Ion Sense

[Simon@FutureProof]

So there's a couple of things here, your absolutely correct the spark discharge is a problem, the ion sense is essentially blind during the discharge which makes things difficult but not impossible.

I don't want to give away too much detail at this stage but we have found a way to determine the transition from spark to flame that allows us to get a better understanding of the ignition delay and remove a lot of that problem.

As Matt notes, I have managed to do some work with the SAAB conventional coils and their CDM module, they have given some pretty solid data and it's looking like we may be able to actually achieve accurate detection on a running engine relatively soon.

[sepp2gl]

@mck1117:

As far we I know, Mazda is doing both of these on their Skyactiv engines.

I will double-check Mazda's Skyactiv system, why they are using IS and which specific functions are inplemented.

We're most interested in #1.

This isn't too big a problem, since there isn't much overlap between the spark and the time we care about sensing cylinder pressure. There are also ways to force a quench of the spark (short out the coil primary) when you know you've delivered enough energy to the spark and want to start sensing now.

How do you find out, if sufficient spark-energy gas been provided?

Saab had two different systems: one using the "cartridge" which was CDI, and one using things that look like normal coil-on-plug with an external module. The normal coil looking ones are not CDI, and are very similar (down to the pinout and coil connector) to the currently sold ones on Mazda Skyactiv.

Yes I know, they used Melco-Coils in the later years.

Having said that, I would not declare it to be technically impossible. But it might be rocket-science for a hobbyist.

Our preliminary testing has shown very promising results

[/quote]
This statement was related to daveblanchard's project, not in general.

In general, combustion phasing using IS is highly challenging esp. at low engine loads, where the thermal ionisation is very low, at the same time the signal is pretty noisy so really locating location of maximum ionisation by means of software can give you some hard times.

@OrchardPerformance:
If a CDI can fulfill all requirements for engine operating points, it can be an enabler for IS technology.

Concerning ignition delay, the most influencing factor might be a high turbulence in the combustion chamber.
But all this highly depends on the combustion system and engine concept.

looking forward to some interesting exchange
T2US, sepp2gl

[bill]

Yes I know, they used Melco-Coils in the later years.

looking forward to some interesting exchange
T2US, sepp2gl

Welcome Aboard!
Looking forward to your help ....
bill

[sepp2gl]

...coming back to Mazda's "Skyactiv engines"...
Skyactiv is a brand-name, that stand for a couple of different types of engines with various technologies.
- Skyactiv-G is a gasoline engine with cylinder-deactivation, a very high compression-ration of 14:1 and for that they need to apply a piston with a bowl.
I guess, they use Miller/Atkinson-cycle to adjust CR14:1 at high load operation.
- Skyactiv-X is agasoline engine with homogenious-charge-compression-ignition HCCI for partload operation, which is supported by a spark to avoid
misfire operation under certain part-load condition and generally in high-loads. So you need a spark-plug anyway.
For HCCI-operation the tight control of combustion-phasing is crucial. As the engine runs homogenious-lean and ignition is initiated by pressure and
temperature in the combustion chamber, there are numerous parameters, that affect the activation energy. Consequently they need kind of combustion
feedback, which can be either Ion-Sense or combustion pressure sensors. Ion sense might be lower in cost, if it is "good enough".
Maybe Skyactiv-X also includes Skyactive-G technology for some reason.
One of the most challenging things with HCCI is the mode switch between HCCI in part-load and "normal" spark-ignition in full-load, which must not
degrade driveability (e.g.jerky ride) of the car.
- Skyactiv-D is just their name for their current Diesel technology.

So the name "Skyactiv" as such does no represent a certain engine technology, it is just a brand-spefific name.
T2US, sepp2gl

[mck1117]

All skyactiv engines with spark plugs have ion sense. I've personally scoped the ion sense line on my Skyactiv-G 2.0 in my ND2 Miata and it looks as expected.

[sepp2gl]

Can you tell, what functionality is covered by IS on Skyactive-G?
Is it more basic misfire & knock detection?
Or does Mazda use advanced functions?

