Knock Sensor tuning/monitoring

[Dan [Sac]]

I know our knock sensor is basically a microphone that listens for telltale signs of detonation and knock.

What I don't know:

*how does it communicate with the ECU? Is it a progressive/variable signal stregth (variable voltage) depending on amount of knock detected, or a simple on/off mode?
*If it's a simple on/off mode, why can't we hook up an LED to it like the DSM crowd and have ourselves a very reliable knock warning light?
*why the heck is the wiring between the sensor and the plug coaxial?

I'm asking because I want to dyno tune the rice....err, racer soon, and from the dozen of threads (been reading up back on the old Sephiro, Vseved, ProGTR, Mikeseli discussions of yesteryear) some say that timing advance works, some say it doesn't, some say it works only in DIAG. mode but you lose dynamic timing adjustments, etc...having a knock sensor LED or other knock sensor-driven warning light would allow for fine tuning the timing under load on the dyno. Even more timing control can be had with a MS unit - yes, you can use the dyno wideband O2 for finetuning it, but still a knock LED would be invaluable for on the fly tuning.

The alternative would be the $500 knock sensor (name escapes me at the moment) that Andrew Sword has on his boosted car - this allows for superb individual cylinder timing adjustments, but it's pricy at the moment for me (unemployed since June).

I want to max out power output with the stock ZE cams before I go farther. The car is a race only vehicle, so idle quality is not an issue, nor is power below 4k RPM. It has no VRIS, no EGR, no PS, no A/C...etc - the car weighs 21xx lb dry.

So yeah, the knock sensor - I'm going to hit up DSMtalk to see where their knock LED hooks up and see what we can glean from their experiences - meanwhile, anyone with good insight into our knock sensor's workings, please share the wealth :)

[Dan [Sac]]

FWIW - here's how it's done on DSMs (2nd generation 95-99 models - i'll post the 1g procedure when I find it 90-94):

Schematic: http://maritime.dsm.org/tech/vfaq/knockled/KnockLED.jpg


How to build the knock LED

It's easy! To build the knock LED you need:

A. any LED (you can get those from any Radio Shack)
B. a 600-ohm or greater resistor (you can also get this from any Radio Shack). Note that this resistor value is very important; if the resistor value is less than around 400 ohms, then even as the ECU tries to close the BCS to reduce boost, the BCS may not close properly.

You'll need to be able to solder the resistor to the LED and then access the ECU to tap its BCS signal. I won't waste space here to discuss how to access a 2GT ECU, but you can easily find this information by searching Tom Stangl's VFAQ page or the Club DSM website. Refer to the circuit diagram mentioned at the beginning of this page if any of the following description doesn't seem clear to you.

1. Solder one end of the resistor to the anode (positive end) of the LED. I recommend choosing a resistor that's between 600ohm to 1200ohm. Keep in mind that the higher the resistance, the dimmer the LED will be because less current flows through the LED.

If you don't know which end is anode, do this simple test: solder the resistor to either lead of the LED. Then, connect the other end of the LED to ground while connecting the unsoldered end of the resistor to battery 12V. If the LED lights up, that means you got it right and you can proceed to the next step. If the LED doesn't light up, remove the resistor and solder it to the other LED lead and re-do this test. Don't worry, you won't damage the LED even if you connected it backwards. Take a look at this diagram if you didn't understand the description above.

2. Now solder the cathode of the LED (the unsoldered, negative end) to ground (any good body ground will do), while tapping the unsoldered end of the resistor into the ECU's BCS output signal. The BCS signal wire is a yellow-red wire and it should be the 11th pin from the left in the top row when looking into the 26-pin ECU connector (two rows of 13 pins each). There are four ECU connectors, but only one has 26 pins total so you can't miss it.

That's it! You can now put everything back together and enjoy your knock LED!

Expected behavior of the knock LED in a 2GT

The knock LED should turn on and flicker for a few seconds everytime you start the engine, and should turn on and stay on for a few seconds after you shut off the engine. I theorize this is because the ECU is just setting the "default" or "fail-safe" condition of a closed BCS. Otherwise, the LED should stay off.

