

MegaSquirt?

russian239

MegaSquirt what? Does it support this trigger? Which version?

OneBadTurboCRV

You are in the wrong place and time for this kind of information. Nobody really cares anymore since most EMSs have been figured out YEARS ago and most systems are built on proven foundations - the factory Honda OBD-1 ECU. The place to go would have been www.pgmfi.org if there are any knowledgeable members even left over there. A lot of them even went on to make their own EMS.

Freemananana

iTrader: (3)

I've stopped suggesting this site to people. As far as I'm concerned, that website is dead. For whatever reason, good or bad, the people who were doing grassroots tuning have left. It's just a lot of dated software.

russian239

You should think bigger There is only as far as you can go based on the OBD-1 ECU. Time flies, eventually a better proven foundation would be needed.

Muckman

iTrader: (2)

Yes its been done, and done well.

AEM EMS
Motec
MegaSquirt is more DIY but has all the functionality you want.

Are you not aware of these systems? They will support any and all triggers and inputs/outputs you want.

russian239

Oh, I am absolutely aware that there are plenty of standalone EMS systems around.
Please remind me which of these is open source? I am sure you would not make the mistake on calling MS open source.

GhostAccord

To be honest I don't think there are many on here that spend much time on open source engine management and scratch built ECU's.

If you are looking for sensor input data MegaSquirt sites would be a good place to go. Even though it's not "Open Source" Quite a few of the people who use it, share a lot of the background electronics information. These users cover a very wide base of automotive OEM's specs.

spAdam

iTrader: (1)

I love seeing DIY stuff, who cares if its been done before.

Nothing wrong with starting out with a DIY project, a lot of the go-to companies in this industry were started by noodling around at their own workbench. It's a great learning experience and who knows what you'll come up with next. I'm in to see the final project.

Build something that runs on a more capable processor than the dinosaur obd1 ecus (and maybe even has at least limited obd2 capability) and you'll have something, if you ask me.

I've found from building my sim that the stock ECU isn't super picky about the shape, it's looking for the zero crossing in the right phase relative to the other sensors. I was able to keep it happy by pushing a set of 5V square waves (properly timed of course) through an rs-232 ship to amplify them and impart the zero crossing.

Since your ECU will probably want a TTL signal, you can probably run through a comparator (as the stock ecu does) or maybe even run through the logic converter as I did, but in the other direction.

Here's what you are looking for:

Pattern:





Shape:



Good luck.

russian239

You'll see it here This thingy has already started four engines in three countries, it just needs a lot of extra polishing and more driving miles put on it.
The chip we are using is 168Mhz - that's probably 5-10 times faster then what was used back in the 90s. Hopefully the extra performance would allow us to add extra capabilities. This board has a CAN chip on in so OBDII-via-CAN is totally possible.


We are using that max9926 - it's a purpose-build chip for VR signal conditioning. I am not an EE so I am not sure what's the point, all I know is that it has better reviews then LM1815. That black&white picture is the TTL signal produces by that chip - it only has two channels so we would probably need to add a third one in the next edition.

Thank you, and really appreciate the pictures - that's huge help.

russian239

Another question if that's OK: CLT and IAT sensor data? I need resistance at three temperature points to build the curve. Asking @ http://www.msextra.com/forums/viewto...p?f=71&t=55191 as was suggested did not really help

The sensor part numbers are 37870-PJ7-003 and 37880-P0A-A02 if that would help.

spAdam

iTrader: (1)

You can estimate it from a factory service manual if yo have access to one. They're fairly linear actually, but the best thing to do would be to set it up with a voltage divider circuit and do grab some data points (5-7 should do). Do a polynomial curve fit in Excel or whatever and create a lookup table for an 8-bit ADC to save processor. No sense calculating it every time by.

Here's the graph from the FSM, this should get you in the ballpark:



IAT is the same iirc.

russian239

Perfect, that should do it for a start! Thanks again.

russian239

I think this picture is misleading a bit - looks like resistance scale is not linear, thus the nearly straight line is actually a curve once you get to linear scale on the axis.

I'm using the 'fancy' http://rusefi.com/articles/measuring...art-Hart.shtml and yes, it has already hit me a bit - the 'log' operation is reeeeallly heavy. Since these temperatures do not change that fast, I will probably limit this math to couple of times a second. Couple of times a second I can afford it


As for the table lookup, that's fast but that cost you memory

russian239

So with my board I have that weird noise issue: with ignition ON position and engine stopped, I have the following ~40Hz signal/noise on the CKP/TDC VR output. As a workaround, I've found that I need to disconnect TCM plug A and the noise would be gone.

Makes me question if any of the TCM plug A wires carries some digital signal? For now I will just have this plug off, but at some point I would need to turn it on.

russian239

The board is officially functional and I need a better base tune:

spAdam

iTrader: (1)

Cool progress.

You're right, it's not very linear. I thought I remembered it being more so but then when I pulled the snip I saw otherwise.

There's no real reason to poll the temp sensors at more than .5-1Hz in my opinion. As you stated, they just don't change that fast.

As far as costing memory, it's always cheaper than processor. Lookup tables also work nice and fast if you're relying on your processor to tick through operations quickly. The obd1 Honda ecus use plenty of lookup tables (think ECT & IAT comps) to save processor resources. You also don't need as much resolution as a constant calculation will provide. If you use a 16 bit ADC to convert the temp sensor you'll have 256 values which is still significantly better than you need to get things running correctly, and you can approximate both sensors from the same table.

russian239

turns out there are some routing bugs on the board - SD card is not routed properly, that would be something to fix in the next revision. the board looks pretty sweet with an LCD screen:



Now a question: what are the characteristics of my stock MAP sensor? I believe it is linear, so I just need two points to build the line.

GhostAccord

MAP sensor output voltage of 3.0v = a gauge reading of 0 in. Hg
MAP sensor output voltage of 0.5v = a gauge reading of 27 in. Hg

russian239

Could it be the other way around?

What is ambient pressure? About 29.8inhg? 100.914 kPa?
I read 2.62v when I power my MAP on a bench.

GhostAccord

Ambient pressure would depend on your geographical location vs sea level.

29.92 in. Hg is sea level.
Pressure decreases 0.368 inHg every 100m above sea level. (15°C and 0% relative humidity)

I'm not 100% on the variance of the voltage/pressure. All of my old Crome datalogs are corrupted when I open them in the newer version. All I am going by here is the data that Honda has provided for diagnostics and troubleshooting the MAP sensor. Hopefully someone who has logged actual data will chime in.

russian239

Does anyone know anything about the wires between ECU and TCM?

I am using the diagram: https://i.imgur.com/eMzLNjz.png https://i.imgur.com/CNoS46a.png

1) it shows a BRN/WHT wire, it's called 'shift acknowledgement' on the TCM side and FAS on the ECU side. what does 'FAS' stand for? Digital/analog? What's the protocol?

2) Up/Down shift signals are called AFSA & AFSB on the ECU side, what do these acronyms stand for? Digital/analog? What's the protocol?

3) Baro out? There is no dedicated Baro sensor, so I guess that's something made of MAP?