Over the past decade, Holley EFI has become a cornerstone of modern performance. Whether it’s the ease of use and enhanced drivability provided by carburetor replacement systems like Sniper and Sniper 2, the sophisticated capabilities that Terminator X and Terminator X Max deliver, or race-ready systems like HP and Dominator, there’s an EFI system for just about every build scenario out there. One core feature ties all of these products together: Holley EFI’s self-learning capability.

“There are effectively two core elements of the self-learning technology – you have what we call ‘closed loop,’ and then we have ‘learn,’” explains Holley EFI Product Manager Harrison J. Alford. “In OEM parlance, these are essentially short-term and long-term fuel trims.”

The Two Sides Of Self-Learning

Closed loop handles real-time corrections based on air/fuel ratio readings provided by oxygen sensor feedback and other active data inputs. The learn function, meanwhile, builds a continuing, ever-changing long-term correction, gradually refining the tune as it collects more operating data.

“Over the course of thousands of miles of use, the system is inevitably going to see a wider range of driving scenarios,” Alford explains. “As parts wear, there are changes that become constant. So that learn aspect, or long-term fuel trim, will notice repeated actions and their results. For example, if it consistently sees that every time the engine is operating in a certain part of the fuel map, the closed loop function was adding 5% more fuel, over time the Learn function will recognize that and gradually adjust the tune itself to address that, which essentially eliminates the need for the closed loop function to make that real-time adjustment in the future. Once a tune is dialed in, it shouldn’t really need to make dramatic changes in real time.”

This self-learning approach provides some big advantages over traditional fuel tuning, which requires manual adjustments to the base fuel map.

“I think the biggest benefit is that it can make the tuning process much, much quicker,” Alford notes. “If you were to manually tune the entirety of the fuel map using what we call ‘open loop’ tuning, you’d have to try get the engine into all of these different operating scenarios manually, either on a dyno or out on the street, to populate all of the different parts of the map. That would take significantly more time to modify cells individually, test the changes, make additional tweaks, and so on. And the end result probably isn’t going to be as comprehensively dialed in. With the closed loop system, we still recommend getting the vehicle into as many different areas of engine operation as possible – drive it uphill, downhill, in traffic, on the highway, etc. The difference is that the self-learning part of the system will automatically make corrections to maintain the target the air/fuel ratios that are programmed, rather than requiring you to do all of that work manually.”

This effectively eliminates a major pain point in EFI tuning: Troubleshooting potential issues with a tune.

“If certain parts of the tune are imperfect and the closed loop feature is active in that area, it will often catch those problems and automatically correct them so there are no hiccups. In fact, you might never know that there was an issue to begin with,” he says.

“And another big advantage with this technology is on cold starts. Self-learning will ‘fill in’ gaps, even if the base map and/or the cold start calibration isn’t already tuned to address those. It will modulate the fuel to try to hit the correct air/fuel ratio, regardless. And we can run closed loop at any engine temperature – the system won’t switch over to the ‘learn’ aspect until the engine temperature reaches a predefined parameter. By default, that’s around 160 degrees Fahrenheit. The thought process behind that is that the ‘learn’ element is like re-jetting a carburetor, and you wouldn’t want to re-jet a carburetor based on how the engine ran cold.”

The Role Of Sensors

In order to make these adjustments to the short-term and long-term fuel trims, the EFI system relies on data measured from several different sensors on the vehicle. As such, it’s critical that these sensors continue to provide accurate data to the system.

Wideband Oxygen Sensor

“The main one that’s used for closed loop and learn is the wideband oxygen sensor,” says Alford. “This sensor monitors the post-combustion gases in the exhaust stream and reads the leftover latent oxygen content is in those gases.” That information is relayed back to the ECU, which in turn makes adjustments to the injector pulse to achieve the target air/fuel ratio based on the data from the oxygen sensor.

“So the sensor might be telling the ECU that at this certain point in time, the engine was running rich or lean. The ECU takes that information, compares it to target air/fuel ratio, and makes corrections to the fueling according to that information in order to get everything as close as possible to that target. And all of this is happening almost instantaneously.”

Shop wideband sensors here.

In order for a wideband oxygen sensor (AEM FAE sensor pictured) to provide the most accurate data possible, the exhaust system should be free of any leaks, excess fuel, or oil content. The sensor should ideally be installed at a downward angle, preferably 45 degrees, in the exhaust pipe about six to eight inches after the merge in a header or manifold collector for a non-turbocharged application. With turbo applications, Alford recommends placing it 10 to 12 inches after the turbine outlet for best longevity.

While the wideband oxygen sensor is a fundamental component of an EFI system’s self-learning functionality, there are other sensors that play important roles in its operation as well.

MAP and TPS

“Most of our systems are going to run what’s called a ‘speed density tune,’” Alford says. “This uses the Manifold Absolute Pressure (MAP) sensor Holley EFI 5-Bar MAP sensor pictured) as one of the main methods to determine engine load. That sensor is critical for the base fuel map and some other ‘modifier’ tables that can come into play. The Throttle Position Sensor (TPS) is another key one as well because it’s telling the ECU that we’ve made a change in throttle angle, or we’re going to be giving the engine more air, so adjust accordingly. In many cases, the throttle position sensor in a speed density tune is used for transitional fueling, essentially playing the role that an accelerator pump would on a carburetor when rapid or more aggressive throttle changes are made.”

Shop MAP sensors here.

Shop Throttle Position Sensors here.