[sepp2gl]

@mck1117:
...any reply?
What do you mean by "...looks as expected"?
What was your expectation?
kr, sepp2gl

[Simon@FutureProof]

In this case "as expected" would be in regard to the conventional ion sensor systems as used by SAAB etc.

image.png

Very similar to what we saw here with the SAAB system.

https://github.com/rusefi/rusefi/wiki/Saab-Trionic-8-Combustion-Detection-Module-on-Mazda-Miata-running-rusEFI

The irritating thing with the Mazda coils is we/I was not able to get a decent signal out of them off the car. Still unknown as to the reason why.

[sepp2gl]

@OrchardPerformance:
Looking to the diagram, the black trace is the typical shape of an ion-sense native signal, with the blue trace being combustion pressure as reference.
It is important to know, that you don't get this perfect shape under all conditions of engine operation: mid/high load operation should be easy, low load or idle conditions I wouldn't rise my hand for it. In any case it is not a trivial task to isolate the characteristic values out of the signal.
Especially with the differentiated signal (magenta), you might get extreme issues with noise.

I went into the GitHub-Link. It is hard to analyse the observed signals, if you do not know any engine operating conditions or signals taken.
At least the yellow trace looks like a native Ion-Sense signal, but heavily disturbed.
As the ion signal is very sensitive, it should be processed with high care to noise cancellation.

[JohnP]

The ion system for Harley Davidson was made by Delphi, but they went back to knock sensing in 2017.

"...the switch from ion-sensing knock sensors to cylinder-mounted ones dramatically increases the range of individual cylinder spark control the ECU has over the firing of the twin plugs, allowing as much ignition advance as possible. “So, you’re not creating heat by retarding the spark to keep yourself away from knock,” explains Bozmoski.

Since knock increases exhaust gas temperature (EGT) and in extreme cases can damage the engine, the Milwaukee-Eight’s ignition timing can be retarded right up to the point of the first signs of predetonation, which has no sound, and then be walked back until it stops. Managing ignition timing to the ragged edge of detonation delivers not only more power with less heat, but it also burns cleaner for fewer emissions..."

Apparently, HD says knock senors are more sensitive. This is the caption in one of the photos:
"Knock sensors mounted to each cylinder allow for the least amount of ignition retard possible without detonation."

Full article is here:
https://www.motorcycle.com/manufacturer/harley-davidson/2017-harley-davidson-milwaukee-eight-engines-tech-brief.html


Visteon is a Ford spinoff. They published a paper in 2008 on ion sensing for a "large displacement motorcycle engine", and showed their system was more sensitive than knock sensors.

It compares knock and ion sensing to in cylinder pressure sensing, as a reference:

Knock detection for a large displacement air-cooled V-twin motorcycle engine using in-cylinder ionization signals.pdf

[sepp2gl]

@JohnP:
very interesting paper from Visteon.
The findings were mostly well-known state-of-technology.
If you only look at the pure detection of knock-occurrence, there is a lot arguments for ion-sensing.
There are three alternative solutions know: vibration-based (knock sensor), ion-current-based, combustion-pressure based.
They all have their pros and cons to be considered.
But after all it is a commercial decision about which way to go.
As criteria for the decision, there is always
- product-cost on top-priority (at least for mass-production) followed by
- development and investment cost
- technical risk (new, technology is always considered higher risk), good/bad experience/reputation, is the "old" solution sufficiently good or not?
- availability of components (second source available?)
- support by EMS supplier (can be a highly political issue)
- business model (especially the depth of OEM-involvement into EMS-development; Make-or-buy)
- timing, can it be developed on-time for SoP?
There must be good reasons, why there is still knock-sensors as mainstream solutions on the market.

The criteria are differently ranked in importance, depending on if we target for mass production or small-scale production.
So for small-scale we can find different solution selected.
Harley-Davidson might have had good reasons for doing ion-sense and they had good reasons for going vibration-based now.
And of course they found a good advertising strategy to claim "It is better now".

