Fuel Pressure Filtering

Hello,

I've got a k20 running on a Haltech Nexus R3, and added several extra sensors when I implemented this new ECU. One of them is fuel pressure. The default filtering on fuel pressure in the base map Haltech provides seems to be quite low (2ms)...I think...and it seems to have impacts to outside parameters like the duty cycle.

When I look at a WOT pull, I can see that the fuel pressure jumps around a lot because of this, since the injectors are firing and whatnot:

(Note: The sample rate on pressure differential here is lower than the fuel pressure and duty cycle. I assume it is equally jumpy.)

Is it typical/common to use a lot more filtering than this? Something like 25-100ms seems like it would produce a much smoother, much more representative of true average, value for pressure differential, dead time, fuel quantity, etc. values.

-Matt

This is something you should try for yourself. You may not need to increase the filtering much at all to see a significant change. Try 4, then 8, then 16mS and see what effect that has on your graph, and the resulting usage in determining injector pulse width, % DC, etc. You want to use the least amount of filtering possible so that you don't mask issues like pressure drop-outs (wiring or sensor issues), that could affect the calculations and/or performance.

I don't mind measurements that represent what is really happening. Perhaps a mechanical solution (larger fuel rail, pulse damper, different pressure regulator) should be deployed.

You've made a good first step -- looking at the data and asking questions...

Thanks for the helpful info! Testing it is. I was mostly curious because of the amount of filtering that the MAP signal gets. I've got a big Radium fuel rail with a vacuum referenced ramper on it, fed with -6AN hose from a Walbro GSS342 pump through a big Radium -10AN filter and regulated by a Bosch-style Radium regulator. I don't think there's a ton I can do mechanically, apart from switching to a different style regulator, but I'm not sure it'll help.

I'm sampling this at a really high frequency. If I zoom way in, I can see the individual pulses. They're not dramatically bigger than the pulses in the intake manifold as the cylinders go through their individual cycles, but maybe that's typical for injectors this size (around 800cc/min at this pressure):

I guess I'll make some small adjustments to see where things go. I kind of assumed that this type of input would ideally be as smooth as others used in the same calculations, like MAP...which is more or less used for similar purposes. Keeping the filtering low enough to not delay the signal, but high enough to smooth it out, seems reasonable though. Given that this isn't diagnostic data, but inputs for fueling, is there a case for it not to be as smooth as MAP (which gets a lot of filtering)?

--Matt

If you were to look the differential pressure (fuel pressure - MAP), does that look smoother? In your graph, there is about a 10-15 kPa range in unfiltered MAP, and about the same in the fuel pressure. Injector flow is determined by the differential pressure, and I'm sure there is not much difference in flow over a 15kPA range (~2psi).

If I sample it at the same rate, it bounces around by 10kpa as well. It's just smoother above because I had the onboard sampler capturing it 1/50 as often (20Hz instead of 1kHz) like the others. The flow rate bounces around by about 10cc/min as well.

When I had no fuel pressure sensor, and these values were calculated rather than derived, the duty cycle was super smooth through a pull like this. I think I just need to re-test with the filter increased incrementally like you suggested to see if things smooth back out.

Here's an example of the duty cycle on my old Elite with smaller injectors over roughly comparable conditions:

--Matt

Filtering was the issue here, but not for the reason I thought. The jump from the Elite series to the Nexus series changed the unit of the filters. While I had the same number for map sensor filtering, and a comparable one for fuel sensor filtering, one is in some proprietary Haltech unit and the other is in milliseconds. Once I got the filters set to what Haltech shows in the base map, the duty cycle (and resulting lambda figures) improved dramatically!