However, the 2GT ECU seems to always cycle the BCS on and off rapidly when the engine speed is between 2400 - 3700 rpm regardless of load/throttle position. You can see this as the knock LED rapidly fickers on and off in the said RPM range. Don't worry, you're not getting knock just by blipping the throttle in neutral. 8^) I suspect Mitsubishi engineers put this "feature" in the ECU code to soften the torque curve as boost kicks in, thereby reducing the sensation of that dreaded "turbo lag." I know this is somewhat stupid for an enthusiast, but I guess for the general public this is a desirable feature.

In the 2-3 years since I've installed the knock LED, I've never seen it light up under any conditions (except in the RPM range mentioned above) until just very recently when I was at the racetrack. After a few runs my car probably experienced enough knock that the ECU has decided to close the BCS (thus lighting up the knock LED). This is easily confirmed by the fact that my boost is only limited to 10psi even though I have a manual boost controller (just a bleeder valve) and removed BCS restrictor. The knock LED stayed on until I reset the ECU.

If your knock LED isn't behaving as described above, you may have made a mistake in one of the steps so please double-check that your resistor-LED combo will light up when connected to the battery, and that you have tapped the correct ECU wire. You could always reset the ECU if you suspect it's already keeping the BCS closed too.

That's all, folks!

Warren Tsai

[Dan [Sac]]

Ah heck, rather than clutter this thread, you can find all the 1G (DSM 90-94) information on how to do the knock sensor LED here:

http://www.vfaq.com/mods/wayback/ledmod_files/howto.htm

Now I hit the fatty shop manual to see what it tells me about our stock sensor :tup:


Edit - the way the DSM sensor works is that when the sensor detects knock, it signals the BCS (boost control solenoid) to reduce boost, thus reducing knock. They are intercepting this signal at the ECU and adding the LED with the proper schematic and they 'see' what the knock sensor sees.

[cyborgzero]

Dan,

I have done some work on the knock sensor..

The cable is coax that is lined with teflon. That is so that, when the car moves, there isn't any rubbing of the insulation on the two wires that might cause a voltage between them. Rubbing a surface causes a voltage potential to be generated and they are trying to alleviate that by lowering the friction to nearly nil. Also too, coax cable doesn't have a significant change in capacitance when its moved because the distance between the two conductors remains constant.

Ok, as for the next thing, there is a signal conditioning op-amp that usually amounts to a filter fed into a window comparator. The filter will smooth out the knock to alleviate false signals, but it also could cause too many knocks before the ecu reacts to the knock. The reject frequency is important there. Our knock setup is fairly primitive. Then, that signal is fed to a window comparator, which determines if the pulse is significant magnitude to be considered a knock or if it is a normal pulse from combustion.

THEN *shew* ;) the ecu compares the spark timing event to the knock event. If it is a TRUE knock event, then the knock will usually happen BEFORE the spark event. Now THIS I don't know if our ECU does. This is the GOOD way to do knock detect, but back in the day, not real common.

Ok, back to what we were doing. :)

We have to find the knock window comparator on the ecu board. With the early ecus 93-94, I am fairly sure that it uses the old school style knock detect and most of the detection is done with discrete components. On the later 95 and up, I am thinking that they *might* be using some simple dsp routines since a bit of progging space is MUCH cheaper than discrete components.

Once the feed that goes to the ECU main processor is found, its fairly easy to set up an opamp to drive an LED to do what you want.

There are some schematics on the inet that use about 10 dollars worth of Radio shack parts. With the really good diy circuits you can tune the knock pulse voltage threshold, knock pulse width, and you can compare knock timing vs. spark timing.

[Dan [Sac]]

Thanks Rob - that's the kind of info I was looking for. I'ma let it sink in for a few hours :D

Andrew Sword's knock sensor is a J&S Safeguard model...I remember him posting that it was a pain to install but that it is worth the price tag and then some.

[Dan [Sac]]

Found this while perusing the J&S website:
http://www.jandssafeguard.com/educatingmario.html

Although not as promising as the DSM's Palm Pilot data loggers, this appears to be a cheap alternative to a full MS for those that want to just monitor, and not control, vital engine data.


Here's a neat little product we came across that plugs into the back of a Gameboy Advance. Among other things, it lets you look at signal waveforms, like you would with an oscilloscope.

You can set it up to display as many as four channels of data, such as O2 voltage, MAP or TPS sensor, or even ignition trigger signals. It's not a data logger, but you can see stuff happening in real time.

