Starting to adjust EAE and decel settings please comment
#43
As I'm editing my EAE profiles while going through my datalog. I remember you (Brain) say ing that you adjust both sucked and adhere settings.
If I want to tune out a rich spot under acceleration at a given MAP reading. Would you increase the sucked from walls AND decrease the adhere to walls at the same time for this MAP reading? Currently I have only been adjusting the sucked from walls. Then I was going to go back and do adhere to walls, then realizing that this seems flawed.
If I want to tune out a rich spot under acceleration at a given MAP reading. Would you increase the sucked from walls AND decrease the adhere to walls at the same time for this MAP reading? Currently I have only been adjusting the sucked from walls. Then I was going to go back and do adhere to walls, then realizing that this seems flawed.
#44
Boost Czar
iTrader: (62)
Join Date: May 2005
Location: Chantilly, VA
Posts: 79,501
Total Cats: 4,080
If it's rich on account of EAE, you will need to alter adhere at the same time....you need to watch the gauge and see when it's increasing fueling based on EAE, the initial stab or over time. Otherwise it might just be the fuel map. This is why 5th gear makes this easier to tune, limits a lot of other variables.
If your adhere to walls tells the MS a lot of fuel got stuck, but the sucked from walls tells the MS little is being removed, then it might increase the PW over time after the initial throttle change.
EAE is ALWAYS active.
If your adhere to walls tells the MS a lot of fuel got stuck, but the sucked from walls tells the MS little is being removed, then it might increase the PW over time after the initial throttle change.
EAE is ALWAYS active.
#45
If it's rich on account of EAE, you will need to alter adhere at the same time....you need to watch the gauge and see when it's increasing fueling based on EAE, the initial stab or over time. Otherwise it might just be the fuel map. This is why 5th gear makes this easier to tune, limits a lot of other variables.
If your adhere to walls tells the MS a lot of fuel got stuck, but the sucked from walls tells the MS little is being removed, then it might increase the PW over time after the initial throttle change.
EAE is ALWAYS active.
If your adhere to walls tells the MS a lot of fuel got stuck, but the sucked from walls tells the MS little is being removed, then it might increase the PW over time after the initial throttle change.
EAE is ALWAYS active.
#47
I am absolutely useless at tuning this. I can't find the correlation between my EAE% = 105% and how the translated to adjusting my adhere and sucked from tables if my afr is rich by 10% (this is hypothetical situation) and my adhere value is 20% while my sucked from is 8% This makes no sense. I went back to the original table and it feels better with it turned off. I did tune my regular AE last night, and have nice crisp throttle response. yay
#48
Adheres to wall is calculated as a percent of each squirt. Sucked from wall is a percent pulled from the puddle that has collected on the wall...
The size of each squirt is very small compared to what sits in the puddle on the walls of the intake port.
So what you end up with is a larger percentage of a small number (adhere to walls, getting added to the puddle on every squirt) eventually equalizing with a small percentage of a big number (getting sucked off the walls on each intake event). You can see the amount (in usec) that the code thinks is on the walls of the port by looking at the wallfuel variable.
When EAE is at "100%" it means that the amount going into and out of the walls has equalized (wallfuel is not changing) and the algorithm is not increasing or reducing the amount of fuel in a squirt to compensate for anything.
When EAE is over 100% and won't come back down, it means that with your settings, you have told the code that more fuel is getting added to the walls than ever comes off, so it's constantly trying to compensate by squirting more fuel on each squirt. In other words, you've created settings that make it impossible to get back to equalibrium (in this case, wallfuel would be constantly increasing).
The most typical reason for getting into this situation is trying to tune for quick blips before tuning slow throttle movements. You keep changing added to walls to get better response, but it keeps not helping b/c the problem isn't the amount of fuel, it's how quickly the fuel is getting added after the transient starts. The solution there is to get it working right on everything but quick blips, and then add some standard TPS-based AE to handle quick blips.
The second most common problem is that VE is not actually tuned properly yet, and the user is trying to compensate with EAE settings. That won't work because EAE's purpose is to make sure the VE-calculated amount of fuel is actually getting into the cylinder.
Ken
The size of each squirt is very small compared to what sits in the puddle on the walls of the intake port.
So what you end up with is a larger percentage of a small number (adhere to walls, getting added to the puddle on every squirt) eventually equalizing with a small percentage of a big number (getting sucked off the walls on each intake event). You can see the amount (in usec) that the code thinks is on the walls of the port by looking at the wallfuel variable.
When EAE is at "100%" it means that the amount going into and out of the walls has equalized (wallfuel is not changing) and the algorithm is not increasing or reducing the amount of fuel in a squirt to compensate for anything.
