Knock-window...it is time
 

Will this ever happen?

In MSIII yes 100% definitely, I think it's scheduled for the code 2.xx release on MSIII. The coders are writing a new clock angle algorithm thingy (way over my head) which will allow this to happen.

in MS1/2 no never ever ever.

I believe it has something to do with your sig quote.

MS3 alpha code has some awesome knock stuff going on. When Ken gets his angle clock code done, it should get even better.

How about 2D rpm filtering and user-adjustable frequency filtering?

What is knock window?

There is already a sensitivity vs RPM function.

That would be a hardware function of the knock conditioner. Keep in mind current MegaSquirt hardware does not include an onboard conditioner--You have to use an external box such as KnockSenseMS or a GM ESC module.

Looking for knock only at specific crank angles when it can occur. It helps to eliminate false knock signals. It can also be used to determine exactly which cylinder is knocking.

I plan to offer this on my next generation of Megasquirt II ECUs.

Unless of course, I have say in this. ;)

Most definitely. The frequency will be fixed based on whether this is a 1.6 or a 1.8, and there weill user-adjustable RPM filtering.

Just

:bowdown: :bowdown:

Now that's disappointing. I was under the assumption one of the spare analogue MS3X inputs would be able to be used to analyse the knock signal.

Relying on a KnockSenseMS is pointless, it doesn't rpm filter so you just get noise induced knock BTDT. No amount of MS trickery will stop a poor input signal.

Shame.

That's pretty darn cool. Does this exist in any EMS's out there, factory or aftermarket?

6.5 kHz bandpass filter is appropriate for the miata. And then said knock window.

I have a schematic for a simple bandpass filter somewhere. One could maybe add it on board a MS.

EDIT: Realized this is for MS

ProEFI has this...

yeah a good amount of the more expensive EMS solutions have support for knock windowing.

AEM has it and AEM is fairly inexpensive.

I meant more than megasquirt costs, which I paid about 300 after I did mine up

I use a J&S safeguard on my miata for knock detection and retarding per cylinder
Got it for a good price so I jumped on it.

This thread has made my upgrade to MS3 worth it. Thank You!

For people who don't know how knock-window works:
https://www.miataturbo.net/attachmen...ine=1315368354

An analog input reading the knock signal won't work. It doesn't have enough sampling speed to use an analog input for knock processing.

The knock sensing will use some more sophisticated hardware - I wouldn't count out user specified frequency ranges or other features.

The windowing not only ID's which cylinder is knocking, it filters out noise that occurs outside said window.

? So how do other EMS's analyze the signal? Does the analog voltage signal go through an A/D converter and then through a DSP chip before feeding into a closed loop "burned in" timing trim?

I'm not trying to derail the discussion from MS; I'm just curious about how this stuff works.

That "filter" could be called infite I guess :).
Why bother listening for knock when there can't be any.
So it only listen for knock in the small time windows, otherwise it just don't care (smart way to deal with noise).

Would it work with less exact crank angle sensors (e.g. OEM) or would it require a 36 tooth wheel or similar?

Btw have anyone tried Fourier Transform analysis of knock (as a chemist I'm quite familiar with NMR, which in principle is analysis of the squeel molecules sends out after you kick them)?

No, it just goes like this:

Knock signal -> bandpass -> Amplitude Detector -> windowing -> software

Up until the windowing, it's just like an AM radio, which was invented, oh, like 100 years ago. I can design one using a single 8-pin op-amp package. The windowing will be triggered by the software and can use a "CMOS switch".

Here is another solution for anyone that wants knock windowing

http://cgi.ebay.com/ebaymotors/Knock...item3a69ed9532

SafeGuard with individual cylinder knock retard was released in 1991. Features included constant energy dwell control with internal igniter and adjustable rev limit.

The detector employed a knock window, but this feature was not disclosed until about 2003.

The processor was an HC11, running at 2MHz bus speed. All the code had to fit in the internal 2k EEPROM.

The detection and thresholding algorithm is elegant, wish I could show it.

I was doing a spark energy test earlier this year, and fired up one of the first gen units, circa 1995. It made 90 mj firing a plug, 109 mj dumping the coil into a high voltage zener string:

http://www.youtube.com/user/strokema.../6/wHtI1FOBKJ4

Fuck windowed knock detection. We need ion sensing capability. Why has that never progressed?

Because those that have the technology try to keep it for themselves? On the other hand, the MS-II had things like nitrous control, but no A/C control. The reasoning? Nitrous is used by drag racers, who help get the name of the ECU out more than the average Joe and his POS daily driver that has A/C on it.

