LMX2572: Fractional Spurious

Hi,

In response to your questions:

(1)  The LMX2572 can use dithering.  To implement this, use a larger fraction.  Two methods to do this are:

a)  Instead of 1/10, use something like 1000000/10000000 and set MASH_SEED0=1

b) Instead of 1/10, use something like 1000000/10000001

(2)  There are two approaches to spurs.  One approach is "Randomization" or  "Dithering" described in (1) above.  The intention here is to spread out all the energy.  The other approach is the opposite of randomization, which is what you are doing when you are finding that 61 is the optimal seed.  For fractional denominators less than about 400, you will see that you can play tricks with the MASH_SEED to optimize the spur, but the settings are specific to MASH_ORDER, and the PLL fraction.  In other words, you have two directions to go for spurs:  Randomize or Seed.  Since your fractional denominator is 160, you might find that the seed approach is more effective, but also more homework as the seed will change every time your fraction changes.

(3)  The Fden is 32 bits, but it is programmable.  So you don't have to make the fractional denominator 32 bits.  You can program it directly to 160.  Many competitors parts force a large fractional denominator, but this part has programmable denominator.  If you choose to do a 32 bit denominator, this is because you would want ultra-fin frequency resolution or you want to go the "Randomization" approach with a larger fraction.  At some point, when you have a large irreducible fraction (note MASH_SEED=1 makes fraction irreducible), PLLatinum Sim models this as a noise floor, as this is what it looks like on the noise analyzer.  Tecnically, it is really spurs that are very close together, but it shows up as noise.

(4)  Through randomization, you should be able to eliminate completely the first sub-fractional spur at 250 kHz, 750 kHz, 1250 kHz, it will raise the noise floor doing.  Using MASH_SEED can reduce this spur without increasing the noise.  Primary fractional spurs at 500 kHz, 1 MHz, 1.5 MHz, … might not be impacted by randomization.  I disagree that the total spurious energy of the MASH will be the same based on setting, even with an infinite loop bandwidth.  For instance, if you choose fraction of 1/2, then with 0% dithering, you will have much less total fractional energy.  Also, for the MASH_SEED, I have seen that there are settings that seem to reduce all the spurs;  just because it reduces spur at one offset does not mean that it pops up at some other offset.

(5)  For your case, your fraction is 160.  So you are probably better off simply programming the fraction of 160 and optimizing with the MASH_SEED, if you are willing to do this.   You might consider using the 3rd order modulator.  I say this because you will have the 1/2 sub-fractional spurs at 250 kHz multiples offset for both the 2nd and 3rd order modulator.  Don't use the 4th order as it will introduce a 125 kHz sub-fractional spur.

Spurs are very difficult and complicated to model and have many causes.  This is why most PLL tools choose not to model fractional spurs at all.   For this tool, it is intended to give good insight and understanding, but bench measurement as the amplitudes of spurs are difficult to predict.

Regards,
Dean