MS3x Knock Module Settings Thread
 

Since my settings were requested by a couple people, I thought I'd start this.

I pulled some of the settings from msextra. I derived my own value for bandpass frequency based on something. I think it was either a formula Jason C posted or somewhere found on the internet.

Integrator time constant is the default I think.

Scaling is simply based on what range I thought "noise" looked appropriate. What I interpret as "obvious knock" are spikes at least 2-3x times taller than the "noise" in the chart at the end.

Car: 2001 Miata NB
Sensor: Stock 2001 miata NB sensor
Setup: big turbo, motor internals and head are stock .

https://www.miataturbo.net/attachmen...ine=1345559000

https://www.miataturbo.net/attachmen...ine=1345559000

I have not really touched the following values:

https://www.miataturbo.net/attachmen...ine=1345559000

https://www.miataturbo.net/attachmen...ine=1345559000

And those settings result in this:

https://www.miataturbo.net/attachmen...ine=1345559000

A+ would read again.

y8s, is your rev limit 8000rpm?

nope, but I like to tune above my rev limit because I don't like ambiguity

Your log shows 8000rpm and 200kPa max.

Oh that. I set the maximums to round numbers or numbers that make the 50% line meaningful (for AFR the dotted line is 14.7). Note also my TPS goes to 101. And volume to 11.

I find that having the graph trace sit right up against the edge of the graph limits makes it hard to read.

So how much noise do you get at peak RPM/boost? I am getting a lot of noise after 5500rpm or so.

a lot.

I haven't really spent much time tweaking on this thing. I suspect some improvement could be had by actually doing something with the windowing but... I haven't driven my car more than about 50 miles in the last month or two.

Here's a random example. I haven't decided if I should call these moderate spikes "noise" or "knock" or neither. I think the way to go would be to get a passenger with det cans and a computer to mark the logs as soon as he heard something and then match up the spike levels with the marks and adjust based on that.

https://www.miataturbo.net/attachmen...1&d=1345580776

Similarly, spending 5 minutes to look at the window settings and make some educated guesses where knock occurs in our motors would be good.

How are you detecting individual cylinder knock? Is this just a setting that is listening for different frequencies that is cylinder dependent? I see the per cylinder settings, but what is the gain, and why are they all the same?

EDIT:

I'm an idiot, it would just be based on rpm and the time window in which the vibration is detected.

Is the gain supposed to be a factor that takes into account the latency of sound traveling through the block for the different cylinder distances? i.e. the furthest cylinder would need a higher gain so that the amplitude of the peak is the same as the amplitude of a peak in the closest cylinder for a knock event of the same magnitude???

I think the gain is simply a linear scaler.

and yes, the individual knock is derived from the window relative to the spark for that cylinder.

I can tell you right here and now that det cans are useless above a certain RPM due to noise. I just can't listen ANYTHING above 6000rpm on all B6/BP engines that I've tuned so far, built or stock. So no wonder the knock sensor readings spike above that.

Please excuse my naivety if this is stupid...

BUT, does it matter? if AFR's are good and timing is static from that point to redline, and you're not knocking at 6000rpm, then you really won't be at 6500/7000 rpm due to the higher piston speed?

Depends on the timing map. I generally add more timing as engine speed increases to hit the ~17* ATDC sweet spot, so there is the probability of knocking above 6000rpm. Also between 5500 and 6000rpm it is also quite noisy, so listening for knock there is not exactly easy either.

17deg ATDC sweet spot? What's that

Do you mean BTDC?

Fair point on the increasing timing. I do the same, but not by a particularly large amount, to be honest it may as-well be flat it's that slight an increase...

Our engines will produce the most torque when cylinder peak pressure happens around 17* ATDC.

Ahh, I thought you were saying the goal was to hit ignition at 17BTDC :-)

How fast does combustion usually occur in our engines? Im trying to figure out how to calculate a timing value for a given rpm that yields the 17 degrees atdc.

There's no answer to that, it depends on AFR used (richer burns slower), fuel, injector type etc etc.

Exactly.

Ultimately though I suspect nobody is knocking above 6000 rpm much. you can probably just set your thresholds relatively high and go back to your normal day.

Anyone had anymore time to play with this? Used a set of det cans to double check what gets picked up?

What input should we be using for the knock module on the MS3x for the incoming signal? I didn't build my MS3x, so I'm not sure how it is wired.

I want to make sure I don't pick a random input, then find out I have a conflict.

EDIT: Please confirm

I am just running the K1 wire from the knock module to a spare port on the DB37 so that I can use a connector on my DIYBOB which is where the wire from the knock sensor will connect.

The knock module can be wired to any unused pin on the lower DB37 or the DB15 on a V3.57 board. The I/O it uses are mostly the LEDs, so you can't use "LEDs spark" mode with it.

you dont have anything on the LED outputs. So bring your knock sensor in one of the spare inputs and run it to K1 and done. Dont forget the wire on the daughterboard youll also have to add.

