So my 100% stock FD is now slightly less stock. The factory muffler had a couple rust holes in it, so I ditched it and put in a Racing Beat muffler that I ordered from Goodwin.





It fits perfectly and sounds good. Not much louder than stock (well, at least not much louder than stock with a hole in it), but deeper.

Signed up for Laguna with GGLC on July 18th, anyone else going?

--Ian

 

In the NB or the FD?

Do they have spots open? I'm not a member, but it would be really nice to get some more seat time.

 

I'm taking the NB. The FD is not going to the track.

I think beginner is full, but there were intermediate spots the last time I checked.

--Ian

 

IS THAT A STEEL PPF? wtf?

muffler looks good. kinda shiny indeed.

 

Yeah, the FD has a welded steel PPF. They're somewhat infamous for cracking under heavy use.

--Ian

 

Wow, it's been a month since I posted here.

So a few weeks ago TK and I did some VE tuning after the valve spring replacement. The table was overly rich through most of it, VEAL pulled about 5-8% fuel in most of the table, except where it added 2-3% in the top-right. I hope that means the valve springs fixed the valve float and improved VE up there, but will need the dyno to say for sure.

Went to Virginia/DC for a week to visit my wife's sister and take the kids to the Smithsonian. I love the Air & Space museum.

Today was the GGLC track day at Laguna Seca, prepped the car yesterday and hauled it down early this morning. I came in off the first session and noticed that the LCD display panel was showing that AFR was flatlined at 22.4 -- uh oh. I pulled the data out of the DL1 and graphed it, to see that it had failed on lap 8, showing a pattern where the richest value it could display got progressively leaner and leaner over the course of about 15 seconds, before it flatlined for the rest of the run. It looks like this:



So the wideband sensor packed it in. OK, that sucks but in theory the car is properly tuned so if I set it to run open loop I should be fine, right? Well, while poking around the data I noticed something else worrying:



This chart is a little hard to read, but it's fuel pressure vs lambda vs time for lap 6. Look at 53 seconds into the lap, see where the fuel pressure drops down and lambda spikes up? It's at WOT and 15 psi there, and it leaned out to 15:1. Ulch, I turned pale when I saw that. Something in the fuel system is *not* working right, and I wasn't going to be diagnosing it at the track, so the car went up on the trailer and I came home. :(

So the reason I'm posting DL1 charts here instead of MegaLogViewer is that my MS3 doesn't have an RTC in it, so data logging to the SD card in it isn't very useful because it can't timestamp the logs correctly so it's almost impossible to figure out which one is which. However, last year I installed a CAN-bus-to-DL1-serial-protocol converter, so I can send a dozen values from the MS3 to be logged on the DL1 along with the accelerometers and GPS data. Once I got home I exported the run into a CSV file and loaded it into MLV to look at it a bit more closely:



So a few weird things about this graph. 'BoostPressure' is actually MAP, but it's in bars instead of kpa, so '2.0' is 200 kpa (15 psi boost). I don't have actual RPM (something wrong with the way that's getting logged, so in the trace it's flat at zero for the whole log), so I'm calculating it off the GPS speed assuming 4th gear. I lay into the throttle at 4300 RPM, boost and fuel pressure come up, AFR/lambda is a tad lean at 12.2:1 (should be 11.75). Then fuel pressure starts to sag, AFR gets leaner and leaner over the next 2 seconds, peaking at 15:1 (!), before whatever the problem is spontaneously fixes itself at 5000 RPM and it snaps back to 11.5:1 for the rest of that straightaway.

The fuel pressure regulator is manifold-referenced, so in theory it should always be manifold pressure + about 32 psi. The 100Hz logging of the DL1 shows me the lag in the pressure value, so there are a bunch of transient errors in the chart, but most of them are a spike high, followed by a spike low. The error in this graph is deeper, longer, and only goes in the negative direction, there's no high spike before it. There are a whole bunch of these in the single log, getting more and more frequent and more severe as the session goes on.

