Charge Temperature Approximation

[TC951]

Hi guys,

I attempted to calibrate my Charge Temperature Approximation Table today with my G4X fitted to a 2.5L turbocharged 4-Cyl and experienced some odd results. 

Preface:

I had already calibrated my fuel table to achieve a lambda within +/- 1-2%  of target across most of the table.  When I originally did this I held IAT around ~40C while calibrating each cell.  I paid no attention to ECT at the time. 

Today I set out to do some testing and the results I found were quite different from what I expected…. So I'm wondering what you guys think of this…

Test conditions for screen shot below:

• RPM = 2000
• Charge Temp Approximation Table = 0%
• Load = ~115kpa (WOT)
• Inlet to turbo was ducted near the headers to increase IAT throughout the test

Based on the Help section, as IAT (and charge temperature while the table is set = 0) increases, lambda should trend richer due to less air density.  However, the opposite occurs…

My result:

What am I missing here??  Why would Lambda trend lean instead of rich as IAT increases??

Additional clarification:

IAT = TMAP sensor welded in thin wall aluminum intercooler pipe, 2-3 inches before throttle body.

TA3 = TMAP sensor welded near back of aluminum intake manifold

Dyno = Dynapack hub dyno

When performing the same test but with the Charge Temperature Approximation table set to 100%, lambda trends much leaner:

 

PC Datalog - 2022-07-7 5;46;31 pm_ChargeTempSetting_2.0krpm_WOT_0%_IntakeAtHeaders.llgx

[Adamw]

Notice that with your charge temp approximation at 0% the lambda is heading leaner as charge temp increases.  Whereas in the lower pic with the charge temp approximation at 100% the lambda is heading richer as the charge temp increases.  So your correct value is somewhere between these two examples.

This is because in the first test for your 20°C rise of IAT the approximated charge temp has risen also by 20°C.  Based on rough ideal gas law numbers the ecu would remove about 8% fuel for this 20°C rise.  

In the second test, you have a similar 20°C rise in IAT, but this time because your charge temp is biased more towards coolant your charge temp has only raised by about 9°C, so the ecu would have removed only about 3.5% fuel.

I cant see the numbers well enough but it looks like the lambda in the first test increased by about the same amount as the lambda decreased in the second test, so your correct value is probably somewhere about half way between these two.

Yes it is a pain in the arse that changing the charge temp approximation messes up the fuel map, but there is really no way around that.  

[TC951]

Hi Adam,

OK I think I was allowing myself to get confused by the large change in lambda between the two tests and the lean/rich amount between the two instead of focusing on the change in lambda error from start to end of test.  

 

I've made some changes to the ~20kpa area at 2.0krpm and am able to get a good result.   As you can see below, with a charge temperature approximation value of 45.5, the lambda error is now consistent throughout. 

 

However, while making a similar change at -30kpa/2000rpm, I am seeing some odd hunting as highlighted in the red dashed lines below.  Have you seen this before?  Lambda error is swinging up and down by ~2%

[Adamw]

Really need to see the actual log to get enough detail, but by eye the inj PW looks stable (ie fuel flow per intake event is stable), provided RPM & MAP is stable (air flow is stable) then you would expect a stable lambda.  Possibly a little on the lean side so you have intermittent incomplete combustion events on the leanest cyl.  Remember the Lambda is the average of all cylinders, you can potentially have 3 that are 5% richer than average and 1 that is 15% leaner...  

[TC951]

The log was too large to attach.  Here is a link if you'd like to check it out: 

https://drive.google.com/file/d/14E5bUdFSaN62AlWowxGzevfSLWd3lYaQ/view

I've double checked and I can't see anything obvious.  It does look control related to me but again nothing is sticking out... 

I don't think the swings are frequent enough to be cycle-cycle variation and I would think if the individual cylinders were offset they would stay that way, meaning the offset shouldn't shift periodically like it is in the data.   Anyway, if you see anything that sticks out let me know because I'm curious to learn what might cause that. 

 

[Adamw]

Nah I dont see anything obvious, there is a jump in MAP and consequently inj PW right on the 6min mark but after that it is pretty stable.  Is it more stable if you richen it up?

[TC951]

I had the pull the car off the dyno for now.

I ended up calibrating a few other area's in the map with good success including some higher RPM low load, higher RPM med/high load, and idle, so it was just this lower/mid rpm low load area that had that hunting.   I also went back and quickly re-calibrated most of the fuel map before pulling the car off the dyno, so I'll be sure to report back what I find when I revisit hopefully soon.  

[mapper]

I would check also injection timing. Try to optimise it first before going to charge estimation table. Maybe you sit just at a inj. Timing the engine is not happy. 

