Jet engine Ignition system Build with LT375 / TL494 / ZVS driver Lockup

[Amper]

Hi,

i have another one of my stupid little problems.
At im building the ignition system for my jet engine and basically everything is fine and dandy.

The high voltage part is a 1uF 2kV foil cap thats discharges through a GDT into the spark plug, the cap is charges from an old ccfl inverter transformer via a normal rectifier which works great. The power required is roundabout 25-50W @ 24V input. For this reason and the small form factor i choose to use the old zvs circuit exactly in the way shown in the picture with a few variation on values and parts types.

This essentially works great from the current limited power supply but occasionally the oscillator locks up especially at startup or after some ignitions because they are violent enough to disrupt everything.

Reducing the current to allow steady operation in this moment i could measure that both gates of the fets are stuck to 2V (roughly threshold voltage) and the supply voltage drops over the fets coming from the transformer.

I tried adding asymmetry to the system to stop it from having stable states but this only made things worse. Transformer impedance and resonant frequency also seem to have some effect on the likelihood of a lock.

My Problem is that the two ignition boxes will later be supplied directly by 24V by a self resetting fuse but they will not be monitored. This means when the engine is in startup and the ignition should fail the ECU will keep doing its thing and inject kerosene which will lead to a hell of a mess and a large cloud of fuel mist in the air. Its wont hurt anyone but at 2l/min rate of fuel i dont want this to be a regular thing even in testing.

Did someone see the same problem in the past? Is there any known reason and or solution to it?
An obvious one would be to just go for a tl494 but space is limited and im not sure if i can cram all of it into the tiny box i have available let alone shield the driver from the intense EMI of the ignition just cm away.

cheers!



https://youtu.be/lPanMl8RUJk

https://youtu.be/3oVrPL6lryI

[fourtytwo42]

To me it looks too symmetrical! In other words something needs to bias it so that one fet is more likely to turn on at start.

[Amper]

wtf is wrong with this forum...

I tried adding asymmetry to the system to stop it from having stable states but this only made things worse. Transformer impedance and resonant frequency also seem to have some effect on the likelihood of a lock.[/qoutoe]

[alsetalokin4017]

To me it looks too symmetrical! In other words something needs to bias it so that one fet is more likely to turn on at start.

I have a lot of experience with that particular circuit. I've even got multiple copies of that same schematic, along with some variations. It can be used for wireless power transmission, flyback transformer driving, induction heating, etc. It's a Royer oscillator, aka Mazilli, and a lot of the "alternative energy" aficionados like it.

I think the answer is just the opposite: the circuit, especially the power section, has to be as symmetrical as possible so that one phase doesn't want to dominate and prevent oscillation by turning on hard and blowing up its mosfet or both of them.

When I first started playing with this circuit I would apply power slowly by ramping up the input voltage. This was fatal: oscillations often wouldn't start and a mosfet would quickly blow. I found that power has to be applied abruptly, by switching on hard, and the power supply has to not sag much when the connection is made.

The "must be good" capacitor in the schematic sets the resonant frequency by forming the LC circuit with the transformer primary. This capacitor will get hot, it must be a good quality poly film capacitor and in my own experience a stack of parallel caps to build up to the correct capacitance works better than a single large capacitor. In fact this capacitor is critical.

You must also realize that there will be very high currents in the LC part of this circuit. Lots of reactive power.

[coppercone2]

wtf is wrong with this forum...

I tried adding asymmetry to the system to stop it from having stable states but this only made things worse. Transformer impedance and resonant frequency also seem to have some effect on the likelihood of a lock.[/qoutoe]

This made me laugh.

[Amper]

Yes, turn on time is very very important. I have built it in many variations before for heating and all kids of HV supplies. Its a very very nice circuit powered from the right source but i had the lockups in almost all sizes and variations at some point, some worse than others, some never worked at all. I just could never figure out the reason and while there is differences between setups i could never reproducably change some certain thing like quality of the resonator, symmetry or ratios of impedances of different components. Thats why i hoped that maybe there is some known thing like "make the RDSon of te fets 5 times the dc resistance of the coils" or " your input inductor needs to be at least this and that size"

But i guess that solution doesnt exist...

