Jespers Mercedes 190 2.3-16v Cosworth.
The car is initially built as a normally aspirated car but pretty much everything for a turbo engine is gathered too.
The VEMS unit in this car is custom built on a v3.1 board with a custom cooling solution to make it easy to service.
Trigger problem:
The car had an extremely noisy trigger signal, after a lot of investigation it seems like the 50ohm antenna lead was far from a good match. The proper cable should be 600-1000ohm and with the 800ohm sensor impendance the resonance wasn't dampened much.
The new setup will use suitable cable and will be terminated at both ends with the right impendance.
The shape of the trigger signal was also very unfortunate, a 50 degree long vane trigged a negative pulse when reaching the sensor and then a positive signal when leaving the sensor. Acording to the scope pics in the service manual 90% of the time the sensor doesn't send a signal out. That means that it is very sensitive to noise during this time as the chip is armed and waiting for a trigger for 130degrees!
Possible solutions:
Alternative1:
- Adding a 1k termination resistor inside the ECU.
- Using a voltage divider with 10k and 1k(also replace the terminator) to gnd to bias the signal toward 0.25V (don't forget about the sensor resistance.)
- Using cable with the right impendance, telephone cable is about right but we need something that is more rugged.
Alternative2:
- 1k termination, the correct wire.
- A series capacitor between the terminator and the ecu.
- A 100k-200k/10k bias network on ecu side of series capacitor
Alternative3:
- 1k termination, the correct wire.
- a HP network built from a 7u series capacitor and a 1k resistor to ground. A 10k pullup to 5v is also connected after the capacitor. This will work as a terminator (not for low frequencies, but termination is not needed at low frequencies.)
Alternative4:
Could we just snub the high frequency part and forget about the impendance matching?! Possibly combining it with the above solutions.
Stock waveform and possible solutions:
Obviously, there is a series capacitor somewhere, way too low value for the given frequency. Eliminate the cap, or make it bigger and it'll be OK.
To confirm this, you can measure resistance or with diode mode (over the ends of the cable) and find that there is no galvanic connection.
Transmission line talk from IRC:
_jorgen: I have a problem with a Mercedes with a VR sensor with 800ohm resistance. It false triggs as a mf. The guy used 50ohm antenna coax for the signal wire. Could that be a bad idea?
GRMracer_: type of cable, freq range etc.
GRMracer_: yes.
GRMracer_: its actually better to use phone cable for that
GRMracer_: its 600phms
GRMracer_: 600ohms
GRMracer_: esp at trigger frequencies that are near audio.
GRMracer_: phone cable is twisted pair.
_jorgen: I was thinking about that but as I only based it on the fact that the ohm and frequency range was similar I wanted to check it out a bit more
mcel: not in a way that's needed
_jorgen: Do you mean telephone or microphone cable? :)
GRMracer_: coax is considered sheidled as if used properly all the E-fields are contained in the cable.
GRMracer_: I meant twisted pair thats in the wall of your house
GRMracer_: for the telephone.
GRMracer_: twisted pair is good.
GRMracer_: twisted pair that is sheiled is better
_jorgen: Microphone/audio signal cable should be perfect as the good ones are twisted pair with external ground shield. The impendance should be about right too.
GRMracer_: yes.
GRMracer_: it is defiantely much better
GRMracer_: you had a big mismatch with the coax
_jorgen: And the PA type cables are almost indestructable.
GRMracer_: 50 ohm --800phm
GRMracer_: 800ohm
GRMracer_: bad reflections in the cable.
GRMracer_: as it the signal bounces back and forth .. that portion that gets thru.
_jorgen: This may be worth noting. In most applications there is only a very small window where it can be trigged because of the ARM threshold in the VR chip. In this case there is a long period of time where the signal is at 0v.
_jorgen: Two 50deg long vanes 180deg apart.
_jorgen: The scope image is very strange
_jorgen: The scope image from the service manual
_jorgen: Negative peak - flat for 40 degrees - positive peak - flat for 120degrees.
_jorgen: Should I terminate the wire with 1k on the input of the box too?
GRMracer_: it depends on what the impeadance of the input to our box is.
_jorgen: At least 10k
_jorgen: Not acceptable...
GRMracer_: nope.
_jorgen: I think that a 10k to 5v to raise the signal from ground a bit is a start and then a 1k to ground.
GRMracer_: it be better to capacitively couple it first
_jorgen: Because the raising edge of the signal that arm the chip is 130 degree or so before the negative edge that trigg it.
GRMracer_: otherwise your pumping DC down the line to the VR sensor
_jorgen: Would that be a problem?`
- GRMracer_ doesnt know..
GRMracer_: so we need to know what is Fmax we need to pass.
_jorgen: 300Hz or so
_jorgen: A 0.3u cap to ground when the signal first enter the box.
_jorgen: 1k in parallel with the cap to terminate the signal.
GRMracer_: we need a PI filter input... low pass followed by a high pass.
_jorgen: Inside of that filter& termination we can put the lowpass, inside the lowpass we can bias the signal.
_jorgen: This start to look a bit like the input in the Audi boxes...
_jorgen: We still need the 1k resistor as a terminator
GRMracer_: and 6d phase shift is 1 decade away from 3dB point
GRMracer_: this discussion needs moved to wiki very soon.
GRMracer_: so we can make it better.
GRMracer_: and I'm off by a factor of 2pi
GRMracer_: i did the cheat RC time constant
GRMracer_: let me do some stuff....
GRMracer_: remember were not dealing with a sine wave.
GRMracer_: its esentially a squarewave.
GRMracer_: to keep signal looking like a square wave we need at least 3x base freq
_jorgen: Maybe the lowpass isn't needed?
GRMracer_: im looking..
