I'm ready for a knock sensor
 

After about two weeks worth of fiddling, I have most of my fuel-related settings fairly well dialed-in. Feedback from the wideband made this a really, really easy process.

I'm ready to jump in and try optimizing the spark table (running MSPNP table right now), but I'm afraid to do so without being able to hear for knock. So my options right now are:

1) Buy knocksense MS.

2) Build knocksensing circuit from MS/Extra website.

I searched but couldnt find useful info. Apparently no one here has built the knocksensing circuit?

I'd like your opinions on both options. How well did it work out for you? Were you able to tune the sensor to actually hear for real knock?

Additionally, once my turbo goes in I'll be having an EGT input to the MS as well, so I hope to tune on-boost spark table using that.

one of our australian members with a megasquirted NB built the knock circuit. can't remember his name for the life of me.

I love my KnocksenseMS. I have it wired in with a second potentiometer to filter out noise and it works great.

Also, remember that the best way to tune spark is dyno time.

That's true, but I'd like to have atleast a "90% there" tune before wasting dyno time. I believe the PNP settings are ultra-conservative?

After the ping I started hearing last week I'm strongly considering getting one as well. The price is right.

Maybe you are thinking of Reverant, but he's not from Australia. I've built the knock circuit but never bothered to test it. :)

I think reverant did do this, but there was another guy... almost positive he was an aussie.

Not ultra-conservative, but there is a bunch to be had, especially in adding timing post-torque peak.

I've been doing some reading over the afternoon and really, knock sensing seems like a sub-optimal setup unless we copy the OEM implementation of it, which was found in the post-99 engines.

The other thing I have come to know is the B6T engine came with some kind of knock sensing setup. I'm sure that will give better results for us 1.6 guys provided the output signal from the knock amplifier module is something that MS can use.

Anyone who has made Knocksense work perfectly please tell me I'm wrong about this whole thing.

AussieDriver? ;)

Anyone care to comment on this? It shouldn't be hard for someone who has a 99 ECU and a DMM to trace the circuit. I could do it if someone lends me their 99 ECU.....

This is of course, assuming, no one has been able to get good performance from the Knocksense.

The other option, of course, is to build something wacky like this:

http://forums.neons.org/viewtopic.php?t=295450

Its a little caveman-like but it could work while tuning the spark maps, atleast.

As Savington said, we have his working well. We used to tune and listed for knock, but its kind of hard with the turbo and exhaust noise, doesn't help that he usually has an exhaust leak as well. ;)

Now, when we are tuning I never hear knock, but every once and a while the sensor goes off and pulls timing real quick, it's really a good idea just for added safety. I am about to install a MS in my miata, and will be purchasing a KnocksenseMS for it since it's worked well on Savington's.

I (don't think) I've had any issue with false positives on my knocksense. I ran the regular one for a while, and upgraded to the KSMS when I got my MS-II.

for the most part, when I expect it to go off, it does. When I ran a front lisence plate (covering the intercooler), I got knock. I get 'knock' when I rev higher than I tuned the KS to ignore. I get knock on messy, grinding shifts.

The downside is I've never 'heard' knock, so I don't know if I'm missing some, or if this is all false. Pulling timing makes it go away so it seems real enough.


My only comment on the stock knocksensor is that the miata hardly uses it. They ignore it above some fairly low rpm (4 or 5k) because the motor is just too noisey.

The problem is the miata can have bad knock which is quieter than normal engine opperation at the top end. You want two things, one should be easy, one is just plain hard:

1) A knock threshold dependent on RPM. The MS-II extra supports this, MS-I might. But you still need a detector circuit which puts out a signal proportional in volume to what the actual sensor sees.
2) A knock 'window', so you only listen for knock ~10-20 degrees around the ignition event. Then you ignore other noises, and maybe 90% of the noise is ignored.

High priced knock sensing solutions use the later methods. In the mean time, there are a variety of <$5 chips that put out a proportional signal, but no one has yet spearheaded the movement to come up with a part, make it work, and lay out a board. If I ever do this, you'll be the first to know, and likely Jean from JBPerf will put together something nice out of the design.

This is precisely the reason why I think many OEMs have switched to ion sensing instead of a piezoelectric knock sensor. Its head-and-shoulders more accurate. I'm looking into the possibility of rigging up a cheap ion sensing solution for datalogging purposes only. Writing the algorithm for closed loop control is beyond my present capabilities.


I was looking at a lecture from automotive sensors class and found this (att). It seems to match what you're talking about in (2).

The chip you're talking about. Does it happen to be this one by any chance:

http://www.chipcatalog.com/Datasheet...DC9E458B54.htm

I don't think it's the exact chip, but it's certainly of the type, there are many of them.

Remember, you don't need to do a closed loop control, it's enough to build something with a signal the MS can see, even if it's simplified.

NASIOC probably has the routines so you can see how the stock knock sensors work as Subaru ECU software is opensource. And the Suby ECU is quite advanced. I'm sure it's been discussed quite a bit in the tuning forum there.

Or if you can find the schematics for the TurboXS knocklite. It doesn't have a signal output, but it does allow a run to redline to determine engine noise when you set it up and sensitivity is controlled via a button on the top. Plus it uses the common Bosch knock sensor.

Frank

I've built the knocksensing circuit and it does what it's supposed to do: basic knock protection. With proper tuning, you shouldn't be knocking anyway. I'm getting false positives between 3400 and 3800rpm, even on decel, but on the bright side, I've never heard knock during tuning that wasn't logged. For $5 in parts, you just can't beat it.

Jim

Well that's the whole point. How to get the proper tuning without spending hours worth of dyno time?

In that case, I think I'm gonna try this. The false positives dont bother me as much as the non-reported negatives do.

Agreed. Plus if you inadvertently get a tank of bad gas (low octane) the knock sensor will let you know.

A link would be really helpful. NASIOC is like the freakin' manhattan of car forums. Too crowded.