EFI Tuning - How to Process Description

[jpnovak]

Recently there have been a large number of questions about installs, build threads with EFI swaps and yet, there was not a good collection of information in one spot. This post is to conceptualize the setup and tuning process of a stand-alone system. I told myself a long time ago that I would write a dialog to help answer most of the questions I get on a regular basis. This is what I was able to put down on “paper”.

This is a compilation of emails, texts, and posts I have sent to many of you on the BBS. So, It may seem familiar to some. If you are new to EFI or if you are considering EFI swap for whatever reasons this should help you.
I will not cover all the basics because there are other resources available. The books by Greg Banish as well as all ECU tuning manuals are a great place to start your Journey to EFI. If you prefer the video method then I highly suggest that you pay for and subscribe to HPA tuners. Their library is unmatched.

I will also say that you should not try to DIY an EFI system without some concept of electronics; including wiring, circuits, computers, and also concepts of how an engine works related to fuel delivery and ignition timing. None of what is in this post will make sense without some basic background.
The examples I give will be based on a Megasquirt platform but the process is the same. I have used the same techniques on MS, AEM, Motec, BigStuff, ECU Masters and others. Some of the parameter names are the same and some are different with the same function. This is not meant to be an exhaustive tuning guide. But, this will get you to a knowledge level that will not only get the car running but allow you to self-tune and have the car drive very well.

As you consider an EFI system, make sure you choose an ECU that your tuner understands. Maybe that is a shop - Every shop has a favorite. Maybe that is you. In this case, be willing to spend hours over weeks and months learning the hardware and software functions to take advantage of EFI; there is no shortcut to DIY, you have to commit to putting in the time.
Disclaimer: I have spent more than 15 years doing EFI systems. I researched for a few years before I pulled the trigger. You can do serious damage to an engine if you get things wrong. In a high-performance build you can do serious damage when you are on the edge of doing things wrong. So. Don’t F-Up your engine.

However, if you get it right EFI can do things that CIS, Motronic, Carbs and yes, even MFI can’t do. I will not be doing a comparison of those systems here. There are 20+ years of archives you can search in Pelican to find comparisons and reviews. There are also a ton of online resources that can be found with a simple web-based search.

Before you get started with tuning your EFI system it’s absolutely imperative that all of the mechanicals are dialed in and operating as they should. The most important thing you can do to set yourself up for success is balancing your ITBs. If your throttle bodies are not balanced, no amount of tuning will make your car run well and you will be chasing your tail. This guide will not go into how to balance your ITBs or set up the linkages as each system varies. Do keep in mind that balancing the ITB’s is not done at idle but also at 3000 and up. You want your car to drive well, not just to idle well.

The other key thing to remember is that tuning EFI is so much about making a single change, looking at the data and then refining. The beauty of EFI is that you have copious amounts of data at your fingertips. You don’t need to listen to or smell the engine to know if it’s rich or lean. Tuning EFI systems takes time and getting a really good tune without a dyno or professional is absolutely possible if you approach each step methodically and driven by the data.

These will be links to individual posts for quick navigation.

Terms and Definitions
Load Models
Sensors
AFR Targets
Ignition
Road Tuning
Idle Tuning
Cruise Tuning
Acceleration Enrichment
Fuel Scaling
Maximizing Resolution
Datalogs and Analysis

[jpnovak]

Terms and Definitions

First let me define some terms that I will use. I can update this list if there is any that I missed. I will try to put a little definition to each term as well.

EFI - Electronic Fuel Injection
ECU - Engine Control Unit
CIS - Bosch Continuous Injection System
MFI - Bosch Mechanical Fuel Injection
ITBs - Individual Throttle Body or Independent Throttle Body
Common Plenum - Single throttle body manifold with runner arms to each cylinder.
CLT - Coolant Temperature Sensor
IAT - Intake Air Temperature
RPM - Do I need to?
MAP - Manifold Absolute Pressure
BTDC - timing Before Top Dead Center - You Probably know this.
PW - Pulsewidth - refers to the injector open time in milliseconds
TPS - Throttle Position Sensor
VE - Volumetric Efficiency
MBT - Mean Brake Timing
MAF - Mass Air Flow
O2 - Oxygen Sensor
WBO2 - Wide Band Oxygen Sensor
AFR - Air Fuel Ratio
AE - Acceleration Enrichment
VEAL - VE Analyze LIve - A Megasquirt Tuner Studio “auto tuning” algorithm
TS - Tuner Studio - Megasquirt Tuning Software
MLV - MegaLog Viewer - Megasquirt Datalog Analysis and Graphical Viewer
MS - Megasquirt - An EFI ECU
AEM - automotive Performance Parts supplier that makes an EFI ECU
Cell or Bin - specific spot on a tuning table. Defines load and rpm condition for tuning.

[jpnovak]

EFI Setup

Let’s start our discussion related to EFI setup. Step 1. Read the manual for the system you want to choose. Step 2. Read the manual on a comparison system. Step 3. Did you really read the manual? Look at software functions. Look at hardware capabilities. Look at wiring requirements. They are all similar in the fact that wires connect the ECU to sensors, collect data, calculate fuel and timing and then operate the injectors and ignition. But there are many differences. Remember I said that shops have a favorite system. Listen to them. They have knowledge and experience to make it right the first time.

Hardware: Modern ECUs (Engine Control Unit) are all fully capable of nearly going to the moon. Really this is true. A modern ECU has a microprocessor that can control more and calculate more than spacecraft from the Space Age. This also means it can be easy to make some uninformed choices about the engine operation. I am a firm believer in the KISS principle. Keep It Simple Stupid. Any EFI ECU can easily replace a CIS mechanical lever arm or Motronic Barn door. They have faster response times and features that are nearly endless. Now, Do you need all those features? My opinion is most likely not. And so, Do you need to pay extra for a system that has features you will not implement? Only you can answer that question.
Most ECU brands have a tiered ECU system. From Mild to Wild. Plan your system carefully. You may not need or want to pay for capability that will never be used. For example, an Old 911 does not need a full CAN configuration to run VVT, Transmission, etc. An old 911 really only needs minimum 4 sensors to run quite well.