It gets its power from the Gameboy's 3.3v battery, which means that for signals greater than 3.3v, you need to use a voltage divider (big deal, two resistors) to cut the signal down to size.

The company is educatingmario, and the product is called the Captigate.

You can get the new Gameboy Advance-SP from Target for $99.

[Joe Bialy]

Here's a bunch of knock info. All from the 'VEMS' website. This is a really nice EFI you build yourself, like MS- more complicated than MS, but more capable too.

http://www.vems-group.org/wiki/index.php?page=DetonationDetection

[b4tn]

You can set up Knock sensing with timing retard on the megasquirt :D

[Dan [Sac]]

I think for starters I'll just start monitoring the sensor with a voltmeter to see what is going on...maybe I'll induce mechanical knock with the engine running (tapping the block with a hammer) and see if it changes the signal (i'll have someone watch the timing as well)...then we can check to see what the treshold is, if it's measurable...really an archaic way to do it, but we gotta start somewhere :D

Does anyone know/remember if the SAFC had timing outputs? Is there a way to monitor timing at all?

[veltpak6]

Dan,

I have done some work on the knock sensor..

The cable is coax that is lined with teflon. That is so that, when the car moves, there isn't any rubbing of the insulation on the two wires that might cause a voltage between them. Rubbing a surface causes a voltage potential to be generated and they are trying to alleviate that by lowering the friction to nearly nil. Also too, coax cable doesn't have a significant change in capacitance when its moved because the distance between the two conductors remains constant.

Coaxial is a type of braid-sheilded cable. The braid protects from ElectroMagnetic Interference (EMI) and Radio Frequency Interference (RFI). The ignition system creates a lot of this at the same rate as detection and could easily trigger false alarms, had they not used coax. It has nothing to do with rubbing wires.


Ok, as for the next thing, there is a signal conditioning op-amp that usually amounts to a filter fed into a window comparator. The filter will smooth out the knock to alleviate false signals, but it also could cause too many knocks before the ecu reacts to the knock. The reject frequency is important there. Our knock setup is fairly primitive. Then, that signal is fed to a window comparator, which determines if the pulse is significant magnitude to be considered a knock or if it is a normal pulse from combustion.


Simply, Mazda and Denso specified a frequency gate that allows only the specific frequency of knock (specific to that engine and knock sensor location) to pass and trigger any retardation.


THEN *shew* ;) the ecu compares the spark timing event to the knock event. If it is a TRUE knock event, then the knock will usually happen BEFORE the spark event. Now THIS I don't know if our ECU does. This is the GOOD way to do knock detect, but back in the day, not real common.


There is no way, even to this day, that an ECU will predict when knock will occur. It has to detect it first. And yes our ECU constantly updates knock events and advances the timing to its optimum. Its called fuzzy logic.


We have to find the knock window comparator on the ecu board. With the early ecus 93-94, I am fairly sure that it uses the old school style knock detect and most of the detection is done with discrete components. On the later 95 and up, I am thinking that they *might* be using some simple dsp routines since a bit of progging space is MUCH cheaper than discrete components.


What does that even mean? Do you suspect the 93-94 has an external knock control unit and the 95+ have an internal knock control unit? Theyre both internal. Old school knock circuits are from like 89 GT's which were in fact external. Even 90+ used internal "digital signal processors"


Once the feed that goes to the ECU main processor is found, its fairly easy to set up an opamp to drive an LED to do what you want.


Good luck man! This is no DSM. I'm not going to say its impossible, but you would spend far too much time/effort trying to map around the ECU PC board to figure out a spot to tap an LED into. Basically its not going to happen, esp. for a simple LED.

I remember those debates of rotated Disty's and timing advance. Maybe some revisions of ECU allow it, maybe some dont. I don't know. All I know is that the crank angle sensor never lies, and if an ECU was programmed to keep a timing curve, it can do it. If you want maximum advance for the track or the dyno, just dummy up the KS so it stays quiet. If youre running 91+ octane you really should not be worried. On the other hand, in the very slight chance you hear "crackle crackle" then you obviously know what to do. I could never justify spending $500 dollars on an individual cylinder timing retard system... ever... unless it was track only, stand alone FI. :lol:

[Probe 2C]

Our KS is simple.

