RomRaider

How sure are you that your injector latency is correct?
http://www.miataturbo.net/showthread.php?t=56061

Have you attempted this experiment with yours? Everyone over there is getting 2 to 3 different deadtimes. I'd say the method might work in theory but isn't robust enough to use in practice. Just my opinion.

I just tried it, I couldn't duplicate it

and the data they have, Map/AFR my values are way smaller than what they show...I dunno, wishful thinking...

I was using the latency specs from injector dynamics, and it idles just fine in open loop mode

Can anyone post the instructions for setting up the transmission pin for map switching, I thought it was in this thread, but I just went back through 30+ pages and couldn't find it. Carbibbles, could you update the first post with this when you get a chance. Thanks.

It should idle just fine in open loop. You aren't try to hit a target AFR, the ECU is just opening the injector with no corrections.

ID's specs are just a starting point. If you run e85 you will likely need to increase dead time due to density differences. e10 would likely require some amount of change as well. The specs provided are based on straight gas and are an average of many injectors I would assume.

In closed loop you are trying to target a particular AFR. After this injector fires, the O2 sensor tells the ECU if it is too lean/too rich. The ECU adjusts the next injector pulse based on this. If the dead time is off, the next injection will be too lean if the dead time is too short or too rich is the dead time is too long. Do you kind of see how your dead time being off will lead to over/under compensation in closed loop? This is especially noticeable at idle, because dead time is fully half or more of the total open time of the injector.

Hydraems has a way to see if dead time is off on their website. Basically you have 2 maps, one normal, the other exactly 10% richer. When you switch maps, your AFR reading should change exactly 10%, if it is off, then your dead time is not right. I am trying this method out right now and will post up results tonight...

Another method I had tried was to let the car idle, logging AFR error and engine load. Add/subtract a few ms from dead time and log again. The plot of error and load, with the smallest swing in AFR error would be a better choice for the dead time value.

Sounds like a good way to do it. Let me know how it works.

This is how I did it and the initial results.
map1: normal map, forced open loop.
map2: map1 * 1.10 (10% more fuel), forced open loop.
I used defroster switch, so not sure how much electrical load changed readings...
As straight and level piece of road I could find, set cruise control.
Started logging. Alternated map 1 and map 2 every 5 seconds.

Threw everything into excel and scatter plotted afr vs time to see how much noise I could filter out.
Plotted map1 data afr vs time as series 1, plotted map2 data afr vs time as series 2.
Added trendlines to both. The intercept 'should' be the average afr.
afr map1/afr map2 should give you 1.1 if your dead time is correct.

How well the data can be filtered will have a pretty big impact on the results...

Im going to go through the thread soon and pull out all the random posts to clean things up a bit. I will see if I can make a wiring diagram and post it in post#2. Its pretty easy though, find the TID pin/wire (black/red I believe), then snip it and connect a length of wire to the pigtail that is now coming out of the ECU. Run the new wire to a terminal on a switch, then run a wire from the other switch terminal to ground. Basically grounding the ECU pin activates the function and open circuit to the ECU pin (no ground) de-activates the function.




I think the method you guys are using to find injector latency would work, but it needs a 10%(or whatever percentage) multiplier applied directly to final pulse width. As it is right now, there are compensations that could be messing with the results if you are just adding 10% to the VE map. I could add a "latency test mode" feature though.

Very interesting!

Some numbers I got from your data:


Does the final injector pulsewidth = injector pulsewidth + injector latency?

If so, it looks like a 8% change in final inj. pulsewidth results in a 15% change in AFR. Does that mean the oxygen sensor scaling is slightly off?

no, 15.53/13.53=1.148 or 14.8% change in fuel.
1.148-1.1= .048 or 4.8% too much fuel. since it was only supposed to be adding 10%. I am going to knock 4.8% off of latency and see how it looks after that.

BTW make sure you guys always log "(direct)" or extended ("*") parameters. The Injector pulse width (direct) value does not include latency, but the ssm parameter does (not as accurate though).

What other comps might be applied? There should be a minimal number in open loop though, right?

I understand that, but to get a 14.8% (I said 15%...) change in fuel your injectors only fired for ~8% longer...assuming your latency was 1.05ms.

2.424 + 1.05 = 3.474
2.698 + 1.05 = 3.748
3.748/3.474 = 1.0788 or a 7.88% change in injector pulsewidth

Also, simply reducing latency by 4.8% is not the answer because by doing that you are not reducing the difference in final injector pulsewidth by 4.8%.

There are a bunch of random little compensations that are in play regardless of CL/OL status. If you want to PM me your ROM, I can add the "latency test" feature to it so you could test it out. If it works well then I will include it in the next version.

You would want to wire up the TID switch, the defogger can cause slight changes in fueling because of electrical load.