Now, Let’s assume you have chosen an ECU and you have gone through the process to wire it up and have it in the car with a laptop connected. Where do you start next? Ideally, you have obtained a base map that will help get the car started. If not, ask around and someone is likely to reach out and help. This is where our journey begins.

Once you have the car running you have to start the tuning process. How often have you heard or read… “you just have to go tune it”? pretty often. What does that actually mean? There are a few basic concepts that we have to cover first. That is entirely the question this post is intended to answer.

Fuel delivery: ECUs call fuel delivery in different names. Ultimately, this is a calculation of the amount of time an injector is open to squirt fuel. All ECUs have a primary scaling factor that is set by the size of the injector. In Megasquirt this is called required fuel (req_fuel). In AEM and Motec you input the injector size. EMU calls it the Injector Scaling Factor. Think of this value as a global multiplier. You can adjust the entire fuel table up or down by changing this value.

Volumetric Efficiency: There is a secondary scaling factor that is what you adjust based on engine operating parameters. Let’s call this a VE value. VE is the volumetric efficiency and it represents how much fuel an engine can process given the amount of air it intakes at a given load and rpm. I will come back to the tuning of fuel delivery/VE later. Let’s talk about engine load.

Engine Load and Tuning Models
There are two primary methods to calculate engine load, and a third that I’ll only touch upon. I will be using concepts to describe the table setup that allow you to maximize both resolution and dynamic range. You want the Load axis to match how the engine operates. All ECUs need to determine the engine Load - or how hard is the engine working, as a function of RPM. All of the primary tuning tables will typically have Load as the Y-axis and RPM as the X-axis. They do differ in the axis being ascending or descending. Most of them are also adjustable in table view so that you can make it easy to understand. I typically like a cartesian type axis where a minimum value is in the lower left hand corner and the axes are ascending in value - just like I learned in elementary school.

Let’s define what Load actually means.

Speed Density: Speed density uses a MAP (Manifold Absolute Pressure) Sensor to determine how much vacuum an engine draws. This engine pressure defines the engine load. The MAP sensor can also work the other direction and read how much pressure is being made under boost. That’s for you (us?) turbo guys. Think of it this way. The engine is pumping air through. When the throttle is closed this pumping action acts like a vacuum and the pressure drops. When the throttle is wide open the pressure should read atmospheric. Partial throttle reads in between. I am not going to cover MAF (Mass Air Flow) in this post because this is a much less common method to implement.

Scaling for MAP based tunes is important. In this case we will quickly talk about dynamic range. Most ECU initial setups will want to run a Load axis on a normally aspirated (NA) car from zero (closed throttle) 0 to 100 kPa, where 100 is near atmospheric pressure in kiloPascals (kPa) (full throttle). There are two problems with this. First, no engine will ever draw vacuum down to zero. The minimum vacuum will occur right near the peak torque rpm when the throttle is snapped closed. This is when the engine is operating at its peak efficiency. This minimum MAP reading will vary mostly based on the cam overlap. Most stock cam engines will see 15-20kPa min values. Many performance cams will only show 30-40kPa minimum MAP values due to lobe overlap and the resulting pressure loss. You must have the engine running and driving and then use a datalog to find the minimum. This will be the lowest MAP reading on your fuel table. And it may change as you continue to tune the engine. Second, maximum Atmospheric pressure varies on weather and altitude and cylinder filling efficiency. At altitude, your car will not draw 100 kPa, It may only be 90. So, if your table goes to 100 you are leaving dynamic range on the table. The MAP scaling should be set from engine vacuum minimum value to the maximum atmospheric +5%. Why the extra 5%? This is to account for changing conditions or high performance engines. An example, I have a friend that has a blue car and his car will pull above atmospheric across peak torque. Yes, its efficient. Yes, its fast. (105% VE )

Below is a datalog trace from a particular engine. This engine has stock cams. On the left side you can see the global min an max values. The load (green) range is 23 to 98 kPa. So, you want all your tables to scale from 20 to 100kPa. Once you know your min and max values for your particular engine config then you change the tables to represent these min/max value. This gives you the maximum dynamic range for your configuration. I will come back to resolution after we discuss other load calculation methods.





Alpha-N: Alpha-N is a method of tuning that is based on the position of the TPS (Throttle Position Sensor). The sensor is usually calibrated from 0% when throttle closed to 100% throttle open. The big catch here is that the ECU needs to be specifically calibrated to know what is open and closed. All ECU software has a method to calibrate. Now, the other part of this is that you need to make sure your linkage is adjusted properly so that you get physical WOT, not just a number on a screen. This is no different than setting up a 3.2 Motronic car where the WOT switch is not activated due to poor linkage adjustment. Yes, it's noticeable when setup and calibration are wrong. ( the electric sensor has to match the mechanical butterfly or throttle plate perfectly)

Dynamic range on Alpha-N is simple. 0% is throttle closed. 100% is throttle open. Nothing more, nothing less. But… Resolution scaling is everything in alpha-N. The percent change in position can have really big differences in both MAP and total airflow depending on how big a throttle is relative to port size. Also, ITBs are way different than common plenum, single throttle body setups. That is a topic for another time. Just like with speed-density I will come back to resolution.

