RomRaider

Long time back I installed a K&N Typhoon Cold Air Intake on my 2004 WRX Wagon. This is NOT the shorty intake you usually see installed. This is the long bastard that snakes through the fender to where the snorkus used to be. The filter sits behind the fog lamp and in front of the RF tire.

I new to open source tuning and thus have been taking it slow and reading whole bunches. Lots of it from this site.

Because I have the stock 420cc injectors installed and the Typhoon already on I decided to tackle the scaling of the MAF Sensor table first.

With the help of RomRaider CL MAF tools, Airboy's spreadsheet and Bad Noodles spreadsheets I have scaled it to the following:


This is after a cumulative CL cruising and idling data collecting sessons of ~3 hours and 9 WOT pulls for OL data. I have a Innovate LC-1 wideband feeding me AFR data from the downpipe.

I couldn't get any data for the ranges below 1.13 volts or above 4.30 volts. The 4.30 volt change you see above is just from smoothing.

Two things I noticed: One, the ranges from 1.25 volts to 2.19 volts (inclusive) were off by over 5%. In one case up almost 16% correction had to be applied (1.41 volts). Two, in the WOT (OL) range of my data the K&N flowed almost like stock. From 2.89 volts to 4.18 volts the intake flowed about 0.2% more.

Next for me is to do a few more OL data collections to confirm MAF scaling, fiqure out what's up with the areas around 2.66, then install DW 750cc injectors and scale for them.

I was thinking of applying an adjustment of 0.15% for all the areas above 4.3 volts, inclusive . The errors seem consistent leading up to that area. And since I know I have a different intake on I think it's a good hypothesis. Plus side is that I'm telling the computer there's slightly more air coming in so it will add slightly more fuel.

Any thoughts?

Closedloop fueling stopped around 2.5volts. You need to scale from 2.5v and above using a wideband. Use the Openloop MAF scaling tool.

I'm onto my fourth iteration in an attempt to nail down this bastard. I chased my tail on iterations one and two. The g/s errors just kept jumping all around. With attempt three and four I took the advice of the knowledge here and moved my WO2 reading backward in time ~0.33 seconds (or a two cell shift up if your logging about 6 data points a second). This in an attempt to estimate the AFR in the cylinder where the actual combustion occurs instead of downwind where the Bosch wideband sensor reads it in the downpipe. I'm using Innovate's LC-1 with that Bosch.

Logging iteration three had 9 runs with a total of 613 data points .
Logging iteration four had 12 runs with a total of 666 data points.

So now with less than 1.0% error everywhere I could get logs I present:

I noticed that there are a lot of areas of the MAFv range that are underutilized. I took some logs of me just driving cruising around. Idiling at lights. 75 on the highway. 45 mph in town. Just driving like I normally would but without 'getting on it'. Took all that data and did a frequency analysis of the voltage range. Noticed five things.

1. MAFv logged never ends in an odd number.
2. Voltage from 0.00 - 1.08, inclusive, was never logged.
3. 1.10 and 1.12v were logged just once each in 4,332 points of of data.
4. Idle happens most from 1.20-1.26v
5. Voltage from 1.30 - 1.76 (0.46 range) logged only 538 points. Comparatively voltage from 1.20 - 1.28 (.08 range) logged 1,213 points.

Knowing this I decided to rearrange the 48 MAFv table columns around the most used areas for CL and skipping the underutilized area. The extra columns where then filled in in the high end adding more solid points of data for OL.

Here's my breakout with tested g/s <1% AFR error for 1.08v - 4.32v. My setups not big enough to pull more air than that.