converting a P.S.D trace to a time domain signal

[Declan]

Hi all
A project I am working on at the mo uses a 1MHz pulse train of 1% duty cycle to switch on and off a carrier of the order of 6GHz. I have been able to recover the sinc spectrum of this signal on a hp8562a spec an. The recovered P.S.D trace is made up of 601 data points. From this trace I was hoping (by the use of the inverse fft function) to recover the time domain pulse so as to measure its rise time, fall time, width etc. I was hoping someone had tackled a similar problem before.
Is what I am trying to do at all possible from a simple PSD trace?
If so should I use the real or complex inverse fft function?
Can I recover a time axis without any phase info?
I know I ask a lot but am under pressure for time and
any help would be much appreciated, no matter how small!
P.S I am working with LV 6.02, win XP and talking with spec an using GPIB.
T.I.A
Dec.

[Andy_Hinde]

I don't believe this is possible. Before you calculate the power spectral density, you must first have a power spectrum. When you convert from a FFT to a power spectrum, your phase information is lost. With magnitude information only, it isn't possible to go back to the time domain from the frequency domain.

One solution would be to use the PXI-5620 frequency domain digitizer with the included Spectral Measurements Toolset. The PXI-5620 is a 14-bit, 64 MSample/sec digitizer which you could use to digitize the raw time domain data. The Spectral Measurements Toolset can be used to turn this data into FFTs, power spectrums, and power spectral density traces, in addition to performing several types of spectral measurements upon them. In addition, since you would
be acquiring the data with a digitizer, you would have access to all the raw time data and could make any number of time domain measurements with built in LabVIEW functionality. With 32 MHz of Nyquist bandwidth on the digitizer (one half of 64 MSample/sec), you could capture all the odd harmonics of the 1 MHz pulse up to 31 MHz which should be enough to capture the pulse adequately.