How to measure the frequency response, phase noise and analog bandwidth?

[Voso]

I have an arbitrary waveforum generator m8195a from keysight. The data sheet of the device includes a graph for the frequency response, phase noise and a rating for the analog bandwidth (25Ghz). However, it is not indicated, whetther these two parameters (frequency response and phase noise) are for the whole device, or the output amplifier or the output of the device. So I want to know this at first. Moreover, I am interested in knowing how to measure these parameters.

data sheet can be found here:

https://www.keysight.com/de/de/assets/7018-04509/data-sheets/5992-0014.pdf

I know from my studies that to measure the frequency response of a circuit (RLC for example) I apply a sweep sin wave to the input and measure the output, then find the ratio of output to input for the bandwidth of interest. I can also use an oscilloscope to do this and use frequency response analysis tool to get the bode plot showing both gain and phase. Is it possible to do the same to the AWG? How?

For phase noise I think I can use a spectrum analyzer to measure it, not sure though

For analog bandwifth I need to find the frequency at which the output voltage drops by 3db, but I have no idea how to do this. Normally I would use a generator to measure the bandwidth of an oscilloscope, by giving a signal as an input to it, then increasing the frequency of the signal on the generator until I see the voltage on the oscilloscope drops by 3db. But I dont know how to do this for an AWG or if its doable.

Any suggestions?

[Kleinstein]

The frequency response is quite straight forward: the generator sets the frequency / test signal and another meter (power meter, scope, thermal RMS converter or spectrum analysator measures the amplitude). In this case the difficulty is to have a good meter for the amplitude and good cables to know that the drop in the responde is really from the generator and not because of the scope, power meter or cable.

The phase noise could be with a SA, if the sensitivty is good enough. Phase noise is also usually more than just 1 number and depends on the delay / frequency offset. With a good signal source it may also need a good instrument to check. There are other tests to check phase noise, at least at some frequencies. A long cable can be used as a delay and one can than check the phase difference at both ends of the delay.

[Voso]

By a good scope or meter, does it mean I need one with a frequency response that is flat over the advertised analog bandwidth of the generator? or at least better than the generator? Because the frequency response I get on the scope will be a combined response of the whole system, isn't?

[Kleinstein]

The frequency response would the the combined response of genrator and meter. At the very high frequenies the cable can be a factor too.

I am not sure a scope would be the right instrument, at least it can be expensive. Chances are a thermal converter may be best choice. It is slow and not very universal, but can have a good frequency response.

To some point one may have to trust the specs / calibration. If one has a suitable instrument at hand it is still good to check. If not one may just accept the specs.

[Berni]

That is one REALY high end signal generator so to verify these specs you will need some seriously high performance gear. I don't even want to ask how much that fancy signal generator costs.

The specs you see in the datasheet are most likely of the whole system combined, since what is useful to the user is what comes out of the coax connectors on the front. So all that you need to do for frequency response is connect something that can measure AC amplitude up to 25GHz and step the signal generator trough the whole frequency range for sine output. The thing doing the measurement can be a oscilloscope, spectrum analyzer or RF power meter (In the order of most to least expensive). Of course whatever you are measuring with has to have more bandwidth a good deal higher than 25GHz as you want the signal generator itself to be the limiting factor. Any cables you use to connect them also have to be characterized and compensated for as even fancy expensive microwave cables have a lot of loss at 25GHz.

As for phase noise, you would typically use a high performance spectrum analyzer to observe a continuous sine wave output and look for any modulation sidelobes on it. Modern spectrum analyzers have a specific feature for doing this and will generate a phasenoise vs frequency plot. As before the spectrum analyzer doing this has to have significantly better phase noise than your signal generator.