Here you go:
http://vvnet.fi/ville/ion/DIY-Ion-Sensing-2.pdf

Could you take a look and see where the really big hurdles are?
Just getting things to the point where we can start measuring ionisation profiles with a small community of MS users could be enough to kick things off. Dunno how realistic that is, but auto-tune for spark and fuel based on current conditions would be nice.

or AMAZING!

Is the idea that you used the wideband, the knock window,a dn the theoretical point of diminishing return to autotune spark? It sounds liek the knock detection is the most complicated part. There seemed to be plenty of theory in one of Hustler's threads about tuning spark and where the point of diminishing return is located in the spark table.

I'll donate to a megasquirt 3 compatible PCB design.

I've got easy access to 12V, regulated 5V, ground, and whatever comes from the factory harness connected to the sensor inside the MS3 box. keep the overall height to under 3/8" and square footage to less than an inch and a quarter square and it should fit easily.

Input would be the sensor itself, output would be an analog signal strong enough to clearly distinguish knock from noise and place the threshold line appropriately. I think it's 0-5V but I'd have to check with the dev dudes. Ken tells me the analog portion may not even be operational yet.

Windowing is all done in software.

Whattaya say? :)

The idea is that you measure when peak cylinder pressure occurs and you tweak ignition timing so that the cylinder pressure does the maximum amount of work during a cycle. The result would be more torque, which is what look for while tuning spark on a dyno.

The knock sensing capability you get more or less for free as the whole ionisation profile goes to hell during knock and is easily detected.

I dunno how accurate you can get this stuff, so maybe you would still need to calibrate on a dyno. But perhaps tuning/correlation could be done quicker, or it could be used to monitor and correct for changes in circumstances: fuel burn rate (octane, temp, additives), humidity etc.

Saab was running ION sense a long time ago. BMW is using it on most cars in their new range. Maybe it is to far of a stretch if even windowed knock detection is beyond our capabilities.

I would also donate money to this project. Brain is working on my MS3x at the moment. If this is something that could be implemented quickly, maybe he would be willing to use my MS3x. That would eliminate the need to buy a whole MS3 test mule.

The downside is that you would be relying on my inexperience and I would need some hand holding for testing. My wallet is open to give donation.

Even if it didn't end up being the end all be all of spark tuning. It would be nice to have the autotune feature with a pair of electronic det cans as back up. Spark is increased until knock is detected. Then based upon a little theory the spark is retarded by a predetermined number of degrees. Those that want to be more aggressive can do as they wish.

I'm not sure if this is possible, or detectable. BUT here goes. If it is based on peak cylinder , is the MAP sensor sensitive enough to notice the change in pressure if the engine is held constant but the timing is being changed? So if I were able to get on a flat piece of highway in the gear of my choosing to keep the car at a steady rpm and then the timing were adjusted, would you see a difference in map signal?

Where I am headed with this: Could you then have a test area in tuner studio that allowed you to like your pulsewidth for a predetermined number of seconds. Basically a tunerstudio cruise control. Once activated tunerstudio listens for knock while adjusting the timing.

How would the MAP sensor be able to see the pressure inside the cylinder, since the cylinder is sealed and any pressure is vented through the exhaust valve first?

I was wondering if the ouput through the turbo would be slightly increased. Higher cylinder pressure should equal higher exhaust pressure, which powers turbo, which plumbs into intake.

No, when spark timing is less retarded from optimum, cylinder pressures go up and then less energy is in the exhaust. Turbine Inlet Pressure (TIP), goes down. And no, there are too many other variables affecting TIP more, to be useful for spark timing tuning, even in an n/a car.

Fail because you can have too much timing and NO knock.

The one and only ECU that had a "best timing finder" feature is the Adaptronic, and mysteriously, nobody was interested in the feature. Such that they removed it in later revs:

https://www.miataturbo.net/adaptronic-63/anyone-play-ignition-auto-tuning-yet-42882/
http://adaptronic.com.au/forum/index...ic=252.msg7487

The way it does it, is it adjusts the timing several times per second, then detects if the acceleration improved or not. No mean feat.

Someone else has to do the PCB layout because I don't like doing it.

put it in an eagle pcb compatible format and i'll do it... or maybe altium and I'll coax a coworker into doing it.

That is where retarding the timing based on where the theoretical optimum timing is based on knock. If research suggested that knock occurred 5 degrees past the optimal timing configuration, then retard timing by at least 5 degrees after knock is detected.

I know I should probably stay out of this conversation since my expertise is zero, but I'm learning something.

