IW STi Forum banner
1 - 20 of 55 Posts

tshrey

· Registered
Joined
·
580 Posts
Discussion starter · #1 · (Edited)
The boys over on the opensource forums have finally licked this and created a way for anyone to custom tune their car for the intake/fuel resonance that exists in the GR engine. The symptoms are slight but tremendously annoying hesitations/surging in acceleration under light load (0 boost and lower) that occur between 2400 and 3000 rpm.

The reference thread is here:
RomRaider • View topic - How to fix the GR STi 2400-3200 stumbles

You need a recent version of Cobb's accessTuner Race, an AP2 and Engine Compensation spreadsheet that can be downloaded here:
RomRaider • View topic - Spreadsheet to generate fuel trim pivot table (RPM v. MRP)
I would recommend reading this thread as well.

I'm not going to hold anyone's hand and write an exact tutorial on what to do - I believe you should not be changing things in ATR if you don't understand what you are doing. But I will provide an overview.

Open the spreadsheet in the form that you download it. The main table will show you the data from another user. The header columns are the things you need to log with your AP. The only wrinkle is that RomRaider and Cobb call the same data different things - manifold relative pressure is boost in cobb lingo, A/F correction #1 is fuel trim (short) in Cobb, A/F learning #1 is fuel trim (long) in Cobb. The rest are straightforward.

Drive on the highway for between 30 minutes to 2 hours logging the required data. Try and keep the throttle relatively constant for minutes at a time and drive across the whole rpm range that you need data for - you will be driving from 60 to 75 mph. If your highway has long slight hills even better.

Open your log in Excel and rename your headers to match the headers in the spreadsheet (they don't have to be in the same order, but I usually rearrange them because I'm anal). Once you've done that save the file, use the spreadshreet and click on 'get data'. It will ask you where the file is. Once it has your data, click on 'Filter'. The spreadsheet generates a pivot table of RPM vs manifold pressure, showing fuel correction needed at each cell point in the table. This is the data you are going to use to adjust your Calc. Engine Load Compensation B table under Sensor Calibrations in AccessTuner. The labels in the spreadsheet for pressure are in boost, the labels in ATR are in boost/14.7, but the columns correspond exactly with those in ATR. Any cell that has sufficient data will be black, cells with not enough data will be empty or grey. Add the value of the cells in the pivot table to what is in the Engine Load table.

To generate the values for Compensation A, you need to click on the ROM tab in the spreadsheet and change the RPMs that the pivot table uses (Compensation A has additional RPM values at 2700 and 2750). Once you change the values in the ROM table, 'Filter data' again and you will get a new pivot table.

The spreadsheet will also generate some nice 3D graphs. In my case it was blatantly obvious what was causing my stumbles. The short term trim was adding over 25% more fuel at 2400 rpm and light load and removing 15% at 2800 rpm at slightly higher load. After adding the pivot values to my compensation tables and relogging all my trim values are less than 5% and I don't get ANY knock at cruising anymore.

If I'm not clear on something let me know.
 
You are the man!

I've been reading about the load compensation tables on the RR forums and followed these exact threads and was planning to do something very similar to what you have done. So this solidifies my thoughts in my mind as for what needs to be done.

Now I just need to do some logging and get lots of data with those columns only.
 
Discussion starter · #6 ·
Here are some of the tips to take note of:
RomRaider • View topic - How to fix the GR STi 2400-3200 stumbles
RomRaider • View topic - Spreadsheet to generate fuel trim pivot table (RPM v. MRP)

Did you zero out the compensation table before doing your logging? Or just run with the current map values?

This is probably one of the best and potentially easiest things that people can do as far as self-tuning goes.
Leave the compensation tables as is. What you get in the pivot tables are values that are added to the values in the compensation table.
 
Note that even after you've made these compensation table changes, STFT and LTFT (i.e. AF learning and AF correction) may continue to jump around +-5%. It's a lot better though.

I am still experimenting, but I also believe that removing the AVCS overlap (i.e. the 40 deg. exhaust retard AND intake advance) in the low-load, lower rev-range, might also help to make the engine run smoother in some of the same cells where the closed-loop fuel trims are being adjusted. This might further make the engine run smoother since you are eliminating the EGR effect (that it was set from the factory to do in order to enhance emissions performance). ANd I would suspect that it might make the fuel corrections more predictable since the EGR effect is taken out of the equation.