ITB_Mode: Some ECUs can run a blend of Speed-Density and Alpha-N. This is often targeted toward ITBs where the car has limited dynamic range of MAP signal due to high cam overlap. I am not going to cover that in this thread. I have set these up before and can work well. But I only reserve this for builds where you really need it because, IMO, this creates secondary compromises. Often, the load axis can be scaled so that ITB mode is not needed. In addition ITB Mode injects complexity where it may not be needed. To make the choice.. If your car idles with a MAP reading above 75kPa then you are a good candidate for a blended load calculation.

[jpnovak]

Engine Sensors

I am not going into detail about individual sensors. Each ECU has recommended sensors This topic heading is less about actually collecting data and more about what sensors can do for you including Protection algorithms.

AFR: The wide-band oxygen sensor (WBO2) gives you feedback on air fuel ratios. Can’t have too much or too little fuel. Most ECUs can set up limits to shut down an engine if the AFR goes haywire. These ECUs also have closed-loop auto-correction that can monitor and adjust on the fly while in operation. I think that if you tune an engine well then you can run in an open-loop mode. (closed loop the info is returned to the ecu and readjusts… In open loop the loop is broken or disconnected so no info readjustments)

Temperature sensors: The coolant sensor in an air cooled engine could be cylinder head, oil temp, case temp or several other temperature readings. This is used as part of a warm-up process to add fuel when the engine is cold. This can also be used to enrich or change timing if the engine gets too hot.

The air temperature sensor (commonly called IAT (Intake Air Temperature)) is used to make air density corrections. There are different ways to mount the IAT. This is important. I am a fan of mounting the IAT in the intake stream. This means inside the plenum or airbox so that you can measure actual air temp going into the engine. Yes, air heats as it goes into the engine. And yes, it really heats if you are running a turbo under boost. Yes, you have to compensate. So, Why am I explaining this? All ECUs have an air temperature correction curve. This is to add fuel on cold days and take away fuel on hot days. You can also change timing curves when the engine gets hot. If you are running a hotrod engine you will need to pull timing if IATs get too high. This is one protection to help avoid detonation. Your tuner will help with this setting or your ring lands will suffer greatly. Note, default curves in most software base files are not correct for an aircooled engine. (excellent )

Knock sensors are another way to push performance with protection. They can be used during tuning to setup ideal timing curves and can be used to reduce timing if knock is detected. All it takes is one bad tank of gas. Now, available ECUs can use either an external knock sensor amplifier or an internal amplifier. This takes the small voltage from a piezo electric knock module and scales it into a range the ECU can detect. Our cars can have a relatively high valve train noise that can trigger false knock signals. The knock sensor frequency is based on cylinder diameter and the filters must be setup correctly. Read manuals and act accordingly for your system. . ( consult an expert!!!!!)

Knock Sensors: Knock sensors can be installed in an air cooled motor. These can utilized the mounting stud by Cylinder 3 for a throttle plate, the MFI pump boss, or a 964/993 style cylinder head bridge. The important thing to remember is that you set the frequency of detection to match your engine build. The frequency is related to the cylinder diameter. The typical formula is:

Knock frequency (kHz) = 1800 / (3.14 x Piston Dia (mm))

Make sure when setting up the ECU you get the sensor parameters correct. Knock sensor protection will usually retard timing when detected. Knock sensors can be useful when adjusting the ignition map.

Knock sensors can be electronic or analog. The electronic sensors use a piezo electric detector and must be run through an amplifier. Some ECUs have the amplifier built in but many do not. You can also use a microphone or even a “det can”. All can work well and vary in cost and complexity to install.


The summary here is to make sure that sensor correction curves work for you. Not against you.

[jpnovak]

Target AFR Tables

Tuning the fuel tables relies heavily on AFR (WBO2 Sensor) values to determine how much fuel to deliver. All ECUs will reference a target AFR table. Some ECUs running a VE model will use AFR as part of a VE calculation to deliver fuel. AEM for example will allow you to, once the engine is tuned, adjust the AFR table and it will adjust the VE fuel delivery to match. However, most ECUs will adjust the fuel delivery table to match the target AFR. This comparison and adjustment feedback is part of the closed loop running. In my example below for Megasquirt, the AFR table is used as a manual feedback mechanism to setup the fuel delivery table. ( relationship of effective change with afr to ve , when to choose one or the other )

Normally, when I start from scratch on an engine I will simply set the AFR Target table to 13.5 everywhere. For the DIY tuner, This is a very safe AFR and the engine will be responsive and make good power. This allows you to get the engine running and the car driving to move onto the next phase. Once you have a good running engine you can go back and optimize. Now, experienced tuners will skip this test phase and go straight to an optimized AFR Target Table. Here are some key points that you need to consider as you optimize the AFR output. Please be cognisant that these are examples. Every engine is different and will need individual adjustment and tweaking.

The Fuel delivery requirements are broken down into four primary areas; idle, cruise, WOT and decel. This has been covered before. Here are my typical targets. YMMV.

Picture Here not showing up


WOT. I like AFR 13.1-13.3 up to peak torque. Then drop to 12.8 to redline. This should be anything above 92kPa and 80% throttle on most cars.
The fuel delivery curve (VE table) at WOT should look just like a torque output curve on a dyno. There are many engine parameters that have known performance metrics. Look one of these up and you will see bump in the torque curve. The center of this bump is peak torque. Fuel requirements usually drop off after this bump as engine efficiency also drops. ( yes supports my #’s)

Cruise. Highway steady should be as lean as you can run. Most 911s like to be 14.3-14.5 to run smooth. It will depend on timing. The more lean you go, the more timing you need to advance. If everything is dialed you can go to low 15s. You will have to find what conditions you run highway (MAP and RPM range) from gauge or datalog. ( tps and inches of vac)

Transitions. Between the two states (WOT and Cruise) transitions should be non linear gradual. Meaning, drop from 15.2 to 13.8-14 as you are cruising and need to go up a hill or change lanes. Drop down to mid 13.3’s when you really need to accelerate but you are not WOT.

