Ion Sense

[kb1gtt]

How do we not have a thread on ion sense? I guess we should start one as there is much content for this specific topic. We of course would need to start with these PDF's with lots of great details.

http://vvnet.fi/ville/ion/DIY-Ion-Sensing-2.pdf
http://www.aces.edu/~parmega/efi/temp/DIY-Ion-Sensing.pdf
http://forum.diyefi.org/download/file.php?id=201
http://forum.diyefi.org/download/file.php?id=200
http://forum.diyefi.org/download/file.php?id=199

http://www.fs.isy.liu.se/Publications/PhD/99_PhD_580_LE.pdf "Spark Advance Modeling and Control Lars Eriksson"
http://www.ieeecss.org/CSM/library/1998/oct1998/02-EngineFineTuner.pdf "An Ion-Sense Engine Fine-Tuner Lars Nielsen and Lars Eriksson"

Method of Ion Current Detection.pdf "Method of Ion Current Detection for HCCI Combustion on SI/HCCI Dual Mode Engine Guangyu Dong, Zhijun Wu. Liguang Li and Fafa Liu.Yingnan Guo"

A Method of Torque Estimation Utilizing Ionization Sensing Technology.pdf "A Method of Torque Estimation Utilizing
Ionization Sensing Technology in Internal Combustion SI Engines Guangyu DONG, Liguang LI, and Shui YU and Xusheng ZHANG"

Mazda Skyactiv http://www.mcx5.org/ion_sensor-1281.html
BMW v10 https://www.newtis.info/tisv2/a/en/e60-m5-lim/wiring-functional-info/power-train/engine-electronics/ignition/ionic-current-control-unit-bank-2/1qP23f6

I've been tempted to make a circuit for this, perhaps we should consider doing it now. You know give the software folks even more to do

[RasPL]

Some manufacturers are making this from few years, this article can help You
https://www.google.pl/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0ahUKEwjfv4TI68vJAhWEKnIKHXtjD4UQFggwMAE&url=http%3A%2F%2Fg-homeserver.com%2Fattachments%2Fharley-davidson%2F653d1312834419-mit-z%25C3%25BCndkerze-ngk-iridum-ix-dcpr8eix-klopfsensor-deaktiviert-10_in-cylinder_ion_sensing_opportunities.pdf&usg=AFQjCNF_veVQ4EOpxdK3qEEVEt4LmG4XuQ&sig2=p8PQWw07GyEFbUoVJtVTJg

[kb1gtt]

I see it's common you can get OEM igniters from ion sens systems for around $50 to $60. I wonder what those have for internal circuits, and if one of them could ease the design cycle,

Some key notes about ion sense. The coil is special-ish as most traditional coils form a low pass filter which blocks the ion signal. So you would need an ion sense coil just to get the data. I know Harley Davidson has been doing this for some time and those coils can be obtained for reasonably low $. Also I know Harley Davison has a tendency to use the ion sense signal to detect crank angle. AKA no hall and no VR. Any how that's a different thing.

The voltage that's applied across the spark plug during the non spark portion of the crank angle is around 80V to 300V. This supply can be created by catching some of the energy on the primary side of the coil, or it can be created by an external V generator. The signals energy of interest in the ion signal is both windowed after the spark and of higher frequency.

If we roll our own, and if we decide to do more in hardware than software, I would think we would want a high pass filter, followed by an integrating circuit. Basically the high pass filter would chop off the stuff you're not interested in, then the MCU GPIO pin would turn on the energy integrator for the windowed period of time, then after the integrator was done, the MCU would take one ADC reading. If you have lots of high frequency energy, you would see a higher ADC reading, and if not you would see a lower reading. This can be done reasonably easily with something known as a transimpedance amplifier circuit. What I don't know is how we get the ADC to work friendly. We are using many ADC channels's for things like TPS, etc. We only have 7 available after the default stuff. I guess we could require the 176 pin STM32 be used for 8 cyl ion sense applications. The low cost board we are using now, could handle a 6 cyl with ion-sense.

[Rhinoman]

I've designed one ion sense circuit, it was quite straightforward to interface to the circuitry in the coil.

[kb1gtt]

Was that the diode / cap trick, where you use the flyback to generate the DC voltage? How did you capture the analog information? If it was a reasonably straight forward circuit, I suspect you relied on significant signal processing from the MCU. With a little bit of hardware we can make this much taxing on the software.

[Rhinoman]

I'll see if I can find some details. The application was a diesel after-treatment system where the exhaust was injected with highly atomised diesel which was then ignited to burn off particulates and raise the exhaust temperature to make the cat more efficient. It was an unusual design of spark plug but a fairly conventional coil with ion-sensing circuitry built in. The output had some simple filtering and clamping and was fed into the ADC.

[kb1gtt]

I could see an ion-sense circuit looking something like this.