I used to see some low-load, low-rpm FLKC, but so far I haven't after tweaking the AVCS like so.
 
Discussion starter · #8 ·
Worked great for me. Remember to post your comp tables and mods in the romraider thread. Good reference for others!
I don't think they would want it there. The compensation table is what is required on top of the map you have loaded. If you are running a Cobb map that doesn't help people who are running the ROM's from the open source download page.

It probably wouldn't be a bad idea to start archiving them here, gives everyone a starting point. I'll post mine when I get back in front of my laptop.

I'm going to try nil's idea and take out some of the overlap in the AVCS tables.
 
I'm going to try nil's idea and take out some of the overlap in the AVCS tables.
Either it's the cooler weather, or the AVCS cruise mod has reduced cruise knock. I am no longer seeing any FLKC over a few days of logging. Previously, I would almost immediately see FLKC of a degree or two in closed loop operation.

My STFT and LTFT are still +- 5%, but they look like they could use a little tweaking. My average closed-loop corrections are for the most part +-2% which is pretty good, but the standard deviation over tens of thousands of data points is still larger, like 5% in some cells.

Therefore, there must be something else in the equation that is influencing the larger corrections, but no matter. The engine is as smooth as a BMW 6 cylinder now :tup: I can't believe that Subaru/Fuji had to compromise the smoothness and driveability of the motor so much in order to meet emissions. It's very stupid. At least the community has been able to make so much progress in reverse engineering that we can fix their mistakes.
 
Is this issue something everyone should be worrying about? Stock or Cobb stage one for all GR models up to 2011?

Is this bad for the engine or is it just a irritant to have the hesitation there?

Reason i ask is i always have slight amounts of knock learning on occasion while cruising.

Does this happen to both stock and AP?
 
Is this issue something everyone should be worrying about? Stock or Cobb stage one for all GR models up to 2011?

Is this bad for the engine or is it just a irritant to have the hesitation there?

Reason i ask is i always have slight amounts of knock learning on occasion while cruising.

Does this happen to both stock and AP?
Worth worrying about? No. Irritating? Hell yes.

I can't speak for Cobb, but this issue is there on the stock ROM. 08-10. Not sure about 2011.

Slight knock while cruising is probably not a big deal. The knock that is more worrisome, afaik, occurs under high load/boost
 
I had 4 hours of logs that I ran through this past weekend and finally ran the spreadsheet and uploaded the new map and it is AMAZING! I'll post more in detail how I did it using the data I had since it wasn't as easy as I thought it would be.

My car is soooo smooth accelerating from 1500-3500rpm now.
 
Here are my step-by-step instructions I did:

Spreadsheet: RomRaider • View topic - Spreadsheet to generate fuel trim pivot table (RPM v. MRP)

I couldn't open it in MS Excel 2007 without getting this error: "This workbook has lost its VBA project, ActiveX controls and any other programmability-related features", but it worked in Excel XP.

I had to change Macro Security: Tools, Options, Security, Macro Security, Medium.

Logging:

The spreadsheet says you need to log certain parameters, but the data is only collected from: RPM, Throttle, Boost, MAF (V), Fuel Trim (Short), Fuel Trim (Long), and Air/Fuel Ratio. I logged a few more parameters such as closed loop status and MAF (g/s), but I don't think it's necessary. I logged on a road trip with speeds 45-65mph making sure to go through the rev range up to 3400 at constant throttle in 3rd through 6th gears. Be sure to check the Barometric Pressure in live data every so often so you'll know what the atmospheric pressure is to put into the spreadsheet. Here in VA, it was about 14.4-14.5. It's not necessary to 0 the load comp map out. I think it's good to have some positive values in the 2600-3000rpm range when logging.

Data manipulation:

I couldn't get the add data function to work, so I just appended my spreadsheet data together and changed the time values of the 2nd and 3rd log files to be higher than the 1st spreadsheet using 2 columns and adding the max time to the new time just so it was in chronological order. Maybe it's not necessary.

Data import and filter:

- On the input tab, click clear and get data, and select your log file.
- Put your barometric pressure in the box
- Match up the required headers by selecting the bullet options.
- Run the filter

Optimizing the spreadsheet to work with ATR:

- Go to the ROM tab, and put in row and column headers with the same rpm numbers that are in Cobb's Calc. Engine Load Compensation Table A. I added another column with a heading of 19 after 3.83, just to make sure there wouldn't be an error (I think this is only necessary if your logged boost numbers go above 3.83). Fill in the excess 0's.

