At >If "ADC clock jitter" is what Flex claims is the cause of the flat >phase noise then I agree with you that it must be because they have >not analyzed the problem in detail. > >If I had to guess it would be the use of the LVDS receiver for clock >squaring the LO drive to the QSD. I think the F5K uses an AD9959 DDS >where the LO output is filtered by a 9 or so pole LPF. The input and >output to the LPF is probably transformer coupled. I think they use a >LVDS receiver (SN65LVDT34) to convert the differential output from the >LPF to a single ended square wave clock to drive the QSD switch. >These LVDS receivers have a fixed jitter contribution in the range of >20 - 60 ps peak. Since it comes after the DDS (which is clocked by a >low noise 500 MHz oscillator from Valpey Fisher) I'd bet that this >flat phase noise is caused by the fixed jitter contribution of the >LVDS receiver. The SDR-1000 had the same kind of problem and it >used a LVDS receiver for the same purpose. > >I did some testing in the past of various LVDS receivers and their >contribution to clock jitter which verified that the lower speed parts >like the 'LVDT34 do contribute an average of 30 ps of jitter. In the >application that we were going to use them for this was unacceptable. >I think 30 ps of jitter is unacceptable in a SDR application also.
I'm not sure that uniformly distributed jitter (which it probably isn't, anyway.. because you're basically thresholding a noisy sine wave, but an assumption of uniform distribution is probably reasonably valid.. the slope of sin(x) at zero is pretty close to linear over a fairly wide range) results in a flat phase noise characteristic (viewed on a classic ssb noise plot). For one thing, the "noise" added to the ideal sinewave isn't gaussian, but is (partially) the phase noise of the source, and partly the analog noise in the thresholder itself. For example, what's the noise distribution of a PFD in a standard PLL... I have to think about this to figure out what it would be. I've actually be trying to do this analysis for a radio design at work, where we're trying to model the stackup of all the phase noise contributions. At some point, though, you bail-out on the models and just build a prototype and measure it. Mostly, the model just tells you where the tall-tent-pole is or whether you need to obssess a bit more about something. _______________________________________________ FlexRadio mailing list [email protected] http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archive Link: http://www.mail-archive.com/flexradio%40flex-radio.biz/ FlexRadio Knowledge Base: http://kb.flex-radio.com/ FlexRadio Homepage: http://www.flex-radio.com/