What spare inputs do I have? There was a wire from the knock module to the daughter board, then I ran k1 to a spare pin on the mainboard DB37, then connected the db37 pin to the knock sensor.

Here is the graph, but it doesn't look like y8s at all, so I think I must have done something wrong.

https://www.miataturbo.net/attachmen...ine=1348963989

maybe this is covered but does the NB miatas use a hardware filter on the signal from the knock sensor, a knockbox if you will?

That's not correct at all - either the knock module is not connected properly inside or your knock sensor is not connected/busted.

How should it be connected? Knock module ---->spare input----->db37----------->knock sensor?

If the above, then what spare inputs are acceptable?

OK, it seems like I am getting a reasonable signal now. The hard part is determining which is knock that needs to be removed. The numerical value increases with rpm, but it is the difference in applitude at each new baseline that I am looking for. i.e. At idle I see about 10-20 as a normal value. Cruising at 3000rpm I see from 20-30 In boost the norm value seems to increase to 50 or more. If I see a value that is 25points higher then my perceived baseline at any point I will start to pull timing and see what happens.

https://www.miataturbo.net/attachmen...ine=1349043331

From the engine to the ecu. Knock sensor ---> knock box ---->ecu pins.

The knock box is supposed to act as a filter and get rid of all that noise your seeing.

That's my understanding.

I plotted knock, rpm, and map using MLV with map beign on the z axis. In just my shore run it appears to show a nice exponential curve as the rpm increases. I believe this exponential trend line could be used as the normal base line for referencing knock spike amplitude. Hence you can go with a rule that says you will retard knock for anyhng that is 25 points above baseline for a given rpm. I don't think it would be correct to just reduce spark anytime the value was above 25 points.

Here is my quick and dirty baseline. More testing will need to be performed to establish firm numbers.

idle: 15
2000: 18
3000: 20
3500: 30
4000: 38
4500: 50
5000: 70

Sounds about right. I'm having about the same numbers on the MS2.

It is totally possible to knock above 6000 rpm.

What are you using as your knock determinate? 1.5x baseline, 2x baseline? I haven't quite figured it out.

I do notice that cyl 3 seems to result in the largest spikes. I pulled some timing out of my map last night, but I'll have to wait until the weather clears up to test it.

Surprisingly I found quite a few spikes out of boost. 2500-4000rpm in the 50-90kpa range.

I pulled about .5 degrees if the spike was 25pts or more above baseline.

This might be helpful for visualizing baseline. I haven't tried it yet.

Re: feature request
by LT401Vette » Mon Oct 01, 2012 5:00 am

In MegaLogViewer you can create a custom field that will average over a set of records.

Custom Field Name: Averaged Knock
Formula: ([knock cyl# 1] + [knock cyl# 2] + [knock cyl# 3] + [knock cyl# 4]) / 4

I get about 13-15 at idle on my knock graph
and 25-35 under boost at 4500
and up to 50 when boost is building and enrichments haven't caught up yet.

On my MS2, I have a "knock threshold" logged value. I put the "knock threshold" and the "DSP knock value" in the same MLV plot, and program the knock threshold settings so that the threshold is about 10-15% higher than the DSP knock value.

Is this custom reverent awesomeness, or something you can spoonfeed a "how to" to the group?

I've been playing with the scatter plots in MLV, and I think I'm able to see some useful data. I'm going to try to use the code above to plot the trendline along with the cylinder I see the most exaggerated knock spikes.

Maybe knock vs trendline vs rpm and knock vs trendline vs map

Enhanced MS2 awesomeness:

https://www.miataturbo.net/attachmen...ine=1349113720

so that's what the MAP signal looks like when you have N/A with ITB...

No ITBs, just a squaretop manifold.

Give me a picture with your cursor of the region the knock is occurring.

I need to find that datalog again first. I have like 100 datalogs for the new version of the Enhanced MS2.

Ok, I know I am asking for a lot, but...........Y8s, Reverent,

Would you two mind posting a bunch of knock scenario datalogs? I'm going to post what I see when I get home. I'm having a tough time deciding what is knock, and what is just noise.

These are both lean tip in knock.

https://www.miataturbo.net/attachmen...1&d=1349400587

You can see in this one that the noise bled over into the next cylinder (or it also knocked maybe)

https://www.miataturbo.net/attachmen...1&d=1349400587

I would say you shouldn't rely purely on level. You have to use a sort of "know it when you see it" evaluation.

Of course if you HEAR the knock (like my lean tip in knock) then you know. I could hear it.

Did you guys take logs with the knock retard turned off, then slowly dial back timing until spikes disappeared? I'm surprised by the amount of timing I am pulling, but it does seem to make a difference

y8s, you have too much timing for that particular rpm/load cell.

