Tom-05-MR

iTrader: (1)

Hey all,

I have a few light bolt-ons, including a modified Buschur MAF pipe and Buschur cone air filter. I recently installed an Innovate LC-1 wideband and I've been working on tuning my AFRs. I've noticed that my actual AFRs are pretty far off from my ECU fuel map AFRs. At high loads, my mid-range is much leaner than the ECU says, while my top end is pretty close to the ECU values.

I think I need to scale my MAF, so I've done some reading about it. I think I understand it, but I wanted to run it by you folks step-by-step before I mess around with anything.

First off, I flashed my ECU to run in open loop all the time so that I can get AFRs everywhere without the ECU trying to maintain 14.7:1 at low loads. Of course I'll turn it back when I get my MAF scaled. I am also set up to log 2-byte load, RPM, and air flow.

I've been logging some fairly calm day-to-day driving to get readings at low-to-moderate air flows. I'll get around to high airflow readings soon.

Here is a graph comparing how far off my actual AFRs are from my ECU's fuel maps:



The numbers are percentages. Negative numbers indicate that my actual AFRs are richer than the map expects, while positive numbers are leaner. The cells are color-coded by actual AFR; cells closer to red are leaner while cells closer to blue are richer. I only used cells with a decent number of data points.

Then, I made another graph of airflow vs. load and RPM:



I just use the raw values for air flow. They're nice whole numbers, and honestly hertz are not any easier to understand. ECUFlash displays the raw values by default, so if you use them you don't have to do any XML tweaking.

Then I made a spreadsheet of all the airflow-%difference pairs, in ascending order by airflow. I also copied the ECU's MAF scale into the spreadsheet. It looks like this, except this pic truncates my readings because there are a ton of them:



So what now? Well, I looked at the ECU air flow values, and my air flow values, and I used Excel's linear interpolation function to figure out what "Result" the ECU comes up with. For example, my air flow reading of 311 is between the ECU data points of 256 and 512, so the result is somewhere between the ECU's results of 152 and 165. Excel's TREND function tells me that the "Result" of a reading of 311 is about 155. Then I applied my AFR differences to the interpolated results to come up with a corrected result. Here's another look at the spreadsheet:



Now I've got some kind of idea of what my MAF scaling should be. Here's a graph of the ECU MAF scale and my calculated MAF scale:



One thing that's very clear is that I need more data. A pretty decent curve is emerging at low air flows, but at higher values it's quite erratic. I can solve that easily enough with some wide-open throttle logs. I also need to determine exactly how I'm going to average my corrected scale and map it to values in the ECU.

None of that's a problem. Here's the real issue: does what I'm doing make sense? I largely hammered out the specifics of my procedure on my own, and I'm worried that I've gone completely off the wrong track. Do I correctly understand what the MAF scale is, what it does, and how to adjust it? Is this a reasonable method for adjusting MAF scale, with a bit more confidence than just logging fuel trims?

I appreciate any input! Thanks for taking the time to read this.

Mr. Evo IX

iTrader: (5)

deleted.

Tom-05-MR

iTrader: (1)

MAF Hz is airflow, converted to a different unit. Charting MAF Hz vs. airflow would be like charting boost in psi vs. boost in bars.

Mr. Evo IX

iTrader: (5)

yea, your right. Sorry. your charting maf vs rpm vs o2. I'm following now. Yes that looks like a good plan. How different are your results from using the fuel trim method?

Do you think the ECU has enough scaling functionality to scale the injectors so the targets are right on at high load and low load? The stock scaling is not accurate at high load on the stock maps. But is that even necessary since it uses the trims mostly to achieve stoich in closed loop. Shoudn't the scaling be biased to that load (idle / cruise).

Tom-05-MR

iTrader: (1)

Sorry for not following up on this earlier. Since nobody jumped in and said "Yes, you're doing it right" or "No, that's unbelievably wrong," I figured I should take some more time to think and gather data. I've been really busy lately, so I didn't get all the wide-open throttle runs I need to smooth out my graph, but I did a few and I'm posting an updated version below.

I think MAF scaling with a wideband will let you correct the entire scale, while fuel trims can only help you at the lowest levels. You mentioned that the stock scaling's not accurate at high loads. Adjusting the top end of the existing scale should fix that. You can even take some low-flow data points and move them, like how people take low load columns in their fuel and ignition maps and use them to go beyond 260. It's true that the scale is more dense at idle/cruise loads, but I think it's a bit too dense. It also has no MAF scaling data for readings beyond 16384, and you can definitely see readings beyond that. You could remove one or more of those low data points and scale the MAF to beyond 16384, or get additional resolution up high.

To explain the hypothetical advantages over the fuel trim method, I'll start off with what I understand about fuel trims. If this is wrong, please, someone tell me so I can get straightened out.

Let's say you have an Evo whose fuel trims have been reset somehow. I don't actually know how to do that, but it's a moot point. Our hypothetical Evo has zeroed-out trims.

