PDA

View Full Version : cheap knock-based EBC



karl
04-15-2010, 04:20 AM
curious if anyone is using one of these- saab APC.

For those who have never heard of the APC, its a control box that every turbo saab came with that controls the factory boost level and curve. It also takes inputs from the ignition coil (for rpm) and a knock sensor and uses that to pull boost back if knock is detected. It runs at 12Hz (would work with most solenoids) and the input for the boost pressure sensor can be modified to accept any MAP sensor.

There are 3 pots on the unit: F, P and K. F sets maximum boost, P controls boost rise rate or gain, and K sets the knock circuit's sensitivity.

I've done some sims on the knock input circuit and it essentially listens to the sensor and attenuates the signal until it reaches a certain amplitude, then increases output of the circuit exponentially. Its not particularly picky about frequency as long as its between 6-8kHz (which covers primary knock frequencies for most of our engines).

The circuit also uses an rpm function to gradually reduce sensitivity at higher rpm to exclude noise.

these are the schematics for the newer 25 pin variant: (they're big pics)
http://img.photobucket.com/albums/v77/scarecrowX/EMS/apcu1_4.jpg
http://img.photobucket.com/albums/v77/scarecrowX/EMS/apco1_4.jpg
i started reverse engineering the redbox one i have before i found those.

basically, the stock boost pressure sensor is a resistive unit, that provides 1/2 of a voltage divider. the F pot and another 909 ohm resistor pull voltage to a 7.35v reference, and as the pressure increases, the resistence of the sensor increases, pulling the voltage further from ground. the voltage range tops out around 650mV. you can see that the reference voltage used on IO2a (opamp wired as comparator) is 618mV (voltage divider comprised of R35 and R46), so the output goes high and the F-circuit controls boost whenever the boost signal exceeds that value. also note that the rpm signal from the LM2907 on the smaller board is fed into this as an averaging circuit via R138 and R37. as RPM increases, this pulls the boost value higher, which is what tapers the boost off. i've read that folks who have tried to remove this feedback have experienced "unstable boost," but i don't have any more info than that- whether they mean that it fluctuates, rises, falls, or what. the RPM signal could also be fed into the reference voltage to prevent tapering and an RC filter on the output of IO2a could help if the circuit oscillates too badly.

pin 24 can be connected to a powered LED that will light whenever the F-circuit is controlling boost, and pin 11, when held to ground, will disable the F-circuit and the unit will only use the P circuit to control boost. i'm assuming this function is to disable F while setting P.

pin 19 can also be connected to a powered LED and will light the LED whenever knock is detected.

to use a normal MAP sensor, the signal would have to be attenuated to a slope whose peak is slightly above the 618mV reference (via potentiometer or opamp) and fed into R41 while disconnecting the rest of the input side of the F circuit.

i'm still trying to get a handle on how the P circuit works.
pin 18 is the internal knock signal. it begins to amplify the signal once input voltage starts to get to around 4V, but it limits the signal below that. i think as long as you have a knock sensor that's tuned to ring at the knock frequency of your block, this will work for any engine, since it's the amplitude it's looking for rather than frequency.

the 2907 produces a reference voltage to weed out noise at higher rpm.

locations on the cars varies. there's a lot of info here:
http://www.900aero.com/applications/phpBB/viewforum.php?f=7&sid=93385c61219531772e07599fc1ade3ae

i plan on running an APC as my boost control this spring. i grabbed a 3-way solenoid valve that i just serendipitously stumbled upon in the u-pull last week, so i have no reason not to try it. i need to build an opamp buffer for the MAP input so i can just tap into the MAP sensor signal without loading it, but that's an easy circuit.

anyone else using one?

saab pressure sensor resistance output vs. boost:

bar Ohm
0: 7
0.1: 27
0.2: 40
0.3: 56
0.4: 62
0.5: 80
0.6: 91
0.7: 100
0.8: 109
0.9: 124
1.0: 137
1.1: 149
1.2: 155
1.3: 167
1.4: 171
1.5: 185
1.6: 194
1.7: 216

i'll have to plug those into LTspice and see what the various incarnations of the F-pot circuit output for voltage vs. pressure. would make using a MAP sensor much easier. i want to be able to control boost up to 300kpaa.

