measure ddr clock seem strange

[Luca]

noisepic do you have a spectrum analyzer? If yes with which setup can you run?
It's still possible to "see" the clock signal, but in frequency domain instead time.

[robertferanec]

Do not worry ... we all went through that

When you mentioned, that memory clock "has vary pulse", I have got immediately suspicious, that there is no enough samples per clock cycle to measure the clock signal properly. That is why I asked about your scope.

[noisepic]

I found myself that I can not evaluate ddr just by oscilloscope, the tektronix system up to 200K$, the probe only has 17K$. That's the reason why I see strange in clk waveform. Suppose that 400MHz DDR clock it mean 2.5ns in period, 1.25ns in pulse width. And we need atleast 125ps rising/falling time oscilloscope (BW>8GHz). So it seem measure DDR3 signal by low performance oscilloscope like a stupid story!

[noisepic]

Do you have a special oscilloscope to measure memory clock? You may need special probes and an expensive oscilloscope - what is your DDR3 speed / clock and what kind of oscilloscope and probes you have?

Thanks Robert, that is ddr3 and i scale down freq to 400 Mhz. My oscilloscope 300mhz bandwidth passive probe only. You think that due to my test equipment issue?

[robertferanec]

Do you have a special oscilloscope to measure memory clock? You may need special probes and an expensive oscilloscope - what is your DDR3 speed / clock and what kind of oscilloscope and probes you have?

[noisepic]

When measuring clockp and clockn of DDR3 I expect, clock like this. But measuring clockp and clockn has vary pulse width, including long pulse and short pulse, my board still work fine. I don't know what's wrong here?