Tuning and mathmatics?

[violacleff]

I'm not very knowledgeable on tuning stand-alones and what not. But I do know that to get the most out of them, you must tune on a dyno. I'm just curious why the tuning process could not be broken down into some kind of mathematical formula that eliminates the need for a dyno. For instance if you were to get a perfect tune on a stock l28et, and you upraded the injectors, turbo, fuel rail ect., couldn't the amount of fuel, air volume and timing be factored in to some sort of equation that takes turbo, injector, cam, and engine size into account and programmed into your ems accordingly? I'm not a physicist or anything but it seems like it would make sense that you could come to a near perfect tune that way. Am I way off in my thinking?

[Pop N Wood]

The theory will get you close. That is how the "open loop" mode works on engine computers. But there is a certain degree of randomness in everything. Not all injectors flow exactly like they should, parts have tolerances, engine heat up and cool down unevenly, not all gas is created the same, and of course air temperature, pressure and humidity are huge variables. Then there is the fact that one guy's idea of a "perfect tune" may be too conservative for someone else.

 

The best way to know if your set up in your vehicle is outputting the most it can is to hook up to a dyno.

[strotter]

When I got into the V8 Z thing, I was determined to use fuel injection, even though I had no experience with it. What really interested me was the computer control - I'm a computer guy from 'way back. As I got into it, I learned about the PROM "tuning cycle" - you drive around and data-log from your ALDL port into a laptop to get data about your motor's behavior under various conditions, then you go home and analyze what you're seeing, then you make modifications to the PROM, document the changes, and back to the beginning with drive around and data-log. What struck me after a while is that, first, everything is (or could be) done on a laptop; and second, that most of the changes you make when tuning, at least in the beginning stages, are trivial and rather repetitive. Did I say "rather"? Sorry, I meant "incredibly".

 

Naturally, being a basically lazy guy, and with some experience in programming, it struck me that a lot of that early "change it and check it" stuff could be offloaded to the computer itself. There have been some limited attempts at "computerizing" the tuning process (in particular smoothing BLM curves with spreadsheets), but I have been thinking more along the lines of a "tuning wizard". Something that you'd run on a laptop hooked up to the ALDL port, as well as to a "Prominator", "Romulator" or some other similar device that would (at least temporarily) replace the ECUs EEPROM with an "on-the-fly-modifiable" device.

 

In practice, tuning on a dyno is great, but The Road, She Is Different. Once you hook into the ALDL port, you have access to enough information to simulate a dyno, but at very little cost per hour. Much of the same information can be monitored directly (AFM, knock, acceleration curves, time-to-speed, a whole bunch of stuff you can infer from the data such as hp/torque, t/c slip, so on).

 

The program might even give the user instructions for things to do to address different aspects of the tune. For instance, "Start the motor and idle", or "Drive at highway speeds on a level surface", or "Accelerate at 100% throttle", so on and so on. All the while, the computer would be adjusting various values and testing them to within certain tolerances - while at the same time recording what changes were made to each value, in response to what input. (I have learned over time that keeping good notes is the #1 most important step in tuning (Thanks to Grumpy for that)). Needless to say, there are *many* subtleties which would be beyond the capabilities of any such program: you'd eventually have to get in there and get dirty. It would also have to include some error diagnostics, particularly in terms of mechanical problems with the motor - false knock comes to mind immediately, but there are many other potential problems it would have to know about. The particular algorithms would need to be, um, complex.

 

The big problem for me is that I'm a Mac guy with a Windows laptop, and I know *squat* about programming I/O on a PC. I'm going to be buying a MacLaptop Pretty Soon Now, so I might actually sit down and do some programming just to see if it's do-able.