If I was intending to apply knock-detection to my motorbike, I might first consider ion-sense.
Why? First of all it can be easily implemented by just electronics in the ignition coil and the ECU. And I could do it on my own.
There is no need for a special bold on the engine block for the knock-sensor, which needs to be carefully selected to allow equal-quality knock-detection for each cylinder.
And I wouldn't even consider to drill a hole into the cylinder head for the In-cylinder-pressure sensor.

But if I was responsible for a mass-product, I would compare the cost of commodity ignitions-coils, igniters and knock sensors with a special coils, igniters and no knock-sensor.

[Simon@FutureProof]

It has been my experience that the piezoelectric knock sensors are able to detect the onset of knock sooner than the ion signal simply because the ion signals seem to be insensitive to that early low amplitude knock oscillation.

[Niklas]

Hello Guys, why do we not use BMW-Ion-Sense Moddule from the V10?

Here some info from TIS.

I think the hardest part of work ist to write the code...

Code: Select all

Ionic current control units MSS65

Functional description, see Vehicle Technology Diagnosis:

110404097
Overview of functions

The DME activates the ignition coils via the ionic current control units. There is an ionic current control unit for each cylinder bank.

The ionic current control units make it possible to monitor each individual combustion and misfire and to detect combustion knock. Here, a distinction can be made by the DME as to whether misfires were caused by a lack of ignition sparks or poor combustion.
Function sequence

Ignition activation:via the signals S_IGZ1-10, the DME controls the final stages for the ignition in the ionic current control units.

    Signal low: charging the ignition coil, 1 ‐1.5 ms
    Signal high: switching the signal to high triggers the ignition spark.

Ionic current measurement:immediately after the end of the ignition sparks, the ionic current control units apply a constant measured voltage to the electrodes of the spark plugs. Here, the ionic current control units receive and amplify the currents that arise (ionic currents).

Via the signals A_ION1 and A_ION2, the ionic current that arises with each combustion is then sent from each ionic current control unit to the DME. The ionic current is 0 -20 mA and is evaluated in the Digital Engine Electronics. This evaluation enables the DME to detect misfires and combustion knocks.

The measured voltage is created in that the spark current that flows during the duration of spark combustion flows into a capacitor in the ionic current control unit. The energy that is stored in the process is applied after the ignition spark as measured voltage to the spark plug electrodes. The ionic current flows in the opposite direction to that of the spark current.

Signal amplification:the Digital Engine Electronics assesses the quality of the incoming ionic current signals and indicates to the ionic current control units via the signals S_IONV1 and S_IONV2 the extent to which the ionic current signals have to be amplified. The amplification factor ranges from 1‐5.

Measurement voltage selection:the Digital Engine Electronics decides depending on the operating range whether 80 V or 160 V is to be applied as measured voltage at the spark plugs. Via the signals S_IONS1 and S_IONS2, the DME sends the measurement voltage selection to each ionic current control unit.
Diagnosis instructions
If misfires are detected, faults are stored in the fault memory. The type of fault indicates whether ignition or combustion misfires were detected. 

Code: Select all

Ionic current measurement S65/S85
The following variables are measured for improved combustion monitoring:

    Spark duration
    Combustion quality

S65

The ignition coils incorporate an additional electronic module which can measure the spark duration and combustion quality.

The two measured values are measured via the conductivity of the mixture between the spark plug electrodes.
S85

The two ionic current control units measure spark duration and combustion quality.

The two measured values are measured via the conductivity of the mixture between the spark plug electrodes.
Spark duration

The spark duration is measured after the ignition spark is triggered. The conductivity is very high for the duration of the ignition spark; the ignition spark is identified by this. The aggregate is ms.

The spark duration varies within a value range of 0.1-1.5. The value drops to 0.1 when the ignition spark fails.
Combustion quality

The combustion quality is measured after the end of the ignition spark. The integral over the duration of combustion is created for this purpose. The aggregate is nSs (nano Siemens seconds, Siemens is the unit of conductivity).

The combustion quality varies within a value range of 0-5. The value drops to 0 when no combustion occurs.

