Model engine CDI easy and cheap

[74Sprint]

I remain surprised that works. Of course, the processor is much faster than the mechanical bits, but even at 0.001 degree jitter (which astonishes me that you can get it that small - I'd have expected just simple slop in the crank bearings to contribute more variability than that!), that's still 10% of your (500RPM) pulse width. It doesn't matter how fast the processor can calculate/time things, if the RPM estimate is wrong, it seems like that'd play havoc with calculating the right value for the delay between seeing the pulse and firing the plug.

Unless I'm completely misthinking something, a 10% error in the pulse width produces a 10% error in the RPM estimate. And a 10% error in the RPM estimate should produce a 10% error in your delay-until-firing calculation. That could be 36 degrees...

Ah, but wait wait wait -- you're not predicting a delay for an almost complete revolution, your pulse occurs somewhere "close to but before" the earliest possible firing advance, so somewhere maybe 45deg BTDC? 10% error in that's 4.5 degrees, but maybe that's not critical?

Your right, I should have mentioned that crank fore and aft play (crank walk) does play a role. Some of the guys complained about this on the cheap China engines and they changed the code do like Eccentric has mentioned and did an averaging of the pulse timing. On the better built engines this wasn't a problem much once the engine got above 500 RPM. I now with the after market race car ignitions with a crank pickup if the sensor is too far away from the crank magnets then crank walk can show jitter at low RPM <2,000. Another problem with those is crank flex and detonation which shows up at higher RPM and load. With crank walk there is also less jitter depending on how the magnet and Hall are positioned. For example if the magnet is placed near the end of say the flywheel and is placed with the south pole pointing outward so it's on the edge of the flywheel instead of on the face with south pointing to the front or back of the engine, then there is less jitter and crank walk it not much of a factor. Unless crank walk is really bad, in which case it means something is really wrong with that engine. Almost everyone drilled a hole/pocket in the flywheel so the south pole was pointing out.

As far as the error is concerned what's the real difference between 500 and 550 to an engine, not much. If we're talking about a hit-n-miss, one can adjust the distance and orientation of the magnet/Hall to manage error or change either the magnet or Hall-Effect. We also found by adjusting the pullup resistor value helped with error a lot along with using a Hall-effect that wasn't to sensitive. The most error I saw was <= 5 degrees on the cheap China engines and it wasn't constant. You are kind of right in your later thinking because most of the time the pickup signal was usually around 45deg BTDC which, had to be entered into the software program. It needed to know where in degrees the pickup was. The more advance timing you need the more the pickup has to be away from BTDC.

Also (starting engine) 200 RPM x 360deg = 72,000 deg. per minute, 72,000 / 60 = 1,200 degrees per second, 1,200 X 1msec = 1.2 degrees, or 10% ERROR 1,200 X .9msec = 1.08 degrees, 1,200 X 1.1msec = 1.32 degrees. The amount of error can be corrected with testing it's a trial & error thingy.

One of the biggest problems to get rid of was the noise from the ignition spark. Things like the negative spike from the ignition coil caused anything from a second spark to burning up parts. Also because it was a CDI ignition we had to use 'snubbers' on the capacitor charging circuit (primary side of Txfrm).

Why do I feel like I highjacked this thread? sorry.

Cheers
Ray

 

[74Sprint]

Very interesting thought. Those changes within the program, even if you are doing some sort of reliable forward-predicting, can be a real bug-a-boo to solve. But as you say, how critical is it?
I remember as a teen ager watching the timing mark on the car flywheel bouncing all around under the timing light, as the advance weights twisted everything around. But the relative BTDC at various rpms was repeatable enough that I never had to spring for a big dollar distributor. There is a big distance between "it works", and "it works perfectly," and somewhere in there, the pragmatist has to say, "it works well enough." It goes along with the old saying, "It's time to shoot the engineers and get on with production." But I totally understand the obsession. It is our nature/curse.
Lloyd

Oh I hate it when the timing bounces around like that. I find it is usually related to worn out timing gears or the cam & distributor gear are worn out. Like the Chevy 4.3L V6, they are bad for wearing out early (razor thin teeth on the dizzy gear).

