Understanding Knock and Preignition: The Combustion Perspective

[arghx7]

You've probably read a lot of stuff on the internet about knock, and maybe preignition. You know that they're undesirable for a spark ignited combustion engine - they can make noise, they can damage your engine if severe enough. We know knock and preignition are undesirable explosions, explosions that happen when we don't want them to happen. What exactly is happening in the cylinder though? Let's look at knock and preignition from a combustion perspective. First we need a brief introduction to combustion pressure and how it relates to spark timing. Then we can look at what happens to cylinder pressure during a significant knock event, and finally we can discuss what serious preignition event looks like.

Spark Timing, Cylinder Pressure, and Combustion Phasing

The plug creates the spark, and then the mixture begins to burn. For simplicity's sake, let's take combustion speed/burn rates out of this part of the discussion. At low speed/low load I can advance my spark very far past MBT (minimum spark advance for best torque) without encountering knock. MBT typically has the peak pressure located at about 11 degrees after top dead center firing. You can see in the image below the combustion getting more and more advanced.



Now retarded combustion has this characteristic "double hump." What happens is, the first hump (looking left to right) is the compression from the piston reaching TDC. My burning is very retarded due to spark timing, but also due to burn rates which are a factor I won't get into. The pressure begins to drop as the piston moves downward, until enough burning begins and the pressure rises again--hence, the double-hump.

Why would I retard my combustion from MBT? I need to reduce knock, and in some cases I might need to raise exhaust temperatures to spool up a turbo or warm up a cat. The more knock limited the engine is, the more of you will see this characteristic double-hump pressure trace as spark is retarded. Retarded combustion lowers the peak cylinder pressure, which may be desirable if I don't want to stress the mechanical components too much (bearings, rods, etc).

MBT combustion has a single hump with significantly higher peak pressure then retarded combustion even if there is no knocking. It's a single hump because burning is occuring much closer to when the piston is at TDC, so the pressure of compression combines with the pressure of the combustion.

When the combustion is advanced is past MBT, you get much higher peak combustion pressures and you can see the characteristic "lean" to the left.You don't have to be knocking to have high cylinder pressure. Under normal circumstances there is no reason to advance combustion past MBT. Now let's look at what happens when I start to knock.

Knocking Combustion Pressure

If the spark is too advanced for the engine speed & load for a given fuel, the gases at the end of the combustion chamber will explode before the normal flame arrives. Knock has different levels of severity. The more severe it gets, the more likely you will hear it audibly, and the higher the peak cylinder pressure will be.



In the image above, the engine is knocking significantly but probably not enough to damage the engine and probably not enough to be heard over normal engine sounds without a microphone. The severity of the cylinder pressure spike is called the knock peak while the amount of "ringing" is called the knock intensity. There are other ways to characterize knock but those are the main metrics that rely on actual cylinder pressure.

This pressure trace isn't going to immediately break anything, but a knock sensor will probably judge it as knock. As I've pointed out in the past, somebody somewhere had to make a judgement call as to what constitutes severe enough knock for the knock sensor to react. The knock sensor can't read the cylinder pressure, only the vibration in the engine. As the engine speed increases, both the engine vibration and the cylinder pressure get noisier.

Severe Preignition

Preignition occurs when an explosion happens before the spark plug fires. This isn't always a major problem, but severe events can happen at high load for a number of reasons. Maybe some deposits broke off inside the engine. Maybe a spark plug overheated. Maybe some debris got into the engine. There are a bunch of causes for preignition, but I want to focus on how the cylinder pressure behaves in the event of a severe preignition event.

Here are three major characteristics of severe preignition:

1) high peak combustion pressure - this is can damage your engine
2) high rate of pressure increase inside the chamber - this is a major factor in the audible noise (metallic sound)
3) high pressure at the time of ignition - this is one of the main ways we know it's preignition and not just regular old spark knock.