Unfortunately, I don't have any record of the knock sensor output in my DL1 log, nor of the knock-based retard. I do have the total spark angle value, though, which doesn't show anything too outrageous. I didn't *hear* any ping, but I'm pretty sure I wouldn't have with a helmet on and the top down at 80 mph.

The fuel system is a walbro 255 with an FM "big fuel kit" AEM FPR mounted where the stock fuel filter goes, non-return -6 AN line running up to the fuel rail. I've never seen anything like this in street tuning or dyno tuning, and the fact that it gets worse over the session suggests to me that it's heat related. With the manifold-referenced FPR, the pump is working harder at higher manifold pressures (trying to put out 60 psi at full boost). I'm still using the stock fuel pump wiring, which suggests to me that I'm overheating the wires, resistance is going up, fuel pump voltage is dropping, and that's what's causing this. I'm inclined to blame the pump rather than the regulator, although it's hard to articulate exactly why.

Thoughts?

--Ian

 

Oh, when I pulled the car into the garage, the AFR reading was back to normal, so whatever failed in the LC-2 spontaneously fixed itself as well. Wonderful.

--Ian

 

Other random data, the car does half a gee of acceleration in 3rd (at 200 kpa), 1.4 gees lateral (brand new RRs) and 1.0 gees braking.

--Ian

 

Thats sketchy. I agree with your attempted fix idea though. Before you do that I would try and reproduce it on the street. So you aren't chasing issues at the track.

 

I blame the pump because it's a more complicated device than the regulator. The pump has gears, coils, brushes (or probably hall effect sensors these days), and lives at the end of a long run of wire, all of which can fail in intermittent ways. The regulator is a ball and spring and should fail less frequently, and in smooth, linear ways when it does.

How much fuel was left in the tank at the end of the session? Could it have started sucking fumes? Maybe you need a swirl pot. Or maybe the low fuel level encouraged the pump to overheat.

 

The tank was about half full at the beginning of the session, so would have been about 1/4 or a bit less at the end. I'd forgotten to fill it up before loading on the trailer, and if I'd done another session I would have needed to fill it up. Miatas are not generally known for having fuel delivery problems at that level, though.

--ian

 

Scatterplot of FP error vs time. The start and end are in the paddock, but it's clearly getting progressively worse while on-track.



--Ian

 

Yeah, I'll go out and see what happens on the street, although since I've never seen it before I'm not optimistic. The first significant excursion in FP error is on lap 4, I really doubt I can reproduce that level of load on the street without getting arrested. At least, if it's related to heat-in-the-wires. If it's fuel level related then it should be obvious.

--Ian

 

OK, maybe there is something to the fuel level idea. I looked at the lateral gees in relation to the fuel pressure issue, and there's a correlation. Mapping it in the DL1 software, it's happening when powering out of turns 5, 9, and occasionally 6 with .3 to .4 gees of lateral acceleration on the car, and the problem resolves itself when the lateral gees go closer to zero. It doesn't happen on every corner, only those three.



--Ian

 

OK, that chart is a bit hard to read. The bottom data trace is the fuel pressure error, calculated as (fuel pressure - ( MAP + base pressure ) ). In theory it should be zero. In practice there's lag in the regulator so a rapid change in MAP will create a transient. Shifting is the major cause here, and manifests as a brief positive error (too much fuel pressure as the FPR lags the manifold going into vacuum), followed by a brief negative error (too little pressure as the boost comes back on). I'm not concerned about that, I'm interested in the longer, sustained negative errors with no positive error before them. I've marked up the graph in photoshop to label them:



(And yes, those laps were slow at 1:53. I was stuck behind a Porsche for most of them. :-) )

So now it looks like fuel starvation in left-handers. I'm also noticing a small effect coming out of 2, nowhere close to as bad as the other three, but present.

--Ian

 

Proper data logging and analysis...makes my heart warm.

 

Oem sock on the walbro?

 

Yes. Well, it was, it's theoretically possible it fell off. Need to open up the tank and take a look.

--Ian

 

This is like Chinese to me.

 

I'll take a general tso and small wonton soup