Things I have learned for the newbies (like myself)
Posted: Sun Apr 05, 2009 7:25 pm
Aircooled VW flat four
2276 cc
Twin dual throat throttle bodies
EDIS
MSII ver.357
Getting there and enjoying the process although it can be frustrating at times. The cool part is that I know more about EFI and tuning then I ever imagined I would. In comparison to some of the more experienced squirters on here - I have a long way to go. Anyway, it seems to me that most of the information you need for the successful completion of a megasquirt process is somewhere between here and the megamanual although it can sometimes be hard to find and really hard to remember where you found it the first time. Also, guys like Grippo and Matt Cramer, take so much of their knowledge for granted (that was a complement guys) that they don't realize that some of the new guys got left behind on some of the explanations. So, here are some simple things that gave me that lightbulb thingy above my head when I figured them out and maybe some other new guys can be helped by these:
The VE table is the basis of the tune. People have asked why you even need a VE table when you have a wideband sensor and can use that to adjust the pw "as you drive". What I have found is that you start out with a VE table that (if yours is like mine) is way off. You also start off with an AFR table that appears to make some sense and that you may be able to make some educated guesses toward changing for the better. In my case, the reading from the wideband is not that stable and if you were to allow too much authority to the EGO correction, you would have your pw adjusting up and down to try to match your AFR table to a ve table that is a mile away and it would never get there (especially in the beginning where you may experience rich or lean missfires sending the afr really lean). The goal is to get the ve table and the afr table close so that EGO correction is minimal and then tune from the afr table. It becomes a circlular exercise going from one table to the other to try to get them close and then try to get the afr where it needs to be. For a long while, I ended up with my ve table so far out that the ego correction was pinned at its max. You need to use that info to adjust individual cells in the ve table to get them close first. The result (I think) should be a stable and accurate ve table that is overridden by a small percentage by the ego correction.
Autotune is a valuable tool but it has its limits. Autotune tries to take the ego correction (which is the difference between the target afr you have in your afr table and your ego reading from your interpreted O2 sensor) and make changes to your ve table to close the gap. So, the best it can do is try to make your ve table match what your afr table is looking for. This is a good thing but, it is only as good as your afr table. Also, I have had trouble setting the parameters of autotune to expand the operating window for initial tuning. The changes in megatune don't seem to "take" all the time and even changing the .ini file didn't seem to matter much. I think this is because there is not only a window which represents a subset of cells in the ve table that autotune will change but there is criteria that allows autotune to change cells only when in a certain proximity of a vertex. Even with some success in changing these parameters I still get a lot of the "outside of the window" messages and the "not near a vertex" type messages when I just wish it would hurry up and change the table. Again, this all gets much better when the ve table begins to match the afr table or vice versa. I have been doing a combination of letting the autotune do its thing and jumping in to manually change the ve table in surrounding cells to speed up the process. The megalogviewer's ve table analyzer works in a similar fashion as the autotune but it does it after the fact. You datalog a drive, load the configuration file you used for that drive and let the analyzer do its thing. It comes up with changes to the ve table that once again, try to get the ve table cells to match the corresponding afr table cells. As in autotune, this is only as good as the afr table. Its probably safer to use since you can do it at the curb or in your house and then re-load the resultant ve table rather than driving around with a laptop on the passenger seat and trying to make changes between wot attempts.
I chased and chased a lean condition at wot and near wot that drove me nuts because the results would change. I would adjust the lean spots out through the iterative process above and then, with the same conditions and settings, they would return in a subsequent drive. What I figured out was that when changing from a carbureted engine to efi, the fuel sloshing around in your tank as it gets low will starve the fuel pump. This revelation resulted in the making of a swirl tank or a surge tank - not sure if these are the proper names but, basically an auxiliary tank external to the fuel tank that reduces the possibility of this sloshing letting air into the supply.
Big WOW! - Although the engine ran better than it did with carbs it still wasn't "right" (don't know where this whole process ends btw). Decided to go back over the initial configuration and start again fresh. In the process I started to wonder (and read some more) about the whole injector staging thing. I initially had selected 2 squirts alternating since this seemed to be correct for my engine. I decided to change to 2 squirts simultaneous and holy $%#* - the engine runs sooo much better. Idles better and runs much smoother through the entire operating range. Even runs considerably leaner than I could ever get it before. I assume it has something to do with the injector opening timing in relation to the intake valve opening.
Anyway,
These are my early findings and although there are many who know way more than I do on this topic I just thought that maybe some new guy struggling down the same path that I am on might benefit.