Below is an image showing three successive pressure combustion cycles for the same cylinder. Cycle #2 has severe preignition. The rate of pressure rise and the pressure at ignition is much higher than the other two cycles (the lower chart isn't crank angle based, don't look at it the same way as the upper chart).



If we zoom in on the preignition event and overlay the ignition coil current trace, we can see cylinder pressure already shooting up when the coil discharges.



Compare that to the image below, which is the same scaling. The coil discharges and then there is a delay until enough burning happens to raise the cylinder pressure.





You can see that there is a significant difference between a severe preignition event, the kind that can definitely damage your engine, and a moderate knock event which might call for retarding spark timing. Of course, a severe knocking event can also cause high peak combustion pressures and noise. However, unlike preignition a severe spark knock event will not have that same characteristic high pressure at the time of ignition and fast pressure rise rate. I hope this clears up some misconceptions about these two types of abnormal combustion.

[Wepeel]

I feel like I should search all posts by arghx7 just so I can thank them all.

One thing I've always wondered - the sensitivity of the knock sensor. I'm guessing there's a whole discipline associated with interpreting knock accelerometers and determining what is "true" knock vs. what is something that might indicate knock vs. when conditions are favorable for knock. There must be some spectrum - our ECUs are designed to pull intervals of timing and it needs to know whether to pull 1 degrees, or 4 degrees... maybe that has to do more with the frequency of knock signal versus the severity, or maybe both.

Also, I've heard tuners say that they've driven around with detcans and can't hear any knock even though the ECU is pulling timing. I know there were some recent findings with the WRX/STI platform that suggested that the ECU could pull timing (can't remember if it was FBKC or FLKC) just based on rapid tip in - maybe rate of throttle opening or pedal position change - something like a preemptive knock control. I know for sure GM had this programmed into their ECU for some cars.

And then there are the cars that can run on regular or premium, where running regular is not a detriment to the engine but premium will have more performance. Not sure if there is a fuel sensor, like the flex-fuel kits for our cars, or if they are relying on knock signals to determine fuel grade. If it was the latter, it doesn't seem like driving the engine to knock would be a great strategy for determining fuel grade, but it would make more sense if the knock sensor got a little "noisier" with lower octane fuel but the engine wasn't necessarily knocking.

[Dezoris]

This is why forums were invented. Thanks.

[arghx7]

Originally Posted by Wepeel

I feel like I should search all posts by arghx7 just so I can thank them all.

One thing I've always wondered - the sensitivity of the knock sensor. I'm guessing there's a whole discipline associated with interpreting knock accelerometers and determining what is "true" knock vs. what is something that might indicate knock vs. when conditions are favorable for knock.

It's more complicated and subjective than you think. Before the knock control system even exists try watching a bunch of grown ass men with a gazillion dollars worth of equipment argue over whether it's really at "knock limit" or not. In a lab some will look at the calculated knock peak, some will look at the calculated knock intensity, some will look directly at the combustion pressure trace on an oscilloscope, some will listen for knock with a microphone. At some point that judgement has to be made--how much is too much--and then the knock sensor can be programmed. The knock sensor literally does not exist yet at this point, and in fact somebody has to figure out where to even put the thing when there are a gazillion other things that need to fit on the block.

Then somebody has to pick a frequency that will be judged as knock (based on a fast Fourier transform chart, wikipedia it) - another judgment call based on the individual engine. There's also a crank angle window where the knock will be measured that takes into account for example noise from the valves opening and closing.

That's all decided before the engine is ever actually in a car driving around.

There must be some spectrum - our ECUs are designed to pull intervals of timing and it needs to know whether to pull 1 degrees, or 4 degrees... maybe that has to do more with the frequency of knock signal versus the severity, or maybe both.

I haven't seen the exact Subaru filtering algorithm and background noise learning. It hasn't been extensively reverse engineered, so I don't know much more than you do in regards to the FA20 engines.