When EAE is over 100% and won't come back down, it means that with your settings, you have told the code that more fuel is getting added to the walls than ever comes off, so it's constantly trying to compensate by squirting more fuel on each squirt. In other words, you've created settings that make it impossible to get back to equalibrium (in this case, wallfuel would be constantly increasing).
The most typical reason for getting into this situation is trying to tune for quick blips before tuning slow throttle movements. You keep changing added to walls to get better response, but it keeps not helping b/c the problem isn't the amount of fuel, it's how quickly the fuel is getting added after the transient starts. The solution there is to get it working right on everything but quick blips, and then add some standard TPS-based AE to handle quick blips.
The second most common problem is that VE is not actually tuned properly yet, and the user is trying to compensate with EAE settings. That won't work because EAE's purpose is to make sure the VE-calculated amount of fuel is actually getting into the cylinder.
Ken
#49
So what you end up with is a larger percentage of a small number (adhere to walls, getting added to the puddle on every squirt) eventually equalizing with a small percentage of a big number (getting sucked off the walls on each intake event). You can see the amount (in usec) that the code thinks is on the walls of the port by looking at the wallfuel variable.
When EAE is at "100%" it means that the amount going into and out of the walls has equalized (wallfuel is not changing) and the algorithm is not increasing or reducing the amount of fuel in a squirt to compensate for anything.
When EAE is over 100% and won't come back down, it means that with your settings, you have told the code that more fuel is getting added to the walls than ever comes off, so it's constantly trying to compensate by squirting more fuel on each squirt. In other words, you've created settings that make it impossible to get back to equalibrium (in this case, wallfuel would be constantly increasing).
When EAE is at "100%" it means that the amount going into and out of the walls has equalized (wallfuel is not changing) and the algorithm is not increasing or reducing the amount of fuel in a squirt to compensate for anything.
When EAE is over 100% and won't come back down, it means that with your settings, you have told the code that more fuel is getting added to the walls than ever comes off, so it's constantly trying to compensate by squirting more fuel on each squirt. In other words, you've created settings that make it impossible to get back to equalibrium (in this case, wallfuel would be constantly increasing).
Shouldn't adhere-to-walls be pretty universal since it is highly dependent on the surface area of the head that the fuel contacts? Assuming an good VE tune. Even with different injectors the req. fuel scales the actual volume of applied fuel to be the same. 1.6 heads and a 1.8 heads should have very similar adhesion percentage.
The sucked-from-walls seems to be the variable.
It seems like controlling this with two variables is unnecessary since you can add and remove fuel by manipulating either. If you get to zero with your adhere settings, then the VE table is probably too rich??? At the end of the day the values for both tables are probably incorrect if you actually measured these quantities. So we are manipulating two variabls to create a single unknown which is add or subtract fuel. Why couldn't we just have two independent AE tables, one MAP and one TPS. The tps would handle the large AE corrections during big throttle imputs, and the MAP based table would take care of the small corrections. Almost like two damped harmonic oscillators out of phase. I guess that would be like a PID algorithm for AE. I'm not trying to bash the model. My lack of understanding is driving me bonkers, and I want to understand it.
The most typical reason for getting into this situation is trying to tune for quick blips before tuning slow throttle movements. You keep changing added to walls to get better response, but it keeps not helping b/c the problem isn't the amount of fuel, it's how quickly the fuel is getting added after the transient starts. The solution there is to get it working right on everything but quick blips, and then add some standard TPS-based AE to handle quick blips.
Last edited by miatauser884; 12-17-2010 at 01:26 PM.
#51
Due to the nature of how a fluid reacts and bonds to a surface. I think a fairly reproducable quantity of fuel will have to be deposited on the head surface before the quantity getting to teh cylinder stabalizes.
Sure, this will include sucked-from-walls, but regardless of how it is timed you still need a certain quantity of fuel to produce a desired AFR.
This quantity is: fuel to meet desired AFR + fuel adhering to head. I think the fuel adhering to the head is fairly constant.
Sucked-from-walls will change with RPM. Therefore adhere to walls should alwasy be greater than sucked-from-walls because some fuel will always remain adhered to cylinder head walls. You can't remove what isn't there.
Adhesion-to-walls will fluctuate with RPM but should be pretty constant from one setup to the next as long as the surface are that the fuel is flowing through is consistant.
#53
EDIT:OK I rethought about what you just said, and now I agree. On non sequential injection you will get significant pooling. Since now there is more fuel availible than just teh amount that will adhere to the head surface.
Does this mean that dividing the default EAE adhesion values will be a better jump off point for tuning? I think so.