I usually tune in the following order:

- use a charge estimation table from a similar engine with same IAT sensor location

- tune charge cooling coefficient (sorry don't remeber exact name right now)

- roughly. Tune variable cam map than fuel map and finally injection timing arround load/rpm points you want to test charge estimation. 

- run charge estimation test. Roughly adj. Fuel map after a big change to bring back Lambda to target. Test and repeat. 

- once satisfied with charge estimation,  start with finetunig camtiming > fuel map & inj. Timing. Finally ign timing

[TC951]

On 7/12/2022 at 5:26 PM, mapper said:

I would check also injection timing. Try to optimise it first before going to charge estimation table. Maybe you sit just at a inj. Timing the engine is not happy. 

I usually tune in the following order:

- use a charge estimation table from a similar engine with same IAT sensor location

- tune charge cooling coefficient (sorry don't remeber exact name right now)

- roughly. Tune variable cam map than fuel map and finally injection timing arround load/rpm points you want to test charge estimation. 

- run charge estimation test. Roughly adj. Fuel map after a big change to bring back Lambda to target. Test and repeat. 

- once satisfied with charge estimation,  start with finetunig camtiming > fuel map & inj. Timing. Finally ign timing

Hmm interesting thought.  I haven't touched the injection timing at all... I left it at single zone and 400deg BTDC.  Certainly makes more sense to make it a map and calibrate accordingly.  However I'm wondering what is the best and most efficient way to calibrate that map.  What method do you use?  

[Adamw]

6 hours ago, TC951 said:

Certainly makes more sense to make it a map and calibrate accordingly.

A signal value is usually fine as once you are over about 30%DC the timing has little effect.  Typically done on the dyno to acheive best torque.  One approximation you can do is with engine running steady at fast idle, adjust the inj timing backwards and forwards until the richest lambda is achieved, this is about your optimum inj timing.  

[TC951]

3 hours ago, Adamw said:

A signal value is usually fine as once you are over about 30%DC the timing has little effect.  Typically done on the dyno to acheive best torque.  One approximation you can do is with engine running steady at fast idle, adjust the inj timing backwards and forwards until the richest lambda is achieved, this is about your optimum inj timing.  

Your comment about 30% DC makes me think about my injector sizing and wonder if I'm sized too large, and what effect that might have on some of the results I saw at low load during charge temp approximation testing as shown above.  At first I thought I might be getting into the non-linear range of the injector curve but looking at Injector Dynamics website I think I'm still way past that range in my "hunting" area.  I'm running ID1050X injectors and I don't get to about 30% duty until I get right around/just above 1 bar of boost.   The max duty I see is typically around 55%.  

So,

1. Considering my car is a partial daily driver, there's a lot of time spent well below 30% duty where I think I could probably benefit from a map over a signal value.  Do you agree, or think I'm being too picky? 

2. Considering I never get over ~55% DC do you think my injectors are too large for my application?  I tried to size them on the larger side to protect for a future higher performance engine build I've been planning, but now I'm wondering if I went a hair too large.  

[Adamw]

8 hours ago, TC951 said:

1. Considering my car is a partial daily driver, there's a lot of time spent well below 30% duty where I think I could probably benefit from a map over a signal value.  Do you agree, or think I'm being too picky? 

I would expect a single value would still be fine.  30% DC means the injector is squirting for 220deg crank rotation.  So moving the timing say 50deg either way is unlikely going to change much.  On very high end NA engines with staged injection, outboard injectors, large overlap etc and injector DC below 30% all the way to say 10000RPM, then injector timing becomes much more important. 

 

8 hours ago, TC951 said:

Considering I never get over ~55% DC do you think my injectors are too large for my application?  I tried to size them on the larger side to protect for a future higher performance engine build I've been planning, but now I'm wondering if I went a hair too large.  

Sounds ok.  ID 1000/1050 are generally ok on petrol with a min cyl size of about 500cc.  There are 3 important considerations.  As far as injection event timing goes, the lower the DC the better.  But in a road car you also need to consider the short PW behavior and not go so big that you will be operating in a region at low loads where the injector behavior is less predictable.  Finally spray pattern, droplet size and atomization can have a big impact on how much puddle is attached to the port wall and also vapor production at cold start.  Larger injectors typically are worse than small injectors from this point of view.  Puddle volume affects transient fuel and vapor production affects cold start.  

[mapper]

It's really depends on your expectation how much you want to optimise your engine. The influence of inj timing differs a lot between engines. I've seen anything between 0.x% and up to 10% torque change by varying inj timing (same AFR). It does also make a good difference to the transient response of the engine and how much accel emrichment is needed. Personally I don't agree that a single value is enough. I do map all cars with full 3d injection timig table and optimise it at all  and load and RPm points.