Also, I've heard tuners say that they've driven around with detcans and can't hear any knock even though the ECU is pulling timing.

If 4 guys in a lab with cylinder pressure transducers, lab-grade microphones and accelerometers can't easily agree on what "borderline" knock looks like or sounds like, it's no surprise that someone driving around with a detcan can't agree with an ECU. Severe knock is often obvious. Very little knock usually isn't that difficult to figure out. Borderline knock - just enough knock to start retarding timing - is a judgment call. As a corollary, if figuring out the knock limit is difficult to determine, then figuring out how much spark retard to get you back to that limit is fundamentally the same problem.

I know there were some recent findings with the WRX/STI platform that suggested that the ECU could pull timing (can't remember if it was FBKC or FLKC) just based on rapid tip in - maybe rate of throttle opening or pedal position change - something like a preemptive knock control. I know for sure GM had this programmed into their ECU for some cars.

That's usually not knock control, it's a torque reduction request on tip-in. There are different terms for it (bump control, bucking control, etc). It's to make a smoother transition when you get into the accelerator pedal. When and how much of that is partly based on some subjective criteria as to what will appeal to the people buying the car.

And then there are the cars that can run on regular or premium, where running regular is not a detriment to the engine but premium will have more performance.

Wellll... it depends on what you mean by a detriment to the engine. If you mean not causing long term damage, yeah that's true for the most part. Except that one time GM released their 2.8 turbo engine in a few SUV's and they would blow up if you put regular gas in them and dropped the hammer enough times (they technically required premium, but how many SRX owners do you think listened to that). Even if nothing is getting damaged, the detriment to the engine is in the quality of combustion, which if done right is not so noticeable to the driver. If it is done wrong, then you can get hesitations and surging because the ECU is pulling so much timing.

[Not sure if there is a fuel sensor, like the flex-fuel kits for our cars, or if they are relying on knock signals to determine fuel grade. If it was the latter, it doesn't seem like driving the engine to knock would be a great strategy for determining fuel grade, but it would make more sense if the knock sensor got a little "noisier" with lower octane fuel but the engine wasn't necessarily knocking.[/QUOTE]

There isn't such a thing as a fuel octane sensor. You can determine the real octane of fuel by sending it off to a lab. It's not always as advertised...sometimes significantly better, sometimes significantly worse, unless it's certification fuel. Now, some manufacturers will model the octane inside the ECU and use that as part of the spark control. Nissan does that--ever wonder why there is so much confusion on how spark timing calculations actually work on a GT-R?

[Calum]

Originally Posted by Wepeel

I feel like I should search all posts by arghx7 just so I can thank them all.

You should read them as well. I don't think the man has made a post that wasn't highly educational.

[solidONE]

Originally Posted by Wepeel

I feel like I should search all posts by arghx7 just so I can thank them all.

One thing I've always wondered - the sensitivity of the knock sensor. I'm guessing there's a whole discipline associated with interpreting knock accelerometers and determining what is "true" knock vs. what is something that might indicate knock vs. when conditions are favorable for knock. There must be some spectrum - our ECUs are designed to pull intervals of timing and it needs to know whether to pull 1 degrees, or 4 degrees... maybe that has to do more with the frequency of knock signal versus the severity, or maybe both.

Also, I've heard tuners say that they've driven around with detcans and can't hear any knock even though the ECU is pulling timing. I know there were some recent findings with the WRX/STI platform that suggested that the ECU could pull timing (can't remember if it was FBKC or FLKC) just based on rapid tip in - maybe rate of throttle opening or pedal position change - something like a preemptive knock control. I know for sure GM had this programmed into their ECU for some cars.

And then there are the cars that can run on regular or premium, where running regular is not a detriment to the engine but premium will have more performance. Not sure if there is a fuel sensor, like the flex-fuel kits for our cars, or if they are relying on knock signals to determine fuel grade. If it was the latter, it doesn't seem like driving the engine to knock would be a great strategy for determining fuel grade, but it would make more sense if the knock sensor got a little "noisier" with lower octane fuel but the engine wasn't necessarily knocking.

