A few weeks ago, EFI Analytics released a new version of their TunerStudio software suite for MegaSquirt EFI. Unfortunately, the update did not include adding auto tune features to the MAF flow curve. This is not an apparent issue for most users, as the preferred metering method is Speed Density, which uses an Volumetric Efficiency table to infer mass flow, instead of metering it directly via a MAF.

Having had not much luck with Speed Density (SD) systems in the past when combined with a larger aftermarket camshaft and turbochargers, I underwent a project to convert over to a purely draw through MAF configuration. This was mainly due to restriction on the exhaust side from the turbo, combined with the larger valve durations, causes the entirety of the air flow through the engine to become chaotic at low RPM. This does not lend itself to be modeled in two dimensions. the end result is an idle that ‘chases it’s tail’, or ‘surges’ around a set RPM as the intake and exhaust manifold pressures vary over time. You can keep adding fuel to the idle load cells, resulting in an overly rich, but stable, idle; but this quickly degrades catalytic converter operation, which is not acceptable.

I used a Bosch ‘slot style’ sensor for the MAF, as supplied on ‘05+ Fords. This was then mounted in a 3” housing, and plumbed to the turbo inlet via a 5’ section of 3” diameter double walled brake ducting, resulting in around 410 cubic inches of transit between the MAF and turbo inlet. No intercooler is being used at this point due to fitment issues. Crankcase evac is non assisted, non valved, and plumbed directly into the turbo inlet.

The first noticeable is that accel enrichment values can be reduced significantly compared to a purely SD configuration. Even with the ‘large’ volume of plumbing behind the MAF, its voltage response is quick enough to handle throttle tip in transients.

Secondly, tuning a function curve is much easier on the nerves than trying to dial in a VE table. Air mass is air mass is air mass, and the voltage conversion is consistent enough to not need constant fiddling.

Since TunerStudio does not natively support AutoTune on the MAF curve, we can ‘fake it’ by using the VE table as a trim table instead of a primary calculation. Under the ‘MAF Settings’ dialog, enable the option ‘Use VE1 as Trim Table’, then open VE1 and set all cells to 100.0. There are multiple references on the internet for slot MAF sensor transfer curves; i started with an off the shelf OEM curve for a 77mm / 3” slot sensor. Engine fired right up with the stock curves and settled down into a nice, stable, idle without too much fiddling. Megasquirt supports the ability to have closed loop spark advance at idle, but this was not needed. a few slow sweeps from idle to 3000rpm ended up with slight adjustments to the base curve, but nothing surprising or extreme.

Once the engine was warmed up, auto tune was enabled, and did a short jaunt around the block. The VE table is adjusted on the fly based on HEGO feedback; so as more data is collected, the better things run. within a few minutes, the ECU had sorted itself out, and the engine was happy running along under light and medium loads. the resultant VE trim table had a range of values from 85% to 110%.

now here where the number crunching comes into play. with MegaLogViewer HD, you have the ability to define custom data channels, and then fold that into tables. the trick is to create a custom field that applies both the HEGO feedback and the VE1 table feedback onto the original MAF generated mass flow rates. The forumula is as thus:

[MAF] * ([Field.egoCorrection] / 100.0 ) * ( [Field.VE1] / 100 )

Then in MegaLogViewer’s Table generator, set the view to be MAF volts vs TPSdot vs MAF Adjustment. Configure the X axis to match the MAF transfer voltage bin values, and the Y axis to cover the TPSdot values used in general driving. I use a semi log scale of [-250,-100,-25,0,25,100,250]. The resultant table is the adjusted MAF flow curve.

graph

Unfortunately there is not a good way that I have found to automatically copy this back over to the MAF flow curve, so there is a bit of a manual process involved. The good news is that this only takes two or three iterations to get dialed in extremely well.

Now if body work issues were this easy to fix, I’d be in a good spot.