[AndreyB]

Hello Guys, why do we not use BMW-Ion-Sense Moddule from the V10?

Let's give it a try!

What's the part number of that BMW module - is it available on eBay for under $100? Next step would be mating plug part number or sourcing used. Glad to see lack of CANbus between module and ECU that makes integration really easy!

[Niklas]

Hey,

Part Number is 13627834713

Unfortunately not under 100€

Another option is to use coils from M3 E9x with integrated electronic (0v is charging the coil, 5 v is ion current sensing)...


This post has some very usefull info:
https://rusefi.com/forum/viewtopic.php?f=4&t=2239

part number for the coil is 12137841754

[JohnP]

My '03 Chrysler 300M was knocking, so I put one of my Vampire knock controllers on it. It was still knocking, so I connected my scope probes to the knock signal and to one of the coils, and saw I was getting pre-ignition at the spark event. I know, Allen Cline has written that pre-ignition is silent. It's possible it was knocking which led to the pre-ignition, so that may be what I was hearing.

I observed the burn time to be less than 1.5msec at idle, dropping to 500usec or less under heavy load. Why would this short a burn time be a problem for ion sensing?

I did a search and found a patent on arc reduction, by Mecel, of course. I think it expires in 2028. You can see it also incorporates ion circuits:
https://patentimages.storage.googleapis.com/e2/ae/9c/49fc26667a29b9/US7347195.pdf

[sepp2gl]

@JohnP:
If your Chrysler has audible knock occurrence,
this might cause an increase in temperature in the combustion chamber,
and by that a "self-amplification" of heavy knock and finally also to pre-ignition.

I assume also the Vampire knock controller might need some adaptation to properly detect knock for your engine,
if it is based on vibration sensing.

The reason why spark-duration of 500µs might be too long for ion-sense-based knock detection is just the occurence of knock vibration,
which normally is between 5 to 30°CS after TDC.
Assuming, that your spark timing @6000rpm/WOT is 5° bTDC spark duration might end 13°aTDC,
and by that you might miss the spark oscillation to a high extend...

BTW: how much is the compression-ratio and displacement of your engine?

T2US, Gerhard

[JohnP]

Thank you for the reply.

My 300M has the 3.5L engine with 10.0:1 compression ratio. I've had it almost two years, and it didn't knock the first year. Might be my driving habits or choice of low quality fuel. 89 octane is recommended, but I buy it from 7-11 stores.

When I connected my scope, I didn't realize I had pre-ignition, I just saw big spikes out of the knock window. Normally, I have the knock window set to start 24° after the ignition trigger, with a duration of 45°.

So, panicking, I advanced the window 12°, then realized I was seeing pre-ignition. After advancing the window, I saw it was wide enough to detect actual knock near the end of the window.

I made a video of the display with my phone, and have attached a screen capture showing pre-ignition, occuring coincident with the spark event. At first, I thought this was an artifact or electrical interference on my probe, or a missing scope ground, but I am convinced it's actual pre-ignition.

Note you can also see the Vampire has retarded the timing from the previous cycle for this cylinder, when there was actual knock, that transitioned to pre-ignition.

The short spike before the ignition event is where the stock ignition was trying to fire, but the Vampire grounded that coil in time to prevent the spark, then released it later to fire the retarded spark.

The next day I drove through LA, to "The Valley" about 60 miles. Before leaving, I topped up with half a tank of 91 octane. To lessen the load, I drove in third most of the way, at 3000 RPM.

The next day I noticed most of the knock or pre-ignition was gone, and the car felt much stronger. I assume there was a build up of carbon that got blown out. Probably a lot still in there, so time for some additives or stronger measures.

I updated the Vampire some time ago, adding a pulse swallowing rev limiter, employing a large MOSFET. I've been experimenting with ways to reduce the burn time for a project I have been working on, and found a cool way to do it using the rev limiter transistor. That's why I went searching for patents to reduce burn time. It doesn't work as well as the one by Mecel, but there are no added parts. The drawbacks are there are limits, and of course, the MOSFET heats up.