I think there is are tables in the rom where you can program in the "tip in" ignition retard according to the amount of change in throttle position and another table that control the amount of additional fuel is added at tip in according to the percentage change in throttle.

I'm also curious about this tip in knock that the ECU is detecting particularly at higher temp operation. It seem like it's impossible to totally eliminate nomatter what you do with the tip in enrichment or ignition retard. Yet, it seems to disappear when temperatures cool.

[thambu19]

Originally Posted by solidONE

I think there is are tables in the rom where you can program in the "tip in" ignition retard according to the amount of change in throttle position and another table that control the amount of additional fuel is added at tip in according to the percentage change in throttle.

I'm also curious about this tip in knock that the ECU is detecting particularly at higher temp operation. It seem like it's impossible to totally eliminate nomatter what you do with the tip in enrichment or ignition retard. Yet, it seems to disappear when temperatures cool.

First of all big thanks to @arghx7 for the write up. Take a lot of patience to do this.
So most tip in knock is due to two main reasons:
1. The actual engine load is not always determined precisely because most Volumetric Efficiency maps & MAF sensor based load determination is calibrated at steady state operation. What happends during Tip IN and also Tip Out for that matter is a transient air flow problem that most OEMs still try to model using manifold filling models but still fall short in properly estimating load. So when the engine load/airflow is higher than what the ECU estimated the spark timing would be over advanced for the load and the fueling could go
slightly on the leaner side until the O2 sensors correct using feedback.

2. As the load changes the cams have to move to a different position based on their lookup table. Most times these days they actually have to move a lot because OEMs try to use cams a lot for fuel economy (Lowest VE cam timing) and they have to travel a lot to go to their peak power positions (Best VE cams). Now this takes a few tenths of a second to happen so it is pretty fast but still slow enough for a cylinder to knock because the spark was calibrated in steady state for a different set of cam positions but the cams are not there yet meaning the spark and cam positions are not exactly where things were calibrated in steady state.

Some OEMs have calibrations to add and subtract spark based on how far the cams are away from their calibrations. Trust me it is better not to try and calibrate that on the street or on the dyno because it is not worth the effort and better to leave those cals as they are. If done wrong the engine can bog down during transients.

One thing I would encourage everyone to do soon after flashing in a cal is to let the engine learn the background noise. This is usually learnt at low loads and across the engine speeds. So basically 1st gear slow pull to redline a few times. Not WOT once the ECU knows the background noise it will be able to determine knock more effectively instead of thinking the valvetrain noise to be knock.

The amount of spark pull out is usually based on the freqency and severity of the knock signal so learning the background noise level will make the knock sensor detect the right severity.

[thambu19]

Originally Posted by Wepeel

And then there are the cars that can run on regular or premium, where running regular is not a detriment to the engine but premium will have more performance. Not sure if there is a fuel sensor, like the flex-fuel kits for our cars, or if they are relying on knock signals to determine fuel grade. If it was the latter, it doesn't seem like driving the engine to knock would be a great strategy for determining fuel grade, but it would make more sense if the knock sensor got a little "noisier" with lower octane fuel but the engine wasn't necessarily knocking.

I was told that GM does two calibration for their spark one for high octane and one for very low octane and based on the knock frequency and severity the ecu estimates an octane number and interpolated between the two spark surfaces. Not sure if this applies to all engines and not sure if they still do it

[steve99]

Originally Posted by thambu19

First of all big thanks to @arghx7 for the write up. Take a lot of patience to do this.
So most tip in knock is due to two main reasons:
1. The actual engine load is not always determined precisely because most Volumetric Efficiency maps & MAF sensor based load determination is calibrated at steady state operation. What happends during Tip IN and also Tip Out for that matter is a transient air flow problem that most OEMs still try to model using manifold filling models but still fall short in properly estimating load. So when the engine load/airflow is higher than what the ECU estimated the spark timing would be over advanced for the load and the fueling could go
slightly on the leaner side until the O2 sensors correct using feedback.

