Producing Triangle Waves

[nitrate]

Hi all,

I usally don't post much nowadays as i don't like to see the place cluttered up with lots of dead end threads but i really need some assistance here. I'm building my latest classD prototype and all is well. It's only a simple 200W half bridge fed off +-45V supplies. It works well, same as all my other clasD examples except i'm having the same reoccurent problem. The thing is... I've seem many people say in these forums that producing the triangle wave in a carrier based amp is the easy part. Well, either i have a small gap in my knowledge or physics is having a laugh at my expence.

The way i currently generate triangle waves is to use a CD4093, one gate for an RC oscilator ( usally around 250KHz ) and a second gate for a buffer. This squarewave clock i then feed into a 2.2nF cap through a resistor and thus... Triangle!

However this is a very crude way of doing things and there must be better ways out there. Using TTL helps a little due to its low impedance and high speed helping to keep the RF and other wee beasties out of the triangle but your limited to running TTL off 5V instead of split +-5v ( 10V ) that everything else runs off. I've tried all sorts of op-amp arrangments and they all fall down at high speed, even when using fast op-amps like the MAX412.

Anyway that i try always results in a reasonable triangle on the scope but when in use there are little whistling noises and other products inbedded in the wave. Can anyone shead some light on the mysterious triangle generation and maybe post some links or schematics as it must be so obviously easy that i must be overlooking the obvious and i'm finding it impossible to do LOL

HELP!!

Thankyou for reading

[nitrate]

I've attached a file of the offending noises for the inquisitive amongst us. The amp is 200W run in open loop, sound recorded from infront of speaker just after power up. It soon settles down and becomes unnoticable.

[PigletsDad]

I haven't done it at 250kHz, but in lab equipment I used a current feedback opamp as an integrator, in a loop with a comparator with hysteresis to make a fairly clean looking triangle wave.

[stinius]

One fast opamp and two fast comparators should do the job as in the LB-23.

Otherwise look at the attached patents.

Cheers
Stinius

[Pafi]

The problem is not because of the triangle wave. It is an interference, switching of one channel disturbs the other's comparator, and this way they modulates each other. What you hear is the difference frequency between the two oscillators. You should synchronize them! (I bet you actually built two channels. Turn off one of them!)

[nitrate]

Sorry Pafi, wish it was that easy, the proto is running on its own ( single channel ) however there may be some truth in your comments as i always wondered if the two gates in the 4093 may be causing intermodulation with each other ???

Leigh

[nitrate]

Think i'll try the suggested tri-generator that Stinius suggested. I've tried different coils on the board ( torroids and air cores ) just to rule out RF interference being thrown accross the PCB from the output filter or bridge.

For your amusment i'll attach some piccys of the prototype pcb and the bridge and current triangle waveforms. As you will observe the ground is one solide copper ground plane and the track side has been implemented using tinned copper wire tacked down on matrix board, beefed up were needed by copper braid. The prototype doesnt require any heatsinking at all at the moment, dead times are approx 50n/s.

Does anyone think not using a Xtal accurate clock maybe influencing things?? The measured frequency doesnt seem to wobble although it does slowly drift up and down by a few hundred herts over time

Leigh

[nitrate]

.....And the waveforms, you can just see a notch in the bridge waveform were the dead time plays its part. the triangle looks good but there are tiny artifacts in it and its not very linear.


Leigh

[Pafi]

Leigh!

Maybe not the other channel, but something surely interferes with your clock. It can be your PC, monitor, SMPS, AM radio station, whatever... try to find it!

Isn't there an SMPS on the board?

[nitrate]

Sure is an smps on board in the form of the LM switching regulator for the 45 to 9 volt drop to feed the 5v reg. I've tried isolating the +_5v rails and powering them seperate from a battery source but no luck. the only other thing that is switching in the room is the fet bridge so one must deduce that it is somthing to do with that. I'm convinced though that using a low impedance highly linear tri-wave should sort it out. After all if its the bridge interfering with the pwm stage then slowing that down is not an option

Leigh