Is one of the GPS transponders that people are using for lap timers sensitive enough to use it for calcing timing based on acceleration? Probably not. It seems like a speed sensor used while locking in a fixed pulsewidth would work. Question is, where do you mount a generic speed sensor.

Unfortunately, research makes it clear that there's no fixed angle between optimal timing and the knock threshold. Some cars (Subarus are Exhibit A) have the MBT timing ahead of the knock threshold. Some combination could have MBT 20 degrees or more behind the knock threshold - I suspect a NA Miata running on methanol would fall into that category, for example.

That's unfortunate. So is a speed sensor method the best solution?

Yes, a knock window is easier to do in software. The first SafeGuards were a combination of analog and discrete digital, made from '88 to '91. It didn't do individual cylinder control, but it did have a knock window. The knock window was controlled by a PLL clock synchronized to the ignition trigger. Two counter chips set the window start and end points. Dip switches on the front set the counters for 4, 6, or 8 cylinder.

I love doing layouts, but they take a fair amount of time. Good layouts can be works of art. I've found that if they don't look pretty, there's probably a better way to do it, so I start over.

This is part of the Interceptor layout, showing the Altera CPLD chip and the 9S12 processor. The CPLD contains I/O, trigger detection, and retard logic for eight channels. A single line to and from the processor is all that's required to individually control eight ignition channels.

http://www.jandssafeguard.com/layout.png

John, how come you know so much about J&S? Is it your business? ex Employee?

just curious!

Coming to you from World Headquarters.

http://www.welovewhq.com/

http://www.viewnewcastle.co.uk/clubs...nfo-47177.html

:confused:

;)

Not THAT whq. Looks like something from the Great Whole Earth Catalog (pre-internet era).

Thanks for chiming in, John!

So current the SafeGuard uses a similar processor to the MS2? Interesting.

Ben and I have been testing some of the future MS knock detection stuff on the dyno today. We're not going to sacrifice a Miata engine for it, but we've got an old beater Sentra sitting around to try this out on.

Change is hard. I finally made the transition from HC11 to 9S12 in 2007.

Adaptronic method seems best outside of a dyno.

Does it combine it with some sort of cruise control so that any observed acceleration/deceleration is due to the timing adjustment? Obviously we would be assuming this would be done on a level road. Mounting an optical sensor system to one wheel doesn't seem like it would be that difficult. Add an input where you can lock the pulsewidth of the injectors. Pick a long level road, or inclined for boost cells. The boost cells seem like they would pose the most trouble, but they are the most important in my eyes.

I think the optical speed sensor idea might fall apart while you are under heavy acceleration to begin with trying to get into boost. How could one tell if the acceleration was due to timing or accelerator pedal movement.

Ok, so here is an idea.

The software would require you to do record multiple acceleration runs. This would log data that related TPS to acceleration through different rpm points. Done on the same road as the tune and you might be able to pinpoint the data changes due to the timing adjustment alone. This would require the user to treat it like an absolute calibration, meaning Temp, pressure, road used, etc all need to be monitored an held somewhat constant between the baseline data and the adjustment data.

No it looks at the RPM. It dithers the timing continuously, rather quickly (several times per second) at some frequency. It looks for an FM signal with the same frequency in the RPM. The sign and magnitude of that signal tells you if there was an increase in torque from the increase in timing.

Re: KNOCK WINDOWING

I just had a realization.

The time windowing should be BEFORE the filter.

The reason is that a bandpass filter rings like bell when stimulated by a noise with in-band frequency components. And the reason knock is hard to detect is that engine noise has some in-band components. Therefore to maximize discrimination, you don’t want engine noise reaching the filter outside of the crank-angle window where knock isn’t possible.

If the ECU had an output that simply went high when crank angle is “in window”, that signal could be used by the analog circuitry.

Uninformed question;
Could one of the cam wheels be modded so a hall sensor could go high/low during the knock window (if a section of the 720 degree cycle would be good enough, i.e. fix window in relation to rpm/load/TPS etc).
That way you could window the knock signal at the source.

Software is still needed to figure out what to do when,but even with simple knock check, you could stop listening during the times when there "cannot" be knock.

But uninformed, as I said.:)

A spare cam wheel? If you could make your own window, yea.

Reverant could add this functionality, if we could find out what crank angle range knock is valid. Obviously it has to be <90*, else it would be meaningless, unless one were doing per-cylinder knock detection and spark adjustment.

They ditched it? I'm amazed. That was the one really cool feature that made me regret not buying an Adaptronic.

Here's an update:

http://adaptronic.com.au/forum/index.php?topic=252.15