If it is plugged in and grounded it will tell the ECU that all is well. It does not have to be threaded in its place, it does not have to hear a specific range for everything to be ok, and it does not really care where it is.

To signal a trouble code there is an upper ceiling, anything above that (under any circumstances) will flip the code. Doesn't matter if you're at idle or 7,000 RPMs, same level.

My KS went out shortly after my ZE install. Since I had a spare I simply removed the wiring harness from the (still in place) busted one and plugged it into my spare. I ran a ground wire from the threads on the spare KS, sealed it all up and secured it in a place that isn't in danger. Right now it is ziptied & electrical taped near my fender until I take my IM off.

Everything functions fine :)

[K Brice]

Dan Icoppied this info from a thread on PT allong time ago.
I can't vouch for any of the info.

Also SAE 920677 discusses some of the workings of the TKS system but I can't seem to find that document.

So I have an arsenal if info in front of me. I have the ford service cd, and I just acquired a 6" thick ford powertrain control manual. The DSM guys use knock counts to fine tune their engines. Our knock sensors work on sending voltage to the ecu, as the knock increases, voltage increases, im assuming as the voltage increases the timing gets retarded. Now the manual says voltage should be 2.4V with the key off and engine off. now with the engine idling it should be at 2.34V, and going 30mph 2.34V, and at 55mph 2.35V. So I'm thinking the way this works is kind of like the O2 sensors, maybe the ecu is "shooting" for 2.34~2.35V. when there is knock the voltage rises and timing is retarded until 2.34~2.35 is obtained. This is my theory what do you think? If this is true we can maybe hook a high resolution multimeter to it and read it to see if we have knock (voltage rises past 2.35 then drops back). Could be a nice tuning tool, i want to see how it reacts to different octane fuels.

Finally I had this thread bookmarked as having to do with the TKS system http://forums.probetalk.com/showthread.php?t=1701116872&highlight=sae+paper

I do have a couple of the dynos from the SAE paper showing the effect of the TKS system

TKS effect on Torque (http://webpages.charter.net/krawczyk/TKS+TQ.jpg)
TKS effect on igniton timing (http://webpages.charter.net/krawczyk/TKS+timing.jpg)

Brice

[veltpak6]

Our KS is simple.

If it is plugged in and grounded it will tell the ECU that all is well. It does not have to be threaded in its place, it does not have to hear a specific range for everything to be ok, and it does not really care where it is.

To signal a trouble code there is an upper ceiling, anything above that (under any circumstances) will flip the code. Doesn't matter if you're at idle or 7,000 RPMs, same level.

My KS went out shortly after my ZE install. Since I had a spare I simply removed the wiring harness from the (still in place) busted one and plugged it into my spare. I ran a ground wire from the threads on the spare KS, sealed it all up and secured it in a place that isn't in danger. Right now it is ziptied & electrical taped near my fender until I take my IM off.

Everything functions fine :)

you bypassed it essentially. as long as youre running high octane, youre set.

[Joe Bialy]

Her's a bunch of knock stuff I found a while ago. The articles are titled 'Octane booster', but as you read it pretty much centers around the KS and how to use it:

http://www.yarchive.net/car/knock_sensors.html

Velpak6 seems to know how they really work here, and as he said, an LED won't do you any good. Neither will a multimeter. Its only been the past few years that the automakers have been able to even see knock reliably over a few thousand RPM. This is a very complicated signal to analyze if its done right. So complicated there are dedicated DSP chips made just to help extract a good signal. TI and Harris Semiconductor to name two have made these DSP's.

[nevin94 PGT]

Why would you want to tune with the knock sensor? Isn't it kinda like driving by looking at your rearview mirror?

[veltpak6]

Her's a bunch of knock stuff I found a while ago. The articles are titled 'Octane booster', but as you read it pretty much centers around the KS and how to use it:

http://www.yarchive.net/car/knock_sensors.html

Velpak6 seems to know how they really work here, and as he said, an LED won't do you any good. Neither will a multimeter. Its only been the past few years that the automakers have been able to even see knock reliably over a few thousand RPM. This is a very complicated signal to analyze if its done right. So complicated there are dedicated DSP chips made just to help extract a good signal. TI and Harris Semiconductor to name two have made these DSP's.

excellent article!