MJLJ V4 + Knock sensor = possible good idea

[brentp]

I'm noodling on some ideas that may fit nicely into the current design. Let me think about it for a day or so.

On a related topic, it seems the Bosch knock sensors are the most common. Of the Bosch knock sensors, what part numbers are the most generic/universal?

[DannyP]

miniman82, I have a Megajolt-sparked type1 VW. It is putting out about 180 hp out of 2165cc. I have Manton Chromoly pushrods. I had aluminum, but they got a little mushroomed at 6500! Solid lifters, not really any valve noise to speak of, they are set at zero clearance cold. About 5000 miles or so on the new pushrods, 3000 on Megajolt. Hydraulic is fine, but I don't see a problem detecting knock(sensor on the engine case). Probably not for me though, my max advance is 32.5 degrees, no pinging any more since Megajolt, due to the accuracy and the fat spark.

Brent, how close are you on the v3 firmware? As long as I get it sometime this winter will be fine, gives me something to do.

[alexander]

not very sophisticated, but here is some basic info:

from an ebay ad:
BOSCH PART # 0261 231 110

http://cgi.ebay.com/ebaymotors/Mercedes ... ccessories

from the bosch online catalogue:
0261 231 118
0261 231 148
0261 231 153

http://www.bosch.com.au/content/language1/html/4563.htm

brent, as you know from previous debates on this subject, that the finesse of knock detection is in the filtering of the output of the knock sensor. after much delving and consideration, i conclude that meaningful knock detection likely involves putting a knock sensor/s on a motor, attaching a spectrum analyser to the output, making the motor knock, and seeing which frequencies show strong amplitude at that point. then creating a filter to filter out other frequencies. my guess is that the frequencies in question would be in quite a narrow spectrum, related to the dimensions of the motor in some complicated way, and different for every motor. all quite challenging for home experimenters!!

anyhow, just some random thoughts...
regards
alexander.

[alexander]

just labouring the point, as usual, as this topic is of ongoing interest, here is an online article on the subject which may be of interest...

http://autospeed.com/cms/article.html?&A=0348

[brentp]

Just wanted to follow up on this topic.

A way to make the advance correction feature more suitable for knock control is to have a mode where the specified 'correction' value is added periodically to the base ignition advance as knock is observed.

Here's a simplified example.

Possible scaled Aux input values:

0 = no knock
1 = little knock
2 = some knock
3 = heavy knock

sample advance correction table:

0 = +1 degrees correction
1 = -1 degree correction
2 = -4 degrees correction
3 = -8 degrees correction

The peak hold value could be re-purposed as a sampling interval, measured in ignition events, allowing one to tune the reaction speed.

Furthermore, the algorithm would strictly be ignition "retard" oriented: There would never be a case where there correction would cause more ignition advance than specified in the main table, making this mode specific for knock-control purposes.

The above ties into the purpose of the first line in the correction table: the +1 correction at '0' knock above allows the algorithm to gradually restore the advance back to normal as knock diminishes.

Of course, this isn't the only way- just one possible way to improve the controller's capabilities for knock-control.

Thoughts?

[Alexander_Monday]

... you need some method of continuing to raising the AUX input as long as knock is detected, and lowering it when not, both at a controlled rate.

If I am reading your idea correctly it would achieve exactly that.

[NITROPIXIE]

Would this mean that you would never have to program an ignition curve again??? Future upgrade could be an automatic programming ignition curve, obviously you'd have to work out the correct knock signal per engine.

[brentp]

Not really. It's not intended to be a fully 'closed loop control' system like how Saab does their engine management. it's more for handling exceptional cases: accounting for environmental issues, such as unusually high intake temperatures on boosted vehicles; bad gasoline / accidental incorrect octane, and so on.

[Alexander_Monday]

Brent is correct. You can over advance to a point of less efficiency and or power without necessarily having knock.

[NITROPIXIE]

Oops i stand corrected

[Alexander_Monday]

Not really. It's not intended to be a fully 'closed loop control' system like how Saab does their engine management.

Brent,

Interesting, does Saab have a way to control the ignition just by working engine variables, or is it still based on a default map for the particular engine? If they do, I would be interested in how they get the efficiency and power feedback from the engine while not being on a dyno. That would be the ultimate step for MegaJolt to take next.

[brentp]

Actually, I had my wires crossed a bit. As I understand the Saab system, the system constantly probes the upper limit of *boost* control, backing off in the event of knock.

What's cool about the system is you get instant power as soon as you feed it higher octane gasoline. Also cool is the controller for this (IIRC) is a separate unit from the main ECU, like the EDIS module is for ignition.

[stanzeman17]

Being a sound engineer myself, and considering the fact that knock signals are located in the high frequency spectrum, I would think that a DSP platform with parametric filtering and EQ would offer a more accurate way to detect and quantify knock events. You can now easely get mini DSP modules that offer outstanding performances: http://www.analog.com/en/embedded-proce ... index.html

I would like to do some experiments but I am so afraid to destroy my engine (V6 Dino). Can I go wild on a Briggs&Stratton mower engine and measure some knocking on that? Would not be too catastrophic if the engine blows up

[brentp]

Unfortunately it's a particularly difficult problem to 'beta test', as each engine has a unique knock signature and requires testing in a way that risks damage to the engine. That's why I have been skeptical of throwing an existing OEM knock sense module, like the GM unit, onto any random engine and blissfully believe the engine will be protected.

Unusually tough problem indeed.

[avendlerdp]

To revisit this. I'm tinkering with a Phormula knock sensor and it puts out a progressive 0-5v signal depending on knock severity.

http://www.phormula.com/KnockMonitor-KS-4.aspx

Is this going to be as simple connecting the Phomula knock amp to the 0-5v input on the MJ and programing retard levels? This sounds relatively easy. Has anyone done this?