2. As the load changes the cams have to move to a different position based on their lookup table. Most times these days they actually have to move a lot because OEMs try to use cams a lot for fuel economy (Lowest VE cam timing) and they have to travel a lot to go to their peak power positions (Best VE cams). Now this takes a few tenths of a second to happen so it is pretty fast but still slow enough for a cylinder to knock because the spark was calibrated in steady state for a different set of cam positions but the cams are not there yet meaning the spark and cam positions are not exactly where things were calibrated in steady state.

Some OEMs have calibrations to add and subtract spark based on how far the cams are away from their calibrations. Trust me it is better not to try and calibrate that on the street or on the dyno because it is not worth the effort and better to leave those cals as they are. If done wrong the engine can bog down during transients.

One thing I would encourage everyone to do soon after flashing in a cal is to let the engine learn the background noise. This is usually learnt at low loads and across the engine speeds. So basically 1st gear slow pull to redline a few times. Not WOT once the ECU knows the background noise it will be able to determine knock more effectively instead of thinking the valvetrain noise to be knock.

The amount of spark pull out is usually based on the freqency and severity of the knock signal so learning the background noise level will make the knock sensor detect the right severity.

Yep their is a heap of tables for knock control , transient ignition retard ect.

most of the stuff here is reliavant to our ecu
http://www.romraider.com/forum/topic1840.html

When you reset ecu or reflash the FLKC table is cleared as well as fuel trims

The ECU does not appear to start re-learning anything with IAM or fuel trims till the coolant is at operating temp. this seems to apply all the time.

The cams are also disabled after reset/reflash for about 1 minute of idle time.

the FLKC table looks similar to this

[thambu19]

Thanks @steve99 for the table. So I do not have OFT or any other equipment to log data. Well as of now atleast.
What I would like to do is first log data and later start fiddling with cals.
So after a flash the cams are sent to lockpin/endstop until the cam sensor encoder edges are learnt.
When the cams are timed in the plant they set it to appropriate timing and when the engine starts up the first time it learns the position of the cam encoder wrt crank encoder and from now on the ECU can keep account of chain/belt stretch as the engine wears in. Most chains stretch most in the initial few thousand miles and then it does not stretch a whole lot. The issue is when we reflash the chain stretch information is lost forever and the cams are not exactly where they were set before the stretch so you may find that the VE adaptions are slightly different before and after first reflash of the car. Okay all of this is treu but not sure if this applies to this particular engine if subaru has a unique way of handling it.

[thambu19]

Does anyone know what the OFT folks actually did to avoid pre-ignition? Pre-Ignition is pretty rare on a non boosted engine except if running at engine speeds lower than 1000rpm and very high loads. Now this entirely depends on the true dynamic compression ratio of the engine (not geometric). I have seen preignition in engine running 600rpm high load but that is only because we force the engine to run there for sake of calibration purposes.

What I am wondering is are we calling extreme knock as pre-ignition here?

Is it possible that these events are only seen during track day sort of events with high coolant temp and IAT and non-existent in normal circumstances?

If it is indeed PI the DI injection could be partly to blame for leaving carbon deposits. Then would it be a good idea to have a track day only cal utilizing mostly PFI to clean up the carbon left behind by DI to avoid PI?

As I understand changing spark timing does not solve PI since PI is independent of spark.

[phrosty]

Originally Posted by thambu19

If it is indeed PI the DI injection could be partly to blame for leaving carbon deposits.

My understanding is that the presence of PI helps clean off any deposits from the DI system.

[thambu19]

by PI, I meant Pre-Ignition

[Ultramaroon]

bump for a good read