[intabits]

I've used this circuit for experimentation with induction heating.
To avoid replacing MOSFETs should it fail to start correctly, I powered it up in two stages.

First via a large resistor of about 2-3 Ohms (nichrome wire in oil), which will delay MOSFET destruction long enough to reset and restart.
Then, if startup is good, throw a switch to short out the resistor, so as to run at full power.

With a bit of ingenuity, I'm sure that process could be automated, with relays, MOSFETs, etc. 
Might be worth thinking about.

[Amper]

Yes, since i have such low power compared to other setups i thought about just using a linear current regulator mounted to the housing but im not really happy with it. I couldnt fit any relay in there, probably it would be possible with a third fet but its a bit of a bodgy solution. When the 494 board is finished i will see if i have to revisit the zvs driver... For larger high power ones in the future i might go with the relay solution though.

[alsetalokin4017]

How much power do you need?

Here's the same circuit driving a flyback transformer:

[alsetalokin4017]

Yes, turn on time is very very important. I have built it in many variations before for heating and all kids of HV supplies. Its a very very nice circuit powered from the right source but i had the lockups in almost all sizes and variations at some point, some worse than others, some never worked at all. I just could never figure out the reason and while there is differences between setups i could never reproducably change some certain thing like quality of the resonator, symmetry or ratios of impedances of different components. Thats why i hoped that maybe there is some known thing like "make the RDSon of te fets 5 times the dc resistance of the coils" or " your input inductor needs to be at least this and that size"

But i guess that solution doesnt exist...

It does, but nobody seems to really want to listen to it.

[Amper]

But what is the solution? I havent heard it yet. I need only 20-30W of HV but in a very small package, you can see it in the videos i posted. The box also has to house the emi filter so i dont influence the sensors.

[alsetalokin4017]

You probably need a different topology then.  The Royer oscillator needs symmetry and high-current design and is designed to drive a low-impedance load, like a few turns of transformer primary.  And it sounds like you don't really need the power levels it can deliver.

You may want to consider using 555 or CD4093 based ignition coil driver topologies.

[Amper]

Yes, though those are much to little in power. I cant fit any irnition coil in the given space. The limit is a box of 43x45x15mm for filter and power electronics excluding the ccfl transformer that is already in place.

Thinking of it the cheap ccfl driver boards allways run on some kind of oscillator very very similar to the zvs. They have to little power for my application but they show that it has to be possible to get it starting and running with almost perfect reliability using no active components besides the two fets.

[langwadt]

Yes, though those are much to little in power. I cant fit any irnition coil in the given space. The limit is a box of 43x45x15mm for filter and power electronics excluding the ccfl transformer that is already in place.

Thinking of it the cheap ccfl driver boards allways run on some kind of oscillator very very similar to the zvs. They have to little power for my application but they show that it has to be possible to get it starting and running with almost perfect reliability using no active components besides the two fets.

ir2153 ?

[alsetalokin4017]

Yes, though those are much to little in power. I cant fit any irnition coil in the given space. The limit is a box of 43x45x15mm for filter and power electronics excluding the ccfl transformer that is already in place.

Well, the output power comes from whatever you switch with the switching elements, not from the oscillator per se. In the Royer (Mazzilli) you are powering and switching from the same source. In the 555 or CD4093 oscillators, you still will be driving a switching element (a mosfet or two) and it can have as much power throughput as you can manage.
You can fit a CD4093 based oscillator in the little box. You need to heatsink your switching transistor (even in the Mazzilli) so it needs to be outside that tiny box anyway. You are using the CCFL transformer to boost up to the HV necessary for your spark ignition, is that right?

Thinking of it the cheap ccfl driver boards allways run on some kind of oscillator very very similar to the zvs. They have to little power for my application but they show that it has to be possible to get it starting and running with almost perfect reliability using no active components besides the two fets.

