Individual cylinder lambda vs EGT - Interesting results

I had the opportunity recently to gather some data on our 86 as I was testing Link's new CAN-Lambda modules pre turbo. This gave me the opportunity to compare the lambda values to the EGT values. It's pretty common in drag applications to rely on EGT for individual cylinder fuel trimming and I'd say that most tuners would be working on the assumption that when the EGT is equal across all cylinders, that the lambda is also equal. Here's some data showing this is clearly not the case.

The data starts with no individual cylinder trims applied and you can see quite a dramatic difference between the 4 cylinders. The group above shows the exhaust lambda measured post turbocharger for reference.

Next we have the data with the individual cylinder tables adjusted and cylinder lambda within about 1% - Note that at this point the individual cylinder lambda data isn't pressure compensated and that's why the reading is much richer than the collector lambda.

Finally we have the EGT data collected with the lambda equalised showing around a 28 degree variation from hottest to coldest.

It's worth noting that the net effect on power from these changes was zero although I didn't try making any adjustments to the ignition timing during these tests.

very interesting, what was the split on the EGT's before you balanced the cylinders using AFR's?

i know EGT's are slow to react but they are definitely better than nothing, case in point.

FSAE car running CBR600 a wide-band in the collector, car was tuned to safe AFR, but seemed down on power but the team didn't have time to check everything so ran the engine.

after the event EGT's were added same engine dyno it was discovered one injector was bad, that cylinder was effectively an air pump and the other 3 were pig rich to compensate. this all happened with an experienced (in the industry not just FSAE) tuner.

shows another weakness of single lambda set ups.

I don't have a screen shot from the first run unfortunately. The split wasn't too much different to what we can see after, but the individual temps were obviously not quite the same.

Yes the EGT sensor is much slower to respond than the likes of lambda, however I've used them in drag applications to decide which cylinder was suffering from a misfire. This saves a lot of time compared to the typical trial and error approach when you don't know specifically which cylinder is exhibiting a problem.

Great find Andre!

Consequently do you woud say that tuning with EGT makes the individual lambda worse than without any cyl. trim applied?

You haven't used pressure compensation. Could there be a chance that the dynamic pressure in each runner can differ enought to influence the Lambda readings significantly?

In this particular instance the EGT split was pretty comparable before and after I'd adjusted the individual lambda to match. While the pressure compensation could possibly affect the individual cylinder lambda readings, normally we would compensate each lambda with a single pressure sensor in the collector anyway so it's still possible to have a variation in the pressure at the individual sensors depending on when the lambda was sampled.

In this instance though I was seeing significant variation in lambda under light throttle cruise conditions when the exhaust pressure (I was still measuring it) was close to 1.0 bar.

Interesting results. We're you using open tip EGT sensors and how far away were they from the manifold inlet face?

These sensors are enclosed tip - I've used the exposed tip sensors in the past and found they offered a pretty short life span (although while they were working they tend to respond quicker and read a little hotter). They are mounted around 25 mm from the header flange. The Subaru head is a little unique in that the exhaust ports are not symmetrical and hence while all the sensors are mounted an identical distance from the header flange, it's likely that the distance to the exhaust valve is actually a little different. Sometimes it's just not possible to achieve a perfect result though.

Interesting results Andre. Does a single value trim cylinder trim work adequately across the whole load range?

Or do you really need at least a 2D trim table for each cylinder?

I have the option to run EGT per cylinder on my ECU and a % gain per cylinder, I guess its not necessarily very effective?

In the case of the Motec M150 I have a separate 3D trim table for each cylinder so I can be quite specific with my adjustments. You'll find that if you're trying to trim cylinders off EGT alone that it's all but impossible to correct the EGT at part throttle/load so they are most commonly used just under WOT where a 2D table will be effective.

I only have the ability to use a % gain to per cylinder fueling with the MaxxECU so 2D or 3D trim tables are not possible yet.

From your data above we can't accept a higher EGT is caused by leaner AFR?

I guess what I am trying to ask is whether EGT data is useful at all for trying to balance per cylinder AFRs?