Ray

 

[willray]

Also (starting engine) 200 RPM x 360deg = 72,000 deg. per minute, 72,000 / 60 = 1,200 degrees per second, 1,200 X 1msec = 1.2 degrees, or 10% ERROR 1,200 X .9msec = 1.08 degrees, 1,200 X 1.1msec = 1.32 degrees. The amount of error can be corrected with testing it's a trial & error thingy.

If I understand the error (in degrees) you stated a few messages back, the error is in the measurement of the width of the pulse, not in the the motion per msec.

A 10% Error in the measurement of the width of the pulse, unless I'm completely misthinking, will produce a linearly proportional error at any other width you use that measurement as a scale for: So if you use a width measurement with +/10% jitter error to calculate how long you have to wait for the crank to rotate 45 degrees, it could really be rotating anywhere between 40.5 degrees and 49.5 degrees. That's enough to move an intended 3deg ATDC starting timing, to firing 1.5deg BTDC (which could be spelled "ouch"), or to 7.5deg ATDC, which is probably OK.

Of course, all of that is entirely "on paper". You've done it and it obviously works, so either I've gone sideways somewhere with the on-paper thinking, or some combination of practical reality works to make all of this a tempest in a teakettle!

Will

 

[74Sprint]

If I understand the error (in degrees) you stated a few messages back, the error is in the measurement of the width of the pulse, not in the the motion per msec.

A 10% Error in the measurement of the width of the pulse, unless I'm completely misthinking, will produce a linearly proportional error at any other width you use that measurement as a scale for: So if you use a width measurement with +/10% jitter error to calculate how long you have to wait for the crank to rotate 45 degrees, it could really be rotating anywhere between 40.5 degrees and 49.5 degrees. That's enough to move an intended 3deg ATDC starting timing, to firing 1.5deg BTDC (which could be spelled "ouch"), or to 7.5deg ATDC, which is probably OK.

Of course, all of that is entirely "on paper". You've done it and it obviously works, so either I've gone sideways somewhere with the on-paper thinking, or some combination of practical reality works to make all of this a tempest in a teakettle!

Will

I think your over thinking this. Everything is either based on measured time or clock counts. Clock counts are more accurate at 16Mhz. Unless you have severe vibration there is almost no jitter to speak of. All the R/C ignitions I have seen usually have less than 1 degree of jitter, depending on how it is mounted. As for the pulse width we are talking about nano seconds to micro seconds in possible jitter. As for the pulse width I WANT to see the pulse width shrink with RPM. The Hall-Effects I use are very accurate and repeatable in results. I'm going to be putting my degree wheel, made from an old DVD drive together with my grand-daughter right away. Using a 13,000 rpm DC motor and the degree wheel (Blank DVD) has 2 magnets, so Ill see how that goes and I'll try to simulate vibration. Wait for the movie hopefully later today. I hope to show the pulse width changing with RPM and that the leading edge point/degrees never changes. Ah crap I forgot, I have to make a new timing light, I borrowed parts off the old one.

Ray

 

[bluejets]

Bit of free time so whipped up a mod circuit using PaulC idea but different bias arrangement and also dc block cap.
If one needs to use a remote Hall effect 3144 type to reduce size of the pickup hall as apposed to the KY-003 module, one could simply remove it from the KY-003 module or use a descete part and add any indicators themselves.
If using the module, it has all the necessary components on board except for the 0.1uF capacitor.
Relay i used was a small 12v unit, details on the drawing but just about any small coil would suffice.
This particular one was chosen due to coil resistance of around 700 ohm.
I cut a small hole in the top of the case and filled it with epoxy, probably not necessary but did it all the same.
My opinion would be that this type of coil would produce a better pulse due to it's inbuilt core whereas the reed coil has an air core.
As best I could tell with my little DS150 scope, trigger pulse is around 8-10V.

View attachment 128785

View attachment 128786

Currently preparing a JLCPCB board to cover this arrangement as there have been a few calls locally interested.
5 boards with approx. 15 panelized on each, so plenty and at that rate relatively "next to nothing cost per board".
Will advise when I have completed the arrangement.