Off-throttle Decel. I am usually at AFR16 for off-throttle. These are usually the lowest 2 rows in VE table. If you are running a full fuel cut, the AFR can be 18-20. This can be challenging to catch idle depending on what RPM you bring fuel back into the tune. ( no need to go here)

Idle. Most 911s are happy to idle about 13.5-13.8AFR. This is timing dependent. You can go leaner if your injector dead time is dialed. There are times when the injector size is too large and the small pulse widths for idle get really short that it forces a more rich mixture.

Now, here is the most important part. Make sure the Load axis on the AFR table mimics the fuel table. Otherwise, you will not be able to adjust with the same resolution. Meaning, Don’t have an AFR table scale from 0-100 when the fuel scales from 25-105. They need to match or be very similar. You may also notice that your AFR table is 12x12 and your VE table is 16x16 making an apples to apples match impossible. That’s ok, just get the basic scale similar. ( pics and demo)

[jpnovak]

Ignition Maps

Initial timing maps are easy to get set up. Look up a timing curve from the factory manual for your engine build. There are many curves available and they are safe to operate. Program a curve that matches the distributor advance vs rpm across the entire engine load. This will get the engine running and driving. This is just a starting point that will allow the engine to run at full throttle and related to rpm.

Once you have adjusted the fuel trim and have the engine running and driving reasonably well then you can look at options to optimize timing. There are different techniques here. Knock sensors give great feedback. Det-cans (look up detonation cans) connected to a stethoscope or ear muffs can be easily made and work well. Both of these have to be installed while assembling the EFI system and not every install has them. They are not required but just another tool in the toolbox. ( more later)

Here are some Basic rules to follow when adjusting your advance curve. You can use idle advance to tune your idle speed. Chugging or bucking at low speeds means you have too much timing. If the engine runs hot, you do not have enough timing. You can run quite high timing advance under lean load conditions (off throttle and cruise). Advancing the timing during off-throttle will help the car slow down due to more engine braking. If the car feels like its fighting itself you have too much timing. If the engine feel lacking power and lazy there is not enough timing. The leaner the mixture, the slower it burns because the fuel molecules are further apart and therefore the earlier you have to ignite the mixture with advanced timing.

Ideally, this is the one time that using a load Dyno with power absorption readout is super useful. You can manipulate timing on the fly and see power gains/loss. MBT (mean Brake timing) is easy to find in this condition. Timing can make the difference between an engine with a lazy or urgent response.
I will say that timing adjustments may change as you optimize fuel quantity. And, the converse is true. The fuel trim may change (AFR) if the timing is adjusted. Let’s explore that briefly.

With someone else driving the car you can change the timing. Often you can watch the AFR as you advance and retard the timing. Typically, you will find that if you start with the timing conservative (retarded) the AFR will lean out as you advance the timing in a given load cell. The AFR will stabilize lean. This is usually within a degree of MBT optimal timing. Now, you have to do that in the car under controlled conditions for almost each cell in the engine operational space. Fun times.

[jpnovak]

Road Tuning Techniques

Let’s now talk about the fun part - Driving the car. I mean, this is why we all bought them. And completing an EFI install just makes them better. Above I have described most of the background that is required to tune an engine. Time to put this to practice. Now, I will say that the process of tuning an EFI engine is very similar to tuning CIS, MFI, or Carbs. The same principles apply. Try not to overcomplicate this process, nor overwhelm yourself. KISS.

Now, An engine will be lazy when it's too rich. An engine will misfire, surge and be unhappy when its lean. The idea is to get between these two extremes and enjoy a crisp, responsive, power-producing State of Zen. Think of EFI as tuning with a keyboard instead of wrenches and screwdrivers. You are not adjusting the fuel pressure in the WUR for CIS. You are not removing and resizing jets on carbs. You are not adjusting flyweights or rack stops in the MFI pump. Just a keystroke here and there to match a fuel delivery to an AFR value. Live this mindset and tuning life becomes much easier. Also, the power of EFi is to give the engine what it wants - EVERYWHERE in the operating space.

There are two primary methods to tune an EFI system. Live and Datalog. I will cover both. Let’s start with live tuning.

You have the engine started but need to work on the fuel VE table. Often you need to get the engine to idle on its own. Remember our discussion about the fuel global multiplier. (in the section for fuel delivery ) In MS this is req_fuel parameter. I will raise and lower this parameter (or injector scaling factor in other software) first, before making changes to the fuel table. This is an easy way to add fuel everywhere. Even a base table has a curvature that is close to what is needed to get through the first few steps of tuning. Now, this process of adjusting the fuel global multiplier works really well if you have a starting map from a known good running engine.

At this point the engine should idle on its own. The first step is to turn off all acceleration enrichments. You can do this by setting values to zero or setting trigger setpoints to levels high enough they will never trigger. The purpose here is steady state operation in a very controlled manner. We are not concerned with changing throttle at this point.

The tuning process Starts with no-load fuel sites. In this case, you are simply increasing the rpms of the engine with the car in neutral and sitting still. This will give clues as to how much fuel is required to actually drive. The process is straightforward. With the laptop open, tuning software running and the fuel adjustment panel open, you will slowly rev the engine. As the rpms rise above idle you need to watch the AFR. This early in the game it is doubtful that the engine will perform at the target AFR. So, you will need to manually adjust the fuel trim until the AFR is in range. We are targeting 13.5 for this exercise. Slowly raise the rpms while adjusting fuel all the way up to approximately 3500 - 4000 rpm. This will create an arc shape across the VE table. In the example picture below the no load points are marked with yellow.

