TL494 based plasma tweeter

[Xkrik]

Hi!
Through last year I've done some experiments with plasma speakers, so i decided to make a stereo version for my high school project. I've made the schematic based on this plasma speaker.
I have serious problems with the modulation section. At first i would like that the flip-flops on IO1 and IO2 are working asynchronously as shown on attached image. I'm using the flip-flops to limit duty cycle to 50%. When i turn on the modulation section it's random whether the flip-flops are running synchronous or asynchronous! This fact is making serious problem for the power section, the transformers in synchro will make huge peaks. I don't know if TL494 is capable for doing this, but i hope it can.
The second problem is if the T3 and T4 are switching load (even a dummy 15Ohm resistance load!) it sometimes disables the flip-flop function! So it doubles the frequency, the GDT cannot handle such high duty-cycle and the power section blows. I haven't found any rule when it occurs but it depends on the duty-cycle and frequency. However it's also random on which TL494 this occurs. I've detected one frequency when it doesn't depend on the duty cycle and everything works OK.
ΔV ripple of 12V source on switching single 15ohms is about 0,8V, maybe the problem is in source, but when I've changed C41 (100uF filter) to 1mF or 10mF it didn't fix the problem
I've looked on gate drivers which can stop DC coupling on GDT, but there are hard to obtain in shops here.
Although the modulation section works poorly, the sound from tweeter is great!
Thank you for help

[newvirus2008]

Are you driving push pull with the flip flops inside the TL494?

I think yuo re getting your 50% by not using one set of pulses (one o/p txr is open)

It is difficult to avoid DC coupling if you plan to use the forward converter. The transformer may saturate if driven with standing DC.

Try to use a Half Bridge Driver like IR2110 or LM51xx.

[Xkrik]

Yes, I'm driving push-pull with flip flops inside and I'm not using one set of pulses. However the transformer saturates only when the flip flop is oddly disabled.
I have just found that is not possible to make internal flip-flops run always asynchronously Does anybody know how to do it by using external flip-flops?

[newvirus2008]

What you do by using these flipflops is simply alternate the pulses of the PWM.

It is merely an odd-even flow splitter. I think it ll be enough you pick up the circuit inside TL494 from its functional diagram.

Is it enough if you re able to split the PWM somehow to make the duty cycle 50% outside the TL494?

Here is how you do it outside the controller.
This circuit drops both freq and duty cycle to half of what it was before entering the circuit.

Now, is that what you needed?

[newvirus2008]

See the spectrum below.
Freq is half of before and the baseband signal is clean. (20kHz only, no harmonics)

Green = Input, Red = Output, PWM freq was 1MHz.

[Xkrik]

Unfortunately no I have to alternate pulses from IC1 and IC2 . The alternated pulses must be asynchronous (left and right) as shown on the second image. I have to guarantee it, but don't know how. I think this schematics based on internal schematic of TL494 will make the same synchronization problem. The first pulses are making synchronization

[Xkrik]

This simulation should be better. This is how I want it to work.
However it doesn't fix the problem with synchronization - first pulse combined with soft-start is killer. Damn, maybe the problem is in special soft start for each IC... And on the second IC i should use the opposite output(C2 E2).

[newvirus2008]

I dont actually understand how you get two PWMs.

Anyway a circuit that will alternate pulses is here. This is not based on TL494 internal. This circuit splits the PWM into two phase modulated signals and divides them in frequency and then combines them to form a PWM with alternating duty and half frequency.

So is the previous one. I think this wont have even a problem with delay because both signal paths are likely to be matched very well.

[Xkrik]

Each TL494 is modulating his audio signal. Sawtooth used for modulation is synchronized. And the pulses from IC1 have to be asynchronous to IC2.

[newvirus2008]

Ok ok. got it

The two channels in a stereo system are usually considered independent.

The uncertainity that you see while turn on is merely dependent on the actual signals that re applied at inputs.

But it is possible to adjust so that the pulses dont overlap by providing suitable DC offsets to the Left and right channels.
Try this.

Left channel => 0.33* left audio + 66% Vcc => PWM1
Right channel => 0.33* right audio -66% Vcc=> PWM2

I mean they should look like this when DC coupled and use the same ramp for both:
Irrespective of signal content in channels PWM1 and PWM2 wont NOT overlap at all