 

[paulc]

Currently preparing a JLCPCB board to cover this arrangement as there have been a few calls locally interested.
5 boards with approx. 15 panelized on each, so plenty and at that rate relatively "next to nothing cost per board".
Will advise when I have completed the arrangement.

Good one! I like the original mini relay packaging, neat!

All the bench tests I did here were impressive and I could stretch the spark about 1" and still push 18,000 spark per minute on the lathe. So I thought what the heck for giggles lets hook it up to one side of our 9 cylinder radial's ignition which at WOT runs 3500 RPM x 4.5 sparks / rev = 15,750 sparks / min. Now we do have compression in play now but the spark plug gap is no longer 1". Anyway at high power and RPM the engine would break down, which surprised me. There is also the possibility of interference from the 18 spark plug leads in the local area and I was not sure if this could possibly be causing some false triggering, although this does not happen on our stock Hall effect ignition (kettering) which used the same hall sensor. So I suggest the ignition has hit its limit. I tried a variety of CDI coils, some worked better than others but all broke down at WOT.
There is also the factor that we run a distributor cap with carbon brush etc etc, so even more work for the ignition.

But with all that aside this ignition would be great for a smaller single cylinder or lower cylinder count engine that did not push the sparks /min too high.

My gut tells me that CDI would work best for coil over plug, and so only use one tiny coil and CDI per cylinder and not ask one CDI and coil to do so much work?

All good stuff.

 

[paulc]

You could consider developing a board that packages the small coil and HT lead all built in. Then you just plug in the CDI and connect the Hall lead and the spark plug lead to the engine.

 

[bluejets]

Ok...thanks for the review.
Multi cylinder or at least 9 cyclinder were not what I had in mind during the initial setups nor was it the original intention, however, good to see your ground work.
Most applications are for the average engine builder who would seldom ever go above a single running at maybe 4-5000 rpm at best.
I do intend a little further down the line to try it on my V8 design (still sitting in castings as yet) and even then, not difficult to do a 2 x black box design.
Mostly used for marine application and again, if it runs over 3 to 4000rpm I'd be very much surprised.
My little 25cc 4 cylinder in the Rivierra runs to 10,000rpm but glow and alcohol fuel but again, bench runs not practical use, more like maybe 4 to 5,000 as well.
Cheers Jorgo

No intent to do any coil, ht stuff.....this was an exercise in getting off-the-shelf cheapo stuff working.
Plenty of room for the original design, even with 2 black boxes...

 

[giel]

why not use a rcexel as a base?
there ia a scematic on internet that can use 12 ignition leads controlled with Arduino nano. with easy adjustable dwell angle rev limiter and easy start mode, using only 1 hall sensor and a knocking sensor can be added.

i use it on a joined 2 cylinder zenoah g4k pushing 8.5hp at 17000rpm
can be used as only a controller with coil on plug system ( i use from bosch for honda they are not much bigger than the sparkplug head
or with custom coils that are in the scematic.. if you're interested just let me know and i look it up..

 

[bluejets]

why not use a rcexel as a base?
there ia a scematic on internet that can use 12 ignition leads controlled with Arduino nano. with easy adjustable dwell angle rev limiter and easy start mode, using only 1 hall sensor and a knocking sensor can be added.

i use it on a joined 2 cylinder zenoah g4k pushing 8.5hp at 17000rpm
can be used as only a controller with coil on plug system ( i use from bosch for honda they are not much bigger than the sparkplug head
or with custom coils that are in the scematic.. if you're interested just let me know and i look it up..

Just post the reference you have .....people will do with it what they like.
The more the merrier, someone might get some ideas from it.
Cheers

 

[bluejets]

Submitted the board design to JLCPCB for review before payment and there were a couple of issues with the layout.
A bit new to this online pcb stuff so understandable I guess.

Long and short of it, cancelled that one layout and presented another for review and it was acceptable.
Panalized out at 2 columns x 4 rows x 5 panels so still cheap as at less than AUD$10.00 boards cost and a bit over $10 for postage (20days).
So around 50c a unit.