Ref Pic 3



Now, most VE tables have some indicator of what has changed. MS shows red/blue as you change fuel, They also put a yellow marker dot in the corner of changed cells. AEM changes color, Motec puts a red marker on the VE map. Either way, Look at your changed cell range and then compare to the adjacent cells. There is usually a consistent shape to the changed cells being higher or lower than the adjacent cells. The goal now, is to blend the adjacent cells to match the ones you changed based on actual AFR feedback. Meaning, If your no-load arc is lower than the surrounding and adjacent cells you must lower the rest. If the no-load arc is now higher than the surrounding areas you need to raise them up to blend.
Here is the same fuel table after blending. I raised up the rest of the table to match the no-load sites.

Ref Pic 4



Actually tuning the engine. The following process allows you to to drive in a certain pattern while datalogging. Then analyze the datalog and make adjustments.

The tuning process of the the fuel tables is for steady state conditions. In this case, we are not concerned about transitions and therefore these transition corrections must be turned off. This will allow you to setup a more accurate tune. So, Make sure you turn off AE Correction. To turn off AE, set the TPS or MAP threshold under “time based AE” as high as possible. I think 1000 is the maximum. Make sure you also have EGO correction turned off. Now, this means the car may be temporarily difficult during throttle transitions. This is normal. We have to tune the transitions after the car is running.

Many ECUs have an auto-tune function. It is useful to get the car up and running where you can drive and have the computer adjust the mixture for you. In this example I will use the tuning functions for Megasquirt. Let’s define this process. VEAL stands for VE Analyze live. It is also called VE Analyze live. It is not called AutoTune even though most people refer to it as such based on the functional name. EMU Black also has an auto-Tune function. AEM has a datalog conversion macro that runs in Excel for fuel trim adjustments. Motec has a key command to correct fuel trim at each individual table cell. They all work by taking the actual AFR, comparing it to a target AFR, calculating a percentage offset between the two and using the offset to adjust fuel trim up or down as needed. Of course, there are error tolerance and other correction factors involved so that the PC doesn’t make too many mistakes.

Once the car is warmed up you can select the tab for Tune Analyze Live. Then click the Advanced settings tab on the right side. Use the filters to block out certain parts once the car is running well; select a range of cells and then right click to lock them from any changes teh software may want to make. For example, once the car idles, block out the idle cells (for example, min RPM 1100). Also, start with the “Cell Change Resistance” pull down menu to Easy to start with. Now you are going to drive the car.

Ref Pic 5

[jpnovak]

The drive: You need to find an empty road free of most traffic. Click the start button on VEAL. Get the engine up to third gear or so. Lock in an RPM that matches a fuel column on the screen. For example, 2600. You should see a cursor highlight on the load cell at this RPM value. Use the screen value and then reference to the car tach value. Now, use your left foot to load the engine; this means depressing the brakes while driving. As the brakes slow down the engine, give more throttle. Balance the amount of brake and throttle to hold that rpm. Keep adding brake effort and throttle effort until you get to WOT (Foot on floor). If you can watch the screen in peripheral vision, you will see that the fuel will adjust the rpm column around 2600. If you do this slowly enough you will have enough datapoints to have the AFR match the target AFR. Now, repeat for the next rpm band 3100. Keep working your way up the table. Don’t forget to pause and let brakes cool between columns. ( good)

Below is a datalog trace of this process. The RPM (white line) is pinned at 2600 +/-50 rpm. The load (green line) is the MAP signal and it goes from 70kPa cruise to 97kPa WOT. Notice that in the middle of the transition there is a lean peak marked with a cursor. This is exactly why you tune this way. The start of the curve the AFR is pinned at 13.5. At the end of the load run the AFR is pinned at 13.1. Both are as they should be. However, the middle transition between these two points is very lean and needed to be tuned to bring the AFR into a safe zone.

Ref Pic 6


Ultimately, this process will get to be quick. Meaning, it should take 45 sec or so per column. In the example picture above the load run took about 26 seconds.

Use higher gears (4th) for the lower rpms, use lower gears (2nd) for the higher rpms. Keep going until you have covered most of the fuel table. Yes, you have to hit those 6500 and higher spots (you may need to set your max RPM to 7000 in TunerStudio and on your MSD). They should go quick. Then you can click “stop”. When done, make sure to hit “apply” and ‘Save to Controller”. Now “save as” a new tune file. I always use Save As because it time stamps the file. It is easier to go back compared to “current tune”. I never know if I am current or not. (good)

After this process of using the computer tune for your process you should review the VE table. Often manually smooth the table in between tuning “runs”. This allows me to fill in areas that I have missed or areas the computer does not interpret correctly. What does manually smooth mean? The fuel table should look like a very smooth waterfall. If you have jagged peaks, valleys or other abruptness in the fuel delivery, this is not correct. This is usually the sign of something else wrong. Could be the process, Could be the setup, Could be the operator.

Now we move on to the “pulls” Start VEAL again, Do some WOT pulls. I usually use 3rd gear for these. Go all the way to the rev limit. HMM. Did you set the rev limit correctly? This should be part of the initial configuration. Below is an example of an RPM pull. If you look at the data trace you will see the white lines with linear upsweep to a maximum value (redline). You will also see the following trace on the VE table to the right with a quick vertical upsweep followed by horizontal taling as the engine reaches maximum load and the rpms are climbing. We can also follow along the AFR and see that it goes super rich at the point of the cursor. This means there is more adjustment to do in the fuel delivery table.

Ref Pic 7


Now, change the cell change Resistance to “normal’ and repeat. And then change to “hard” and repeat. At this point the car should be driving very smoothly with no major rich or lean spots.

Ref Pic 8


At this point you are ready for the fine tuning process. This involves the datalog function and data analysis. I find that this method gives a smoother running engine in the low- mid- load ranges.