I'll see if it's still working and post a video.

[sepp2gl]

Hi John,
to be honest: with a CR of 10:1 I would recommend 95 octane fuel in any case.
With the screenshots shown, I can see spark-voltages of kind of timing.
How do you detect knock or pre-ignition in these screenshots?
Concerning the Vampire-controller: I am completely unfamiliar with that unit.
Can you give me a link to a description/manual?

What do you do with the said MOSFET?
kr, Gerhard

[JohnP]

Thank you for asking about the Vampire. The first version was two channels, for four cylinder waste spark, in 1998. I made a small run of 100, primarily for the Mazda Miata. I got sidetracked for some reason, and dropped it. Then in 2006 I designed the eight channel version, primarily for the Ford COP ignition.

I made a prototype but never tested it on a vehicle. I knew the concept worked, as I had made them for the Miata. I did the board layout, had two printed, built one. I powered it up and saw the oscillator was running.

Shirley and I went to the SEMA show (booth in a briefcase), and showed SN 001 to Ford Racing. Naturally, I failed to mention I had no working code for it, so I was relieved when they said all their dynos would be busy for the next few months.

This was my first design using the 9s12 processor. I spent the next few months porting the code from an older version running on the HC11 processor. By the time I had the code working, in June of '07, Ford Racing had lost interest, saying there wasn't enough margin in it for them. I was offering 30% off list.

At the time, the only other option to control timing on Ford COP ignition (besides standalone ECU), was an MSD four channel CDI. TWo were required. To use on Ford, MSD made "Tach adapters" to prevent the ECU from setting ignition codes. And it was a wiring mess, as you had to cut the eight coils signals and connect the tach adapters.

With The Vampire, no wires are cut, and no tach adapters are required, as the stock ignition remains in place and does most of the work. The Vampire goes active only when it retards.

It includes an onboard MAP sensor if you want to retard in proportion to boost, a proportional nitrous retard function, and individual cylinder knock retard. And a staged or top end rev limiter, depending on how you set it up. In addition, a row of Knock-Finder LED's on the front panel show which cylinders are being retarded.

As I said above, I am exploring the possibility of using the rev limit MOSFET to reduce arc length, for use with ion sensing. But that part of my code stops working when I connect a real coil and fire a plug. It works fine when being triggered by coil simulating inductors.

[JohnP]

Video showing Vampire rev limit action, with experimental arc length reduction (not fully working).

[JohnP]

I got it working, ran the test, it works well, but the MOSFET reached 172°F in just a few minutes, running at 2900 RPM (6.8msec between spark events in six cylinder mode). Could possibly use a monster external module with heat sink, to quickly warm up the the engine.

Always a danger to post stuff as someone could patent it then block you from using it. Or if it's really good, just mass market you out of existence.

For the good of mankind (LOL):

[JohnP]

i was out testing for a couple hours, learned a lot. it's not pre-ignition. Note the signal is ALWAYS coincident with the spark, and it's not organic looking. It's always the same.

Then I remembered the bundle of wires I tapped into. The knock wires are just a twisted pair, not shielded, and they run parallel with all the coil negative wires. Touching them, even. And what's worse, I laid out the pcb for a "noise blanking transistor", but always leave it unpopulated. I'll fix that and try again this evening.

I have some video to go through, I'll post something after I take my cat to the vet. Nothing serious. Time to get her staples removed.

[JohnP]

Eighteen minute video from 3AM last night, showing how I come to the conclusion the large signal is ignition noise, not pre-ignition. As my tag line says, thinking is hard work, so take that into account as I slowly mull things over out loud.

I am pretty sure the artifact that looks like an inverted dwell signal on the knock trace is related to the large current going out the ground pin of the Vampire, when it is retarding.

I'm sure this brings up fond memories for Gerhard from forty years ago:

[JohnP]

I spent the last few days working on the noise blank problem. Upon close examination, I see my "noise blank pulse" happens about 20usec too late. I guess that's why I don't populate the board with the noise blank transistor. It didn't work.

