Forum » General Tuning Discussion » MLVHD - Transitional Areas - Best Filters?

MLVHD - Transitional Areas - Best Filters?

General Tuning Discussion

Forum Posts



Tech Articles

Discuss all things tuning in this section. News, products, problems and results. 

= Resolved threads

Page 1

ECU: AEM Infinity

Car: S2000

I've used MegaLogView HD successfully for low-rpm (850-5000) part-throttle tuning & all of WOT throttle tuning, however, transitional areas (30kPa-80kPa) at higher rpm (6000-8000) are just not accurate.

MLVHD suggests I am seeing 0.70-0.90 LAMBDA in those high-rpm 30kPa-80kPa areas however if I get into a steady state at say 6500RPM & 50kPa I am really at stoich or much leaner.

I've done the following and it made no difference:

- Accel enrichment is disable

- Decel cut is off

- DataLoggin @ 20Hz, so hit counts are astronomical and long data logs are monstrous (1GB +)

Also, MLVHD filters that I've created:

1. Differentiate between VTEC & non-VTEC data

2. Filter-out any gathered data with throttle rate above 50 (again, accel-enrichment is OFF)

What other filters would you recommend for improved data accuracy? I have a lot of gathered data/hits to spare.

Many Thanks,


I think you need to filter on RPM not changing rapidly as well. You are likely getting the data when the throttle is closed (rate not changing), but the RPM is dropping on overrun -- that can be filtered out. In fact, you might just filter when throttle is greater than 50% to make sure. Also, pay attention to how many samples are in a cell. You need a lot to have to outliers ignored and be confident it is the correct change.

Sometimes you need to focus on the graphs and find a good sample point, where you are steady state at a fixed RPM & load for a few seconds. Adjust based on that Lambda.

Thanks David,

I've tightened down my throttle rate filter even more. Filter is now below -10 and above 10, and all cells that didn't seem correct even though they had 100s of hits... had disappeared.

Now there are blanks, thus confirming it's the data WB was capturing in off-throttle/overrun rich condition, maybe? I think that's what you meant when you said "on overrun".

Looks like we can't blindly trust logs even if they're vast, with filters and 100K of hits... which I didn't so that's good.