LF347 High Frequency Triangle Wave Generator

[GrumpyMailMan]

Greetings
I've been trying to use two LF347 to generate triangle wave at 250 kHZ for a PWM.
But from the Tina simulation i could get is around 100kHZ and 160kHZ higher then that the waveform are distorted.

I've read a topic from other member in this fourm, but one thing I don't understand is how does the circuit starts up when there is no input to the circuit.

I understand the standard equation for this circuit is f=R2/(4*R3*C*R1), but it seems that over 100khz this is not valid.

Mainly
1. How can i fix the circuit so it can generate 250KHZ triangle wave without distortion
2. Why does the equation not valid when the frequency is over 100 KHZ


My possible guess
1. due to the slew rate?
2. noise feedback into the opamps


The data sheet given did not provide much information on hight frequency operating.
Wonder if anyone chould help me with it.

 

[ronv]

Yes, I think way to slow. Try it with a high speed comparitor like the LM119.

 

[GrumpyMailMan]

But the thing is the restriction is only allow with LF347 opamp...

 

[ronv]

Ahh. The plot thickens.

Try this one with your LF347.

 

[audioguru]

The datasheet of the LF347 shows that its slew rate limits good output above about 80kHz.
The timing capacitor is only 10 puffs! But the wiring capacitance and the capacitance of the IC might also be 10 puffs or more so the frequency calculation will be much too low.

 

[MrAl]

Greetings
I've been trying to use two LF347 to generate triangle wave at 250 kHZ for a PWM.
But from the Tina simulation i could get is around 100kHZ and 160kHZ higher then that the waveform are distorted.

I've read a topic from other member in this fourm, but one thing I don't understand is how does the circuit starts up when there is no input to the circuit.

I understand the standard equation for this circuit is f=R2/(4*R3*C*R1), but it seems that over 100khz this is not valid.

Mainly
1. How can i fix the circuit so it can generate 250KHZ triangle wave without distortion
2. Why does the equation not valid when the frequency is over 100 KHZ


My possible guess
1. due to the slew rate?
2. noise feedback into the opamps


The data sheet given did not provide much information on hight frequency operating.
Wonder if anyone chould help me with it.

Hi there,

If all you want to do is generate a PWM signal you dont need to generate a true triangle waveform. A sawtooth would do it also for example, but more to the point is that you dont need a perfect triangle or sawtooth, all you need is a waveform that varies in width as you follow the amplitude of the wave up or down.

One such easy to generate waveform like that is a simple capacitor charging through a resistor. That generates an approximation to a triangle...a triangle with curved sides. If you want a better approximation of the triangle you can simply limit the range used by resetting the cap sooner. What you get in the long run is a PWM which is pretty good and you dont need an op amp just to generate the wave, just a few comparators and maybe a NPN transistor to reset the cap.

 

[GrumpyMailMan]

The main goal for this project is to use this OpAmp to generate 250 Khz triangle wave and compare with audio input signal to create PWM.

I have tried three different sets up but still unable to reach the goal of 250 KHz....... driving me crazy :S

 

[MrAl]

Hello again,

Ok i thought that might be the case. So you have to use the 347, no problem.

The 347 data sheet shows a slew rate of 13v/us and at 250kHz ramping up once and down once per cycle means a ramp time of 2us. The 347 should be able to do that without too much difficulty as it can slew 26 volts in 2 microseconds, and you only have to slew maybe 20v tops in 2us. So it's not the slew rate for the triangle. The slew rate on the square generator however may limit the sharpness of the corners and that could affect the triangle wave too. If it slews 13v in 1us that means it takes a little less than 1us to slew 10v, and that means the square wave wont be a square wave it will be a pseudo square wave with sides that ramp up and down rather than shoot up and down. That will distort the triangle wave output as the time integral of this kind of pseudo square wave is not a triangle wave. Since the ramp period is 2us and the square wave slew time is roughly 1us, that's very significant. The ramping part of the square wave will integrate into a curve rather than a ramp, so the 'triangle' output will have ramping sides with rounded top and bottom rather than nice sharp corners.
That could be the problem but of course a spice simulation is in order.


What is strange though is that the circuit you show in your first post has plus and minus 10v supplies yet your 347 is biased as if there was only a single positive supply. I think you should fix that unless you really need it that way for some reason. Are you looking for a positive only triangle output?

 

[audioguru]

I think the square-wave is worse than you think because its opamp saturates which uses extra time.

 

[ronv]

The trick is to keep the triange amplitude small so that the total swing of the square wave is not important. Try the circuit I posted.

 

[JimB]

Why do you need 250khz for PWM?

That sounds excessively high, what are you trying to do?

JimB

 

[Roff]

I think the square-wave is worse than you think because its opamp saturates which uses extra time.

Audioguru hit the nail on the head. Due to internal compensation capacitor recovery time, the Schmitt trigger delay is on the order of several microseconds. The actual delay depends on the amount of input overdrive.

 

[GrumpyMailMan]

the reason for 250 khz is the source is 20 ~ 200 hz we want the triangle to compare at 10 times max frequency so we set around 250~300k HZ but still not sure how to get around the slew rate to get the target goal......due date is like next week (sad face)

 

[Roff]

the reason for 250 khz is the source is 20 ~ 200 hz we want the triangle to compare at 10 times max frequency so we set around 250~300k HZ but still not sure how to get around the slew rate to get the target goal......due date is like next week (sad face)

Ten times 200Hz is 2kHz, not 250kHz.

 

[GrumpyMailMan]

sorry type 20k hz

 

[Roff]

You're going to have to use comparators. Common op amps are too slow, and not really suited for this application. An LM393 should be able to do 250kHz.

 

[GrumpyMailMan]

Problem is we can only use LF347.....

 

[Roff]

This is apparently a homework assignment. Post the assigned problem as it was stated to you.

 

[ronv]

Did you try the circuit in post #4?

 

[MrAl]

hi again,

ronv:
The trouble is the slew rate always seems to mess up the triangle.

Grumpy:
Are you allowed to use a couple of transistors with the 347's too? If so, you can probably use two transistors to boost the slew rate of the output and get a fairly nice triangle wave. The slew rate would be more dependent on the transistors rather than the op amp and so would be very high.