Yes, it is possible to get the Mazzilli started with perfect reliability, if you respect its requirements! Symmetrical layout, high-current paths for the LC tank, low impedance in the output L, high current chokes, and a hard turn-on from a power supply that won't sag.

[Amper]

The IR2153 looks really nice! Iwill definetly get some and play a little.

@alsetalokin4017
Yes, its for boosting 24V system voltage to 2kV for igniting the engine.

I know how oscillators and half bridges work or i would have posted this in the beginners section. 555 and cd4093 ar not ideal at all here because i dont have an ignition coil style of transformer that is happy to be driven by only one transistor. I need resonance by push pull or it will not deliver enough power through the small transformer. Thats not really possible when talking about fixed frequency generators like the 555 that maybe wont even ave nice 50% duty, so id have to go through another bunsh of logic gates and a gate driver at least which will make things much more difficult then needed. In that case using dedicated chips like the 494 would be a much better choice and thats what im planning atm since it will give me voltage and current limit as a nice bonus.

Heat sinking was not necessary with the zvs up to over 100W of input power the TO220 has enough sink for the tiny amounts of power lost in modern low resistance fets. Thats why i like the idea so much, switching wit non resonant frequency will increase losses significantly.

Mine is adhering to all your criteria but still locks up. I have very good foil capacitors, short distances to the transformer, all wires are the exact same length, all parts are even from the same lot. Very low impedance did not have any effect in my case, i had induction heaters not start to oscillate even though they were made with copper tube and proper induction capacitors. The problematic thing is the hard turn on which might be no issue when you operate the thing by hand, you have a regulated power supply coming from the grid that is very predictable  and when something goes wrong you just try again. In my case i have two of them sitting on the side of a jet engine with pumps, sensors, starter motor, ill wear ear protection and have to monitor all kinds of things, i cant afford to care about two zvs driver starting up or not because maybe the battery supplying them is not at the right voltage or impedance. in ccfl inverters its also no problem to get them running, i never had them lock up ever. There must be some thing that can be done passively.

Attached is a screenshot of what im planning as a backup right now. Obviously the layout is not done yet, just routed roughly to see if i can fit it all on a single layer aluminium pcb.

[TurboTom]

Hey nice job and very good firing rate.

Are you using it on a TS-21 JFS? The original exciter box provides much less energy than your proposed 2 Joules, maybe 1/10 of that. I'm not sure how long the original spark plugs will do under these circumstances.

Having a lot of experience with high energy ignition exciters myself, I'ld recommend to either find a proper switching spark gap tube or utilize a semiconductor switch for the high voltage (several SCRs in series with a trigger transformer perform quite well). My own experiments with these surge arrester tubes (even the bigger ones) proved mediocre at best. Apparently the gas in these discharge tubes gets trapped in the electrodes, reducing pressure and thus ignition voltage within comparably short periods of time. This means, at a constant charging rate, spark energy decreases and firing rate increases.

I haven't got any experience with these ZVS drivers to power the step-up tranformer. Since the ignition exciter usually has to operate over quite a considerable range of supply voltages (while the starter motor is energized, it pulls down the battery voltage to 10V easily on a 28V system and a cold battery), I prefer a current-control flyback topology with charge voltage limiting. This, combined with a semiconductor switch and a high-current series ignition coil fires any spark plug at very high efficiency.

If you'ld like to experiment with the mentioned concepts a little bit, I may be able to provide more in-depth information.

Cheers,
Thomas

[Amper]

Oh wow, i didnt expect to run into someone who knows this stuff this exact 

Yes, its for my TS-21 in case you havent yet you can look at one or two of my other videos on youtube to get an impression of what im trying to do.

