I've been pouring over App Note 150-2, among others, and I'm trying to understand where the pulse desensitization factors come from.
Specifically, I'm looking at the factor for line spectra which is given as:
20log(teff*PRF)
I understand that when the PRF doubles, the spectrum contains half as many spectral lines, each line containing twice as much power. That only seems to account for half the desensitization, however.
This term seems to be following a voltage law, rather than a power law, but the plots are labeled in dBm-- which implies that it is, in fact, power.
Specifically, I'm looking at the factor for line spectra which is given as:
20log(teff*PRF)
I understand that when the PRF doubles, the spectrum contains half as many spectral lines, each line containing twice as much power. That only seems to account for half the desensitization, however.
This term seems to be following a voltage law, rather than a power law, but the plots are labeled in dBm-- which implies that it is, in fact, power.
Apologies if I'm wasting your time on naive questions-- I thought I'd read that the analyzer was capable of calculating power from voltage and impedance so I was assuming a difference between dBm and dBmV...
If I pulse modulate a carrier, and then reduce the duty cycle by a factor of 2 by increasing the PRF by a factor of 2, I expect half the power to be delivered and for the display to show a change of -3dB.
The alphaL factor in app note 150-2 suggests the displayed change would be -6dB, which would indicate that power was reduced to a quarter of the original.
Of course, if dBm is really dBmV then all this is moot.
What you are saying makes sense to me for amplitude modulation, but not for time modulation. In the case of pulse modulation we'd be delivering all of the voltage (and thus all of the power) for half of the time...
I think it's the difference between:
((.5*V*sin(wt))^2 * teff * PRF) / ((V*sin(wt))^2 * teff * PRF)
for AM and:
((V*sin(wt))^2 * teff * (.5*PRF)) / ((V*sin(wt))^2 * teff *PRF)
for pulse modulation.
Thanks for sticking with me on this.
Is that what is happening? The analyzer is summing then squaring rather than squaring then summing?
If that is the case, then understanding the analyzer is critical to understanding how to relate my measurement to my real signal.
Can you help me to understand the signal path and where the integration and squaring are occurring, and how I can determine the integration time?
If what I'm imagining is correct, then I would expect two side effects to show in the system:
1) For a continuous pulse train, I would get the same reading using the peak or average detectors?
2) If the integration time is, say, 10/PRF and the signal consists of a repeating pattern of 5 pulses and 5 blanks, then my peak power would be underreported? Issuing 5 pulses at twice "peak" power would report the same as issuing 10 pulses at peak power?
(If the RBW is too wide to see each line separately, then the dBm value will not be correct, since it will include power levels of multiple lines. This is covered in chapter 3 of the app note.)
I was mistaken. The correct way to look at it is to transform the pulsed RF signal to the frequency domain and look at the effects of duty cycle on each frequency component. I hope this will get you going in the right direction. My apologies for the previous error.
As an example, (I happen to be working on this at the moment), the single-sided frequency components of a periodic pulse (no RF) are: a_n = 2*A*d*sinc(n*d), where d is the duty cycle, for n=1 to infinity (integer). Power is proportional to a_n squared.
Hope that clears things up.
Thanks again.