Thanks :)

Let's consider the implications of trimming the fuel to achieve consistent lambda across all cylinders: If for example we need to remove 5% fuel from one cylinder in order to achieve a consistent lambda, this would infer that we have less air entering that cylinder. If we're combusting less air and fuel then it would be sensible to assume that this would result in a lower combustion temperature and finally a lower EGT. So on this basis it's reasonable to expect that if we achieve consistent lambda, our EGT are likely to be different and vice versa. Of course the combustion charge isn't the only aspect affecting EGT but it's a large consideration.

So on this basis is EGT relevant at all? In my opinion yes. We're really only going to be concerned with individual cylinder trimming typically on engines producing very high specific power levels (with the noted exception of our relatively lame 86... That's all in the name of science). As the boost and specific power levels climb, the window that we can operate in safely gets narrower and in these cases EGT will let us know if we're about to end up with a molten pool of scrap. In this case I'd argue that it's more critical to maintain safe EGT, even if this is at the expense of a cylinder potentially running richer than the others.

If we are interested in efficiency, economy and emissions then I'd say lambda is the key metric, but for high boost/high specific power levels and perhaps methanol fuel I'd be more inclined to watch EGT as I feel this is a better guide to ensuring engine reliability - Look for example at any Top Alcohol drag engine and it will be equipped with individual cylinder EGT sensors but never a lambda sensor to be seen.

nice writeup ;-)

if we use individual EGT on a high boosted engins, how we know where is the limit in temperature?

I tuned a 650hp 4G63 with 4 EGT's and made a big 3D table for each cylinder. It is, as you mentioned Andre, nearly impossible to get equivalent EGT's at low load. It took up to 25% trim at low load (i don't left it there!) I suppose there are coming other effects into the game at low load. Consequently I ended up the low load range with smaller adjustments than what was needed for equivalent EGT's .

Something other really important was, the EGT's need 10-25 seconds to settle (have a look on the attachement!). What I also found, the needed correction can differ alot from cell to cell. For example WOT at 5000 or 5500 rpm can make a difference of 4-5%. EGT can be be tuned perfectly in steady state, but they differs still slightly on a WOT run.

Later i added cylinder ign trims. The effect to EGT can than nicely been seen on the gauge.

All in all i spent alot of time and hold the engine on many different load points until the EGT were stabilised and made the adjustments only than.

It's to pity to hear now that it was potentialy not worth to do this sophisticated procedure...

I'll be seeing the individual EGTs next week on a M800 equipped Supra with a GTX42 ... I'm not really fond of holding that particular car steady state at 6000rpm WOT and wait for the EGTs to settle.

Let's hope a ramp run in 5th gear will give them enough time to get to some sort of equilibrium :)

can't wait actually :)

Thanks guys the detailed replies :)

KGPerfomance

for sure, i wouldn't do that too! ;-) I made most Steady state tuning with boost control turned off. Means around 500 ponys. ahh and it's a Time Attack car with a serious driver. Means car sees regulary 20min WOT stints around the track during training. So it must hold mid boost for that time. Done 12 Maps with different boost, ign strategies on this car, but that's another story...

Maybe it's just me, but I think the layout of a Subaru engine might influence the EGT.

Cylinders 3 and 4 are obviously on the back, so less chance to be cooled down by the air, and perhaps the turbo is in front of the cylinder n°1 hence the higher EGT compared to the n°3 ?

Adrian, I didn't mean to question your steady state tuning, I was assuming you didn't do those things either :)

I will have boost control turned of also, but the wastegates are equipped with 1.5bar springs ... that'll make things fun I guess :) :) :)

@Ludo I agree that the boxer arrangement would likely influence EGT. As mentioned earlier, the exhaust ports on the Subaru engine are also not symmetric from the front to rear cylinder which means the distance from the exhaust valve centreline to the header flange is slightly different too.

@ Adrian understanding the EGT limit for a particular combination unfortunately often comes from expensive experience. As I've said before in a couple of webinars, the ultimate EGT reading will vary based on so many factors - Fuel type, compression ratio, ignition advance (or degrees from MBT), EGT type and sensor location. I've noticed for example a difference of something like 50 degrees C between an exposed tip EGT sensor and an enclosed tip sensor.

