Last semester I took a projects course offered by my colleges engineering program and I had designed and built a coil on plug conversion for my 99 Infiniti G20. Stock these use a distributor ignition system. The goal wasnt to produce more power, but to eliminate the mechanical aspect in the ignition system. The protoboard worked great and I was working on making my own pbc via etching and when complete and hooked up, I had a short and fried my pcb. I found and corrected the error in my schematic and decided to get some pcbs printed for quality purposes. On the bench the board tested perfectly and sequenced exactly like my protoboard and used all the same components. So when I hooked up the printed board, the car would not start. What I observed is that the board was able to identify TDC of the #1 cylinder, just as its designed too, but upon the spark event, the board glitches and is thrown out of sequence. I thought the glitch was isolated to my printed board so I installed the protoboard and the same thing happened. It seems that when the spark event happens, one or all of the IC's are affected even though all the input signals from the cam sensor are not distorted.

I have tried multiple coils with the same result. First being GM LS2 coils. These use either an IGBT or mosfet, but draw 20ma of current. Then second set of coils are the nissan p12 coils, these use an IGBT and draw 20ma as well. The third is the VAG coil which use a mosfet and draw around 600ma.

I am not sure if I am getting some sort of feedback from the spark, or inductive kickback from the igbt or the mosfet. I'm not really sure. I'm not looking for anyone to solve this problem as this has been a very complicated project, i'm just looking for some fresh ideas on how to approach this problem.

 

Welcome to AAC!
I'd suspect the supply decoupling and spike suppression arrangements.
Can you post a schematic of your coil drive circuit?

 

Welcome to AAC!
I'd suspect the supply decoupling and spike suppression arrangements.
Can you post a schematic of your coil drive circuit?

This is an older one but hardware wise its the same. The 7805 has a .33uf cap from 12v to gnd and .1uf from 5v to gnd. Each IC is decoupled with 50v .33uf caps.

Just to add, the coils I am using are logic coils which all use 5v pullup triggering.

 

Your circuit is clearly being affected by the emf pulses generated by the ignition events.
You are going to need to shield your circuit, filter its supply, and will also probably need to use multiple layers to shield its traces

 

reverse engineer the circuit housing of a auto that uses electronic ignition. A few dollars and a few minutes at a wrecking yard will get you plenty of examples to examine. metallic housings and ferrite beaded leads plus caps(sometimes a feature of a wiring feedthru) that have low impedance in the low MHz RF range will be present. Some hacksawing or dremel cutting wheel work may be needed to gain inside access.

 

Last semester I took a projects course offered by my colleges engineering program and I had designed and built a coil on plug conversion for my 99 Infiniti G20. Stock these use a distributor ignition system. The goal wasnt to produce more power, but to eliminate the mechanical aspect in the ignition system. The protoboard worked great and I was working on making my own pbc via etching and when complete and hooked up, I had a short and fried my pcb. I found and corrected the error in my schematic and decided to get some pcbs printed for quality purposes. On the bench the board tested perfectly and sequenced exactly like my protoboard and used all the same components. So when I hooked up the printed board, the car would not start. What I observed is that the board was able to identify TDC of the #1 cylinder, just as its designed too, but upon the spark event, the board glitches and is thrown out of sequence. I thought the glitch was isolated to my printed board so I installed the protoboard and the same thing happened. It seems that when the spark event happens, one or all of the IC's are affected even though all the input signals from the cam sensor are not distorted.

I have tried multiple coils with the same result. First being GM LS2 coils. These use either an IGBT or mosfet, but draw 20ma of current. Then second set of coils are the nissan p12 coils, these use an IGBT and draw 20ma as well. The third is the VAG coil which use a mosfet and draw around 600ma.

I am not sure if I am getting some sort of feedback from the spark, or inductive kickback from the igbt or the mosfet. I'm not really sure. I'm not looking for anyone to solve this problem as this has been a very complicated project, i'm just looking for some fresh ideas on how to approach this problem.

Some years back I made a basic ignition coil driver with a power MOSFET and not much else.

It was a 900V 9A MOSFET from a scrap 19" PC monitor PSU, it was wired "grounded gate", which means drain to coil LT, gate to +12V and source to the points. The main problem I was trying to solve was wetting current for the points on a small motorcycle with limited generator output - the current in the source (and therefore the points) was the same as the current supplied by the drain to the coil. Grounded gate switches *VERY* fast and the voltage across the points never gets over 12V, so energy lost arcing the points is pretty much nothing.

The downside was high-ish RDSon with a high voltage MOSFET and the cold cranking voltage only just made VGSthr. Low RPM improvement was minimal - but it could red-line in every gear.

If you don't have any points, you can just casc-o-de it by driving the source with a low voltage-high current MOSFET with ultra-low RDSon.