The project moves forward.
Paul
2276 cc
Twin dual throat throttle bodies
EDIS
MSII ver.357
Getting there and enjoying the process although it can be frustrating at times. The cool part is that I know more about EFI and tuning then I ever imagined I would. In comparison to some of the more experienced squirters on here - I have a long way to go. Anyway, it seems to me that most of the information you need for the successful completion of a megasquirt process is somewhere between here and the megamanual although it can sometimes be hard to find and really hard to remember where you found it the first time. Also, guys like Grippo and Matt Cramer, take so much of their knowledge for granted (that was a complement guys) that they don't realize that some of the new guys got left behind on some of the explanations. So, here are some simple things that gave me that lightbulb thingy above my head when I figured them out and maybe some other new guys can be helped by these:
The VE table is the basis of the tune. People have asked why you even need a VE table when you have a wideband sensor and can use that to adjust the pw "as you drive". What I have found is that you start out with a VE table that (if yours is like mine) is way off. You also start off with an AFR table that appears to make some sense and that you may be able to make some educated guesses toward changing for the better. In my case, the reading from the wideband is not that stable and if you were to allow too much authority to the EGO correction, you would have your pw adjusting up and down to try to match your AFR table to a ve table that is a mile away and it would never get there (especially in the beginning where you may experience rich or lean missfires sending the afr really lean). The goal is to get the ve table and the afr table close so that EGO correction is minimal and then tune from the afr table. It becomes a circlular exercise going from one table to the other to try to get them close and then try to get the afr where it needs to be. For a long while, I ended up with my ve table so far out that the ego correction was pinned at its max. You need to use that info to adjust individual cells in the ve table to get them close first. The result (I think) should be a stable and accurate ve table that is overridden by a small percentage by the ego correction.
Autotune is a valuable tool but it has its limits. Autotune tries to take the ego correction (which is the difference between the target afr you have in your afr table and your ego reading from your interpreted O2 sensor) and make changes to your ve table to close the gap. So, the best it can do is try to make your ve table match what your afr table is looking for. This is a good thing but, it is only as good as your afr table. Also, I have had trouble setting the parameters of autotune to expand the operating window for initial tuning. The changes in megatune don't seem to "take" all the time and even changing the .ini file didn't seem to matter much. I think this is because there is not only a window which represents a subset of cells in the ve table that autotune will change but there is criteria that allows autotune to change cells only when in a certain proximity of a vertex. Even with some success in changing these parameters I still get a lot of the "outside of the window" messages and the "not near a vertex" type messages when I just wish it would hurry up and change the table. Again, this all gets much better when the ve table begins to match the afr table or vice versa. I have been doing a combination of letting the autotune do its thing and jumping in to manually change the ve table in surrounding cells to speed up the process. The megalogviewer's ve table analyzer works in a similar fashion as the autotune but it does it after the fact. You datalog a drive, load the configuration file you used for that drive and let the analyzer do its thing. It comes up with changes to the ve table that once again, try to get the ve table cells to match the corresponding afr table cells. As in autotune, this is only as good as the afr table. Its probably safer to use since you can do it at the curb or in your house and then re-load the resultant ve table rather than driving around with a laptop on the passenger seat and trying to make changes between wot attempts.
I chased and chased a lean condition at wot and near wot that drove me nuts because the results would change. I would adjust the lean spots out through the iterative process above and then, with the same conditions and settings, they would return in a subsequent drive. What I figured out was that when changing from a carbureted engine to efi, the fuel sloshing around in your tank as it gets low will starve the fuel pump. This revelation resulted in the making of a swirl tank or a surge tank - not sure if these are the proper names but, basically an auxiliary tank external to the fuel tank that reduces the possibility of this sloshing letting air into the supply.
Big WOW! - Although the engine ran better than it did with carbs it still wasn't "right" (don't know where this whole process ends btw). Decided to go back over the initial configuration and start again fresh. In the process I started to wonder (and read some more) about the whole injector staging thing. I initially had selected 2 squirts alternating since this seemed to be correct for my engine. I decided to change to 2 squirts simultaneous and holy $%#* - the engine runs sooo much better. Idles better and runs much smoother through the entire operating range. Even runs considerably leaner than I could ever get it before. I assume it has something to do with the injector opening timing in relation to the intake valve opening.
Anyway,
These are my early findings and although there are many who know way more than I do on this topic I just thought that maybe some new guy struggling down the same path that I am on might benefit.
The project moves forward.
Paul