Thanks for humoring this discussion Brain. It's getting the wheels turnign and I'm enjoying it.
Which brings up an unrelated question. Is a non seq injected car running leaner than the AFR suggests due to this raw fuel entering the exhaust?
#57
This will help a lot.
Shouldn't adhere-to-walls be pretty universal since it is highly dependent on the surface area of the head that the fuel contacts? Assuming an good VE tune. Even with different injectors the req. fuel scales the actual volume of applied fuel to be the same. 1.6 heads and a 1.8 heads should have very similar adhesion percentage.
The sucked-from-walls seems to be the variable.
It seems like controlling this with two variables is unnecessary since you can add and remove fuel by manipulating either. If you get to zero with your adhere settings, then the VE table is probably too rich??? At the end of the day the values for both tables are probably incorrect if you actually measured these quantities. So we are manipulating two variabls to create a single unknown which is add or subtract fuel. Why couldn't we just have two independent AE tables, one MAP and one TPS. The tps would handle the large AE corrections during big throttle imputs, and the MAP based table would take care of the small corrections. Almost like two damped harmonic oscillators out of phase. I guess that would be like a PID algorithm for AE. I'm not trying to bash the model. My lack of understanding is driving me bonkers, and I want to understand it.
Shouldn't adhere-to-walls be pretty universal since it is highly dependent on the surface area of the head that the fuel contacts? Assuming an good VE tune. Even with different injectors the req. fuel scales the actual volume of applied fuel to be the same. 1.6 heads and a 1.8 heads should have very similar adhesion percentage.
The sucked-from-walls seems to be the variable.
It seems like controlling this with two variables is unnecessary since you can add and remove fuel by manipulating either. If you get to zero with your adhere settings, then the VE table is probably too rich??? At the end of the day the values for both tables are probably incorrect if you actually measured these quantities. So we are manipulating two variabls to create a single unknown which is add or subtract fuel. Why couldn't we just have two independent AE tables, one MAP and one TPS. The tps would handle the large AE corrections during big throttle imputs, and the MAP based table would take care of the small corrections. Almost like two damped harmonic oscillators out of phase. I guess that would be like a PID algorithm for AE. I'm not trying to bash the model. My lack of understanding is driving me bonkers, and I want to understand it.
On my rotary engine, I tuned EAE with batch injection, and upon switching to properly timed sequential injection, the EAE tune is now completely wrong. When you inject affects how much fuel gets into the puddle and how much comes out. So the ability to change both can GREATLY help tunability. On that particular engine, it runs better if you inject during peak air velocity since there is no hot valve to squirt on for better atomization. This means less fuel is going to go into the walls, and the amount coming out won't change that much percentage wise (But the puddle will be much smaller).
Bruce and Al's original X-tau implementation also had a single variable for what was getting added to the walls, and a 12x12 table of time values for evaporation/sucked from walls time, and very few people could get it working. After I wrote EAE, they looked into why X-tau wasn't working for most people, and it turned out they needed a curve for added to walls, and a curve for what was getting sucked from walls. Their current implementation now has both.
I recommend watching the wallfuel variable closely in the datalogs while tuning, as that can help you see what effects are occurring. I also recommend if you have a stim, get a cheap plastic cooking syringe from the grocery store, and hook it to your MAP sensor, and play with it on the bench to get a clear understanding of how it will work with your configuration.
Finally, using it with sequential injection does make it a lot easier to tune.
Ken
#58
Ken,
Whch direction would you expect to have to adjust the sucked from walls with sequential (assuming a spot on VE table)
I'm starting by reducing the adhere to walls by half since i am running sequential. I think the wallfuel value will be key. I noticed it increasing and decreasing but didn't know the units until you mentioned microseconds. I now should be able to create an approximate percent/microsecond correction scale.
The bad part is that you can't really reproduce a datalog unless tuning on a steady state dyno; so testing is hit and miss.
If given a value in the adhere column, and a value in the suck column. How does it calculate the percent increase? Is there a rule of thumb where ADHERE needs to be at least X amount greater than SUCK.
Whch direction would you expect to have to adjust the sucked from walls with sequential (assuming a spot on VE table)
I'm starting by reducing the adhere to walls by half since i am running sequential. I think the wallfuel value will be key. I noticed it increasing and decreasing but didn't know the units until you mentioned microseconds. I now should be able to create an approximate percent/microsecond correction scale.
The bad part is that you can't really reproduce a datalog unless tuning on a steady state dyno; so testing is hit and miss.
If given a value in the adhere column, and a value in the suck column. How does it calculate the percent increase? Is there a rule of thumb where ADHERE needs to be at least X amount greater than SUCK.