After some initial tuning runs I will review datalogs and look at how the fuel is scaled. Megasquirt will run a VE table but this is not actual VE calculation. It is a math number to calculate injector pulse width. The challenge, especially with ITBs is they are often so efficient that you may run out of room in adjustment. Remember we said we would revisit resolution in the table? To Quote from Independence Day. “hello Boys, I’m BAACK!”

Identifying problem areas using the datalog can be very useful. With some practice you will start to use these logs and almost become obsessed. For example, the MAP signal in the picture below has a sawtooth pattern. This is a tell-tale pattern for “my ITBs are not balanced”. So, go fix the hardware. And yes, it was fixed. Some of you paying attention may see the same pattern in a previous picture. How astute of you to notice.

Ref Pic 9

[jpnovak]

Tuning Cruise.
One of the great things about EFI is being able to lean out cruise sites so that you can get good gas mileage. My old 3.2SS was pushing 300Hp and would routinely get 26-27mpg on trips.
I like to do final cruise tuning manually. With the tuning software running and most importantly on an empty road, open the 3D VE fuel table map. Check the box “follow mode” at the top. This will put the adjuster cursor at the same place as the engine operating cursor.

As you are driving and without looking at the laptop (eyes up front) you can use the short cut keys (shift up/down arrow, or on a mac “<” for down and “>” for up) to manipulate fuel. Lean the car out under light load, On many cars this is in the 50-75kPa range. Do this from about 2500 - 4K. You will also want to bump the timing up a few deg in this area (if your controlling timing too via the ECU) As you approach AFR15 you will find the limit as the car will surge. Surge is not good and too lean.

Now, watch the MAP reading on the gauge Note the map range while normal cruise driving. On my car this is 55-78 kPa. Next, note the rpm range from 45-80mph. This range becomes your cruise island that you can match the lean cruise bins to the lean cruise target AFR. Blend outward from there. ( auto fill)

Tuning Idle
Use timing to set idle speed. Then set mixture to MAP minimum. When you idle rich the MAP reading will be minimal. Start to lean out the mixture until the MAP reading just starts to rise. this will usually be the most stable idle. I find this is around 13.5-13.8 AFR. Each engine is different.

Finally, the car runs great. It's time to lock down the idle. I usually make an idle “island”. All ECUs will use a linear interpolation between datapoints on the load/rpm space. I space the RPM bands to be about 200 rpm below and above the actual idle rpm. This means the idle rpm is in between 2 columns. Do the same for the idle load value. If you are scaled properly these should be rows 2 and 3 on the load axis. Now, your idle load and rpm true value is in the center of 4 points. Make all of these the same fuel value. This means the small fluctuations in engine parameters will not affect idle and cause it to drift. The result is a rock steady idle speed (that means within 50 RPM of your target). I call this the “idle Island”. You can see my example island highlighted below on the fuel table.

Ref Pic 10

[jpnovak]

Acceleration enrichment

After all this tuning the car should be running quite well. Now it's time to make it bark under throttle changes. Crisp throttle response is why you ditched the CIS in the first place. But, ITBs can be tricky to get the AE set right. You already correctly tuned the steady state conditions with AE off. Now it's time to turn it back on.

Acceleration Enrichment is what drives the throttle response. How well does an engine rev when you heel-toe a downshift? How does the car feel when you floor it to change lanes or make a legal pass? How does the car slow when you come off the throttle into a corner? These changing states are what define part of a 911’s driving character. When I talk about changing states I refer to moving the gas pedal (up or down) and having the engine respond quickly to that pedal movement. Think of a race car with that light switch throttle response. Yeah, we all want that and EFI can give you that if setup correctly. Let’s cover some basics.

Most AE Algorithms can control based on the rate of change of either the manifold pressure (MAP) or the throttle position (TPS). Some ECUs allow you to do a combination of MAP or TPS based AE and even allow you to change the percentage weight one vs the other. BTW, For those math guys this is a first derivative calculation.

What types of AE are there? There are 3 basic models that are used; time-based, Accelerator pump and wall-wetting. I am not going to explain each. I find that they can each be beneficial to different builds and configurations. Do your homework, experiment and find the right one for your application. Oh, And read the manual to determine how they work.

How do we trigger AE? I mentioned that AE is based on gas pedal movement since you are requesting the engine to do something different. This can be measured by throttle position or a change in engine pressure because of movement in throttle position. Think of this as 2 ways of measuring the same thing.

There is a lot of background information that says TPS AE is way better to tune because the gas pedal based change is much faster than the pressure based change when considering MAP. The truth is… It depends. On a common plenum single throttle the TPS AE tuning will be faster. This is due to the increased volume of the plenum and the time lag for the MAP to actually change when the throttle is moved. With Most ITB setups, the throttle is very close to the cylinder head intake tract. In this case they can be near equal. In some systems, like an MFI throttle body, where the throttle is effectively sitting on top of the cylinder head intake tract, the MAP based pressure change can be faster than the TPS. The point is, you will need to determine the best AE protocol for your particular installation.

There are accel and decel parameters to control a throttle blip. So, you have to separate accel (foot going down) from decel (foot coming up). They each do different things in the software. The acceleration part (throttle opening) is usually the more important to get right. Decel will control how responsive the engine is coming off throttle and how you perceive the start of engine braking.

I typically tune a TPS curve and a MAP curve separately and then use the slider (blend %) to try in 25% increments until the car drives smoothly. There are times when MAP will respond faster than TPS signal. Either one can be used. Below is an example of the setup from a Megasquirt AE turning table.
Ref Pic 11


One of the challenges with ITB AE tuning is getting rid of the bog - specifically off-idle. When you are working on AE keep in mind that there is an adder for accel and subtractor for decel. Often you can move your foot fast enough with a throttle blip that the engine will first squirt a small amount and then pull a large amount when throttle closes. These will compete against themselves and cause some poor responding throttle. To fix you have to separate what is throttle opening from what is throttle closing.

