celebrevida
Member level 2
I am very new to Frac-N PLL design. This is where you can get Fvco=(N+K/M)Fref, where N, K and M are integers and K < M (and M is 2^m).
The simplest way to implement Frac-N is using a first order delta-sigma modulator (DSM). But it has bad performance due to large spurs. You get a repeating pattern with short sequences so you want to noise shape so the pattern is much longer and more complex to lower these spurs and spread the energy to higher frequencies.
To so so, you can use higher order DSM. But I had a lot of problems finding such a design that wasn't overly complicated. I tried to find just 2nd order design to understand it as a starting point and apparently it is not popular. Third or higher order design is too complex for me to grasp. The articles just don't explain it in a way I can grasp.
After a lot of investigation, I found a very popular method is MASH 1-1-1 or 1-1-1-1 that is simple yet seems sufficiently effective.
Here is a great link on a fourth order MASH 1-1-1-1 design:
This is a great article! It spells out the design and actually explains the math behind it. I actually put together a behavioral model and got it working!
I already know 1st order DSM is no good. But is 3rd order MASH or 4th order MASH good enough?
I realize that this question depends on application and desired specs but wonder if Frac-N PLL using only MASH 1-1-1 or MASH 1-1-1-1 is "good enough" to use? I am wondering if I can get away with simple MASH 1-1-1 or MASH 1-1-1-1 and leave it at that.
The more I delve into it, the more I am seeing much more complicated stuff like dithering, other weird MASH configurations not to mention complicated multi-order DSM designs.
If there are some other techniques I should consider that is simple and straight forward like MASH 1-1-1 or 1-1-1-1 but perform even better please let me know and point me to it.
Thanks.
The simplest way to implement Frac-N is using a first order delta-sigma modulator (DSM). But it has bad performance due to large spurs. You get a repeating pattern with short sequences so you want to noise shape so the pattern is much longer and more complex to lower these spurs and spread the energy to higher frequencies.
To so so, you can use higher order DSM. But I had a lot of problems finding such a design that wasn't overly complicated. I tried to find just 2nd order design to understand it as a starting point and apparently it is not popular. Third or higher order design is too complex for me to grasp. The articles just don't explain it in a way I can grasp.
After a lot of investigation, I found a very popular method is MASH 1-1-1 or 1-1-1-1 that is simple yet seems sufficiently effective.
Here is a great link on a fourth order MASH 1-1-1-1 design:
This is a great article! It spells out the design and actually explains the math behind it. I actually put together a behavioral model and got it working!
I already know 1st order DSM is no good. But is 3rd order MASH or 4th order MASH good enough?
I realize that this question depends on application and desired specs but wonder if Frac-N PLL using only MASH 1-1-1 or MASH 1-1-1-1 is "good enough" to use? I am wondering if I can get away with simple MASH 1-1-1 or MASH 1-1-1-1 and leave it at that.
The more I delve into it, the more I am seeing much more complicated stuff like dithering, other weird MASH configurations not to mention complicated multi-order DSM designs.
If there are some other techniques I should consider that is simple and straight forward like MASH 1-1-1 or 1-1-1-1 but perform even better please let me know and point me to it.
Thanks.
