# Need help on high energy CDI (capacitive discharge ignition)

Hi everyone,

Trying to ignite a gas turbine, I decided to control it with an arduino, OK I could have gone with a NE555 to generate pulses but it's going to control some other things too in the end

So here is my thought : I'd like to control a circuit as the one below. It's basically a DC-DC converter charging a high voltage capacitor which is then grounded by a Thyristor and discharges through a coil (not represented).

Now, here's the deal :

My application needs a lot of energy, quite a lot more than a bike / car for that matter. Basically, a normal piston engine CDI will give anywhere between 50-200mJ and I need around 2 Joules. But also I don't need the high frequency of recharge! 1-5Hz is fine.

Doing the math, a 44ยตF capacitor charged at 300V would store 2 joules (doesn't mean it will give all of them back...).

A couple of questions for you CDI specialists out there :

• Is the circuit pictured correct to charge a cap that's almost 50times bigger than originaly intended? I'm also trying to find out the recharge time to check the max frequency but my knowledge stops there...

• The 22nF and 27R on the high voltage side are there to discharge the cap when powered down, right?

• On all the CDI schematics that I've seen, I've sometimes seen the 1N4007 in // with the power cap. Sometimes not and sometimes the other way around... What it's purpose to begin with?

• Last but not least : I was of course thinking of controlling the thyristor directly from the arduino... Is that safe in any way?
I was, of course, thinking of using an optocoupler but I've heard some bad things about it's rise/fall speed and that in HV ignition switching, this speed was of a prime importance...

But then again, I'm trying to make an indecrutible / as reliable as possible circuit (within my limited skills : -) so I'm open to any suggestions that could make it so!

Thanks for the feedback!

Marc

P.S : SportDevices. CDI Programmable Digital Ignition. to give back to Caesar!

Hi, 1 to 5 Hz, won't work, the 3845 has to run at high frequency.
Google how a switchmode power supply works, the high frequency is used for efficiency and if you need more energy you are going to need efficiency.
The thyristor can be switched at 1 to 5 Hz no problem, it will give the dc-dc time to charge the cap.

Tom......

Hi Tom,

Yes of course, I was talking about the frequency of the spark -> frequency of switching the thyristor.
I know the switching DC-DC needs HF.

Thanks,

Marc

marc426:
A couple of questions for you CDI specialists out there :

• Is the circuit pictured correct to charge a cap that's almost 50times bigger than originaly intended? I'm also trying to find out the recharge time to check the max frequency but my knowledge stops there...

The size of cap that can be charged is only limited by the leakage current of the cap,
more capacitance takes longer to charge.

• The 22nF and 27R on the high voltage side are there to discharge the cap when powered down, right?

They cannot do that as there is no DC path. They look to be a snubber circuit to reduce
EMI, or to protect the thyrister from dV/dt, but not sure.

• On all the CDI schematics that I've seen, I've sometimes seen the 1N4007 in // with the power cap. Sometimes not and sometimes the other way around... What it's purpose to begin with?

Can't really comment on a circuit I haven't seen - there are various topologies possible.

• Last but not least : I was of course thinking of controlling the thyristor directly from the arduino... Is that safe in any way?
I was, of course, thinking of using an optocoupler but I've heard some bad things about it's rise/fall speed and that in HV ignition switching, this speed was of a prime importance...

Opto couplers are fast enough surely! I wouldn't want to control this circuit directly from sensitive logic circuitry, the severe dV/dt and dI/dt values of the discharge are
enough to induce surprising large voltages around the thyristor and high-current
circuitry - optocoupling is one the way to go, otherwise you'll have to consider protection/
filtering circuitry in the gate driver.