Best data logging rate

[tminer]

What would be an ideal sampling rate of ecm data? Something with enough detail to tune properly, but not so much as to be cumbersome. Or would it be better to have two rates, one for general use and the other for special occassions when one really needs to drill down?

Thanks,
Tom

 

[Doug Jacobson]

Their seems to be only a few options for getting data from the ecm. If you go through the ALDL port (Turbo-Link, Scan master, Tech Scanners etc...) The fastest roughly is 1 time per second, With the scanmaster fast update code it's a little faster. Or if you plug directly intot he ECM like a Direct scan you can update 18 times per second, I like the 18 times per second better. Once you go DS you'll never go back.

DJ

 

[bruce]

Originally posted by tminer
What would be an ideal sampling rate of ecm data? Something with enough detail to tune properly, but not so much as to be cumbersome. Or would it be better to have two rates, one for general use and the other for special occassions when one really needs to drill down?

For drivibility tuning 20 frames a sec is fine.
If you really want to see what'd going on during a pass 40 wouldn't be too much.

What would be neat is different sampling rates for different items. TPS, and Coolant could be 10 sec, and the a WB, RPM, MAF/MAP, at 50+ times a sec.

I've seen some 1 msec sample rate stuff and it's obtuse. While you can watch individual cylinder events that's really oem stuff.

What would be a cool addition is an actual torque sensing sensor, and see what the engines really doing.

 

[tminer]

Thanks.

I was hoping that DS was overkill, but I guess not.

Tom

 

[TurboTR]

It depends on the bandwidth of the signal you are trying to monitor, as well as what you are trying to do with it. If you are trying to catch and faithfully reproduce a waveform, ala a digital oscilloscope, then you must sample the signal at least 2x the bandwidth or highest spectral content of that signal. This is known as the Nyquist rate. If for example you have a periodic waveform (and it's not a pure sinusoid), say the switching O2 sensor square wave output during normal cruising then that signal bandwidth is NOT, repeat not simply equal to the switching requency, it can be much higher actually. How much higher depends on how steep the edges are. In general a step-like change is a high frequency event. You have to know some things about Fourier theory to be able to establish what the real bandwidth of a signal is. The problem that arises from violating the Nyquist rate is called aliasing, where higher frequencies become folded down or aliased into lower ones, and the resulting picture is then false. Thus sampled data systems generally need to also employ anti-aliasing low pass filters somewhere in the front end to ensure that the signal bandwidth remains below the desired range. Again, depending on what you are trying to do.

Digital oscilloscopes can put up some very misleading pictures, and the the savy user needs to be very aware of some of the fundamentals

For most of what we do 10 Hz is probably adequate. Again unless you are trying to faithfully reproduce a waveform that has frequency content higher than about 5 Hz. That's the sampling speed myself and many others use with the FAST box for example, and for relatively slow signals like overall a/f ratio, air temp, spark timing, rpm, etc it seems to be plenty adequate to tune with.

Hope that helps some

TurboTR