Mongo

Thanks for clarifying.

Is there still some safety room in these chips? Last chipset I threw in a car (944) busted some rings, and that was with new injectors and an L-Jet AFM. Bad chip, or someone reckless enough to over-advance ignition timing when burning it.

PorKen

In a sense they are 'safer' than stock chips. The new 'Octane-Adaptation' code and the self learning WOT ignition maps try to set the advance level just below the knock threshold of an individual engine. The stock chips' static maps rely on knock retard which reduces power quite a bit.

Fueling is 'better' too, with less chance of lean outs at wide-open or heavy throttle. With heavy throttle or over 3000 rpm it runs open loop in the 13's. Not having the O2 loop pulling it into the 15's plus no injector cutoff on decel makes running at high rpms much smoother. (Smooth enough that I don't mind running in 4th or even 3rd -manual- on the highway if I know I'm going to have to jump ahead, quickly.)

BC

You sent an interim ign chip correct? Last year?

Mongo

Ken, I remember seeing somewhere that these cars like high 12s (12.7 to be exact) at WOT and High RPM loads. Shouldn't the chips be realistically be achieving that AFR rather than 13s?

Also, what about compatibility with wide band O2 sensors emulating narrow band? For example, AEM UEGO and Innovate LC-1?

PorKen

RE: 13's - better said, heavy-ish/high-medium throttle or light throttle/high rpm. Not heavy enough to be considered WOT, which is tuned for mid 12's. In either case it stays in this 13-ish 'Sport Mode' until load is reduced to a level highway load again. The stock chip will go open loop at heavy-ish throttle or higher rpm, but it drops into the cruise map, which is set for high 14's.

I prefer using a stock O2 sensor for the LH input for fewer problems.

PorKen

After an exhausting winter of coding...individual cylinder wot ignition mapping, sorta.

Two groups of four cylinders in 60 column maps (retard/cylinder list @ rpm). Given the semi-permanent 'learned' memory restriction (just 120 bytes), knocking cyls have to go by the worst offender at that rpm, but not all cyls knock at the same time so many have little to no retard. #6 really drags down the curve but it is in the knock prone inner cylinder group 3726. In contrast #4 almost never knocks in the outer group, 5481. (Firing order is 13726548.)

Still boiling it down (again) to be chip-worthy, but it drives oh-so-good. The 'rough draft' graph below might give you an idea of the additional TORQUE. Most previous WOT advance maps have pretty much looked like the red line, which is the ideal (max possible) map for 3726. (The black line is the stock WOT map.) The green line is the ideal for 5481! Turns out those cyls can handle 3~7 degrees more advance - except at low and high rpms where they are roughly equal.

Individual engines learn further by group/rpm/cyl as needed. (Most cylinders in both groups retard more at different rpms in response to knocks so really only #4 uses the green line entirely, for example.) At WOT, the stock time delay based retard has been completely replaced by this rpm based learned system which further increases the average power level with no knocking. (After an initial learning period. Stock based all-cylinder tuning virtually guarantees that at least 6 and 7 will knock every WOT pass.)

One more trip to the dyno, next week, and it should be ready to go. It's been dynoing a steady 315tq/305hp rwsae using something like the red line...'reckon it'll do a little better now, at least fill in the curve a bit.

Ignition advance by cylinder group vs. stock all cylinders

(displayed rpm columns are not exact after rescaling maps for better resolution)

ptuomov

Ken -- Which of your findings are consistent and which inconsistent with my simulation results?

I mean specifically the knock risk plots (maximum unburned mixture temperature) in the below post:

https://rennlist.com/forums/928-foru...l#post12008397

PorKen

Tuomo - I'll look at it in the morning. Flu = tired. (My post above probably does not make much sense. )

Eventually when things settle down I may go through and record knocks by individual cylinders, possibly make base maps for each. That might answer your question better.

PorKen

Note that the knock threshold for me was such that I had to make two base maps or it would overload the limited learning memory.

After repeated WOT passes the inner group would all have knocks at almost every rpm column using a map that the outer cylinders could handle. (#6 can't handle most of the red map in varying degrees, either. #6 likes the stock map.)

BC

So remove the possibility of knocking and your maps will keep the advance you set?

PorKen

Couple of notes from this (overlong) programming/winter season on the EZK...


