The Designer's Guide Community Forum
Design >> RF Design >> merit of pulse swallow counter

Message started by dog1 on Jun 1st, 2018, 4:01pm

Title: merit of pulse swallow counter
Post by dog1 on Jun 1st, 2018, 4:01pm

Hello all,

I want to ask a stupid question: what is the merit of pulse swallow counter compared with, for example, synchronized binary counter (with decoder to reset, so that a step of one is achieved).

I am thinking, maybe it have less power consumption if the P and S counters are synchronized?


Title: Re: merit of pulse swallow counter
Post by vroy_92 on Jun 5th, 2018, 1:45am

For very high speeds of operation, a synchronous divider is a terrible option because it consumes tremendous amount of power. This is because your clock is connected to so many transistors (read:capacitance) that you will be wasting a ton of power for no reason. And if you use dynamic registers, the power consumption will be even higher. A ripple divider is the most energy efficient option but for a very high width divider, it can be hard to meet the timings at high speeds. (think about getting to MHz from GHz range).
A pulse swallow counter is in-between. Only one of the block runs at high speed clock, and meeting the timing is not difficult.

Title: Re: merit of pulse swallow counter
Post by dog1 on Aug 21st, 2018, 2:34am

Thanks V Roy, that's very helpful. so can I conclude that due to power consideration, it doesn't make too much sense to have the P and S counter in the pulse swallow counter syncronized? And that the only syncronized part may be the pre-scaler?

Also, I would like to understand better the advantage of swallow counter in high speed case. If I have a synced div4 counter followed by a ripple counter, would it be comparible in performance (power, speed) but less in complexity than a swallow counter? Or would swallow counter be higher speed? I attached a drawing to show my idea. For example, In swallow counter case, the pre-scaler count 3 or 4.

From power perspective, they are similar. Becasue they both have around 2 bits async and the rest sync.

From speed perspective, for swallow counter, the latest time for the state to change from count 3-> count 4 or the other way round is t1=N x T, where N is a integral (in this case, 1 or 2) and T is the input clock period. Then t1 must be smaller than delay1 and delay2. Whereas in the ripple counter case, T must be smaller than delay3. I think this can be an advantage for swallow counter, because t1 can be larger than T, and the delay1/2 can be similar than delay3. But I am not sure if this is the only reason that swallow counter is prefered in high speed design (eg. PLL design) over the binary structure.


The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.