ZX-Tex

iTrader: (15)

So I have been having problems with a noisy MVSS signal, noisy to the point where the s/n ratio keeps it from being useful, especially in the lower gears. I think some others here have this problem too IIRC. Per a thread on the Adaptronic forum the noise comes from the reed type switch (?) on the Miata transmission sensor. Per Adaptronic's suggestion, I tried running a capacitor between the MVSS input into the Adaptronic and ground; I used one of the Adaptronic grounds.

So far I have tried a 1000nF capacitor and am still getting noise. I also tried a 470 uF capacitor (much larger) and that seemed to kill the signal completely. I am thinking the happy place lies somewhere between these two.

So before I scrounge up a bunch of capacitors and keep trying this, I am wondering, am I headed in the right direction? Or is there a better way to go about filtering this? I know very little about analog filters so I could use some guidance here from an EE type. I think the idea is to average out or dampen the value so the peak-to-peak noise is reduced. The signal appears to rise and fall with vehicle speed, but there is just a lot of peak noise on top of the nominal value.

TurboRoach

iTrader: (3)

That's the easiest way to do it, unless you want to actually figure out the frequency of the noise. Then you could be a little more exact in your selection of the cap.

Post your log so see we can see how noisy it is. I might even do the spectral analysis for you.

ZX-Tex

iTrader: (15)

I suspect the noise might be at a higher frequency than the sampling rate of the Adaptronic log (Nyquist) so there might be some significant quantization error

ILoveOffRamps

iTrader: (2)

I don't suppose you have anyway to know the frequency of the noise?? If so, you can do the math to solve for sizes.
Since you think it's high frequency use a series low pass filter.

Set Desired Frequency.
Choose R or C (whichever you have lying around) Solve for the other.
R=Resistor
C=Capacitor
w=Cutoff Frequency
R= 1/(wC)

Example:
w=3kHz (convert to krad/s) = w*2*PI
C=1uF
R=1/((2PI*3*10^3)(1*10^-6))
R=53 ohm

Not sure if that helps or not.

TurboRoach

iTrader: (3)

Well then what are doing asking for a different different way if you don't think we can determine the noise freq?

ismael_pt

iTrader: (1)

I suck at electronics, but what is the purpose of the BAT54S diodes in the NBB PNP loom? The MVSS has one on it.

ZX-Tex

iTrader: (15)

I meant in terms of the circuit itself. As in, does the capacitor to ground idea have merit, or should I have a resistor in parallel with it, or with the sensor, or consider something else entirely etc. etc.

Sorry, not trying to be difficult. But, according to Nyquist, unless the noise is at or below 2 Hz (logging is at about 4 Hz) then the log is not going to help. Maybe another way to look at it would be to determine what the max possible pulse frequency is from the sensor (6th gear, 7500 RPM) and then try a low-pass filter tuned to somewhere above that, if that is what you are thinking? That is assuming the noise is at a frequency above the max sensor frequency.

I picked up a bunch of varying value capacitors and a set of alligator clips so I can swap values quickly. I'll post up as soon as I have some results.

TurboRoach

iTrader: (3)

That would work. If the noise is below that you won't be able remove it w/o out killing the signal or doing some spectral analysis. Which might be a bit difficult to do in analog.

I know you're not being difficult I was just joking.

I want to try some stuff I've been learning about this quarter, can you post a log for me anyway? I don't have a speed sensor.

If you're going to use a resistor put it in series with the speed sensor and the cap between the sensor and ground. Adding the resistor will let you get away with using a smaller cap.