AnsweredAssumed Answered

Large EVM at low data rates in 89600

Question asked by hudla on Apr 8, 2010
Latest reply on Apr 20, 2010 by tabbott
Hallo all,

I wanted to try the 89600 software Trial and to check the EVM specifications of a ESG generator in our lab (ok, not metrologically correct, but Agilent does it similarly). The generator was set as follows:

carrier freq. = 50 MHz, output level = -10 dBm
mod. = 16QAM, Root Nyquist 0.35 filter, data PN9
diff. encoding = off, optimize FIR for EVM
symbol rate = from 10 ksps up to 5 Msps

The generator was connected to the PSA spectrum analyzer with B7J general demodulation option, external trigger from the generator's EVENT1 output has been used and 4096 symbols were captured with 89600 software. The problem is a large EVM value for low data rates which is much greater than ESG's specifications. For 89600's Measurement filter = Off, Reference filter = RRC 0.35, I obtained following numbers:

10 ksps: rms EVM = 4.5 %
20 ksps: rms EVM = 2.6 %
50 ksps: rms EVM = 1.47 %
100 ksps: rms EVM = 0.90 %
200 ksps: rms EVM = 0.52 %
500 ksps: rms EVM = 0.41 %
1 Msps: rms EVM = 0.43 %
2 Msps: rms EVM = 0.57 %
5 Msps: rms EVM = 1.5 %

For Measurement filter = RRC, Reference filter = RRC 0.35, the constellation seems to be much better and the points are exactly at their positions, however the EVM values are even worse (over 10 %). I obtained similar results for QPSK and different carrier frequency setting. Settings of Range, RBW, Window Type in 89600 doesn't help much. What could be wrong?

The 89600's recordings are large (7 MB each), in case of interest you can find them here (the last number in the filename denotes symbol rate):
http://klokan.sh.cvut.cz/~hudla/89600/

Thank you for any suggestions,
Martin  

Outcomes