I played around some time with smaller capacitors but i didnt expect it to require so little energy considering the insane power of whats shown on you tube. Also there is manufacturers of modern ignition systems ranging from 1J up to 12J so i thought i was in a good region. Though the way im doing it is definitely a problem in the way of having to high currents flowing, the spark plugs dont show signs of erosion yet but i guess if i used my system extensively i might have to change some things. Using an inductive coupler would be nice but im afraid tit would be very heavy and bulky. The purpose of my project is to make the entire system more light and modern wich is also the reason for mounting the ignition boxes directly to the combustor casing. If i used something similar to the original one it would cost me in the range of 1-2kg which is not acceptable.

direct switching would be a next step, maybe even using mosfets but it will not be easy driving the gate at such voltage levels and transitions though in that case even pulse shaping would be possible.

Im very interested, if you can tell me more please do so!! 

Grüße

[Berni]

The ZVS circuit is indeed a very good and simple way of driving a transformer at pretty high powers.

One way to get around the startup problem is to power the resistors that go to the MOSFET gates separately. So you can have your main supply ramp up slowly and once that's up to voltage you suddenly apply voltage on the gates and it kicks in. Alternatively you can apply power to the resistors first. This causes enough power to move across the diodes to start the oscillator up, but can't blow it up as the resistors only provide limited current. Once its already oscillating you can ramp up the main supply as slowly as you want, you can even ramp it up slowly from 0V with a variac and it will just simply provide more and more output power as you turn it up. Its ramping both the gate supply and main supply slowly that kills the transistors.

As for it stopping osculation that is likely because your sparkgap firing loads the transformer too suddenly. The inductor limits how much power can enter the tank circuit in an instant, so when you suddenly short the transformer you suck all the energy out of the oscillating LC tank circuit and so leaves it with no energy to continue oscillating. To fix that you want to add leakage inductance to your transformer. This can be done by correct transformer design or by adding an additional inductor in series between the transistors and the transformer. This inductor will store some extra energy that can't be sucked out by the transformers secondary coils, so there is always some energy left to keep the tank circuit oscillating.

Many people build these ZVS circuits without actually understanding how they work.

[Amper]

I like the additional inductor idea, i think ill give it a go tomorrow. Yes you are right, not many actually know how it works but also it is a lot ore difficult to understand then it looks at first glance. I know it for over a decade now but usually its just "yay it works lets make arcs!"

EDIT: Also i attached an image of my current approach sadly i dont have any tl494 on stock so ill have to wait before  finding out what mistakes my layout has  Also i dont have a god way of dispensing solder paste so its not the prettiest job i guess.

[TurboTom]

Unfortunately, I can't go too deep into detail of my latest ignition exciter design since it was contract work and hence I'm bound by an NDA. What I can tell is that the whole electronics of that 1J+ exciter weighs just round about 160g and the size equals approx. a cigarette box that's 50% thicker than standard. It's completely solid-state.

If you ever disassembled an aircraft turbine engine ignition exciter, you would be surprised how small actually the series ignition coil is. Basically, it's a ferrite rod of a few centimeters length and maybe 10mm diameter with a single layer of 0.3mm enamelled copper wire (well isolated) as the secondary and on top a few turns of copper ribbon, wired as an auto-transformer. For the details, please have a look at the schematic of a replacement I made maybe 15 years ago to revive an ignition exciter of a GTP30-67 turbine engine where the step-up transformer secondary isolation broke down. I ended up with almost all of the original high voltage stuff and made a flyback converter to charge the storage cap. The HV rectifier diode (D5) is rather critical and the mentioned one wouldn't work reliably. Problem is the considerable peak and also high average current that causes it to heat up. After that, when the spark cap fires, the induced plasma across the spark plug is quite conductive so there will be an LC oscillator formed of the storage cap, the ionized spark gap (can be considered as a conductor), the secondary of the series ignition coil and finally the ionized spark plug. The problem that results is that during the negative half-wave of the oscillation, the (already hot) rectifier diode starts to conduct and, depending on the characteristics of the output coil of the flyback transformer, will carry a substantial current. A diode that performs quite well is the Sanken UX-FOB that's normally used in microwave oven with an inverter PSU. These diodes don't sell cheap but they will permanently withstand the abuse...

