LMX2595: Calibration at 25C does not lock at -40C for VCO4

Hi Dylan,

I don't have short answer, I will look into this.

Hi Noel,

Any luck on this one? For context our loop Fpd is 200MHz w/ a 3mA charge pump, we did notice better performance with a 6mA charge pump but the improvement was marginal.

Hi Dylan,

I checked this out in the lab, with the default TICS Pro configuration, I've got the following plots.

The memory trace (light blue) is the phase noise at room temperature, full assist mode with calibration data at room temperature as well. 

The active trace (dark blue) is measured at -40C. We can see that the loop bandwidth is smaller and therefore close in phase noise higher. This is due to Kvco being smaller at cold temperature. This is a normal behavior of the VCO.

Then I change the charge pump to 3mA. Loop bandwidth gets smaller but the overall phase noise is normal. 

In your design, what is the loop bandwidth and phase margin?

If the phase margin is already small (e.g. <40 degree), because Kvco is smaller at cold, then the phase margin will further decrease at cold temperature. If phase margin becomes too small, the PLL may not be stable. 

One more thing, make sure auto-calibration use the following setting for all frequencies.

VCO_SEL = 4

VCO_DACISET_STRT = 300

VCO_CAPCTRL_STRT = 1

Hey Noel, thanks for running that experiment for me and providing those plots

I believe our phase margin is around 47-50, we are running a 200MHz Fpd w/ 3mA charge pump (expanded to 6mA to increase loop bandwidth improved results but we would prefer to keep 3mA)

Our concern however is settling time, since calibrating at -40C causes it to lock at all temp. However, settling time is worse at -40C than it is at +25C and +71C - a lock time at +25C of around 8µs would be 11µs or so at -40C. Additionally, the waveform on the SSA appears to be underdamped somehow, not quite sure why it appears like that or why the performance is much worse. This is jumping from the middle of VCO cores as well, it's not like we are on the edge of one or anything like that

Would you be able to provide Kvco curves over temperature for the cores? From our sims and loop filter design we should be passing at -40C but can't seem to figure out why we are having issues with VCO4

Hi Dylan, 

Let me see if I have data available otherwise will take the data.

BTW, did you calibrate the VCO with the following setting?

VCO_SEL = 4

VCO_DACISET_STRT = 300

VCO_CAPCTRL_STRT = 1

Hi Noel, sorry for the delay

Yes, I just double checked my initial register values and those are all the values we have set for LMX 

Hi Dylan,

Here is the Kvco vs temperature at VCO4.

Thank you for running those tests, I think we will try and resolve this issue by increasing the charge pump to 12mA from 3mA when running at cold

A follow up question to this approach: would changing the value of charge pump used during the auto calibration have any effect on the values the PLL self calibrates to? We are currently using a 3mA pump current, but I wonder if using a 15 or 12mA value instead would improve the performance during Full Assist