24GHz VCO design

Hi, I have built a discrete VCO at 24 GHz, with a tuning range of 1GHz, tuned at the varactor diode. Now I am trying to measure the maximum tuning speed of the VCO. I tune the VCO with a function generator (the tuning voltage is set as a rectangular series pulse or a senoidal signal). As I increase the tuning speed, I see that the output signal (the two peaks) power in the spectrum analyzer (frequency domain) drops. I thought it were the indication that the VCO can no longer follow the high tuning speed. Later I realized that as the minumum sweep time of the spectrum analyzer is 2.5ms, and when I tune at 1MHz, the IF filter of the spectrum analyzer cannot select the signal of the interest properly. Thus I conclude that the drop that I see is due to the attenuation by the IF filter, instead of the power drop in VCO. Then I set the spectrum analyzer in zero span mode, but strangely the output signal still drops at the same range of tuning frequency. (around 3MHz) I also measured with a power meter, this time the output power does not drop at all even at much higher frequency. Finally, I measured with a digital sampling oscillator (able to cope until 34GHz), then at 100MHz of tuning frequency, I see that the shape of the envelope of the signal is no longer in pace with the shape of the tuning signal. My question is, does the output signal of the VCO really drop due to high tuning speed? Because I see in one article, that at high tuning speed, the tuning range of the VCO actually shrinks, but it does not mention that the output power of the VCO drops. For me, this criterion of shrinking tuning range makes more sense, so I would say that the tuning speed of the VCO is around 100MHz. This conclusion is drawn from the observation of the output signal in the oscilloscope, because the envelope illustrates the AM in the output signal, due to the modulating signal. So when the envelope loses its shape, it means that the VCO cannot follow the tuning signal any longer, is it right? One more thing, when I put a wide or narrow tuning range for the tuning speed test, it has the influence on the maximum tuning frequency I observe, right. So how should I include these two parameter for the characterization of the tuning speed of the VCO. Also I am still puzzled by the fact that the zero span measurement does not give the expected result. If somebody works with it, or has worked with it, or has interests in it, please give me some hints. Many thanks