Last year I purchased my first Corvette, a 1996. Since then I have been doing a lot of reading about numerous aspects, most recently a book about tuning. As an intro, I began computerized control of processes in 1980 and attended a number of courses on the topic; the concepts and terminology have been around for a very long time. So my comments are prefaced as someone who understands computerized process control, and trying to learn more about the 1996 Corvette in general.

First, it took me a while to understand that the auto industry appears to have borrowed conventional terminology but completely misapplied it, so I was confused until I realized this. Terms like PID, histogram, etc are not remotely the same in the rest of industry. And the use of terminology is inconsistent. for example tuning and calibration is not the same. But now that I realize that, I have altered my interpretation to suit the auto world.

A significant aspect of tuning is actually calibration, primarily the MAF. Meaning the initial effort is to achieve an accurate calibration of mass air flow measurement by running the engine across the full range of flow, and correcting the results based on measured O2. In this manner the desired air:fuel ratio (Lambda) can be achieved via the fuel injectors. So when alternate air:fuel setpoints are being evaluated to provide additional power, it is suggested to alter the calibrated MAF calibration based on the new desired air:fuel to achieve the desired result based on the Lambda measured via O2.
.
This seems very convoluted to me; since the MAF was initially calibrated to assure accuracy, the inability to achieve different desired air:fuel would appear to be most likely associated with imprecise fuel injection, so why is that not also properly calibrated first? Then any new air:fuel ratios can be evaluated without MAF calibration every time.

Given the short and long term trimming of MAF based on O2 measurement, how does this interact with intended tuning?

As you can see there is much I need to learn, and an specifically interested about how this is implemented on the 1996. If you can provide decent references I would appreciate it, as I have been unable to find any via Google.

 

I'm in a very similar situation. I have a 96, just bought JET DST and trying to learn the basics of ECU tuning. I have degrees in physics and nuclear engineering and have encountered the same problem regarding terminology, and sometimes blatantly incorrect statements from YouTube instructional vids. In the automotive world bullshit runs rampant; just search this forum for the "HP vs Tq" topic and you'll see what I mean.

Anyway, I'm not sure this is helpful for your question, but the PCM file is a only collection of lookup tables. The injector characteristic is an assumption based on average performance of that model injector. If the performance of that injector degrades, how is the PCM to know? A simple feedback loop with the O2 sensor will indicate that something is wrong no matter what the source is, and PCM is able to correct it to a certain point. Obviously a perfect injector calibration set would be best, or even one O2 sensor per cylinder would be great too, but complexity and cost would increase.

 

There are a couple of books by Banish (I have Amazon Kindle version) which provide a decent overview from an engineering perspective. You are correct regarding lookup tables, which seem very archaic and limiting vs utilizing transfer functions and algebra. Instead, if simple models were utilized then the possibility of normal PID with self tuning capability becomes an economic possibility. Walmart sells self tuning controllers for $50, it's not rocket science.

I would be interested in learning your experiences as you progress.

 

You have to remember, this was still in the relatively early days. By 2004 (maybe even earlier) (using this as an example because of my diesel truck), the ECM can control individual injectors. It will add or take away fuel as needed. I monitor these "balance rates" to make sure I am not losing an injector or they are getting dirty.

Each time they added better computer chips, they were able to do more. How many of you understand the progression from 8, to 16 to 32 to 64 bit processing ? Each time the chips got
better, the OS's could do more and do it faster.

 

I understand your point concerning processing speed. But doesn't calculating a required change using an equation require less memory and time than utilizing a lookup table? I am starting at the bottom of the learning curve in the auto application, and was surprised that the terminology and approaches seem very disjointed vs what the industrial world was doing at that time.

 

There are an awful lot of variables to be programmed into an equation. Remember they were only 4 years into dealing with the OptiSpark and what they were trying to do with positional crankshaft timing etc etc, lots going on and a computer can only do 1 thing at a time, it just does it very very fast. Sometimes it is just easier and quicker to fetch it out of a table and easier for other people to understand.

Also easier to debug than a very complicated algorithm.

It takes a lot of processing power to do lots of calculations and remember, it has to be dealing with all the other functions, timing, etc.

Those folks were learning also how to do it.

 

The history is what it is. If you have any recommendations for resources that would help understand the calculations which are utilized it would be appreciated.

 

Auto manufacturers also have the challenge of reusing as much hardware and software as possible across many applications to reduce development time and cost for individual solutions. Not saying that GM couldn't have applied smart controllers, just that there is many factors ($$) that go into why a manufacturer choses one path ($$) over another.

