Spark gap to ground causes large spikes on other voltage rail.

[king.oslo]

Hello there,

I have this circuit. It is attached to this thread. It is a simplified diagram of what is happening. If it is helpful, I can post the full schematic.

I have a transformer which transforms voltage from 15V to ~20kV. The current will jump a spark gap. If the spark gap is narrow enough for the current to jump, I find huge spikes on my other power rails.

However, once the gap gets wide enough, no jurrent jumps the gap, and no spikes appear on the power rails.

These spikes cause my MCU to reset each time. I dont want this.

So I need to find a way to be able to jump the spark gap and fire repetitive sparks.

I tried separating MCU ground and spark gap ground by an inductor, the spark would no longer jump the narrowest of distances. Any idea why the primary voltage refuses to rise with the inductor in place?

Or ideas how I can proceed to achieve my goal.

I came to think, this is probably very similar what happens in cars when the ECU tries to fire a spark plug. How do car manufacturers solve this problem?

Thank you for your time.

Kind regards,
Marius

[SeanB]

Are there any decoupling capacitors across the supply rails, you will definitely need at least 1000uF 35V across the 15V rail right next to the transformer and the FET, along wityh some 100n ceramic capacitors as well to handle the fast pulse edges. Preferably connect the top of the secondary to ground, so the spark energy does not go through the supply rails.

[ConKbot]

So, once the spark gap fires, and it has a nice ionized channel, its low impedance, essentially a short, and your winding is outputting 20kv... right across your 15v rail.  I'm assuming youre going for a flyback mode of operation (energy stored in the mutual inductors... not a "proper" transformer where instantaneous watts in = instantaneous watts out) If so,  lift your connection to your 15v rail on the secondary side, and connect it to ground, or even isolate both if you can/want. 

Depending on your power levels, you may need a snubber across your mosfet (it should be avalanche rated, but if the energy stored in your primary is more than it can handle, you still need snubbers)

Also, remember that you need a low impedance path for your mosfet gate current to flow, and that the flyback requires a sharp dI/dt  so keep that in mind when placing inductors for isolation. 
   

[king.oslo]

The physical location of the transformer when it is installed in the environment which it will be used, makes it impossible to mount capacitors any closer than about 1 meter away. This is the same location as the MCU will be mounted. So I have have added lots of capacitance, ceramic, polyester and electrolytic across the 15V rail at that location.

I also added 4 mH of inductance to keep the noise away from the 3v3 rail. Still I get 40 volt transients on my 3v3 rail.

I have run out of ideas of how I can isolate my 3v3 rail from the noise. Any more ideas?

Another thing: I used to have a schottky diode across the primary side of the transformer to clamp the transient when I swich the MOSFET off. However, in this configuration, the voltage across the spark gap would be much lower, and no spark would jump. Any idea why?

Thanks.M

[Alana]

Where are those big caps on 15V rail someone recommended earlier?

[G7PSK]

Your diagram shows the secondary circuit as going through the primary in order for the output to circuit to complete, can you not tie the lower end to ground and have the spark gap fed from the top of the coil.

[c4757p]

Your diagram shows the secondary circuit as going through the primary in order for the output to circuit to complete, can you not tie the lower end to ground and have the spark gap fed from the top of the coil.

This. Why is the secondary connected to the rest of the circuit?

[mikeselectricstuff]

The problem with sparks is the risetime is extremely fast (typically sub nS), which means that you have very high frequency components, so even a small amount of inductance in ground paths becomes very significant - a piece of wire acts more like a resistor. You will also get capacitive coupling betteen adjacent parts.
You may also get some capacitive coupling back through the  MOSFET.
You need to design layout and grounding very carefully to ensure that stray currents don't go where they will cause problems. 

Looking again at the curcuit, you are injecting the spark current directly into the power supply - the gap should be across the output coil.

[king.oslo]

Hello everyone,

Thank you so much for your replies! It is very kind if you to want to help me out. 