So, A throttle blip should be open throttle and hold open for a few seconds. If you are watching the time based AE panel there is a scrolling graph at the bottom. Watch the TPS_dot or MAP_dot traces at the top. A bouncing ball will move up the graph to indicate how fast the rate is changing. Then glance at the AFR trace on the bottom (usually yellow). open throttle and add fuel until the AFR dips just rich. This should eliminate the stumble. Then add more fuel in the graph until it is crisp. Usually, you have to add accell time duration (increase milliseconds) not just the amount of fuel squirted. Test in the car by driving at light load and then floor it and hold it down with different opening speeds. There should be a snap feeling (no stumble or bog) and a smooth transition to increased load in the car.

What I usually do is to drive the car with the AE dialog open. Something like 3rd gear and going 2K rpm. Here you will first tune TPS AE and then MAP AE. The idea is that you will “accel” the car at different rates. You should trigger AE and then see a bouncing ball go up the curve. You have to try adding fuel (ms adder, Y-axis) and also increasing the time of the injection squirt. There is both an accel time and a time taper. You will open the throttle and hold it open. Then watch the AFR trace line (blue) after you have opened the throttle. It might go very slightly lean but should not have a spike. Add some fuel (ms) until you see the initial spike diminish (delay). Then add time until it stays relatively flat. At this point the engine should have a smooth transition to a higher load before it starts to pick up RPM. Once you get it to stay flat you can use faster throttle openings to move up the curve.

The decel fuel cut is done in the car. Hold throttle open driving and then snap closed. In MS the fuel decel amount is a percentage of the actual fuel delivery. In the example picture above this is set to 60. Meaning I cut fuel to 60% was the throttle is closing. This should help you get the throttle response. You will have to spend the time to experiment on your particular engine to get this right. And I will tell you that AE is the hardest to get right.

[jpnovak]

Fuel Scaling

After some initial tuning you will often see one or both of two things. If the fuel global multiplier is too large, or your injectors are too large, the idle VE numbers will be really small. So small that you will see big swings in AFR just by single digit changes to VE values. If the fuel global number is too small, the VE values will max out. These are digital calculations and can only go to 255. For you turbo guys you may see both at the same time depending on how much power you are pushing under boost.

Below is an example datalog that has had a fuel delivery analysis completed. The cells highlighted in RED were too rich and the fuel has been reduced. The cells highlighted in BLUE were too lean and fuel was increased. Notice in the upper right hand of the table the fuel value is 253. In this case, the fuel has maxed out on the table. This must be scaled back to allow further tuning. Yes, this engine is thirsty and wants more go-juice. How to we make that happen?

Ref Pic 12






Remember I commented before that the fuel map should look like your torque curve on a dyno. Even if you have not taken the car to a dyno I am sure everyone here has seen a power readout trace from a dyno. Visualize this in your head. Peak fuel requirements should match peak torque and they should occur roughly at the same rpm range. . In this example case 4800 rpm should be right about the peak engine torque for this particular 3.2 engine. After our tuning process the max VE value is 253. The challenge is that this will max out at 255 (8-bit in digital numbers). And this means you can’t really add more fuel if you need to. This means you have to scale the fuel table.

Open the tune file in TunerStudio. Then open “basic/load settings> Engine and Sequential settings” and then open “fuel settings > VE Table 1”. To scale you have to mathematically increase “req_fuel” value and decrease the entire fuel table. I would start with 5% change.

So, req_fuel 3.5(starting value) x 1.05(5% increase) = 3.675 so we will input a new value of 3.7. Next we go to the VE Table and select all the bins in the table. Then use the multiply function [ (X) on top header] and multiply by 0.95 for a 5% reduction. Now the maximum table value is 240 and there is room to add more fuel up top (remember the max is 255). It is possible that you have to scale the fuel table several times as you reach a max fuel limit while tuning. When you can complete a WOT pull hitting your target AFRs you do not need to further scale the VE table as long as you are on scale.

Now, I always tune an engine so that the maximum VE value is above 220. This is because I get maximum resolution in the idle range. The ECU nor the engine care if this mathematical calculator matches any real VE values. Now, I will say that some ECUs do care. AEM get’s unhappy if you try to manipulate curves like this. But, there are other things I do not like about AEM and that’s why it's not my favorite. I digress and will not discuss further.

Maximizing Resolution.
This is a topic I get asked about often. My ITB car doesn’t make much vacuum at idle and It's like a light switch from low to high throttle. How do you fix that? You have to apply some non-linear scaling to the load axis. This is because ITBs have increased swept throttle area compared to a single plenum. They also have resonance tuning effects due to equal length runners. Ultimately, this gives an efficiency increase at higher load values that can not be matched in a common plenum, single throttle. Wicked good that’s why we choose them

This is what a plot of MAP vs TPS looks like on an ITB engine. Notice that iit is quite curved. If your bin division on the load axis is divided into even steps you will miss all the fine throttle change resolution. This is true for speed density or alpha-n. We can see that this car makes a minimum vacuum reading at 42kPa. The cams in this particular engine have a lot of overlap and reduce overall vacuum at closed throttle. This engine is also very efficient at moving air so it will see atmospheric maximums above approximately 50% throttle. So, we have to setup tables to operate in those ranges.
Ref Pic 13


Speed density will have a lot of MAP value change with very little throttle opening. Usually the MAP will rise very quickly to 85-90kPa with only 1-5% throttle opening. Then you have the last 90% throttle for the final 10kPa. So, you can have something like 20-kPa jumps in low rows on the load table, Then 5 or 10 kpa jumps in the middle, and 1-2kPa jumps as you cross 90kpa. An example of non-linear scaling is below. Basically, you want more rows where your engine spends the most time; don’t waste table rows down at 10kpa when your car will never see them. As an example, the engine above should have a minimum row for MAP based load at 40kPa.