When you're cruising around in closed loop, the ECU looks at the scaled MAF reading. It decides how much fuel that air needs to hit 14.7:1, and injects it. Then it checks the front narrowband O2 sensor to see if it was actually 14.7:1. If it's lean, it knows that it should actually inject a little more fuel than the MAF reading would indicate. If it's rich, it should inject less. This is your O2 feedback trim.

If the O2 feedback trim indicates the MAF reads consistently low or high, the ECU stores that info in the fuel trims, so that in the future it can hit 14.7:1 right off the bat.

So what's the problem with MAF scaling based on fuel trims? They're calculated based only on closed loop, because that's the only time the narrowband can provide useful feedback. Therefore, they're only based on low MAF readings. However, your MAF scale runs all the way to wide-open-throttle territory. If your fuel trims say your MAF overestimates by 10%, and you apply a 10% correction across the whole MAF scale, you're assuming that your MAF overestimates by 10% all the time.

What if your MAF overestimates air flow at low levels, but underestimates at high levels that you only see in open loop? An unrealistic scenario? I don't think so:



If I force my car into open loop (so the ECU doesn't make corrections based on O2 feedback) I run richer than my fuel maps at low air flows. That means my MAF is overestimating the amount of air, so the ECU is injecting too much fuel.

However, I run much leaner than my fuel maps at higher air flows. In order to get a nice 11:5:1 in the 3000 RPM 240 load cell, I've got to tell the ECU to shoot for 7.8:1. This means the MAF is severely underestimating the amount of air flowing through it.

Why isn't the difference consistent? My theory is that my less-restrictive filter and MAF pipe result in lower air velocities than stock when air flow is low, but as air flow increases, the intake flows much better than stock.

Charlie_B

iTrader: (1)

Id like to bump this for some more input on the subject. Is this the "correct" way to scale the MAF? I recently installed an ETS intake, and my LTFTs are averaging zero after adjusting my maf per l2r99gst's how to. But now I am getting into tuning the open loop areas of my map, and was wondering if having the maf scaled appropriately in those areas would be of any benefit.

l2r99gst

iTrader: (2)

The benefit of calibrating your MAF beyond closed loop conditions would be that you have more accurate load, and thus mass airflow readings. In a perfect world, you should correct your MAF over the entire range. However, most people don't have a base to compare to before their mods. If you do have base data (before your mod(s) that messed up your MAF readings), then you can use that to adjust your MAF over the entire range. You can't simply use a wideband, since the open loop fuel tables aren't true AFR numbers.

In reality, most people simply use the open loop fuel tables to compensate for open loop fueling. It's a bit quicker and easier. The only downside is that your load numbers may not be comparable or 'accurate' compared to other cars with properly calibrated MAFs. If you like to log mass airflow to look at things like compressor maps, etc, then your numbers may be off a bit.


Eric

Jumperalex

Well ... from a purely engineering POV, it is never a waste of time to calibrate your sensors. Much better to have a MAF properly calibrated and then play with your fuel tables, than kludge your fuel tables to account for innaccurate sensor input.

As for your method ... my opinion doesn't mean too much in this context since I have no reputation, but your logic seems reasonable and [insert self licking ice cream cone] mimics what I've been thinking about doing. The only difference is that I was planning to still use the closed loop trims to quickly get that part of the MAF Scaling table under control and then using a wide-band (once I get one) to handle the part of the table that operates in open-loop. I'm just not in the mood to deal with running open loop during cruise and replicate the job of a rather decent feed-back trim system. But I'm lazy like that

Oh and finally, I don't think anyone can tell you if this is "the" correct way as I don't think there is such a thing other than what works, and there does not yet seem to be a true consensus on the topic yet. At least not in so much as completely recalibrating the curve. The most talk I see is just about tweakin idle and cruise and even that tends to involve rather gross adjustments over a band of Hz.

Opps one more ... I think you're on the right track as to why the difference isn't consistent. Modelling and predicting air flow is anything but simple. Consider that not only are we talking about flow at various, and widely ranging, velocities, but we're also talking about the resonances involved in a convoluted system (think about all the bends and the BOV dump) that is feeding an engine which does not drink in air at a constant rate. To think that volume/hz would be constant would just be silly, and to think that a change to the system would cause a single offset change would be even sillier.

Charlie_B

iTrader: (1)

gotcha. Too bad I bought the car with an open element filter and screwy tune, huh? How far off would you expect the airflow readings to be in open loop? My LTFTs never went above/below 10% before making adjustments to the maf scaling.

I have noticed that my WB AFRs are about .2-.3 richer than before the intake, but attributed it possibly hitting higher loads (or higher reported loads due to the scaling being off) and am hitting richer cells in my fuel map.

burgers22

Missed this post first time round, nice work. I think your theory holds up. There maybe be other things influencing the way the KV element measures air flow. It uses the eddies in the air stream to calculate the air flow, a filter of different design could influence these eddies and change the apparent airflow. Different air flow speeds could easily show different variations in the measured eddies, so leading to the MAF scale being of by different amounts at various speed.

MB

silver_evo

Im really interested in this topic. Actually having the Fuel table be the actual afr would be an amazing thing. Ill start logging and playing with excel also.