 

[Stefan-K]

[QUOTE="giel,
if you're interested just let me know and i look it up..
[/QUOTE]

Hi there
I'm interested in this scematic. Where can i find it?

 

[bluejets]

why not use a rcexel as a base?

For starters they cost around $80 AUD.

 

[paulc]

why not use a rcexel as a base?
there ia a scematic on internet that can use 12 ignition leads controlled with Arduino nano. with easy adjustable dwell angle rev limiter and easy start mode, using only 1 hall sensor and a knocking sensor can be added.

i use it on a joined 2 cylinder zenoah g4k pushing 8.5hp at 17000rpm
can be used as only a controller with coil on plug system ( i use from bosch for honda they are not much bigger than the sparkplug head
or with custom coils that are in the scematic.. if you're interested just let me know and i look it up..

Hi There, I'm interested in this. Please send me any links or references you may have.

 

[giel]

Arduino Based CDI single cylinder.. a few hours of reading..


Arduino Car Ignition System... the Carduino!.

here is a car version.. but just google there is a lot to find on the topic!

and the rcxel code is also available on the i net but you will need to have a pic chip programmed by someone or buy a programmer (€5)

http://ot40.com/cdi_ignition.htm
rcxel ignition 1 cylinder is 15€ and 5 cylinder is €50 now on ali expr

one drawback on them is they do not have a programmable dwell angle and knock sensor.. on the Arduino versions there are versions with buttons to preset max rpm adjustable dwell angle
there even is a 2 /4 stroke efi version with piezo injectors so you don't need a high pressure fuel pump ..

 

[Stefan-K]

Hi Giel
The Links to the Websites you Provided are already ancient. And so are the ignitions. And most of the members here should already have seen them. Either they are way too big to fit in a Model engine or they simply dont have what has already been discussed here. Im interested in the RCexl ignition scematic and the Code for the Chip inside. Afaik there is none of IT to find on the i-net. If it were that easy.... just a Look at Google and youre done we would not need the discussion and all the work that the members Do here.

 

[giel]

use a arduino micro version?? need only 6/10 wires to the arduino i included a pic version of the rcexl cdi they analized the dwell angle from rcexl back in 2003 out of .y head if you want to have a 1:1 copy just buy one! at hobbyking they are €25 here in nl you can't build one for that money including hours..

before i wanted adjustable dwell angle i just used a coil on plug and a hall sensor.. and 3xlion. batteries...
adjust the hall sensor until you get the max rpm at your desired throttle setting..

costs €1 hall sensor and magnet and €2 for a coil on plug at the breakers weighs 50 grams with wires and batteries 100 grams and your ready..

 

[Stefan-K]

Really?? What a comment! "Go to Hobbyking and buy one"?
I believe you didn't get me.
It's not about just having or buying an ignition. But anyway, i stop talking about this now, knowing its another worthless comment. Nothing But warm air......

 

[bluejets]

Arduino Based CDI single cylinder.. a few hours of reading..


Arduino Car Ignition System... the Carduino!.

here is a car version.. but just google there is a lot to find on the topic!

and the rcxel code is also available on the i net but you will need to have a pic chip programmed by someone or buy a programmer (€5)

http://ot40.com/cdi_ignition.htm
rcxel ignition 1 cylinder is 15€ and 5 cylinder is €50 now on ali expr

one drawback on them is they do not have a programmable dwell angle and knock sensor.. on the Arduino versions there are versions with buttons to preset max rpm adjustable dwell angle
there even is a 2 /4 stroke efi version with piezo injectors so you don't need a high pressure fuel pump ..

Your link to "CDI Ignitions" is as old as, never worked properly and was the reason behind several more advanced attempts to rectify.
"Carduino" I never saw before but "instructables" have a reputation of failed or incomplete or untested systems.

 

[bluejets]

As a lead on from post #99, I received the newly designed interface boards and made a short video of the working cdi with mods similar to PualC.

Note I did not use his extra transistor or reed coil, instead I used a small relay direct on the hall signal open collector output.

Works well as can be seen.

I'll drop the files for the board in here later.



Cheers Jeff