Anyway, that's basically an insight how ignition exciters had been built in the 1950s to maybe 1980s, of course with varying methods of charging the energy storage cap (electromagnetic trembler, single-transistor blocking oscillator, voltage limited flyback converters of discrete or integrated designs). In early units the high voltage rectifier was just a cold cathode gas-filled diode tube, later silicon rectifiers were used. All these exciters contained a specifically made switching spark gap that was designed to break down at approx. 2kV. The ugly thing about these is that they were usually pressurized with tritium gas to aid ionization (they always contain a "radioactivity" warning sign, yet it isn't possible to measure any radiation externally. That's a good indicator for tritium with its relatively low energy beta radiation that's completely shielded by the glass envelope. These switching spark gaps are more or less unobtainium, and as I mentioned before, conventional surge arresters won't last too long. I once found a (Chinese) company who had these special switching spark gaps in ther portfolio but I quickly decided against using them after I heard the price...IIRC must have been some 500EUR a piece, and the minimum order quantity was 50 since they made them specifically to the required break-down voltage. Once in a while you can find suitable spark tubes on the bay, though...

FYI I attached the schematic of the original TS-21 ignition exciter (two such sections in one box, no series ignition coil in this one but a sense transformer instead).

BTW, I'm quite impressed by your work, I once intended converting a TS-21 to turboprop as well but finally decided against it due to the low efficiency of that engine. It's funny I've also got a Turnigy "Aquastar" brushless motor sitting in a box and waiting to become a starter generator on a KHD/BMW/RR T312 engine. Yet, if this motor will be permanently capable of the 50,500 rpm of the TS-21 gas generator is (in my opinion) at least questionable.

Cheers,
Thomas

[Amper]

I know how the old ones operate but it is difficult to get info on anything not vacuum based.

I havent seen many cut open exciters on the internet so my experience is limited there. My assumption allways was that the output transformer is more of a security feature not to have charged capacitors on any output ports and also limiting the current by inductance. I just wanted to throw in a ferrite rod later and hope it will flatten the pulse enough.

About the discharge tubes i have to agree, i have previously worked with the smaller 8x8mm ones and had some failing so i switched to the large 25kA rated ones. Obviously they are not designed for a long life. How long do you expect them to last? Since im not going to use it in an aircraft in the near future i would be fine changing them from time to time which is actually the reason i mounted it on screw terminals in an accessible spot.

regarding full solid state: Wouldnt it be possible to do away with all HV stuff and just feed the transformer 24V from a very low esr cap? Discharge a stack of ceramic capacitors using some of the modern crazily powerful mosfets might do but even driving it like a normal transformer with a lot of power should be possible at few hundred kHz.

Yes, the TS is not the most efficient engine but its actually better than expected. I got mine 3 years ago and a friend of mine got the second one. Since at the time we were both in the akaflieg stuttgart we had hopes to get our engines flying some day so he decided to build a turbojet from it for the usual glider applications. I meanwhile allways liked the idea of turboprops  and the TS construction is ideal in many ways. The output pretty much exactly matches a rotax 912 which opens up a lot of accessories. At some pint i noticed that the propeller bearing of a thielert 1.7 aviation diesel EXACTLY fit the output flange of the TS2 which made it easier. Sadly at that time due to other more important projects i didnt have the time to continue and my friend decided that he wanted a turboprop as well, so after i obtained two sets of prop bearings he put his engine together and did first runs with propeller in the late summer of 2017 iirc.

Since last summer he is now doing his masters thesis on characterizing the engine. He did full instrumentation on it, i did the electronics and if everything goes well he should be done soonish.

Problems he found was especially high oil consumption but that could be for other reasons and it can be solved.

Most interesting so far is that its extremely sensitive to changes in pressure of the supplied air. It was run in a laboratory with an eddicurrent brake and the gasses were supplied by ducting for flow and exhaust measurement. Even just a very small pressure drop on the inlet of a few dozed mBar reduced the shaft power by half(!). This is due to the fact that the pressure drop over the second turbine is very very small and is directly influenced this way. This also means though that adding a booster stage will increase efficiency dramatically!