GRMracer_: the DC blocking cap will set the minmum freqency.
_jorgen: Yes
_jorgen: We have to keep the impendance inside thatone high or the capacitor will have to be huge
GRMracer_: I'll work on it.. assuming the VR is 800ohm, Tline =600Ohm, Chip load is 10k, term is ~600-1k , DC blocked, fmax is 300Hz *3 for 6d phase shift.
_jorgen: Do you know any links with graphs of what happens in a unmatched t-line?
GRMracer_: http://www.propagation.gatech.edu/ECE3025/tutorials/SinusoidalTlines/sinusoidal.htm
GRMracer_: here as a quick fix put 50 Ohms to ground.
_jorgen: Yeah, but the load would be too much for the sensro.
GRMracer_: its already on the sensor.
GRMracer_: well then say try a 100ohm.
GRMracer_: it will be better..
GRMracer_: 50,100,500,1k
GRMracer_: in that order.
GRMracer_: and see if you can "patch" it
_jorgen: For a signle car that would be ok, but this has to be a solid solution as the system will hopefully be duplicated several times.
GRMracer_: that may work
_jorgen: Transmission lines work like when you tune the exhaust on cars... Interesting. I wonder if there is some good software for calculating transmisison lines that is useful for header calcs.
GRMracer_: yes.
GRMracer_: there is I can write matlab for it.
GRMracer_: its not hard.
GRMracer_: try this as a patch when you get the 600ohm line in it.
_jorgen: Wasn't it two 18ga wires that ended up beeing a 600ohm line when they were twisted together?
_jorgen: (just like MSD does it.)
GRMracer_: any old pair can be made to do it..
GRMracer_: its a function of wire diameter, separation and twists per inch
GRMracer_: try this..
GRMracer_: 7uF in series, 1k to ground.
GRMracer_: acutally 8uF in series 1k to gnd.
GRMracer_: with twisted pair.
_jorgen: That is a very high cuttoff.
GRMracer_: very low cutoff.
GRMracer_: High pass filter.
_jorgen: Engine will not start with that.
GRMracer_: 2Hz 6d pont
_jorgen: Ok, that could work
GRMracer_: f3db = 1 / 2 pi R C
_jorgen: But will that terminate the t-line?
GRMracer_: f3db = 1/ (2 pi R C)
GRMracer_: its a 1k termination at 2hz.
GRMracer_: |Z| = 1/(2 pi f C)
_jorgen: cool
GRMracer_: its a tradeoff
GRMracer_: its goes down with freq.
GRMracer_: 100hz its 100ohms..
- You have been disconnected. Fri Nov 11 22:49:14 2005.
- Automatically rejoining channel...
- Channel re-joined successfully.
- #lt1882: _jorgen GRMracer_ sly_ jrussell_ jrussell mcel sly___ humming
- #lt1882 was created on Fri Oct 07 14:33:26 2005.
_jorgen: Isn't it better to put a resistive terminator outside the decoupling cap?
GRMracer_: no.
GRMracer_: messes up the math.
GRMracer_: and then you need a second one to to the biasing
GRMracer_: the ckt should be transmission line || bias network(vcc-10k-*-1k-gnd) - * to chip
_jorgen: I could use a much smaller cap if I had internal biasing with more resistance
GRMracer_: 1k to 10k means 1% of voltage goes in other directions..
GRMracer_: yes. but I'd have to work the math out..
GRMracer_: you may need a bigger one anyway.
_jorgen: Wouldn't that be a perfect transmission line?
GRMracer_: yes, but you have a complex load.
_jorgen: Not very as the load inside the capacitor would be low.
GRMracer_: ok.
GRMracer_: maybe. can you give me a week or 3 to come up with an optimal SPICE'd solution
GRMracer_: Im going on vacation next tue
GRMracer_: the one I gave you should work.
GRMracer_: for not too terribly high revs
_jorgen: I hope to start the car tomorrow and the 1k resistor outside the capacitor sound like the way to go.
_jorgen: This car will rev 8200rpm or more
_jorgen: I have to get a scope and check the signal with the engine running.
_jorgen: I'll ask him how long the cable was when the car ran too. He had a roll that probably had 10m on it in the shop.
GRMracer_: http://www.play-hookey.com/ac_theory/hi_pass_filters.html
GRMracer_: 4.7uF isnt hard.
GRMracer_: two in series
GRMracer_: whoops.
GRMracer_: two in parallel gets you really clos
GRMracer_: and hell they come in 0603
GRMracer_: use 2 4.7uF caps in series.
GRMracer_: they should do the trick
GRMracer_: http://www.epanorama.net/documents/wiring/cable_impedance.html
_jorgen: I don't follow you now, two 4.7u is 2.35u.
GRMracer_: it you put them in parallel with each other but in series with the line its 8uF
GRMracer_: approx.
GRMracer_: resistors add in series, caps add in parallel
_jorgen: You forget the rule of caps
GRMracer_: ok.
_jorgen: Exactly
_jorgen: It's easy to forget that the VR signal edge is very sharp and high frequency is sent into the t-line.
GRMracer_: yes.
GRMracer_: that why we use a HP filter.
GRMracer_: with a very low f3db
_jorgen: When first thinking about this in the garage I totally ignored it, I thought: No way an antenna leads properties can affect a low frequency signal like this.
GRMracer_: it does.
_jorgen: Yeah, if it was a sine wave it wouldn't
GRMracer_: twisted pair changes from 600ohm @ audio to 120ohm at ethernet
_jorgen: 600 ohms is a typical impednace for open-wire balanced lines for telegraphy and telephony. A twisted pairs of 22 gage wire with reasonable insulation on the wires comes out at about 120 ohms for the same mechanical reasons that the other types of transmission lines have their own characteristic impedances.