That's an old log but I think it is off.

is that to say I'm probably past MBT and/or slowing down my spool?

what does your midrange timing look like?

25-27. I was knocking around 29-30.

I looked a t a log this morning, and the timing I have pulled is making an impact. I'm surprised by the number of knock events I was seeing below 100kpa. I was also seeing some when transitioning from higher kpa to lower kpa. I don't think the timing change was fast enough to avoid a knock spike.

The next question is.....Does one lone knock spike matter all that much when it happens over a fraction of a second.

y8's I'm noticing at high rpm and high boost that my knock signal is approaching what appears to be the max level 102.1% It doesn't look like knock, just an increase in signal strength. If I lower the gain in my settings, will it keep it from maxing out?

Oh, and I have turned off the knock retard so i can log the knock and try to tune it out.

Yes, decreasing the gain will allow you to not max out just by noise alone.

I am seeing this as well. I 2200rpm to 3100rpm at 68-80 kpa I'm down to 25 to 27. I was over 30. 42-50 kpa I'm right at 30 degrees

This seemed to be a big area where I ws getting knock, but I'm slowly tuning it out. I have a feeling I will end up lower than it currently is. Once I feel like I've tuned out the knock I will post a before and after of my ignition table.

I was knocking originally but after correcting the leanness it went away.

Is there a way to estimate mbt. My original tune didnt have audible knock. I checked the pistons when i pulled the head, and i had no signs of det. Yet the knock sensor is definitely detecting something. If u back off the timing until the sensor stops detecting knock. Wheredoes this put you with respect to mbt? It appears to be an uber safe tune that you can flog without fear of blowing up. It probably also builds in a margin of safety if you go a little lean

You should be able to run ~14.7 at 70-80kPa at < 3000rpm without knocking at all. The fact that you were knocking at that load with a 13.2 AFR says a lot.

does this give you warm fuzzies?

the map you don't like:

https://www.miataturbo.net/attachmen...1&d=1349461432


one you might like more:

https://www.miataturbo.net/attachmen...1&d=1349461432

Honestly I haven't messed with my spark map much in the last... 5 years. The original map I had was from FM for the hydra and was ridiculous.

The area u changed is the same place im finding that i have to pull some timing.

Here is running the second map. It's interesting to note that the noise level does not increase with RPM while I'm in boost

This is about 3700 to 5600 rpm. So a little higher RPM starting point, but the initial knock noise is much lower.

I take a lot of datalogs.

https://www.miataturbo.net/attachmen...1&d=1349488977

You may take a lot of datalogs but you must not look at them. If you did, you would see that noise does indeed increase with both load and RPM.

Stop datalogging on arlington blvd or whatever that road is that has 25mph speed limit.

I have my sensor installed and have been logging. I'm at work so I can't attach anything. I had a pretty nasty lean tip in knock, that went away- but I find that I am having to go very rich with the resolution I have in my AE chart. I will have to look at what others have.

I also notice a noise spike when I let off the gas and AE enters into the negative TPSdot. Anyone else see this? It will go from 20 to 40 when I let off the gas and then settle back down to 20, but not if I pulse the throttle.

I'm working on setting up my knock setting on an NB running and MS3X (stock knock sensor)
This is on a mostly stock naturally aspirated engine.

I tried to systematically set up knock. I set the bandpass to the theoretical frequency of the miata bore size.
Then did two WOT sweeps at -5 and -7 degrees timing pulled from the stock MSPNPpro base map.

https://cimg4.ibsrv.net/gimg/www.mia...a6d7335226.png

I assumed that these two pulls were very safe in timing and detestation free so I set my knock threshold slightly above the knock signal of those lines.

I then ran a pull with 8 degrees added at 3000 rpm to try and induce knock. The knock input was slightly over the first two pulls but not the magnitude I had hoped (no audible knock was detected).

https://cimg0.ibsrv.net/gimg/www.mia...5d2474284e.png

I then tried 10 and 20 degrees added at 4000 rpm. Here, you can see a pretty sizable change in the knock signal but still not the magnitude that I had hoped to see.


https://cimg1.ibsrv.net/gimg/www.mia...1b16041731.png

After, I ran more pulls at of the basemap spark. These pulls seemed to be knock free, however, some points lie slightly above the knock threshold I set from the -5deg and -7 deg pulls...



https://cimg6.ibsrv.net/gimg/www.mia...0ab0d02f55.png

I'm not sure If I should accept these as tiny knock events or if I should increase the threshold slightly that the baseline knock lines are below the KNK threshold?


Here is my current settings:

https://cimg8.ibsrv.net/gimg/www.mia...a84f3fb80d.png

https://cimg9.ibsrv.net/gimg/www.mia...914a825522.png

https://cimg7.ibsrv.net/gimg/www.mia...f23d09c538.png