AnsweredAssumed Answered

HP8753 a question about IFBW and averaging

Question asked by b.d.hall on May 2, 2010
Latest reply on May 11, 2010 by Dr_joel
I have some unexpected results using an HP8753ES to make S12 and S21 measurements of a thru connection. The measurements were made to assess the performance of the instrument. A thru is left connected after a 2-port SLOT calibration.

Perhaps there is some mistake in the instructions sent to the VNA. I would be grateful for an opinion as to whether the following commands seem to be correct or not.

A frequency range of about 250 points between 30 kHz and 3 GHz was defined. The VNA stepped list mode is used, set up like this (pseudo code, but the VNA commands should be obvious)

output('SADD;STAR%eHZ;STOP%eHZ;STPSIZE%eHz;SDON;' % (start_Hz,stop_Hz,step_Hz) )

IFBW and averaging are set to specific values and then a large number of measurements are made (about 100 repeats) and the variance of S12 and S21 at each frequency is calculated.

The measurement command depends on whether averaging is used or not

      query('OPC?;NUMG%d;' % average)

Now, with averaging off I observe a continuous variance that is roughly constant across the frequency range and the variance drops as expected when IFBW is decreased (1000,100,10). 

However, when averaging is set to 4 sweeps a discontinuity appears at mid-range for IFBW=10 (but not for 1000 or 100). At lower frequencies the variance behaviour is unchanged but a step-change occurs above about 1.6GHz and the variance can be as much as 1000 times higher. The variance seems to be uniformly distributed up to this limit of about 1000 times the low-frequency value, giving the appearance of a speckled block on a scatter plot.

When averaging is set to 16, the IFBW=10 behaviour exhibits two steps over the frequency range. The first step is now at about 500 MHz and the second at about 2 GHz. Once again a scatter plot looks like a uniform distribution of variance values up to some maximum. This maximum is about 6 times bigger at 2 GHZ than it is at 500 MHz.

When averaging is set to 32, the behaviour for IFBW=10 changes again (2 steps but the 2GHZ step is now at about 1.3GHz) and a step is now seen in the variances for IFBW=100.

Well I imagine that it is hard to visualise the data just described and I do not think that it is possible to post a figure.

What concerns me is that there appears to be an interaction between IFBW and averaging: when averaging is 'too long' compared to IFBW something bad is happening to my data. My understanding is that the 8753 takes whatever time it needs before is steps to the next frequency point, so IFBW and averaging should be independent.

I would be grateful for some advice on how to go about looking for the cause of this effect.

I wondered about setting the VNA sweep time, but then I am in stepped-mode so would this have any effect?