So, I added a simple line of code to turn on the blanking pulse before the ignition event, but it doesn't work. I spent about three days trying to get this simple addition to work, then gave up and asked google Ai, and apparently it's a known issue with using TC7 to control another timer. Google gave me a few things to try, so I will work through them.

I'm using TC7 to control the retard pulse, and at the end it's supposed to control TC3 to stay high another 100usec beyond the spark event.

Here's a video showing what happens:

[JohnP]

This is the reply from google Ai. They made the same mistake about three times until I finally said "You are still wrong". They included my statement to them at the top:

[JohnP]

This is from mid 2022, when I was testing my prototype ion system on a 1990 Chevy Suburban. It's distributor ignition, so I used the ion sensing distributor cap and the processor box. The tentative name for it is the "Ion Probe", but that's too similar to what the aliens like to do to abductees.

The truck belongs to a machine shop I use, and had been sitting since 2012 with a dead fuel pump. I told them I would drop the tank and replace the pump if I could use it to test my system. I ended up also replacing the rusted out tank, the starter, and battery, but it wouldn't pass smog as it needed a new CAT.

The stock location is behind the rear axle, so it always had trouble passing, as it was never hot enough. They bought a new converter but have not installed it, so I could only test it in the parking lot, against the torque converter and brake. I don't think we ever got to 2000 RPM under load.

During the test, I recorded the ion signal from a test point inside my ion processor box, and the ignition signal, for playback on my test bench. I built a proto board interface for the DAT output and played the signals into the processor box, so I could observe the waveforms.

I installed a Digital MSD 6, so you can see two or three ignition events per cylinder, indicating we are at a low RPM.

You can see at several points the unit make good detections. I had the timing way advanced, and the factory knock controller was still controlling.

[mpgmike]

Looking at the Delphi Ionic Sense literature, as well as over a dozen university lab reports on the subject, I don't know how to interpret your scope patterns. Believe me, I want to know and am intrigued. What you showed doesn't match up with what would come up on a Google search on Ion Sense scope patterns. Please teach me more!

[mpgmike]

More specifically, there are conventionally 3 peak waves in Ionic Sense/Feedback of consideration: First wave is the spark event where the spark plug itself causes a major spike; 2nd is the flame kernel formed around the spark plug immediately after the spark event, and the most important is the pressure peak following the flame kernel extinguishing (3rd event). For the purpose of Ionic Feedback Ignition Timing Compensation, the 3rd event is the most important (albeit the least intense), as the center of that peak equates to the Peak Cylinder Pressure (PPP) in degrees of crank angle. Using Ionic Feedback to measure where peak cylinder pressure occurred on the previous combustion cycle, ignition timing adjustments can be made to positively affect the next combustion cycle by altering ignition timing (preferably with a PID filter).

It is commonly accepted that 17-18 Degrees ATDC is optimal for PPP. Thus, monitoring where PPP occurs, an active ignition feedback system can be employed to alter ignition timing to deliver peak performance and economy based on the Ionic Feedback results. With that said, as per my previous post, I am having a difficult time dissecting what you posted in reference to popular topic lab reports. Please -- PLEASE -- spend some time discussing what the waveforms mean pertaining to your hardware.

As a reference, I attached a lab report that illustrates the 1st, 2nd, and 3rd waveform peaks mentioned. The waveform I reference is near the top of Page 2.

A Method of Torque Estimation Utilizing Ionization Sensing Technology.pdf

[JohnP]

There are white papers and then there is the real world. What you posted are stylized and idealized drawings of results obtained in a lab, averaged over twenty samples.

The paper states they used a spark plug as the ion sensor, but the drawings show no connection to the ignition system. I am guessing this was a separate spark plug, used only for the ion signal, and not for making the spark.

What I posted are results from home built equipment on an actual vehicle, with the spark plug doing double duty, as part of the ignition system as well as being the ion sensor.

Yes, the results are very hard to interpret. At first I didn't think it was working, until I scrolled through the recordings frame by frame to see actual knock detections.