what are the techniques in high frequency circuit design to eliminate sharp edges of the clock?

 

What is your definition of "high frequency"?
Why do you want to eliminate sharp edges?

 

what are the techniques in high frequency circuit design to eliminate sharp edges of the clock?

Low pass filter.

Slew-rate limit.

 

what are the techniques in high frequency circuit design to eliminate sharp edges of the clock?

Are you talking about overshoot on the scope? Or vertical edges to the high/low transistions?

 

Are you talking about overshoot on the scope? Or vertical edges to the high/low transistions?

yes i am talking about the overshoots. when i see the wave form of the clock, it has high peaks at the rising edge. i want to eliminate that effect.

 

What is your definition of "high frequency"?
Why do you want to eliminate sharp edges?

as the frequency of interest is 30GHz. i am having some high peaks at the rising edge of clock and want to eliminate that.

 

What are you doing at 30GHz ?

 

as the frequency of interest is 30GHz. i am having some high peaks at the rising edge of clock and want to eliminate that.

At those frequencies, there is a big chance that most of the overshoot is your scope and probe (and inductance of your ground clip) causing the apparent overshoot. Not many people here have measured anything beyond 5GHz. Contact your scope manufacturer for advice.

- What brand and model scope are you using to measure 30 GHz?

What is your application?

 

At 30Ghz the setup you use for measurement is maybe even more important than the signal itself. Wonder what kind of scope you are using that can show the rising edge of a 30Ghz clock.

 

At 30Ghz the setup you use for measurement is maybe even more important than the signal itself. Wonder what kind of scope you are using that can show the rising edge of a 30Ghz clock.

Usually sampling scopes are needed to handle those frequencies.

Can't remember any more than a vague hint at the theory - but they sample at specific intervals and build up a trace of dots on the screen (probably completely different on digital) - AFAIK: they can only display repetitive waveforms.

 

yes i am talking about the overshoots. when i see the wave form of the clock, it has high peaks at the rising edge. i want to eliminate that effect.

Wow. That is one fast signal. I agree with an earlier post that what you are seeing may be a measurement error.

If you can prove to yourself that the overshoot is real then I think the next step is to make sure that you are not getting reflections due to changes in impedance in the signal path. This could be as little as the change in a trace width somewhere.

I have never done anything like this so I am very interested in what you find.

 

Wow. That is one fast signal. I agree with an earlier post that what you are seeing may be a measurement error.

If you can prove to yourself that the overshoot is real then I think the next step is to make sure that you are not getting reflections due to changes in impedance in the signal path. This could be as little as the change in a trace width somewhere.

I have never done anything like this so I am very interested in what you find.

Even a short trace can masquerade as a tuned circuit at that frequency - at logic type switching times; a little bit of resistance in series with a trace can subdue ringing. But 30GHz probably needs a different solution.

 

Even a short trace can masquerade as a tuned circuit at that frequency - at logic type switching times; a little bit of resistance in series with a trace can subdue ringing. But 30GHz probably needs a different solution.

I would sure imagine so. If the clock frequency is 30 GHz, that's a period length of only two to three centimeters. The edges are probably 10x that, so now we are talking about wavelengths in the millimeter range over which you need to treat them as transmission lines -- so even much of the routing within an IC would need to be treated as such. I would imagine that every connection to anything would need to be considered carefully.

 

I would sure imagine so. If the clock frequency is 30 GHz, that's a period length of only two to three centimeters. The edges are probably 10x that, so now we are talking about wavelengths in the millimeter range over which you need to treat them as transmission lines -- so even much of the routing within an IC would need to be treated as such. I would imagine that every connection to anything would need to be considered carefully.

I always thought 2.45GHz (as in microwave oven) had a wavelength of a few Cm - IWHT: 30GHz would be just over the border in the millimetre wave range.

Antennas are often 1/4 wave; its one of the things you might have to consider why. Its heading in the general direction of wave nodes per millimetre - its pretty much waveguide country............

 

I always thought 2.45GHz (as in microwave oven) had a wavelength of a few Cm - IWHT: 30GHz would be just over the border in the millimetre wave range.

Antennas are often 1/4 wave; its one of the things you might have to consider why. Its heading in the general direction of wave nodes per millimetre - its pretty much waveguide country............

I mistyped -- 30 GHz in vacuum would be right on 1 cm wavelength and I meant to say two-thirds to one centimeter (in a circuit). The quick rule I use is that the speed of light is one foot in one nanosecond and since this is about 30 cm, a 30 GHz signal would be about 1 cm at the speed of light (in a vacuum).

 

I mistyped -- 30 GHz in vacuum would be right on 1 cm wavelength and I meant to say two-thirds to one centimeter (in a circuit). The quick rule I use is that the speed of light is one foot in one nanosecond and since this is about 30 cm, a 30 GHz signal would be about 1 cm at the speed of light (in a vacuum).

Can't remember the TS ever mentioning vacuum - if they're using standard PCB techniques; ringing is pretty much a given.

Maybe a stripline (can't remember the name of the other one) - coupling can be better than transmission line, but the layout can get a bit bunched up.