All of the EZK versions, S4-GTS, have diagnostic code which looks to only be for factory testing. Not sure if it's for proving the ECU itself or perhaps for chassis dyno testing of the completed car. Even the early '87 brain with the most compact code (fits on a 8K chip, vs. later 16K) which does not have regular 'Hammer' type diagnostic capabilities/code, has this factory serial code. The code allows externally reading and writing of some variables, including rpm, load, temp2, voltage, knock level, knock sensitivity, etc. It's mostly called from the section used to calculate the advance time and dealing with knocks. There's also a separate counter to keep track of the current cylinder. Since this code is unused, I removed it to recycle it's codespace/variables/bits.


The 'inner' group of cylinders, 2367, pretty much all knock eventually at the same advance level along the red line above but for #6, which requires a few degrees less. 2367 are all fed by the long straight runners. The 'outer' group, 1458, have some variation under the green line, in order of frequency, 1-8-5 with #4 not bothering to knock anywhere. 1458 use the shorter 'noodly' runners. My working assumption is that the lower advance needed/allowed by 2367 is because they are filled better = higher compression. It has been said that the inner cyls run hotter which might beget a cycle of knock, retard, less advance makes cyl run hotter, knock, etc.


It perplexes me why the factory map does not have anything close to the possible advance around 5000 rpm. Granted, there's not a lot happening there, power wise, so it takes a lot of advance to make it knock. Conversely, it takes a lot of advance to move the dyno needle at all in that area.


I found no difference in power output in dyno testing between the RMB I made from a factory muffler and a stock factory muffler, except for the weight. Good, because I've gotten tired of the sound of a RMB and either the small or large mids. X-pipe, small mids and factory rear muffler is my fav, plus I find I actually drive faster when it is quieter.


I like my S4 much more now with the new ignition chip. Combined with last year's fuel programming (Sport Mode, no decel fuel cut, etc.) I can nonchalantly buzz along at higher rpms, on the freeway in 3rd, or toodle along in first from a stop light, at sub-light speed, then floor it for a near silent warp-drive. Maybe it is quieter because more fuel is burned in the cylinder and the energy put into the driveline instead of dumping out into the exhaust and/or the pulses are more even for the resonators to soak up. (Or, my hearing is gone from the RMB and small mids.) Whatever. I just like flooring it all the time, now. Less: bwaaaaah, more: woooosh...

PorKen

Stockish S4s (mine at least) really want to make this same graph! (I can show you a hundred runs...) It can make a bit more (over 320 TQ/310 HP SAE) but only if the engine stays cool and my dyno place does not have a good rad-fan setup. At 5000 rpm it starts to get hot - that's where the graph starts to get bumpy. I have made the chip really aggressive about throttling back the timing by coolant temp or knocks. (At worst, when it's real hot, the graph looks close to the X-pipe graph.)

Two years...of programming has made it super consistent and eminently responsive plus expanded the torque in a few sections, but in the end there were no huge untapped areas. The fueling is an extremely flat mid-12's in almost any situation (90% inj duty at 6000). The ignition timing is custom for each cylinder. There is really nothing left to mine that can be done with tuning.

It feels great to drive, of course. The chip always tries to maintain the highest power level it can, safely. The custom cylinder timing makes it pull like a thoroughbred whenever you hit WOT. It drives so smoothly, seeming effortlessly at high rpms. You can pretty much drive all the time in 3rd (manual) from 20 to 120.


(STanDard), (DIN), and (UNCorrected) show how it feels to drive in PacNW fall weather. (SAE is a B.)


1988.0 S4 manual, S4.S300s V1.1, 24# injectors, SLC X-pipe, stock small mids/rear muffler/air filter/cam timing, WR5DC plugs, 92 octane

Ketchmi

Those chips look like the ultimate compliment to my crossover, almost makes me want an S4 without boost!

PorKen

I thought too late (and after hundreds of runs between S, S3, and S4) that I should be testing in two passes to better show what street driving performance might be. Duh. One for torque at 1500~5000 and a separate one for HP from 4500~6400 with a cool down period in between. All this time I have been fighting the coolant/engine getting too hot at the end of a run to have meaningful HP results.

Oh well. As an aside, you know it's bad when your dyno guy (who knows you by name when you call) asks if you might be dyno'ing too much.

FredR

Ken,

Awesome work. Besides the obvious increments, an often overlooked [comparative] parameter is how the torque curve is extended. If one considers the second [flappy] torque peak at 4250, you are still making that amount of torque 1000 rpm up the curve. Not sure how that translates on a manual tranny but on the auto box quite significant I would say given the way the stock auto box "dies" in top gear. I would guess it is almost like having a 5 speed auto if that makes sense. Beyond that in the dip at 3600 you do not go below the stock torque peak 4250

Stick a good set of headers and a set of Colin's cams on top of that and I wonder what the resulting curves might look like- any offers?

Rgds

Fred