- On the EL Comp tab, add missing data by right clicking on the header for Average of Total Correction and Count of Total Correction Pivot Tables. Use the Wizard, click the back button, and change the cell references to include all of the cells on the EL Comp data page.

- Refresh the Pivot Tables and you should get the results.

- You might need to adjust the line graph to see all of the numbers, but it doesn't affect the Average results, so it's not necessary.

Changing values in ATR:

Like all disclaimers, don't mess with this stuff unless you know what you're doing. Changing load compensation values could potentially harm your engine. But if you compare different maps' load compensation values, you'll get an idea of what is normal for OTS maps. Take a look at the RR thread and you'll see that some people are getting some crazy values for the GR.

- Copy your current map's Load comp A into a blank spreadsheet and then copy the average of total correction into the same worksheet, spacing the 650rpm rows on easily identifiable rows, then add the cells into a 3rd section to get the total compensation values.

- Copy these numbers back into ATR's load comp maps A and B, noting the offset of the B map. Also, because of the fuzzy math, the numbers might be off by 0.39 or so, so you can go through and optimize it further to the closest number.

- Save map, upload to AP, reflash, and enjoy!

Someone try it and let me know what I need to edit.
 
Has anyone tried logging the accel vs cruise parameters and try to build different load compensation maps for accel and cruise? To me, it seems like the different AVCS settings would change the volumetric efficiency to the point that the load compensation accel and cruise maps should be slightly different. However, the stock ones are pretty similar, so I wonder if it is a big deal worth considering.
 
Has anyone tried logging the accel vs cruise parameters and try to build different load compensation maps for accel and cruise? To me, it seems like the different AVCS settings would change the volumetric efficiency to the point that the load compensation accel and cruise maps should be slightly different. However, the stock ones are pretty similar, so I wonder if it is a big deal worth considering.
Which parameters are you referring to?

Yes, volumetric efficiency is a function of AVCS parameters.

A good way to solve this problem is to set the AVCS tables identically for cruise/non-cruise. ;)
 
The normal parameters for running the load compensation spreadsheet that need to be logged are AF Correction 1, AF Learning 1, AF Ratio 1, Boost, MAFv, Throttle, and RPM (as mentioned above).

I thought about making two maps with all-cruise AVCS and all-accel AVCS, but as we know, the timing and AFR values are also slightly different between accel and cruise. Didn't the old v1.02 maps have the cruise knock because they had the advanced accel timing on the cruise AVCS?

I also thought about logging the accel/cruise switch and then parsing it all out, but that would be a pain, and there would be so little accel data compared to cruise data.

Problem is, I still notice hesitations, but usually when it's in the accel maps.
 
The normal parameters for running the load compensation spreadsheet that need to be logged are AF Correction 1, AF Learning 1, AF Ratio 1, Boost, MAFv, Throttle, and RPM (as mentioned above).
Right, I wasn't sure which parameters you were referring to. YOu might also be interested in logging atmospheric pressure, and closed-loop fueling target. CLosed-loop fueling target I think will allow you to know when the target AFR is being achieved (by comparing the target CL AFR to the AFR sensor reading, see what I mean?) And when they are roughly equal within some margin of error, you know the A/F trim and so on is close to correct allowing you to recalculate compensation at that cell.

I thought about making two maps with all-cruise AVCS and all-accel AVCS, but as we know, the timing and AFR values are also slightly different between accel and cruise. Didn't the old v1.02 maps have the cruise knock because they had the advanced accel timing on the cruise AVCS?
I Don't have experience with Cobb maps, but couldn't you set the cruise timing and AVCS identically to non-cruise?

Problem is, I still notice hesitations, but usually when it's in the accel maps.
The stumbling could be from transitions from cruise to non-cruise. Or it could be your timing. What's your cruise/non-cruise timing look like? If there are big differences, when the ECU transitions from cruise to non-cruise and vice versa, it could be adding/removing quite a few degrees in a very short amount of time which would be felt as a hesitation/stumble.

If you can post a screenshot of your compensation table.
 
I will have to try this with my tune after my warranty runs out
 
1 - 20 of 55 Posts