SharkTuner/SharkPlotter Question- Standard Deviation
#1
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
This is something I really never paid much attention to. What is reasonable? Currently, I am getting a fair number of cells that read like: Ave AFR 14.7 Sdev= 0.9 N= 576. Seems high to me but I am basing that from a position of great ignorance.
Thanks, Ralph
Thanks, Ralph
Last edited by Ralph Newman; 07-16-2019 at 03:23 PM.
#2
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,496
Received 1,638 Likes
on
1,069 Posts
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
Short story: That is a very well-tuned cell.
Medium story: 14.7 is the target air/fuel ratio. std-dev of 0.09 means that ~68% of the time the air/fuel ratio is between 14.61 and 14.79.
This is about as good as it gets unless you throw away LH/EZ[KF] engine control.
EDIT: above applies for non-idle, non-WOT data.
Medium story: 14.7 is the target air/fuel ratio. std-dev of 0.09 means that ~68% of the time the air/fuel ratio is between 14.61 and 14.79.
This is about as good as it gets unless you throw away LH/EZ[KF] engine control.
EDIT: above applies for non-idle, non-WOT data.
#4
Rennlist Member
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
As you can gather there is quite a scatter of results and the greater the sample volume the more accurate the evaluation is going to be.
As to why you have such a relatively high range I have no idea unless the supercharger itself creates a degree of scatter due to the nature of the beast.
Time for Jim to chime in I suspect.
#5
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,496
Received 1,638 Likes
on
1,069 Posts
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
Correct your original post.
Now that I actually think about it, 0.09 std dev on a stoichiometric cell is impossibly-good unless the data is filtered.
You still running the stock narrow band O2 sensor, hot-wire Mass-air Sensor and LH/EZK?
Under steady state operation the stock system sweeps the a/f ratio across the target. So a wide std dev is expected for ‘cruise mode.’
Now that I actually think about it, 0.09 std dev on a stoichiometric cell is impossibly-good unless the data is filtered.
You still running the stock narrow band O2 sensor, hot-wire Mass-air Sensor and LH/EZK?
Under steady state operation the stock system sweeps the a/f ratio across the target. So a wide std dev is expected for ‘cruise mode.’
#6
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,496
Received 1,638 Likes
on
1,069 Posts
#7
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
Correct your original post.
Now that I actually think about it, 0.09 std dev on a stoichiometric cell is impossibly-good unless the data is filtered.
You still running the stock narrow band O2 sensor, hot-wire Mass-air Sensor and LH/EZK?
Under steady state operation the stock system sweeps the a/f ratio across the target. So a wide std dev is expected for ‘cruise mode.’
Now that I actually think about it, 0.09 std dev on a stoichiometric cell is impossibly-good unless the data is filtered.
You still running the stock narrow band O2 sensor, hot-wire Mass-air Sensor and LH/EZK?
Under steady state operation the stock system sweeps the a/f ratio across the target. So a wide std dev is expected for ‘cruise mode.’
Trending Topics
#8
Rennlist Member
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
I can buy into a small increase due to bigger injectors operating on light loads but not those numbers Ralph reports unless he has some absurdly large injectors.
For sure it would be interesting to see if others have had similar experience
When I changed out my injectors to 30lb items over the stock 19lb injectors I did not notice any obvious difference but then I made no effort to analyse to the nth degree either.
For sure it would be interesting to see if others have had similar experience
When I changed out my injectors to 30lb items over the stock 19lb injectors I did not notice any obvious difference but then I made no effort to analyse to the nth degree either.
#9
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,496
Received 1,638 Likes
on
1,069 Posts
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
Ah! Open loop numbers! In that case then, yes, the stddev seems too high. Jim/John will need to opine on them as my recollection and data for open loop is both distant and small.
#11
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,496
Received 1,638 Likes
on
1,069 Posts
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
Edit: above is true for low-to-mid load cells.
#12
Rennlist Member
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
You are not running a supercharger, yes? Therefore your fueling requirements do not have a base offset requirement for turning one that would drive the closed-loop sweep wider due to what I assume (based on black-box testing) are hard-coded sweep step parameters.
Edit: above is true for low-to-mid load cells.
Edit: above is true for low-to-mid load cells.
An SD of 0.9 when running on open loop/high load sounds way too high to me but not having had personal experience with a supercharger that is more an intuitive type of thing- not established fact. I fitted the bigger injectors in anticipation of iftting a supercharger [hence my interest] bu never got round to it- the Cayenne Turbo S put paid to that project!
