Customer update:
Thanks for the response. Actually the engineer from T.I. has reproduced our results, only the discrete 5 MHz lines aren't showing up in his case (we don't know why). However, the noise floor is a problem. Please look at the attached Word document in which I've pasted his spectrum picture (blue trace), together with a plot (yellow trace) that he sent previously--see the high noise floor between 120 and 1200 MHz?
(Please visit the site to view this file)
The blue trace has the same shape as the yellow trace between 200 and 900 MHz, and the noise floor is being generated by the CC1125. If we amplify that noise floor by 20 dB, it exceeds the -49 dBm allowable limit. Can you ask him to set up his analyzer the same way, but then send the SIDLE command to the CC1125--the noise floor between 120 to 1200 MHz will drop down to the noise floor of the analyzer (and being an Rhode & Schwarz should be at least 15 dB lower).
Why does the CC1125 generate this broadband noise between 120 and 1200 MHz?
I did some experiments with the CC1125's frequency synthesizer calibration. FS_VCO2, FS_VCO4 and FS_CHP calibrated values only change by 1 or two codes between room temperature and +85 Celsius, and the part seems to still work. How concerned should we be with the calibration errata? We would like to avoid the manual two-step calibration. What happens if the calibration value comes out slightly wrong due to the errata? I can't find anything in the documentation that states what kind of performance degradation would result from lack of calibration, mis-calibration or wrong calibration values.
Attached are more test results.
t (Please visit the site to view this file)
What it comes down to is we need the spurs to be at least -80 dBc down (because +30 dBm - 80 dB = -50 dBm and the limit is -49 dBm). Actually, -85 dBc is required to give some margin.
The best we've been able to get with the CC1125 is -74 dBc at certain power output settings.
The closest spurs occur at +/-40 MHz (for 40 MHz XOSC) so the closest spurs to the desired band would be at (902 + 40) = 942 MHz and (928 - 40) = 888 MHz. We will try using a SAW filter between the CC1125 TX and the exciter for our P.A.
The SAW filter we tried gets rid of all but one spur, which is at the carrier frequency minus 21.4 MHz.
(Please visit the site to view this file)
We see the -21.4 MHz spur in two different spectrum analysers. But when we generate a signal from an HP signal generator and feed it to the two analysers, we get the same -21.4 MHz spur!
The best we can figure is that both our analysers (of different brands) use an internal SSB mixer with 21.4 MHz L.O.
The SAW filter in the CC1125 TX path to our P.A. is probably the only way to go.
Also the 920 MHz (23 times 40 MHz XOSC) is at -60 dBc to -70 dBc (depending on the TX frequency selected). It stays there and doesn't change frequency (of course), and we haven't decided what to do about it yet. We know we can't use that frequency, but aren't sure if we can notch it out sharply enough to pass the TX conducted spurious test.
We are still interested in data from the Skyworks 1-watt reference design.