SNAA413 August 2024 LMX2492 , LMX2572 , LMX2594 , LMX2595 , LMX2615-SP , LMX2820
TI Synthesizers are operated in integer and fractional modes. This application brief focuses on the fractional mode of the Phase Locked Loop (PLL) operation. Most of the spurs that appear at the output are modeled in PLLatinum sim tool developed by TI.
Figure 1 shows the LMX2820 PLL block diagram which includes a ΣΔ Modulator, also called as MASH.
During the fractional mode of operation, the ΣΔ Modulator is engaged. Output of the ΣΔ Modulator varies the feedback division values. Range of the variations at the output of ΣΔ Modulator depends on the Modulator order. Figure 2 shows the MASH engine and placement of MASH_SEED.
For a particular Fraction input and MASH_SEED, the PLLatinum sim tool predicts and models the pattern of the MASH output. Export the pattern from the PLLatinum sim tool as shown in Figure 4. Column X in spurs sheet of the downloaded excel provides the sequence.
Fractional Mode: NUM/DEN is non-zero. There are multiple orders for MASH operation. This App brief covers the following topics:
As shown in below in Figure 5, first order MASH gives IBS (Integer Boundary Spur) at 20MHz with magnitude of -62.1dBc/Hz. This spur magnitude can be optimized as shown in the rest of the App brief.
When MASH_ORDER is changed from first to second, IBS spur at 20MHz are reduced from -62.1dBc/Hz (Figure 5) compared to -73dBc/Hz (Figure 6).
IBS (Integer Boundary Spur) at 20MHz offset is not impacted by the MASH_SEED variation.
Sub-Frac Spurs (0.5×Fpfd/DEN=0.5×(200M/10)) at 10MHz that are generated from MASH are reduced when MASH_SEED optimization is used. As shown in above Figure 8 and Figure 9, when MASH_SEED is changed from 0 to 100, dominant Sub-Frac Spur at 10MHz are reduced from -57dBc/Hz to -65.9dBc/Hz. The PLLatinum sim tool predicted the vanishing of Sub-Frac spurs with MASH_SEED of 100 but in silicon there is reduction of 9dBC/Hz for the major Sub-Frac spur. So the direction in which PLLatinum sim tool predicts is matching with the silicon.
Example 1
The following list of figures shows second versus third order and spur optimization at 6.42GHz:
Third order MASH can improve spurs over second order MASH as shown in Figure 10 and Figure 12. PLLatinum sim predicted a Sub-Frac spur at 10MHz magnitude reduction from -62.8dBc/Hz to -80dBc/Hz (change of 17.2dBc/Hz) when MASH_ORDER is changed from second to third order. Silicon is showing a change from -59dBc/Hz to -73dBc/Hz(change of 14dBc/Hz).
Another option to reduce the spur is to reduce the gain from the charge pump. Lets see in the PLLatinum sim tool how the roll-off at these offsets look when charge pum gain is changed from 15.4mA to 4.2mA.
Figure 14 and Figure 15 shows the roll-off variation due to charge pump gain.
There is ≅11dB variation in the roll-off (10MHz offset) from the PLLatinum sim tool. Figure 16 shows the silicon result with charge pump gained reduced to 4.2mA. Comparing Sub-Frac spur (10MHz) between Figure 16 (-74dBc/Hz) and Figure 12(-84dBc/Hz), there is close to 10dBc/Hz variation. This closely matches with the PLLatinum sim tool prediction.
The following Figure 17 and Figure 19 show the silicon plots comparison for second versus third order and spur optimization at 6.02GHz.
For example 2, when MASH_ORDER is moved from second to thrid order, the PLLatinum sim tool(Figure 18 and Figure 20) predict a spur magnitude change of ≅17dBc/Hz (-62dBc/Hz to -79.3dBc/Hz) for the Sub-Frac spur at 10KHz. Silicon shows a spur magnitude change of -15dBc/Hz(-57.47dBc/Hz to -72.66dBc/Hz).
Gain from the charge pump is reduced to 4.2mA and impact on the phase noise is shown in Figure 21. When this option is exercised, make sure that the phase margin is greater than 45 degrees.
There is another way to reduce the spurs as shown in Figure 22. Using a larger nonequivalent fraction is an effective way to reduce spurs. Compared to Figure 21, the magnitude of the spur is less in Figure 22.
In the third order, the denominator is offset by 1, as shown in Figure 23.
Another example for the slight change in denominator impact on the spurs:
Following Figure 24, Figure 25, Figure 26 shows the phase noise plots for the cases shown above. Offsetting the denominator by 1, gives better improvement compared to without offset and MASH_SEED as 1.
Using a large nonequivalent fraction is effective way to reduce spurs. If the user does not want an the inexact output frequency, use MASH_SEED = 1, however MASH_SEED = 1 is not as effective as the larger, nonequivalent fraction.
Points to note when changing MASH_SEED:
For purposes of spur reduction, there is no point in making the Seed any larger than (PLL_DEN) - 1
Impact of MASH_SEED Mod PLL_DEN. For instance, if the fraction is 3/10, then MASH_SEED = 1, 11, 21, 31, … 10n+1, … all have the same impact
Simplify fractions if MASH_SEED = 0
Fraction (PLL_NUM/PLL_DEN) of 10/100 has the same spectrum as 1/10 if MASH_SEED = 0, but not if MASH_SEED=1
Use predividers and a multiplier in the reference path to change the spur offset location.