On 21/07/2012 16:28, Alexander Chemeris wrote:
Hi Srdjan,
On Sat, Jul 21, 2012 at 4:33 PM, Srdjan Milenkovic s.milenkovic@limemicro.com wrote:
Hi all,
As far as I am aware, you are currently discussing how to improve LMS6002 PLL phase noise. Below are some inputs from my side which may help.
Yes, here are some pictures of phase noise we have at UmTRX right now: http://lists.osmocom.org/pipermail/umtrx/2012-July/000030.html
It looks like they're 5-12dB higher then data I saw in your temperature measurement report. First thought is that this could be due to a clock source. Did you use your EVB board for those measurements?
Yes, we used Lime EVB, 30.72MHz TCXCO. However, using 26MHz instead of 30.72MHz TCXCO should not affect phase noise so much (5-12dB). Do you have an alternative 26MHz TCXCO with better PN? As you quite rightly mentioned, you are probably limited by TCXCO PN at the moment.
Apart from playing with PLL registers we have two additional options.
- Use clean TCXCO which provides 4 times higher reference followed by
divide by 4 to generate PLL reference clock. Recently, we experimented with 30.72*4 MHz TCXCO. Dividing its output by 4 before going into LMS chip we have got improvement of 6dB in phase noise plateau region.
- Current PLL loop filter has been designed to cover the whole LMS
frequency range hence using kind of mid value for Kvco. We can customize the loop filter for a particular band.
Well, we're working with GSM and it has only four main bands: GSM-850: 824.2–849.2 UL / 869.2–894.2 DL E-GSM 900: 880.0–915.0 UL / 925.0–960.0 DL DCS-1800: 1,710.2–1,784.8 UL / 1,805.2–1,879.8 DL PCS-1900: 1,850.2–1,909.8 UL / 1,930.2–1,989.8 DL
It would be great if we could optimize for the use in these bands
Let us first try to recover 5-12dB PN as mentioned above. Redesigning loop filter will not help at all if we are limited by TCXCO.