Where are those big caps on 15V rail someone recommended earlier?

They are between the inductor, and R5 on the schematic. I cannot place them closer to the transformer than this because of the physical location of the transformer in the application.

Your diagram shows the secondary circuit as going through the primary in order for the output to circuit to complete, can you not tie the lower end to ground and have the spark gap fed from the top of the coil.

I am afraid not. It will be impossible to rewire the transformer, you see, I want to use an automotive coil  Alternatively, I have to use a different transformer, but for a number of reasons, I would very much like to use an automotive coil as transformer, and all the cheap ones are wired this way.

The problem with sparks is the risetime is extremely fast (typically sub nS), which means that you have very high frequency components, so even a small amount of inductance in ground paths becomes very significant - a piece of wire acts more like a resistor. You will also get capacitive coupling betteen adjacent parts.

I thought so too. So it should be possible to isolate the high frequency component on the 15V rail from the 3v3 rail (which is derived from the 15V rail) using inductors, right? I tried this, it helped. After the inductors, the 15V rail only rings about 4V_PP. Unfortunaltly, my 3V3 rail dips all the way to zero volts. How come my regulator, which is a LDO type drop to zero volts, when the 15 rail only drops to about 11 volts? Should be enough headroom, right? If I could prevent my 3v3 dripping, my problem would be solved, I dont care about noise on the 15 volt rail, as it is used for nothing else but powering the voltage regulators. Does it exist any LDOs which don't dip to 0V like the one I am using?

Looking again at the curcuit, you are injecting the spark current directly into the power supply - the gap should be across the output coil.

I tried this. I drove the coil using a battery, and the circuit using my PSU. However, they were ground coupled. I got similar problems when I did this. In the actual application, the whole thing has to run from a single 15V rail.

This must be pretty similar to what happens in an automobile, right? How do automotive manufacturers manage to clamp the voltage from the spark plugs? The MCUs in my car seems to have no problem with 4 spark plugs firing tens of times each second.

What are your thoughts?

Thank you for your time.

Kind regards,
Marius

[c4757p]

I am afraid not. It will be impossible to rewire the transformer, you see, I want to use an automotive coil

Ahhhhhhh! I wondered why you would connect it that way. Now I get it.

I thought so too. So it should be possible to isolate the high frequency component on the 15V rail from the 3v3 rail (which is derived from the 15V rail) using inductors, right? I tried this, it helped. After the inductors, the 15V rail only rings about 4V_PP. Unfortunaltly, my 3V3 rail dips all the way to zero volts. How come my regulator, which is a LDO type drop to zero volts, when the 15 rail only drops to about 11 volts? Should be enough headroom, right? If I could prevent my 3v3 dripping, my problem would be solved, I dont care about noise on the 15 volt rail, as it is used for nothing else but powering the voltage regulators. Does it exist any LDOs which don't dip to 0V like the one I am using?

Seems weird that it would do that. 1) Which LDO is it? 2) What is the rise time of the oscilloscope you are testing the rails with? Perhaps you still have a tall spike that you're not seeing.

[king.oslo]

Seems weird that it would do that. 1) Which LDO is it? 2) What is the rise time of the oscilloscope you are testing the rails with? Perhaps you still have a tall spike that you're not seeing.

The scope has 3 ns rise time. You could be right. The LDO is my beloved friend, the LP2951. <3

[c4757p]

15,000 volts is a lot of volts. There's plenty of room for a lot to be left over after it couples across the circuit. And it doesn't matter how much capacitance and inductance you add if the rise time of the spike is above the SRF of the components - then your inductors become capacitors and your capacitors become inductors.

