More Data

[grippo]

I went through the code and found that I had already taken care of both % and ms (for example, tps accel) enrichments to the fueling so that when you go from wasted to sequential they take care of it for you and you shouldn't have to re-tune. It does this by always calculating for sequential and then correcting at the end for wasted. So this corrects for accel as well as VE, etc. I know it doesn't appear to be working for you, but that may be because of 1 vs 2 squirt per cylinder differences, as I said previously.

I also made a run with your msq with and without x-tau and datalogged the transition. Without x-tau, the pws for all the cylinders did the same thing, namely they stepped up to topen + 2 x (pwasted - topen). With x-tau enabled and with special code to capture just the pws for cylinders 2 and 5 I saw pretty much what I was expecting and have attached the plot. The vertical axis is PW in microsecs, the horizontal is time. The black line is engine status and is 1 until it switched to sequential and jumped to 65 at about 40.5 secs. Spare5 is the red curve and is the pw for cylinder 2; the yellow is for cylinder 5. As expected, cylinder 5, which has just had pw/2 injected in wasted and gone 6 tach cycles without any more now gets over 8 ms injected for a tenth of a sec or so then quickly drops down to a steady state of 4.95, the same pw as with no x-tau. This extra fuel should easily cover the bogging you were seeing without having to change topen (I left it at 0.8 ms throughout). The curve for cylinder 2, which all of a sudden got blasted with 2x the pw got a lot less fuel during the transition, but also quickly flared into the same 4.95 ms steady state pw as cyl 2.

This can all be tuned in x-tau to enhance or decrease the effects shown.

[TheMonkey]

Al- great work. i've been wanting to swap to x-tau because i like the concept. nice that concept fits well with the wasted->seq transition.

On the other stuff.... I put in a new voltage regulator, and I think that has taken care of the stumble problem. I wonder if the injector drivers were dumping current when voltage popped up so that they were not opening?

Also- I swapped the EGO correction algo from simple to smith/pid and the correction oscillations went away. This will sure make tuning better.

[TheMonkey]

With regard to testing pw vs wasted / sequential - I would first do it in the driveway where things are much more controlled. Plus I want to first see if it is a function of rpm, then we can see if it is a function of load.

I just did a quickie in the driveway. it's only one test, and prob not enough data points, but i think it's the right way to look at things. it was just a very slow increase to 3000 rpm in both wasted and sequential. i found that i actually needed to bump up to 0.4ms topen in order to get the range from 1000->2000 rpm the same AFR.

anyhow, interesting pattern develops > 2200 RPM. just ran test once... prob need to confirm this is repeatable. ego corrections turned off.

[grippo]

This is a good way to do it. Did you use 0.4 ms topen for both runs ? What would happen if you used 0.8 ms for both runs ? (You may need to raise ReqFuel to get it to run decent in sequential, but then use the same ReqFuel in both wasted and sequential. What I am getting at is we know 0.4 ms can't be a reasonable open time unless the double squirt in wasted is changing the injector open time because the electrical system can't handle 2 squirts at once - I doubt that. The reason I am asking is that the only thing that is different in wasted and sequential is the double squirt and I would expect single big or 2 small squirts to produce the same mixing at high rpm. I was even thinking if the injection start time being used is bad for your engine, then possibly wasted improves things because at least one of the 2 squirts occurs at a different, possibly better injection time. But this should become less of an issue at high rpm where everything gets swirled around. So maybe if you used 0.8 ms for both runs, you would see an afr difference at low rpm and no afr difference at high rpm.

[TheMonkey]

This is a good way to do it. Did you use 0.4 ms topen for both runs ? ...

Y

& turns out that this is definitely repeatable... at 2300 revs it separates. even after i richened up the higher revs on the map. i did it longer with more data points, and just past 3500 revs. it maintains same AFR through 2300, then about 0.5-0.7 AFR leaner in sequential all the way up to 3500.

...So maybe if you used 0.8 ms for both runs, you would see an afr difference at low rpm and no afr difference at high rpm.

actually, as topen gets bigger (0.8ms), then sequential AFR is ~leaner~ toward idle (intuition would suggest same thing).... so i suspect that it would start out with a sequential lean gap up to 2300 RPM, and the gap would actually grow after that. anyhow, it would be worth looking at effect of topen to these data plots.

right now, injection delay is set to 15%. pretty much arbitrary starting point... i think that was leftover when i tried to move calculations around trying to discover what was causing the 'phase shifting'.

i think it would be particularly interesting to look a data plots like the above graph by changing injection timing delay and look at the effect of the burn. question is: how do you determine what is best?

[grippo]

Scott,

Can you plot the exact same data in your previous graph but use PW instead of AFR. I just want to verify there are no surprises in the PW calculation.

[TheMonkey]

Scott,

Can you plot the exact same data in your previous graph but use PW instead of AFR. I just want to verify there are no surprises in the PW calculation.

Sure... from same data set... the PW is on the left axis. I started thinking about the fueling calcs though.... the second variable of fueling is MAP. So... it may be that at 2300, it starts to become 'unhinged' and self fulfilling; actually getting different VE direction from the fuel map. On the right hand axis (in light blue and yellow), it shows MAP. Map starts to unhinge by about a factor of 1 kpa. this test really should be done by flattening out the VE vertically in this range on the fuel map so it gets the same VE instructions at every RPM.

remember the video game when you were a kid... 'Asteroids'? if you stay in the middle and just spin, the game is pretty easy to control. as soon as you boost and start to travel, it's a game changer. very hard to control.

[grippo]

The PWs look reasonable. I thought about tclose time which really needs to be subtracted out of topen. When you do that maybe 0.15 ms isn't too unreasonable. If you used that number then you should get that wasted and sequential AFR match at high rpm, and wasted is lean at low rpm, the leanness due to poor mixing of the 2 squirts - maybe one going partially out the exhaust. But this is all speculation.

[TheMonkey]

Al-

i have been thinking about, and taking a close look at the code. this formula: [ pw_wasted = topen + (pw_seqtl - topen) / 2 ] seems to me the correct representation of how it should work.

as i have read through some of this post again, it sounds like you were thinking that of topen variable is changed, it shouldn't change the transition AFR from wasted to sequential. BUT, that formula changes the wasted pulse width as topen gets changed. SO it makes sense that adjusting topen will make the transition leaner/richer. most importantly, as i think about this... it really must be adjustable otherwise how would it be adjusted into balance if it didn't start like that.

for instance, the formula could be: [ pw_wasted = (topen / 2) + (pw_seqtl - topen) / 2 ] which would maintain a constant wasted PW as topen changes, but the transition would not be adjustable.

anyhow, i guess my point is that i like the formula, i like the behavior, and it seems to be doing exactly as commanded. and i agree... the very low 'net' topen time is probably that low because of the closing time.

most importantly.... i was really putting my foot into the car today, and i give the sequencer a huge thumbs up! been very easy to tune the fuel. drives sooo nice. have prob a couple hundred miles on it without ANY hiccups. if i can stop driving it for a bit, i'll fool around with the injection timing delay and do some of those data plots.

[cmonref]

the formula could be: [ pw_wasted = (topen / 2) + (pw_seqtl - topen) / 2 ] which would maintain a constant wasted PW as topen changes

This equation reduces to: pw_wasted = pw_seqtl / 2 which sounds right since wasted injects only half of the sequential amount, twice as often.

One day I'll be having the fun you are, but you will have paid the frustration price of working out the bugs. Many thanks.

Brian