CTS and IAT

Other sensors that contribute to a self-learning EFI system’s operation are the coolant temperature sensor (CTS) (Holley EFI Sensor 534-10 pictured), which is mainly used for cold start and warm-up enrichment, and the intake air temperature sensor (IAT), which measures temperatures changes in the air coming into the engine and makes fueling adjustments based on that information. Note, the IAT has no effect on Closed Loop or Learn, its primary use id during a start-up. The CTS is also mostly start-up focused, but it does allow the system to enter Learn at certain operating temperatures.

“Engine speed [RPM] is obviously another critical data point for the ECU as well,” says Alford. “That’s typically delivered to the ECU by a crankshaft position sensor reading a crank trigger, or a trigger reluctor like you’d find on an LS or a Coyote V8. Or, in some cases, you can also get that from a compatible distributor that has a magnetic or Hall effect pickup. Basically, the faster the engine speed is, the faster the closed loop system is allowed to work because there’s a shorter amount of time between cylinder firing events, and a larger mass of exhaust gas is flowing past the oxygen sensor. If there’s a rich or lean condition at a higher engine speed, we want the ECU to be able to react to that very quickly.”

Shop Coolant Temperature Sensors here.

Understanding The Algorithm

Now that we’ve looked at how Holley EFI’s self-learning technology establishes short-term and long-term fuel trims as well as the sensors used to accomplish those tasks, Alford turns our attention to aspects of its operation where some folks get tripped up.

“So you have your base fuel map, and as the system learns while you drive, it’s going to be making corrections and ‘learning’ whatever deviations from the target air/fuel ratio that it finds with the closed loop. And as the system does that, it’s going to be populating a Learn Table based on where the engine has gotten into different load and RPM ranges. You may have some areas where it’s taking fuel out, and in that case the Learn Table will show a negative percentage value in the software – for example, it might say ‘-12.4%.’ That means that in that particular area, or in that particular cell, it has learned to remove 12.4% fuel in order to consistently hit our target air/fuel ratio.

“And on the other side of the coin, you may also see positive numbers in, for instance, the wide-open throttle areas, where it might be just a little bit lean. So it might add, say, 6.8% fuel. And so it notices that it had to consistently add 6.8% fuel to hit the target AFR. Some folks get confused here because they see this percentage and they think it’s essentially just a 0 to 100% amount that’s measuring the system’s learning progress. They’ll call up the Tech Line and say, ‘How long does it take the system to learn? I’ve been driving my vehicle for a while and it seems like it’s stuck at 15% learn.’ What that’s actually telling you is that the system has added 15% fuel to that area of the fuel map. If it ever gets to 100%, that basically means that the system had to double the fuel map to hit the target AFR. You don’t want to see that. The smaller the numbers for closed loop and learn, the more dialed in the tune is.”

When to Adjust Closed Loop and Learn Settings

Alford also points out that if you disable the system’s closed loop function in the software, it will disable the long-term learn function as well. Although most folks won’t need to take that into consideration because they’ll leave the closed loop and learn elements of the system operating under their default parameters, he cites a few scenarios where you may want to disable these features, or take control over the specific RPM ranges where they operate.

Shop Holley Sniper 2 EFI conversion kits here.

“We do have some customers who use the closed loop feature with the ‘learn’ part of the system turned off. And you can set different values that will automatically enable or disable closed loop and/or learn. If you want to, you can have different values for each, so you can have it operate in closed loop up to a certain RPM, and then once the engine speed gets past that, it then goes into learn. Or maybe you’re in a situation where your cold start tune is totally dialed in, but once the engine gets up to a certain temperature, closed loop can fire up, and then at another certain point it can go into learn.

Shop Holley EFI Terminator X and Terminator X Max EFI systems here.

“The RPM values are typically used in more extreme high-performance and racing applications, especially in situations where you’re using a really aggressive camshaft with a lot of overlap and/or a really short length exhaust, like open headers. One of the biggest things that can trick an oxygen sensor is ambient oxygen, and in those situations a cam like that can create negative pulses at low RPMs after an exhaust event that try to draw atmospheric air back into the pipe, which can cause the oxygen sensor to get a false lean reading. If closed loop is enabled in that situation, it will try to compensate for a lean condition by adding fuel that the engine might not need at that point. And it’s a similar situation with open headers or short exhaust systems, as the short length makes it easier for the exhaust system to draw atmospheric air in and trick the oxygen sensor into thinking the engine is running lean. And the best way to get around that is by setting a minimum RPM to enter closed loop.”

If you’re data logging to troubleshoot the issue, he recommends looking for big swings in the air/fuel ratio between rich and lean at low RPMs that begin to stabilize as the RPMs go up. That’s a good indication that this issue is occurring, and to address it, you’d set the RPM parameter for enabling the closed loop function at roughly the point where those big oscillations between rich and lean calm down.

Self-Learning Is Powerful—Not a Band-Aid

Alford also adds that while self-learning can address small issues and even mask some tuning problems, it’s not intended to be a Band-Aid for mismatched parts – or a substitute for a solid base calibration.

“You need proper fuel pressure – if you have constant variations in pressure, the closed loop system is going to chase that wherever it goes and likely end up swinging back and forth between rich and lean conditions as a result. And while the closed loop system will cover up tuning sins to a certain extent, at the end of the day, it’s not designed to compensate for sloppy tuning. If you have a base map that’s way out in left field, closed loop is going to do its best to keep things in line, but it’s never going to operate as efficiently as it would with a proper base calibration. If you keep those things in mind going in, you’re going to have more success with the system.”