i'm also not certain on knock sensor compatibility, since some sensors have one side connected to ground and the saab sensors are completely disconnected from ground. the saab units would be easy to install on anything though too, since they use a through-bolt to bolt to the block.

the circuit simulation works ok with one side connected to ground, but polarity does matter (it only works with pin 16 connected to ground and pin 17 as signal), and the output is significantly higher than with it decoupled from ground, so the K-pot would have to be adjusted to compensate.

the internal knock signal for sensor voltage in stock saab configuration is: (all values are at 6.4kHz for the 90mm bore of the saab engine)

sensor/circuit
3V/0.180mV
4V/24mV
5V/120mV
6V/220mV
7V/330mV
8V/440mV

and the circuit voltage using a grounded knock sensor is:

sensor/circuit
3V/6mV
4V/120mV
5V/260mV
6V/400mV
7V/500mV
8V/650mV

i'm up to a pretty decent understanding of how everything works though. some parts of the APC are ridiculously overcomplicated, and some are brilliantly simple. the knock detection side of it is quite clever.

the circuit at the lower right corner of the top schematic i posted is the knock sensor check circuit. one of the opamps on the bottom schematic (I04, LM224N, pin8) takes the basic noise from the knock sensor and turns it into a digital signal. that signal is AC coupled to the base of T14 in the top pic, lower right corner and acts as a charge pump into the base. when there is engine noise on the knock sensor, the charge pump turns T14 'on' such that it pulls one of the main solenoid drive circuits (I03, pin6) noninverting input low, allowing the APC to control boost. if there is no noise from the knock sensor, T14 stays 'off' and that input stays high, and the APC disconnects the solenoid and reverts back to wastegate pressure.

the section of the circuit at the top right corner of that same page is the voltage reference regulator. this could have been accomplished with an adjustable regulator rather than the 15 or so discreet components used here.

it seems that I03 pins 8 and 9 are a comparator that takes boost pressure against a reference (~0.4bar) and above that reference, adds voltage to the internal knock signal, making the circuit more sensitive to knock at higher boost, even as the RPM signal decreases the sensitivity as revs climb.

the pin 19 output (knock lite) could also effectively be used to pull an input to whatever ECU you're using (IAT, for example) to ground to simulate very high IAT values. this could be set to both retard timing, add fuel, or both. it would also be very easy to log as knock, showing up as high spikes in IAT.


it seems that it is possible to avoid the boost taper too.

info here: http://www.saabcentral.com/forums/showthread.php?t=49981

removing one leg of R36 and feeding it a reference voltage instead of the RPM-based voltage allows for steady boost to redline (since the RPM signal no longer adds to the boost signal), but tuning becomes a bit finickier. i'd be willing to bet much of the boost taper is due to the tiny turbos on those saabs as well.

Luni
04-15-2010, 04:24 AM
karl I think youre the only mofo on here with the electronics background required to do this.

Id say nuke could do it, but he isnt active anymore and isnt into this scene much.

Go for it and post your results. reading your post is very interesting. That thing is cool how it works.

Naemion
04-15-2010, 04:28 AM
So where is the flux capacitor going?

joe's gt
04-15-2010, 04:48 AM
karl, I'm so glad your posting on this board now. personally, I'd probably have to reread that a few times to understand it, and I'm an engineer. klapa on here I think has a good electronics background, but I don't know of anyone else. Nuke's a frickin genius but he rarely comes on here anymore.

idk man, as an ME I didn't even come close to that in depth with electronics stuff so I had a really hard time following. Definitely post your results, but someone like me would probably find it easier to use a manual one or buy a used apexi or greddy controller.

Would make for a cool project tho. Post it up in the My Projects section if you choose to go through with it.

Shadowlife25
04-15-2010, 09:40 AM
Great post Karl.

I subscribed to the thread so I can follow your progress.

Night_Wolf
05-10-2010, 05:37 PM
karl I think youre the only mofo on here with the electronics background required to do this.

I built a robot once! Operated all by itself around an obstacle course. But that was was easy electronics lol

karl
05-10-2010, 06:34 PM
Robot would be beyond my abilities at the moment, heh.

I'm working through some other quirks in the car right now, but I hope to have this on the car by june. Id like to try it on the bench first too.