#59
I'm starting by reducing the adhere to walls by half since i am running sequential. I think the wallfuel value will be key. I noticed it increasing and decreasing but didn't know the units until you mentioned microseconds. I now should be able to create an approximate percent/microsecond correction scale.
The bad part is that you can't really reproduce a datalog unless tuning on a steady state dyno; so testing is hit and miss.
If given a value in the adhere column, and a value in the suck column. How does it calculate the percent increase? Is there a rule of thumb where ADHERE needs to be at least X amount greater than SUCK.
Code:
/* Calc actual amt of fuel to be injected in next pulse */ SQF = ((unsigned long) ((unsigned long) (tmp_pw1 - SOAtmp) * 100)) / (100 - AWC);
So the answer is that there's no simple answer for how it calculates how much to inject.
The percent increase is just the amount EAE calculates to inject divided by the original pulse-width.
I also forgot to add ... the amount getting added to the puddle can change based on MAP. Higher air pressure means fuel has a harder time evaporating/atomizing. Another reason to have both adjustable. Both change based on the physics of what is happening.
Ken
#60
I thought I would update this thread since I have once again started tuning this now that I am on the MS3. It still hasn't gone smoothly, but this is due to me not being diligent about checking my settings. IF you are going to tune EAE, then before you start, check the following:
Turn off other accel enrichment and set the EAE rpm/CLT tables to 100%.
I did a 100 mile trip and tuned along the way. I didn't turn off accel enrichment and I didn't zero the rpm table or the CLT table. If you perform the EAE adjustments at 3k rpm in fifth gear, then this really isn't going to be a problem. Just adjust the table so the 3k rpm is 100% on the rpm table and adjust your EAE tables accordingly.
I've read multiple threads that say you don't really need normal AE if EAE is dialed in. I'm finding that this is 100% true. I was amazed this morning at how well my car drove with normal AE off. I didn't miss it at all. Granted, I wasn't stabbing the throttle quickly.
Things I've noticed: Off idle stumble appears to be EAE related and not AE related since you are mostly moving the throttle slow and steady. In the past I would dial out the of idle stumble with AE.
I have not confirmed this next bit, but it looks promising as an indicator if your off idle EAE fuel is dialed in correctly. If you autotune and you see that you are getting large enrichment spikes off idle (i.e. 1200 rpm) there is a good chance that your EAE is not tuned properly.
It's not nearly as difficult as I expected since like the acceleration wizard, the EAE windows show your last throttle movement, so you know where to adjust. A small, quick, lean spike after the accelerator is depressed seems normal.
Make sure your afr table is smooth around the 3k/map area where you are cruising. If you are in an AFR table transition zone, it will be hard to determine if you AFR is fluctuation due to EAE or just trying to get to the afr table value.
I like tuning one table at a time. i.e. adhere to walls since I can focus solely on the afr associated with pressing the throttle, then tune the throttle lift AFR.
I think the EAE has one of the largest impacts on drivability for a DD, other than idle.
Turn off other accel enrichment and set the EAE rpm/CLT tables to 100%.
I did a 100 mile trip and tuned along the way. I didn't turn off accel enrichment and I didn't zero the rpm table or the CLT table. If you perform the EAE adjustments at 3k rpm in fifth gear, then this really isn't going to be a problem. Just adjust the table so the 3k rpm is 100% on the rpm table and adjust your EAE tables accordingly.
I've read multiple threads that say you don't really need normal AE if EAE is dialed in. I'm finding that this is 100% true. I was amazed this morning at how well my car drove with normal AE off. I didn't miss it at all. Granted, I wasn't stabbing the throttle quickly.
Things I've noticed: Off idle stumble appears to be EAE related and not AE related since you are mostly moving the throttle slow and steady. In the past I would dial out the of idle stumble with AE.
I have not confirmed this next bit, but it looks promising as an indicator if your off idle EAE fuel is dialed in correctly. If you autotune and you see that you are getting large enrichment spikes off idle (i.e. 1200 rpm) there is a good chance that your EAE is not tuned properly.
It's not nearly as difficult as I expected since like the acceleration wizard, the EAE windows show your last throttle movement, so you know where to adjust. A small, quick, lean spike after the accelerator is depressed seems normal.
Make sure your afr table is smooth around the 3k/map area where you are cruising. If you are in an AFR table transition zone, it will be hard to determine if you AFR is fluctuation due to EAE or just trying to get to the afr table value.
I like tuning one table at a time. i.e. adhere to walls since I can focus solely on the afr associated with pressing the throttle, then tune the throttle lift AFR.
I think the EAE has one of the largest impacts on drivability for a DD, other than idle.