#14
Addict
Rennlist Member
Rennlist Member
Join Date: Oct 2003
Location: Gone. On the Open Road
Posts: 16,496
Received 1,638 Likes
on
1,069 Posts
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
Was the data gathered on a chassis dyno or via 'open-road?' If the latter, the vast majority of the data with a 14.7 target will be for low load cells.
We also do not know the overall tuning strategy. For a boosted application there are various tricks that need to be applied for high-load operation. And 14.7 should not be the target for high-load cells.
Bottom line, I agree that for a high-load cell, open-loop or otherwise, a StdDev of 0.9 is curious.
#15
Nordschleife Master
![Default](https://rennlist.com/forums/images/icons/icon1.gif)
If one were to start debugging this, I'd take the logs and see how persistent the AFR is in those cells. Does the AFR vary around the mean in a quickly mean reverting, highly transitory way? Or is the AFR variation persistent such that for the next fifteen minutes it reads high and then an hour later it might read low for antother fifteen minutes.
Some ideas based on my personal experience:
- The O2-MAF sensor adaptation doesn't work very well when the injectors get big and the MAF flows more air. I would disable the O2-MAF sensor adaptation algorithm. If that thing gets confused, AFRs can go all over the place.
- The LH doesn't have a fuel temperature compensation. If you have Bosch 044 (or, worse, and aftermarket "performance") fuel pump, the constant recirculation of fuel can increase its temperature. The fuel density goes down and AFR goes up. This is a slow moving error.
- Are the fuel dampers there? If the fuel rail driven by the LH doesn't have sufficient fuel damping, that creates resonant slightly angled but mostly vertical "stripes" in the Sharkplotter AFR visualization. It will be impossible to get rid of those errant AFRs within a cell, because the resonance is so sharply determined by (rpm, pulse width). It's a fools errand to try to tune the LH without very well dampened fuel rails.
- Is the WBO2 consistent with the NBO2 sensor in the vicinity of 14.7 AFR? If it's not, then you may be chasing a non-existent error. The WBO2 grounding etc. needs to be done in a way that is conducive to accurate measurement.
- Is fuel pressure in the rail constant when referenced relative to the intake manifold pressure? The LH assumes it is. The fuel heating and aftermarket "performance" fuel pump cavitating lead my car to have a permanently installed mechanical fuel gauge in one of the fuel rails.
- Can the cell high std. cell be accessed with or without the WOT sensor triggering? I don't recall off the top of my head if Sharkplotter adjusts for this, but if it doesn't then one point richer with WOT relative to part throttle isn't crazy, given how the LH ads the WOT fueling vector on top of the regular map.
- Finally, if you're about to drive off the spec range of the Bosch MAF sensor, then it stops measuring air flow past certain air mass flow. Unfortunately, this level of mass air flow where the sensor goes out of range depends on the operating environment.
Some ideas based on my personal experience:
- The O2-MAF sensor adaptation doesn't work very well when the injectors get big and the MAF flows more air. I would disable the O2-MAF sensor adaptation algorithm. If that thing gets confused, AFRs can go all over the place.
- The LH doesn't have a fuel temperature compensation. If you have Bosch 044 (or, worse, and aftermarket "performance") fuel pump, the constant recirculation of fuel can increase its temperature. The fuel density goes down and AFR goes up. This is a slow moving error.
- Are the fuel dampers there? If the fuel rail driven by the LH doesn't have sufficient fuel damping, that creates resonant slightly angled but mostly vertical "stripes" in the Sharkplotter AFR visualization. It will be impossible to get rid of those errant AFRs within a cell, because the resonance is so sharply determined by (rpm, pulse width). It's a fools errand to try to tune the LH without very well dampened fuel rails.
- Is the WBO2 consistent with the NBO2 sensor in the vicinity of 14.7 AFR? If it's not, then you may be chasing a non-existent error. The WBO2 grounding etc. needs to be done in a way that is conducive to accurate measurement.
- Is fuel pressure in the rail constant when referenced relative to the intake manifold pressure? The LH assumes it is. The fuel heating and aftermarket "performance" fuel pump cavitating lead my car to have a permanently installed mechanical fuel gauge in one of the fuel rails.
- Can the cell high std. cell be accessed with or without the WOT sensor triggering? I don't recall off the top of my head if Sharkplotter adjusts for this, but if it doesn't then one point richer with WOT relative to part throttle isn't crazy, given how the LH ads the WOT fueling vector on top of the regular map.
- Finally, if you're about to drive off the spec range of the Bosch MAF sensor, then it stops measuring air flow past certain air mass flow. Unfortunately, this level of mass air flow where the sensor goes out of range depends on the operating environment.