[time-nuts] 35601A as stand-in for 11848A ?

John Miles jmiles at pop.net
Sun Jan 6 22:55:46 EST 2008

> how _exactly_ are you turning the dBc/Hz
>  values you
> >get from (say) the 8561E into your ~-120 dBc/Hz  numbers?  Let's hear the
> >step-by-step procedure on that, and the  problem should reveal itself.
> Hi John,
> by selecting a 1Hz RBW on the 8561E.

And subtracting (LNA gain + 6 dB - about 2.5 dB), right?  Don't forget that
the 8561E is also reading out in dBm, so you have to take the carrier
amplitude out of that reading as well.

Not to belabor the two points at all, but you *really* want to read the app
notes, and you probably want to use either the HP 85671A package or my
host-based clone of it.  This is not something you want to do manually if
you can avoid it; too much room for human error.

> The APII analyzer should
> have a 1Hz  BW
> normalization as well, although I checked in the user manual
> again and it
> just specifies "dbm" without regards to RBW. How does one
> calculate RBW from an
> FFT? Is it the width of a given FFT bin? I will check this again,
> and try to
> increase the number of samples.

Yes, the RBW for the 10*log(N) equation would be the Nyquist bandwidth
divided by the # of bins in the FFT, aka the Hz per bin.

However, FFT window functions also have different equivalent noise
bandwidths; see http://www.bores.com/courses/advanced/windows/10_end.htm for
a table.

> But even with say a 10Hz RBW I would expect the noisefloor to only rise
> 10xlog(1/BW) or 10dB, much less error than what I am seeing.

If you're not taking the LNA gain out of the picture, that would explain it.

-- john, KE5FX

More information about the time-nuts mailing list