Ion_Sense_Draft.PNG (28.78 KiB) Viewed 23396 times

From left to right, you have battery, coil and typical arrangement for spark. Then diodes D1, D2 cap C1 and resistor divider R1, R2 which is what creates the ion-sense signal. Then you have a high pass filter from cap C2 and resistor R3 in the middle, which is followed by an integrator circuit. That analog switch is held low to keep the cap C4 discharged, then when you want to integrate the input energy, you open that switch and the op-amp will capture the energy, rectify it and store it in the cap. Then on the far right is a voltage follower, which allows for the ADC to capture the signal.

MCU operation would be, "integrator_reset" = 0V, until the start of the ion-sense window. When you start the ion-sense window, "intergrator_reset" = V+. Then when the sense window is done, you take the ADC reading of "Analog_Signal". After ADC has been read, "integrator_reset" = 0V in preparation of the next capture, and you start the cycle over again. The ADC reading indicates the cyl pressure. If the cyl pressure is to high, back off timing etc to decrease the pressure.

[Rhinoman]

I'm limited by confidentiality agreements as to what I can publish, however there is an applicable patent here:

https://www.google.com/patents/US8547104

The ion sense circuit is a resistor to convert the current to a voltage and some filtering. Given the amount of processing speed that we have available it shouldn't be necessary to integrate the signal, it depends on what information you want to extract. An alternative is a smart IGBT with current feedback such as the FGR3040CS (part number from memory!).

[kb1gtt]

You can pick up these ion sense coils from Harley for about $40 or OEM for about $65 Mine is marked Delphi 28058097 with a serial number of aE141028 48

http://www.ebay.com/itm/Harley-Davidson-Ignition-Coil-28058097-/351540572325?hash=item51d97370a5:g:~PgAAOSwuWVWEt1f&vxp=mtr

I toasted mine in part because I didn't get the matching connector. I see lots of connectors listed under ebay for about $10. But I don't know what one is the correct one or if any of them are the correct one. http://www.ebay.com/sch/i.html?_odkw=harley+delphi+28058097&_osacat=0&_from=R40&_trksid=p2045573.m570.l1313.TR0.TRC0.H0.Xharley+delphi+.TRS0&_nkw=harley+delphi+&_sacat=0

I could reasonably easily have soldered wires to the connector, then filled it in with hot glue. I didn't do that and got a bunch of arcing in the connector. I didn't notice it until it was toasted. See attached schematic that I created for it.

From the ohms meter, I find this coil is just 2 coils, it did not appear to have the diodes, but it would have the coils in such a way that you can obtain the ion signal.

[Rhinoman]

I posted the wrong link above, the coil was 'similar' to this one with a built in 5V analogue output:

https://www.google.com/patents/US8461844

[kb1gtt]

Nice. I think I should patent ion-sense but with the coil painted blue. The first article to me seemed like a real patent, it had a genuinely unique idea. However the second post was just the other patent, but using an op-amp as a buffer to the MCU. Some times what the USPTO allows boggles my mind.

Any how, the question is if we should try to solve the capture in hardware or in software. I know software is limited on time resources, so I'm tempted to make it easier for software folks by doing some of it in hardware. Doing DSP kind of stuff can be taxing on software resources.

[Rhinoman]

We do very little in hardware these days, except a basic anti-aliasing filter. The advantage of software filtering is that it has a constant time delay rather than a phase shift that varies with frequency, that means its easily compensated for when coding control algorithms. the other advantage is that you can buffer the signal so you can examine the raw signal rather than just the output from the hardware. The STM32 should be fast enough to be able to cope, is it a scheduling issue?

[kb1gtt]

I would have to leave the comments about scheduling up to russian. I know he has many software items he's working on and ion-sense is a bit lower priority for him right now, as there are other more basic features that need attention. I know that if I can drop the software development load by adding a pinch of hardware, that would make it easier for him to bump it up in priority. Also I know that if someone details the steps that need to happen for it to be implemented that also increases the potential for him to prioritize it. Most of all if some one has hardware that allows it to be tested in the real world, that's also increases the chances of him giving it a try. I'm kind of hoping I can entice his interests if I show the parts don't cost very much, and make a small PCB that would allow him to have a prototype in his hands. However first things first, figure out what should be pushed, then make that PCB and get it out there for software folks to start working with.

As for scheduling, I know we have a tick kind of timer. I don't know if that would allow for a proper bandwidth of the ADC signal. I suspect the current ticks are happening at to low of a rate to capture the higher frequency signal of a knock or peak pressure of an ion-sense signal. However that's also a bit of a guess. I would think for a software solution, we would likely need to develop a way of getting CPU cycles for the ADC processing that doesn't currently exist. It would likely be a significant effort for software development.

I guess we should detail how you could process the ion signal to detect things. Like in window, compare ADC's keeping min and max values. I guess we can figure out if we have hit a peak or a low, then the time between to detect if we see knock, but we would probably be better suited with an FFT bin or similar, as a detection of that frequency.