Just to give you an idea, I was involved with the tuning of the Heat Treatments R32 GTR drag car for a number of years (has run as fast as 8.4 and 199 mph). This ran on methanol fuel and at the time I was tuning it we were running around 50 psi. My 'happy place' for EGT on this engine was around 750 deg C.

As another example I'll use the EVO 9 drag car that my old business built and I tuned. This ran VP Q16 fuel and ran 42 psi boost in 4th gear. I was comfortable with EGT in the 900 deg C range on this engine. This might not be strictly useful but hopefully illustrates that there is a little more to it than just saying XXX EGT is safe.

Understanding the latency of the EGT sensor is definitely another important aspect - Compared to lambda sensors the EGT sensor is positively pedestrian in its response. In a drag application (which is still about the most common place we rely on EGT sensors) I tend to ignore the data until the engine is in the higher gears in the deep end. At this point we tend to see the readings from each cylinder kind of reach an equilibrium and we can make sense of the data.

Let's consider the implications of trimming the fuel to achieve consistent lambda across all cylinders: If for example we need to remove 5% fuel from one cylinder in order to achieve a consistent lambda, this would infer that we have less air entering that cylinder. If we're combusting less air and fuel then it would be sensible to assume that this would result in a lower combustion temperature and finally a lower EGT. So on this basis it's reasonable to expect that if we achieve consistent lambda, our EGT are likely to be different and vice versa. Of course the combustion charge isn't the only aspect affecting EGT but it's a large consideration.

something just came to mind when reading that over...

if we have a cylinder that we are putting less fuel in as is getting less air, should the timing for this cylinder trimmed also?

In a perfect world, the answer is simply yes.

But in reallity it's not that easy.

In an non knock limited engine, i suppose it's unlikely to see obvious gains from individual cylinder timing if you sitting already at MTB.

There are many other variables in a knock limited engine. Timing is limited by the knock limit of each cylinder. This limit depends on many factors, some of them are not equal on all cylinders. For example combustion temperature, static compression (influenced by port design, exh. manifold etc.) can vary depending on engine design.

Consequently i would suggest to tune tables in this order:

- fuel table and ign table

- adjust ign table a few degree back from knock limit

- tune fuel cyl trims by individual lambda

- tune idivuall ign trims, use an Audio knock detection and datalogging (for example Plex knock monitor)

Really it depends how much time you want to spend on the dyno (and obviously what instrumentation you have on the engine to start with). Yes we may find that the ignition timing per cylinder might want to change based on the combustion charge, however you're getting to the point of diminishing returns as to what you can expect to gain. The torque curve tends to plateau over perhaps 3-5 degrees of ignition timing near to MBT anyway so you're not likely to see any dramatic changes in torque.

The more powerful use of individual cylinder timing trims is in the situation where you have one or two cylinders that are particularly prone to knock. In this case if we advance the timing on the remaining cylinders we can often see a worthwhile gain in power/torque.

nice noted.... thank you guys

can someone tell me like whats the pros and cons of tuning with EGT vs lambda like i need all the tips and tricks guys im new to all this and i just started learning all these stuff...

Hi Andre, Thanks for the write up regarding this topic. I build and tune my self.. I use the Cobb Accesstuner software on a 2005 Subaru STi I bought brand new! I am dialing in a PTE 6466 turbo and put the tune on hold until I get the per cylinder bungs AFR bungs installed. I am curious if you have any more data that may show some kind of pattern for the "typical" adjustments required to balance the AFR. Thanks again for the good work. I attached the "Nutritional Facts for my build for the fun of it.

I may have missed it (skim read) - but a couple of other factors may be at play when EGT and AFR don't co-incide.

1/ Heat soak and/or the cooling around the cylinder that may mean differing amount of heat drawn out of the exhaust gas.

2/ Scavenging of the cylinder as residual exhaust gas may reduce the net charge density, and hence the heat released during combustion.

Both those issues may be more of a problem with 'log' type exhausts.