600V MOSFETs are much more common and easy to get hold of, but I'd suggest feeding the gate 12V via a resistor and clamping the drain to gate with a 600V zener (I believe such things exist specifically for electronic ignition) - if the drain spikes above 600V; the zener turns the gate on and it clamps it.

 

@ian field Ah, you were using a capacitive ignition system on your bike. How you had it setup is a lot like how the mega squirt driver board is designed. It uses DPAK drivers to control the ignition coils that draw too much current that the megasquirt cant handle.

@Kermit2 Great idea! I never considered the beaded leads. I'll have to look into this more. Thanks for the help!

@cmartinez I think you are right. I dont know a whole lot about PCB design, but I have been making and printed numerous designs to see what works best. I'll take it under advisement when I redesign my board. 4 layer over kill? Maybe top layer as power plane and bottom as ground plane? My latest pcb is 2 layer and the bottom layer is the ground plane. I hoped it would fix the issue, but still no luck.

 

@cmartinez 4 layer over kill? Maybe top layer as power plane and bottom as ground plane?

Are you going to try to make the 4-layer PCB yourself, or are you going to use an external supplier?
If you could make it yourself, I'd probably use the top and bottom layers as ground planes, and the middle ones for all the traces and power. Normally, in a 4 layer PCB you'd use the two middle layers as ground and power planes, respectively. But in your case, you want to shield your circuit from EMF pulses... so maybe a 3-layer circuit (if there's such a thing) would suffice.
Generous use of decoupling caps (0.1 µF) for every chip is recommended, as well as a few ferrites here and there... it all depends on your circuit and its layout.

 

Are you going to try to make the 4-layer PCB yourself, or are you going to use an external supplier?
If you could make it yourself, I'd probably use the top and bottom layers as ground planes, and the middle ones for all the traces and power. Normally, in a 4 layer PCB you'd use the two middle layers as ground and power planes, respectively. But in your case, you want to shield your circuit from EMF pulses... so maybe a 3-layer circuit (if there's such a thing) would suffice.
Generous use of decoupling caps (0.1 µF) for every chip is recommended, as well as a few ferrites here and there... it all depends on your circuit and its layout.

Ive been using OSHpark. But I need to see if eagle supports more than 2 layers. Thanks for the help with this. Hopefully through due diligence I can get this working properly.

 

@ian field Ah, you were using a capacitive ignition system on your bike. How you had it setup is a lot like how the mega squirt driver board is designed. It uses DPAK drivers to control the ignition coils that draw too much current that the megasquirt cant handle.

.

The only capacitor anywhere was similar in value to the original points capacitor, but moved to directly in parallel with the LT winding.

It probably would have worked without even that capacitor. The capacitor mainly delays the Dv/Dt so the points can part quick enough to quench the arc that would otherwise turn some of the energy into waste heat.

At the end of the day - the capacitor also stores some energy that can be put toward producing a fat blue spark, so I decided to leave it in. The capacitor reduces the peak voltage slightly at certain RPM, but the fast clean switching of the grounded gate MOSFET meant the peak voltage capability was considerably higher to start with - its highly likely that spinning the engine over without a plug gap clamping the HT lead, the coil probably would have suffered breakdown.

 

@ian field You really seem to know your way around automotive ignition systems.

I am glad this site has changed its policy on automotive related topics. I will post again when I have done some more research and more tests. Thanks all!

 

I took apart one of my spare ECU's and traced out the ignition driver circuit. The diode are not the same as the ones on the board simply because they were obsolete and finding accurate info on them was difficult. Also, the trigger input voltage has not been verified, however I do know the output voltage to be correct based on my own measurements of the ignition circuit. The only thing not shown is a single ferrite bead on the input from the MCU. The transistor is also not exact as it too is obsolete.

So this circuit is used to drive a single IGBT power transistor built into the distributor.

 

I took apart one of my spare ECU's and traced out the ignition driver circuit. The diode are not the same as the ones on the board simply because they were obsolete and finding accurate info on them was difficult. Also, the trigger input voltage has not been verified, however I do know the output voltage to be correct based on my own measurements of the ignition circuit. The only thing not shown is a single ferrite bead on the input from the MCU. The transistor is also not exact as it too is obsolete.

So this circuit is used to drive a single IGBT power transistor built into the distributor.

The 1N5820 diodes in your diagram are shottky-barrier - is that based on a DMM diode check Vf reading of the originals? IIRC: the 5820 has a PIV of only 20V.

Don't know about cars, but some motorcycles use entirely electronic ignition advance - the pickup coil feeds into a network that accentuates the faster rise time as RPM increases.

 

@ian field I just used those diode as they worked in multisim. I do still need to figure out which diode will work best. These cars use electronic ignition advance and it does use the coil collapse as a "check" function. Meaning it wants to see if the coil fired, however it is not needed to run and it is often bypassed when doing upgrades to the ignition. The advance is all controlled through the MCU based off of the cam sensor.