LMX2820: LMX2820 phase synchronization for two device after changing frequency

Qiaonian,

Our team is currently out of office for the US holiday. Please expect a response by next Monday.

Thanks,

Kadeem

OK, thanks for message.

Hi Qiaonian,

This looks weird to me, can you try direct VCO output instead of doubler output? 

Thanks for your suggestion, I tried to disable the VCO doubler calibration and it works. It seems that recalibrating the VCO doubler affects the phase of certain special frequencies. However, I don't understand why the VCO doubler calibration affects the phase of the clock. And I have two more questions, 1) I want the frequency configuration bandwidth to be 13.7~20.4Ghz. If I only calibrate the VCO doubler once at the beginning, for example, when set chip to 17Ghz, we calibrate the VCO doubler, will there be any problem with this process? 2) I found a regular phase jump. When I change the frequency in steps of 1Mhz, I found that there is a phase jump about 5 to 8 points. This jump is about 10 degrees. What may be the reason for this? For example, the following figure is the scan result. The figure shows the change of phase with frequency (frequency Ghz, phase degree).

Hi Qiaonian,

One doubler calibration data is not enough to support your frequency range.

You can try this: do doubler calibration at 15.5GHz, use this data to cover 13.7G to 17GHz. Do another doubler calibration at 18.5GHz, use this data to cover 17G to 20.4GHz. 

Phase jump is probably due to VCO_CAPCTRL change because of VCO frequency change.

Hi Noel,

I am currently using the automatic VCO calibration operation. Is there any other way to solve the phase jump problem?

Hi Qiaonian,

We can use full assist to make the configuration of every device same. However, this involves a lot of registers read and write. 

Could you repeat the phase vs frequency test but read back the VCO parameters at each frequency? Purpose is to confirm if the phase jump is due to big VCO parameters change.

Hi Noel,

The following results shows the values ​​of the three registers I read as the frequency changes.

(frequency Ghz, phase degree, R74 for rb_VCO_SEL&rb_VCO_CAPCTRL, R75 for rb_VCO_DACISET)

Hi Qiaonian,

This is normal, when the device selects a different VCO core, then the CAPCTRL will be very different. 

Hi Noel,

I see that not all cases will have a phase jump after selecting a different VCO core, and even if a phase jump occurs, the phase may return to the correct track when the frequency is changed next time (the VCO will be recalibrated), is this because of the recalibration or sync?

I found that there are also many cases where the three registers above do not change, but there is still a phase jump. It seems that VCO core different selection is not the only reason.

Qiaonian,

I note that this is very high frequency, around 18 GHz.  This has a period of less than 50 ps.  I see that you are not using instant calibration.  So when you calibrate, which is activated during SYNC, you might be choosing a different value for VCO_CAPCODE or VCO_DACISET.  This causes the charge pump voltage to be just a little different and 180 degrees on a 50 ps signal is only 25 ps.

VCO_CORE is the strongest factor, but you might want to also force VCO_DACISET and VCO_CAPCODE.  Also realize that instant calibration uses the same values each time.

Hi Dean,

Thanks for your reply. You are right, the frequency that i am using is very high. In fact, in general I observe a maximum to minimum phase change of about 3 degrees, for example, after multiple frequency settings and calibrations at 17Ghz, 3 degrees is about 0.5ps, it works very well. But for phase jump cases, there is a 10~15 degree change. This may be caused by the change of VCO_DACISET and VCO_CAPCODE as you said. I have tried to force them to the same value, and it seems to have no improvement. I will also try the instant calibration mode, which seems to require a forced calibration and initialization at 5.65 GHz.

Hello,

The team is out for the holidays. Please expect a response by Jan 2 at the earliest.

Thanks,

Kadeem

Happy holidays!