I want to measure how much noise is a certain oscilloscope responsible for. Up until now we (my team) have used the approach of inserting a square wave and measuring the random jitter RMS. This method assumes there is no noise from the signal generator (minimal noise), which is not good enough for us.

  • Other than that is this a correct approach? If not what is wrong with it?

  • Is there a way to evaluate the random noise generated from within the scope itself without using an external signal?

P.s. To answer the questions in comments below, we are trying to evaluate an oscilloscope but we have a small noise budget aloud in the system while measuring (I cannot disclose other details). We are currently considering DPO70000SX series with 33GHz and 200Gs/s. Our goal at the moment is to measure the RMS random jitter in the time domain which is the standard deviation on the total jitter distribution.

Any other suggestion on how to evaluate this oscilloscope are most welcome.

  • \$\begingroup\$ How do you factor in noise from the signal generator supplying the square wave? \$\endgroup\$
    – Stanley Pawlukiewicz
    Commented Nov 9, 2017 at 7:33
  • 1
    \$\begingroup\$ Havatok, can you clarify what you're after? Are you looking for noise in the amplitude (you mention RMS) or within the time base (some have mentioned atomic clocks)? Is it in X or Y axes? \$\endgroup\$
    – Paul Uszak
    Commented Nov 9, 2017 at 15:34
  • 1
    \$\begingroup\$ Forget rubidium and caesium clocks -- they have good long term stability but short-term jitter (like phase noise above 1Hz) is dependent on the onboard PLL. This means a fixed frequency temperature-stabilized oscillator will provide lower jitter at a much cheaper price. Now, what is the scope bandwidth? At what frequency do you want to measure? And also, are you interested in cycle to cycle jitter? Or period jitter? Or phase noise? Or... other? \$\endgroup\$
    – bobflux
    Commented Nov 9, 2017 at 15:45
  • 1
    \$\begingroup\$ "I want to measure how much noise is a certain oscilloscope responsible for" - why? \$\endgroup\$ Commented Nov 9, 2017 at 19:01
  • 2
    \$\begingroup\$ That's way above my league LOL. Considering the $250k "starting price" of such equipment, discussing with one of the manufacturer's engineers about your specific requirements should be an expected part of the sales process, if only to be able to select between the myriad of options... Consider talking to a Tek field applications engineer (I think this is the correct English translation?) about your setup, it's their job to make sure you can do your job ;) \$\endgroup\$
    – bobflux
    Commented Nov 10, 2017 at 9:17

1 Answer 1


Using a square wave signal will not get your desired results unless you can control the square wave very precisely, which is not easy. An easy way to measure noise in such a system is to use a sine source, with as little noise as possible (but not overly so), use a power splitter and connect it to two inputs of the scope. Then sample the inputs and store the raw samples. Using cross correlation, you can eliminate the noise of the signal source you have used and get the noise of each channel. But be aware that there are certain noise sources within the scope that are equal on all channels and thus will cancel out when you do cross correlation. An important one of these noise sources is the noise of the sampling clock, which is correlated between channels.

If you want to measure the sampling clock jitter, you have to start with a known good signal, which has a noise that is lower than the jitter you expect from the scope. As I guess you want to measure in the GHz range, this will be most likely be a special DRO or if you cannot find a DRO with low enough noise a CSO. Unfortunately, a CSO is prohibitively expensive to buy, unless you are at a university that owns one. Alternatively you can ask your national metrology institute whether you can stop buy for some measurements.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.