DISCLAIMER: This is not a "how-to", this is documentation on how I used information on the interwebs to solve the 2400-3000 rpm low load acceleration stumble on my 2011 STI. YOU are responsible for any changes you make to your maps.
The GR stumble is an issue that is not detrimental to the car, just annoying to deal with. It happened to me just about everyday when driving normally. Under light load, partial throttle (at about 2800 rpm) the car would hiccup or stumble for a moment. After doing tons of searches on here, IWSTI and Rom raider, I took existing information and eliminated the stumble on my car. The result is beautifully smooth, low rpm, partial throttle acceleration.
I read, and re-read every post in the following threads...
1) The spreadsheet located here
3)Windows Based Excel (I used Excel 2007 on a virtualized version of windows XP on a mac)
4)Basic excel data manipulation abilities
5)Accesstuner Race (free download from cobb's website)
6)Ability to Datalog using Accessport
I logged the following parameters:
A/F Correction 1
A/F Learning 1
A/F Sens 1 Ratio
Also the "live" parameter I made note of while logging was the barometric pressure. I need that number for the spreadsheet. I logged for at least an hour or so of highway driving. I went through most of the rpm range from 2000-4000 in 4th, 5th, and 6th. I dwelled for a while at each rpm for each gear. I spent a couple minutes around 55mph, 60mph and so on. I needed enough data points for the spreadsheet to get enough of a sample to size to make the averages.
1) Once I had my log, I renamed the headers of each column to EXACTLY the same as the spreadsheet. I saved in csv format.
2) I opened the romraider spreadsheet, made sure I enabled macros, and clicked "clear and get data". I located my modified header csv file. I entered the barometric pressure into the box. (for me 14.5 psi).
This is where it gets a bit complicated, but once I was in the files it made sense...
3) I opened my Stage 1 OTS map from cobb (Stage 1 93 octane boost select). I opened the sensor calibration folder and opened the table labeled "load compensation (cruise)"
4) I copied the first row (the headers) and copied into a blank spreadsheet. I then made the following conversion to each number. Note the header title is a pressure in BAR. I called the header number X. The barometric pressure I logged was Y. The conversion is: (X*Y)-Y. The conversion makes a new row of headers. I copied that row as new headers into the tab labeled "ROM". The ROM tab is just used for the pivot table. Now the pivot table that the spread sheet generates matches the numbers in my cobb table.
5) I went back to the input tab and clicked filter. After the calculation was complete, the pivot table generated is on the EL comp tab. Its the top table labeled average correction.
6) notice the pivot table is full of bold, grey and empty cells. The empty cells had no data, the gray cells didn't have enough samples for an average, and the bold cells are valid data. I didn't worry that not all the cells had data, most did, I think it might be next to impossible to get data for EVERY cell.
7) I copied the entire load comp cruise table into a new spreadsheet, rpm titles and headers included.
8)Next to it on the same spreadsheet, I copied the pivot table, I made sure to line up the pivot table on the appropriate rpm row. The pivot table doesn't have data for every corresponding cell on the cobb table. It's only concerned with a certain portion of the table. So I made the pivot table equal in rows and columns as the cobb map, i simply made every empty cell 0, and every cell with a gray number 0.
9)So now i had two tables, with the same headers and row labels, the exact same size, side by side. One table is the table from the cobb map, one table is the pivot table with a bunch of zeros added to make it the same size as the cobb table.
10) I made a third table under the cobb table, same headers and rpm labels. I made the first cell equal to the value of the first cell on the cobb table PLUS the value of the first cell from the pivot table. I then dragged the formula across the row, then down the columns. (please reference excel for dummies lol).
11) I copied the new table into access tuner race (values only).
12) I Repeated the process for the accel table. (get the new headers, do the conversion, paste new values into ROM table, filter data, get new pivot table, add pivot table to existing map, paste new map into access tuner).
13) I saved the map with a new name
14) I flashed the new map, went for a cruise and voila! NO STUMBLE.
I think this may be an iterative process, I may perform all these steps again to see if anything changes, or just datalog again and look at the graphs in the spreadsheet to ensure there are no outliers anymore.
Hopefully this documentation will help others diagnose and fix the same issue on their car. FYI, this was the first time I changed any of the tables in access tuner and it couldn't have been easier.
I'm still re-reading the reference material to determine exactly why this eliminates the stumble.