[kb1gtt]

I also wonder, does CDI and ion-sense exist at the same time? I like the multi-spark during combustion of CDI, but I also like the detonation detection of ion-sense. Can both be obtained?

[AndreyB]

I might stop by a junk yard this Sunday, I wonder if there is anything I should grab there.

I would love to use OEM hardware and I would love that OEM hardware to be as smart as possible.

http://www.speedsterclub.nl/bibliotheek/technische%20documentatie/csi.pdf maybe?

Also empty page to dream about well-structured info http://rusefi.com/wiki/index.php?title=Hardware:Ion_Sense page

[AndreyB]

http://www.speedsterclub.nl/bibliotheek/technische%20documentatie/csi.pdf maybe?

Probably not The cassette design would not fit in case of a different cylinder spacing. Any OEM coil-on-plug integrated designs?

[AndreyB]

2TC Designs 2 years ago
I'm usingan isuzu 00-02 Rodeo ignition module, which has ion sensing capabilities on 3 of the coils. Driving the ignition module with an Arduino, and supplying a pullup to the Ion Sensing channel so it can be read by a scope.

[AndreyB]

I'm usingan isuzu 00-02 Rodeo ignition module, which has ion sensing capabilities on 3 of the coils.

http://i.imgur.com/644SH1P.png

[Rhinoman]

I also wonder, does CDI and ion-sense exist at the same time? I like the multi-spark during combustion of CDI, but I also like the detonation detection of ion-sense. Can both be obtained?

Some of the early systems used CDI, I have seen one paper that states that ion sensing only works with CDI but that may not be true nowadays. I'm not convinced that multi-spark solutions offer any advantage.

Realistically it would be four months before I would be able to try this on a vehicle so its not a priority for me.

[kb1gtt]

See Delphi's pencil coils. They work for a range of 35mJ to 190mJ. I'm not sure where to find them or what they have for internal schematics, but I'm sure they are an OEM option for many engines.

http://www.delphi.com/manufacturers/auto/powertrain/gas/ignsys/igncoils
http://www.delphi.com/docs/default-source/old-delphi-files/0de42fb3-7a89-4d22-a0b7-2be357ceaf32-pdf.pdf?sfvrsn=0

Delphi_Ion_sense2.png (57.44 KiB) Viewed 23346 times

Delphi_Ion_sense.png (55.99 KiB) Viewed 23346 times

Saabs ionozation module seems to range from $50 used to $300 new. Again not sure about the schematic, or what's inside.

https://www.google.com/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#tbm=shop&q=saab+9-3+ionization+module

[AndreyB]

This marketing material does not mention any specific part numbers unfortunately, how do we translate this into actual hardware on hands?

I am interested to implement this code but I am tired of being the person soldering the first board and trying different resistor and capacitor values, it's just not my field. How do we get from where we are now to a youtube video showing ion sense signal an oscilloscope?

[Abricos]

For example I have distributor ...it will work???
or like spark plugs flooded with oil or gasoline or alcohol ???

[Rhinoman]

The coil is special-ish as most traditional coils form a low pass filter which blocks the ion signal. So you would need an ion sense coil just to get the data.

Does this mean that you have to use CDI? do you have any specs for the Harley coil L and R?

[kb1gtt]

I would expect that CDI may or may not be ion sense friendly. I would suggest using a coil that is specifically used on an ion sense application.

All the specs I have for that Harley coil are what I measured and wrote on the graphic.

[AndreyB]

First draft by @ available at https://svn.code.sf.net/p/rusefi/code/trunk/hardware/ion_sense/

[Rhinoman]

I was going through some papers yesterday and found another patent application. I looked it up:

http://www.google.com/patents/US20110210745

I'm going to talk to a Harley dealer and see what I can find out about their coils, most companies were closed over the Christmas period.

[Rhinoman]

First draft by @ available at https://svn.code.sf.net/p/rusefi/code/trunk/hardware/ion_sense/

I would highly recommend the use of an automotive quality regulator, the LM7805 isn't suitable for use in an automobile, except maybe as a post regulator.

[kb1gtt]

With some luck you might get hold of a broken Harley ECU. The magic of their system appears to be inside the ECU.

[kb1gtt]

First draft by @ available at https://svn.code.sf.net/p/rusefi/code/trunk/hardware/ion_sense/

Very nice, very nice. I agree about the 7805, however that can be easily changed later on down the road. For now getting this up and working would be a great step in the right direction. I'll try to look it over and start studying it.

[g68h50]

First draft by @ available at https://svn.code.sf.net/p/rusefi/code/trunk/hardware/ion_sense/

I would highly recommend the use of an automotive quality regulator, the LM7805 isn't suitable for use in an automobile, except maybe as a post regulator.

thanks for the comment. regarding schemes with a charge from the ignition coil - capacitor is quickly discharged and will not give an accurate picture of the ION current. to get a beautiful picture need a current mirror with external power.
excuse my English =)