LMX2492: High level of wandering spurs in CW, low when ramping

Hi Szczepan,

At 9650MHz output (no ramping), will the wondering spurs appear if you make the denominator = 1000?

Hello,

After choosing the Default Mode 2016-6-17, when I switch the denominator to values like 1000, 2, 16777214‬ and the numerator to the half of denominator the high level of wandering spurs does not occur (spectrum measurements give the same results for these 3 cases). To compare - every trace corresponds to LMX2492EVM configured to 9650 MHz CW:

Orange trace - max hold, peak detector, denominator different from 16777216, numerator set to half of denominator

Yellow trace - single fast sweep, peak detector, denominator set to 16777216, numerator set to half of denominator

Blue trace - max hold, peak detector, denominator set to 16777216, numerator set to half of denominator

In fact, I have also noticed that when setting denominator to 16777216, TICS Pro automatically sets bit RAMP_EN in register 0x3A to 1. With different denominator values it holds it set to 0. When I tried to program LMX2492 from my MCU to denominator = 16777216 without enabling ramping functionality, the PLL simply ignored the fractional part of N divider.

Hi Szczepan,

So the problem is actually due to TICS Pro which automictically enables ramping whenever you set the denominator to 2^16? In other words, you don't have the wondering problem anymore if you make the programming through your MCU? 

Hello Noel,

Programming using MCU did not solve my problem. Actually, programming denominator to 2^16 without enabling the ramping functionality results in ignoring the fractional part of N divider by LMX2492 which then works as in integer mode.

Hello Noel,

Thank you for your answer. I have tried to set DEN to 16777215 and NUM to 8388608 (after choosing Default Mode 2016-6-17 in TICS Pro) which gave roughly 9650 MHz CW. Level of wandering spurs was the same as earlier for DEN = 16777216 and NUM = 8388608. Please see the attached LMX2492EVM spectrum measurement results.

TICS Pro configuration after choosing Default Mode 2016-6-17:

Orange trace - max hold, peak detector, DEN = 16777215, NUM = 8388608
Blue trace - max hold, peak detector, DEN = 1000, NUM = 500
Green trace - max hold, peak detector, upchirp ramping on (RAMP0, 9650 - 9651 MHz, dly enabled, 1000 us duration, RST enabled)

It can be seen that spurious level for ramping enabled is lower than for CW with DEN = 16777215, NUM = 8388608. I am wondering why this phenomenon occurs and how I can optimize phase noise if the spurious levels are different for ramping and CW.

Hi Szczepan,

Could you try DEN=16777210? I am sorry I have to let you try many experiments as I do not have lab access at the moment.

I have never seen this kind of moving spurs before, we have to do some experiments to find out the source. Could you also try with a 3rd order modular (FRAC_ORDER = 0x3)? 

During ramping, the PLL is in fractional mode and the NUM will be continuously incremented in order to sweep the VCO frequency. Since DEN=2^24, the fractional engine is highly randomized, discrete fractional spurs will be spread over the spectrum, as a result, the phase noise gets worst. We can try with different FRAC_ORDER and PFD_DLY values to improve the phase noise. 

Hello Noel,

I have tried some experiments, here are the results.

1. Configuring DEN=16777210:

TICS Pro configuration after choosing Default Mode 2016-6-17:

Yellow trace - single fast sweep, peak detector, DEN=16777210 & NUM=8388606
Blue trace - max hold, peak detector, DEN=16777210 & NUM=8388606
Orange trace - max hold, peak detector, DEN=16777210 & NUM=8388605

For DEN=16777210, NUM=8388605 (exact 1/2) spectrum measurements give the same results as other DEN, NUM sets which give exact 1/2 fractional part of N divider.
For DEN=16777210, NUM=8388606 spectrum measurements give the same results as for DEN=16777215 & NUM=8388608.

2. Comparing levels of wandering spurs for DEN, NUM settings which give the fractional part of N divider very close (but not equal) to 1/2:

TICS Pro configuration after choosing Default Mode 2016-6-17:

Blue trace - max hold, peak detector, DEN=16777215 & NUM=8388608
Orange trace - max hold, peak detector, DEN=524287 & NUM=262144
Green trace - max hold, peak detector, DEN=16383 & NUM=8192
Yellow trace - max hold, peak detector, DEN=511 & NUM=256

The smaller DEN value, the lower level of wandering spurs. For DEN=511 & NUM=256 spectrum contains many discrete spurs, despite of which it looks similar to cases of other DEN, NUM sets which give exact 1/2 fractional part of N divider.

3. Changing PFD_DLY:

TICS Pro configuration after choosing Default Mode 2016-6-17:

Blue trace - max hold, peak detector, DEN=16777210 & NUM=8388606, PFD_DLY=860ps
Orange trace - max hold, peak detector, DEN=16777210 & NUM=8388606, PFD_DLY=1200ps
Green trace - max hold, peak detector, DEN=16777210 & NUM=8388606, PFD_DLY=1500ps

Increasing PFD_DLY gives worse phase noise close to the carrier.

3. Changing FRAC_ORDER:

TICS Pro configuration after choosing Default Mode 2016-6-17:

Blue trace - max hold, peak detector, DEN=16777210 & NUM=8388606, 2nd order modulator
Orange trace - max hold, peak detector, DEN=16777210 & NUM=8388606, 3rd order modulator
Green trace - max hold, peak detector, DEN=16777210 & NUM=8388606, 1st order modulator
Yellow trace - max hold, peak detector, DEN=16777210 & NUM=8388606, 4th order modulator

TICS Pro configuration after choosing Default Mode 2016-6-17:

Orange trace - max hold, peak detector, DEN=16777210 & NUM=8388606, 2nd order modulator
Yellow trace - single fast sweep, peak detector, DEN=16777210 & NUM=8388606, 2nd order modulator
Blue trace - max hold, peak detector, DEN=16777210 & NUM=8388606, 3rd order modulator
Green trace - single fast sweep, peak detector, DEN=16777210 & NUM=8388606, 3rd order modulator

TICS Pro configuration after choosing Default Mode 2016-6-17:

Blue trace - peak detector, DEN=16777210 & NUM=8388606, 1st order modulator, single ~200msec sweep recorded with spur
Yellow trace - peak detector, DEN=16777210 & NUM=8388606, 1st order modulator, single ~200msec sweep recorded without spur

Wandering spurs are also present when LMX2492 operates with 3rd order modulator, however with reduced level comparing to 2nd order modulator case.
In case of 1st order modulator spur has different spectrum from 2nd & 3rd order modulator cases and it is also more difficult to record it on swept-spectrum SA without using max hold and long sweep times (looks it appears less frequently).

Hi Szczepan,

Your test result proved that the combination of DEN, NUM and modulator order matches the theory. When NUM/DEN = 1/2, there is no randomization at all, so the phase noise is good but spur is high. Different modulator order changes the phase noise characteristic, higher order modulator reduces spurs by scarifying phase noise. 1st order is usually not used, as there may be big phase jump due to the architecture of 1st order fraction engine. 

At OSCin = 100MHz and VCO = 9650MHz, the spurs I am referring to should be located at multiples of 25MHz offset. However, things that puzzle me is your spurs appear in a 10MHz window and they are moving around. At the moment, I don't have any idea what may be causing the issue. If you have new finding to share, I am happy to brainstorm again. I am sorry that we cannot help resolve your problem.