[king.oslo]

And it doesn't matter how much capacitance and inductance you add if the rise time of the spike is above the SRF of the components - then your inductors become capacitors and your capacitors become inductors.

lol

Sounds about right. Can I use some exotic RF component for which my brain is too small to implement, to solve my problem? Perhaps an RF inductor to separate my 3v3 and 15V?M

[c4757p]

If I'm right about fast spikes, you'd need to figure out their rise time first, so you can choose components suitable for operation at that frequency. The problem with that is you'd need a fast, expensive oscilloscope, but good luck finding someone who will let you plug their 500 MHz oscilloscope into your circuit with a 15 kV spark gap. Obviously if you have an assortment of RF parts available you could do it by trial and error, though.

How about triggering it over an optoisolated interface? If you're not triggering it nearly constantly, you won't need too much power for gate drive, so it would be easy to provide separate power to the circuit. A 9V battery would go a long way if you design the isolation interface carefully. I also suspect that 9V battery doesn't give a crap if it gets a big, sub-nanosecond spike once in a while.

Edit: Brain fart. Clearly you have to power the coil itself as well. I might consider still isolating the circuit, though, but using a relay and some big caps. Close the relay, charge the caps, open the relay, trigger the spark. If this microcontroller connects to anything external I would definitely do that.

[SeanB]

Cars do the work of clamping the pulses by having a loody huge capacitor as a supply, AKA as a battery. Emulate with some high value capacitors, and a 33V transorb or ten across the power rails to do clamping, along with some more with RC decoupling to the MCU power rails, and some more before the regulator as well. Place a 340V varistor across the poor FET, and this will still give a high voltage.

[ConKbot]

I am afraid not. It will be impossible to rewire the transformer, you see, I want to use an automotive coil  Alternatively, I have to use a different transformer, but for a number of reasons, I would very much like to use an automotive coil as transformer, and all the cheap ones are wired this way.

Is it an auto ignition transformer currently?  The secondary is connected to ground, not +12/15V, unless you have the +12 and gnd terminals swapped, in which case, if your spark gap goes to ground, youre applying 15-20kv directly across your 15v supply.  If you do indeed have the +12 and gnd hooked up correctly, then its just a schematic issue causing confusion there. 

[king.oslo]

I am afraid not. It will be impossible to rewire the transformer, you see, I want to use an automotive coil  Alternatively, I have to use a different transformer, but for a number of reasons, I would very much like to use an automotive coil as transformer, and all the cheap ones are wired this way.

Is it an auto ignition transformer currently?  The secondary is connected to ground, not +12/15V, unless you have the +12 and gnd terminals swapped, in which case, if your spark gap goes to ground, youre applying 15-20kv directly across your 15v supply.  If you do indeed have the +12 and gnd hooked up correctly, then its just a schematic issue causing confusion there.

Hello, I am pretty sure my schematic corresponds correctly to how an automotive coil is wired. The spark plug is threaded into the engine which is circuit ground, remember?

Kind regards,
Marius

EDIT: Cross-measuring the resistance of the ignition coil I am using confirms that this is true.

[c4757p]

You're correct. An ignition coil is almost always an autotransformer.

[Alana]

I'm not an expert but what if we go to power supply separation the other way around.
Sory for crappy schematic editor but i think its enough to see how it may work. Diode 3A 1000V, caps 2x4700 or 10000uF/35V.

[Neilm]

If you have a lot of reasonably thin wires, it could be that the 0V is moving rather than pickup in the supply rails. I have seen this on a product I designed a few years ago - the spark caused the 0V to rise appearing as a single on reset pin.

The way to fault find this is a fairly boring, repetitive try and try again.

Can you shield the spark? Put the either the spark or the micro in Faraday cage. Then filter the wires going to the transformer. Pay close attention to where currents will flow, remembering that every connection is an inductor.

 In the unit I designed, there was only one 0V with careful layout of the unit ensuring the currents didn't cause problems. The PCB was bolted to the metal work in many locations all of which were connected to 0V. The seams in the metal case were covered either with copper tape or with EMI gasket to provide low inductance connections everywhere.

Neil