AnsweredAssumed Answered

ENA IF band width and band center issues

Question asked by GW on Aug 31, 2006
Latest reply on Sep 14, 2006 by OsamuKusano
I’ve been struggling to test an unusual multi-channel wideband up/down converter (< 2 MHz to > 7 GHz) and needed to accurately measure the LO frequencies before looking at conversion gain flatness. Not having a microwave counter handy I came up with the idea of using the E5071 since it had the high accuracy/stability time base option. I configured the ENA for A(2) absolute mode, shut off the RF, and connected port 1 to the first LO (around 2 GHz) using a probe. Using the bandwidth function in the marker search menu, I read the LO frequency but was troubled by the error. I switched IF bandwidths to get a closer look and was surprised to find that the error changed. Each IF bandwidth selection I tried gave me a different value. Puzzled, I connected the ENA to an 8642B that I had phase locked to the ENA, set it to 2 GHz, and repeated my measurements but with the same results. Thinking the 8642B was not actually locked to the ENA, I connected port 1 to the 10 MHz reference output of the ENA (via pad and DC block). The same error showed up! With the ENA measuring itself I expected 0 Hz error just as a counter would do. Just the same, I measured the 10 MHz reference with my trusty old 5335A recently calibrated to a rubidium source and got 9.999999248 MHz…not the source of my problems. Now even more curious, I went through the whole range of available IF bandwidths and recorded the IF bandwidth and center frequency error. I was surprised to get the following:

IF BW setting = 100 kHz, measured BW = 47.3 kHz, center error = -840 Hz
IF BW setting = 70 kHz, measured BW = 44.3 kHz, center error = -840 Hz
IF BW setting = 50 kHz, measured BW = 41.5 kHz, center error = -170 Hz
IF BW setting = 40 kHz, measured BW = 35.0 kHz, center error = +210 Hz
IF BW setting = 7 kHz, measured BW = 7.14 kHz, center error = +24 Hz
IF BW setting = 5 kHz, measured BW = 5.06 kHz, center error = -1.56 kHz
IF BW setting = 4 kHz, measured BW = 4.15 kHz, center error = -1.29 kHz
IF BW setting = 10 Hz, measured BW = 9 Hz, center error = -3 Hz

The omitted data (….) was basically intermediate values between the endpoints without inflections. The two big surprises were the large error in bandwidth for settings above about 20 kHz and the large step in center frequency error between 7 kHz and 5 kHz and the smaller one between 70 kHz and 50 kHz. The two attached screen shots show the shift (the only difference in the pictures is the IF bandwidth…also ignore the BW value for the 5 kHz case since the cursor was not placed at the response peak).

None of this would bother me too much except that I need to make measurements on a frequency translating assembly with embedded LOs so that the LO frequency error must either be known precisely or the ENA’s bandwidth must accommodated it. As to the second condition, the ENA having its own band center error was not expected. Also not expected is the use of a Gaussian IF filter shape such that there is always an amplitude error unless the filter is dead center on the signal (using the 5 kHz setting would have produced a 1 dB amplitude error for an on frequency signal).

Am I doing something wrong (I would be happy to know this all goes away because I set something up incorrectly)?
Is the variable bandwidth error and center frequency error a byproduct of FIR filter coefficient quantization error?
The error appears unrelated to the RF carrier so with the ENA tuned to the bottom of its frequency range (300 kHz) the worst case center frequency error works out to be a whopping 0.5% or 5,000 ppm for the 5 kHz setting. Shouldn’t this be mentioned in the instrument specs?
Why is there more than a factor of 2 error in the bandwidth of the highest setting (100 kHz IF produced a 47 kHz BW)?
Why the use a Gaussian filter for frequency offset mode where, more often than not, measurements are made without access to the LO(s)?

P.S. ENA FW is A.06.05