[SOLVED] Why did I get the slow sine curves when measuring clock signal ?

[talking]

Please first check out an attached figure.

Figure1 shows my test setup.
What I'm doing is that a pll-based clock generator provides 100MHz clocking signal,
and my test system is measuring the signal on its 50 ohms load.
And then now, I'm trying to measure the signal in the middle of the signal trace between
the clock generator and the test system using an oscilloscope. The scope has 300MHz signal BW in its spec.

Then, I expected the oscilloscope screen would show a 100MHz clock signals, which have consistent voltage level for each of high and low state of clock, as shown in Figure 2.

But, the strange signals were viewed from the oscilloscope, as shown in Figure 3.
Looks like a very slow sine curves were shown first in a large span of the oscilloscope, and if I zoomed-in the sine, finally the clocking signals, which I'm looking for, were shown.
Why did I get the slow sine curves? Anyone who knows what happens in my results?

 

[keith1200rs]

Re: clock signal measurements

Is it a digital oscilloscope and if so, what is the sampling rate? I am wondering about aliasing with the oscilloscope sampling rate.

Keith

 

[chuckey]

Re: clock signal measurements

You do not say what frequency the sine wave is, its not mains hum is it? (50/60 HZ).
Frank

 

[talking]

Re: clock signal measurements

Keith - Yes, one that I used was a digital oscilloscope (recently came out.). The sampling frequency of the oscilloscope was a 300MSPS.

Frank - The slow sine shown in Figure3 was ~300kHz, and the clock signal was a 100MHz which was shown in zoom-in view of the 300kHz sine.

 

[keith1200rs]

Re: clock signal measurements

I think you are having aliasing issues with your oscilloscope sampling rate. You cannot view a 100MHz clock with a 300Ms/s oscilloscope.

Keith

 

[talking]

Re: clock signal measurements

Thanks, let me find an option in my oscilloscope to turn on anti-alias. BTW is there any reason that a scope does not filter out an aliasing?

 

[keith1200rs]

Re: clock signal measurements

Antil-alias filters can only work on frequencies above the bandwidth. I doubt there is a filter you can turn on that will help. Your problem is, I think, to do with the way your sampling aliases with the 100MHz which is lower than the sampling rate. You still get an alias and the fact that the frequency you are looking at is close to an integer multiple of the sampling rate (or the other way round) gives rise to the low frequency "beat". While you can often get away with equivalent time sampling (repetitive sampling), sometimes you just need a fast single shot bandwidth. It depends on your oscilloscope as to what you can do to eliminate aliasing issues. With my 'scope I have "envelope" mode which is quite useful for showing that you have a problem, if not actually eliminating it. One clue to aliasing problems is waveforms which change shape or frequency completely when you change the timebase.

I cannot be 100% sure that is your problem but 300Ms/s for a 100MHz clock is an issue. In repetitive mode you will have 1Gs/s or higher, I would have thought, but that only works for periodic signals where you have a reliable trigger. If you do genuinely have some strange high frequencies on top of a low frequency sine wave then you have no reliable trigger for repetitive sampling and need a fast single shop oscilloscope.

Keith.