Tx signal quality improvement

Hi all, I've found an issue with out Tx configuration of LMS which led to constant phase rotation of the transmitted signal. After fixing this issue I get quite better signal quality on transmit side. I advise everyone to update to the most recent UHD on the host side to benefit from this improvement. The reason for the issue was that we used the wrong polarity of TX_ IQ_SEL. As a result, LMS transmitted I from one sample and Q from the other sample. One of the visible results was a huge I/Q imbalance on Tx which made it impossible to use multi-ARFCN configurations. We've solved this issue in software by setting tx_fsinc_polarity and tx_interleave_mode to proper values in LMS configuration registers. Attached screenshots from our E4406A shows signal quality transmitted by OpenBTS we could have now. Tests are performed with "laurent" branch of Fairwaves, compiled with SPS (samples per symbol) set to 4. DC offset and I/Q imbalance are calibrated. screen2.gif, screen3.gif - constellation and phase error when LO leakage is at the center of the signal spectrum, i.e. with no LO offset on Tx. screen4.gif, screen5.gif - constellation and phase error when DC offset is when LO leakage is shifted to 300kHz away from the signal spectrum center, i.e, with 300kHz LO offset on Tx. screen6.gif - phase error per symbol plots for the same settings as with screen4.gif and screen5.gif. When we implement a true GMSK transmitter we should be able to get even better phase noise parameters, meaning excellent downlink signal quality. But even this values are much better than required by the Standard. -- Regards, Alexander Chemeris. CEO, Fairwaves LLC / ООО УмРадио http://fairwaves.ru