The example below shows the effect of scaling an ITB MAP based load detection. Notice the Y axis has tight increment spacing of the rows above 82 kPa. This particular engine needed more resolution to control the fuel delivery as resonant tuning drives fuel requirements.
Ref Pic 14



Alpha-N is the same but opposite direction. You want 1-2%TPS jumps up to 10 and then increasing step size as you approach 60. 70 to 100% TPS can be one step in resolution.
Ref Pic 15


Spending some time to optimize here will be required for you to fine tune the part throttle driveability, especially with ITBs. Don’t be afraid to make these type of changes. The world is non-linear. Your EFI tables need to be too.

[jpnovak]

Datalog and Analysis

Now, If you have followed the above your car should be running and driving pretty well. I have found that the fine tuning is best done by taking a datalog and doing both manual and algorithm analysis.

Do you have MegalogViewer program? If not, Its time to get it. It usually comes with the paid Tunerstudio bundle. Here is a tutorial to use it. At this point you will be making small changes that are likely finer than VEAL can do. So, A drive with datalog followed by analysis is the way to go. I will say that the MLV program can be used with most other ECU systems. The trick is to know how to change file formats so that the data is present in the correct fields. If you are familiar with data formatting I can answer questions about this.

In TunerStudio (TS) you will start a datalog and then go for a drive. The idea here is to capture lots of data at all 4 corners of the fuel curve. This means, lots of on/off throttle, loaded in different gears and different RPMS. also, Don’t start the datalog until the engine is fully warmed up.

Once you have stopped driving, open your log file into MLV. Then open your tune file using the dialog button on the right center. I will pull down the lower dash section so I can view the VE table and AFR target table along with the run trace. It should look like this.
Ref Pic 16




Once loaded you can scroll through the log file and look for certain areas. You can right click in middle of graph and navigate quickly to certain conditions. The space bar can be pressed while driving to put a time stamp marker in the file. mDatalogs can also be taken using a mobile phone and blue-tooth connection. The MSDroid program works really well compared to ShadowDash. I will not go into details about this now. We can cover later if there is enough interest.

Next, click the VE analyze button. This will let you quickly see where you are lean and rich relative to the AFR targets. Use the large VE analyze button to run the analysis. When finished the VE table will be color coded; Red means it is rich and blue means it is lean. This quickly allows you to spot trouble areas. This is the analysis of your latest datalog. You can see the rich spot right below 4krpm at 87kPa and then going lean with anything needing more throttle. This also shows that you need more fuel up top. And you also need to start pulling to RPM cutout to get that 6500 rpm column.

If you run the mouse and hold over a cell it will tell you what the previous value is and how rich/lean it was. The mid-range is making small changes. For example: 2400rpm@75kPa is only a 1 VE value change. But the top end is really lean. The 4400rpm@100kPa went from 190 to 242. That’s a big change.

Make sure you click “accept new table” and then “save tune as” to commit to new file for loading.
Ref Pic 17

[jpnovak]

So, Manually adjust or use the smoothing function on the VE table editor. I manually toggle. Shortcuts are arrow keys to toggle across the table. Then shift+ arrow key (up/down) (or on a mac “<” and “>”) to add or remove fuel in 1 increment or ctrl+shift+arrow to change fuel in 5 increments. The adjusted cells will have a yellow tick mark on them. Here is my manually smoothed map. Notice I do not adjust a single point in the map. The fuel calculation will always use the 4 closest points and weigh them. So if there is a local high (or low) spot you want to average over that area. For example, a single low spot would be increased and then the neighboring spots would be decreased. I am sure you get the idea. In the picture below the manually smoothed bins are marked with a yellow dot. If you compare the two graphs you can see how transitions are smoothed, holes filled and peaks reduced. Ultimately you want the car fuel table to look like a very smooth waterfall.





Now you can save and load this in the car. Once you go through this a couple times it will take you just 2-3 minutes for the entire process. So, drive and datalog. Focus on one section at a time. Stop in a parking lot. Analyze. Reload. Repeat.

Next, if you want to lean or enrich an area. Change the AFR target value and run the analysis. It will automatically update in that area and you can upload a new tune to the car quickly.

[jpnovak]

I hope you all find this helpful. Remember that this is a process and one that will take time, require patience and sometimes walking away for a minute. But what you will gain is a car that runs and performs to its maximum ability and a greater understanding of exactly what your car wants and needs. This will make you a smarter Porsche owner and enable you to really understand your car.


Please feel free to add your experiences in tuning. You are also welcome to request information about any areas I may have missed. I do consult on tuning and EFI installs. PM or email directly if you want to start a conversation.

yes, I will be working to fix the missing pictures. They are uploaded with links but not showing up in the system.

[blucille]

Originally Posted by jpnovak

Reserved 13

I've got a lot of reservations, but I am thrilled to see this coming together, and Jamie, Thanks for all that you are doing here and everything you do for the community, sorry to say I will not see you this year at HCR, but hope to see you soon

[Porkshop]

This will come in handy for me to decide on fueling system when I get there on my project. Once I start it....Subscribed...

[icarp]

Jamie, thank you for this fantastic effort . This compilation of data is a true resource for us all.
Have fun at the HCR this year as I too, will not be able to make it .

Ian

[FA-18C]

Subscribed - Jamie, me thinks that this has the roots of a phenomenal tech book...

[al lkosmal]

Jamie, Publish the book.
I have compared notes with Jamie for quite some time now......and he is who I go to when I need tuning/debug help.

regards,
al

[Showdown]

Following Jamie's guidance was the way I, an idiot with no automotive experience whatsoever, was able to complete anEFI/ITB conversion and tune it myself.

Mods, make this a damn sticky.