Regarding the starter generator, my friend used the motor of an align T-Rex 700 though its an outrunner and it will rupture the bell at the speed necessary if not using it only as a starter motor. I choose the aquastar because it was the cheapest alternative too the lehner motor i actually want to use. Since im still not sure about the exact configuration i decided that it would be a good idea to first get a cheap motor for the beginning and not ruin one that costs half the price of the engine itself...

The rotor will not survive the 50.5k in stock configuration, the kevlar is really crappy and dray. I rewound it using aviation carbon roving and it should have a burst speed of over 100k now. I tested it to 60k and it was perfectly fine doing that and didnt even require rebalancing.  In terms of bearings i have a little experience with home made rc jet engines and repairing turbo molecular pumps. Using partially home made bearings with permanent lubrication one of my pumps is at almost 100h of operation at 75k min⁻¹ and while the forces involved are much smaller its very possible to use such bearings on the aquastar and get decent lifetime.

A T312 would be a dream to have but i dont have to tell you that they are very very rare compared to the TS and im not yet willing to sell both kidneys for one ; )

EDIT:
Can you give me any numbers on pulse duration and power or is it nda relevant as well?

EDIT2: Habe jetzt erst gemerkt, wer du bist, meinen kumpel kenst du wohl schon, ihr habt öfter mal Telefoniert

[Amper]

Ane heeere we goooooo!

https://youtu.be/j4iKh5aaWNQ

I had to wire a DIP 494 to the pads because i couldnt wait to get the right ones in the mail and to my absolute shock this thing worked on the first try. Amazing. Now i have current  regulation and voltage limit so i cant overcharge the cap if the sparkgap should decide to die. Pulse frequency can be much higher then in the video but for now im sticking to only 20kHz to drive the small transformer because its still only fr4 instead of aluminium pcb. Once i have the right parts ill put everything together and hope for the best.

[Amper]

And another Update:
TurboTom was perfectly right, the GDT caused another set of unexpected problems. It turned out that it works very nicely with a certain spark plug impedance but if it changes due too soot or moisture or some voltage levels dont hit the sweet spot it can start to go into simmer mode not letting the charge circuit do its thing.

For this reason i decided to go full solid state as he recommended. The scheme used now is a boost converter using my original tl494 board just reduced to one 600V IGBT driving the most basic booster. The output charges roughly 15-20uF of film caps which are then switched into a toroidal inductor / transformer with a single winding and a tap going to ground via a capacitor. I got the idea from this great pdf:

http://www.gasturbineworld.co.uk/Small%20Gas%20Turbines%203%20starting.pdf

The result of the bulky test setup is amazing performance and a more inductive pulse that will not eat the spark plug as quickly as the pure capacitor version.I took a video at 50Hz trigger frequency but up to 100Hz was possible and only limited by the speed of the relay used to trigger the thyratron.
Thesound you can hear in the videos btw is not at all representative. This thing is seriously loud and can not be operated in the lab without hearing protection. Its comparable to a spark gap teslacoil.

I will post updates when i continue as there is still some issues to solve.

https://youtu.be/ypEAAUoQ-iY

[spec]

Hi Amper

Just read through your thread: jet engines, Royer oscillators, ignition systems 

I see that you have been having a bit of trouble with the circuit of your OP.

With that circuit, there are issues at the macro and at the micro level.

One of the micro level issues is the gate drive arrangements, so I have quickly knocked out a revised gate drive circuit, as shown on the attached schematic, for you to try out, if you wish:

[Amper]

Hi!

There is news again. After trying a bunch of different designs i now settled on having a small boost converter now using the LT3757. Its pretty much the previous design just using much more compact electronics that fit on one 50x60mm single sided pcb including capacitor and ignition coil.

Ill post the board later, for now here is a video to demonstrate how insanely strong a 2,5J discharge can be.

https://youtu.be/YtumcI1mIns