 

https://www.lt1howto.com/articles/pcmtuning.htm

http://www.lt1pcmtuning.com/tips/

http://translatedby.com/you/the-basi...ning/original/

 

Agreed, a LOT of the "free" (but laced with ads and constant begging for donations!) YouTube content has glaring technical errors. Just because a video has lots of hits/views doesn't mean it's accurate.

We work really hard to make sure our technical training content is 100% on point. The good news is that physics doesn't change, so many of the lessons can applied to any ECU. My books are over a decade old, but still apply to cars rolling off the assembly line today for this very reason. We also have a bunch of video content on our streaming site (see my sig), some of which is even FREE, none of it has ads.

 

Thank you for joining this discussion, and I hope you will consider answering a few questions. After a bit of Google and Youtube research I came across your inputs and purchased your book which was very informative. This and several other apects led me to question:

1. From the factory I assume the ECM is loaded with generic calibrations. You stress that calibration should also lead to smooth response curves such as MAF vs hz but from the factory are these as irregular, unsmooth, hatchy as the examples I typically see? Or is this the result of the long term and short term adjustments done by the ECM?

2. Or is the long term / short term adjustments stand alone loops which do not affect the calibration curves?

3. The calibration example in your book discusses achieving an accurate, smooth MAF relationship based on adjjustments using lambda=1 which is sensible. Later you discuss developing additional horsepower by evaluating richer lambda values, and during that process redefining the MAF curve again to achieve the desired lambda across the the range. On this point I am confused since the first step assured the MAF was accurately calibrated, so it would seem that if a different lambda could not be achieved that this would be associated with an inaccurate level of fuel injection? Why not calibrate the fuel as well, or is that not possible within the ECM?

These and many more questions arise because I have not been able to find a resource which describes how all of these control loops are configured to operate alone and in conjunction with each other ( I personally like flow charts as utilized in typical computer programming). Does this exist, and can you recommend a more detailed resource?

Thanks.... Charlie

 

The factory MAF curve IS very smooth (equation fit, actually!), but is sometimes clipped at the top of the curve for limits in the ECU code. Older PCMs cannot "count higher than 512 g/s" for the airflow variable because it's the limit in the hexadecimal code for that variable. Going past this (in the physical world) like when you add a supercharger requires some creativity to work within the PCM limits like we show in our scaling exercise videos.

You will occasionally see "rough" OEM numbers in things like the PE multiplier because the actual target ratio is calibrated at each point to either LBT or the minimum required for thermal control. This is by law for OEMs, not necessarily what they would prefer. In the performance world, we often smooth this out.
Fuel trims are an after-the-fact correction to what's going on, that are independent of the original feed-forward commands that come from the calibration tables. Ideally, you get your base maps (MAF, VE, spark) correct and there will be almost no short/long term correction required unless something changes.
My actual open loop correction factor is (delivered wideband lambda)/(target lambda), so when the target ratio is 1.00 the math gets really easy. But when you go WOT and target a richer ratio, the same correction method works. If you do this for all points along the MAF curve, some will get corrected when you are at stoich and others during power enrichment, and some in the middle under both conditions. But in the end, it should drive you to one continuous curve. I do all of that with the trims disabled so I can concentrate on the base table(s) without trying to hit a moving target while trims are active. After I'm done, the trims are re-enabled but usually don't move much.

While I don't have flow charts, I do outline the process in the videos with more detail than I was able to put in the books. Don't be turned off by the fact that they were done with LSx engines because the basic GM software was pretty much the same from your C4 LT1 all the way through the new C8 LT1.

 

I'm really interested in this thread as I look to potentially tune my '94 LT1 six speed.

If the lookup table is large, and the equation simple, then the calculation itself could require less memory than a lookup table. But, a lookup table is generally used because it is an extremely efficient way to reduce the computation time. A lookup table allows the CPU to index into a table and get a result within a few cycles (sometimes one, sometimes a few depending on the architecture). A calculation on the other hand, requires the processing of tens to hundreds of low level instructions to execute the instructions for that calculation and also to fetch the desired data from memory, in order to achieve the same result. There's tremendous flexibility when a calculation is coded, but it always comes with the trade off of reduced computational speed. The CPUs in the late model C4 are extremely slow by today's standards. Using a lookup table for a defined set of input/outputs was an excellent choice for the time and budget.

Joel

 

Thanks very much for your reply Joel, I will study your videos for answers before quizzing more.