From: blockedofcourse@foo.invalid   
      
   > You know the count of a ripple counter when everything has come   
   > to a halt, including the carry chain. That is slower than the   
   > first flipflop.   
      
   But you may not *care* about the actual "current count".   
   Especially if the modulus of the counter is fixed and   
   not reloadable.   
      
   > Even decoded outputs will feature transient wrong results.   
   >   
   > Look ahead carry has been invented, just for this.   
   > An LFSR is hard to beat but will very probably also need some   
   > combinatorial delays.   
      
   Usually, the delays are at the input to the first stage   
   (assuming a single LFSR counter, not a set of counters   
   of different moduli, cascaded to obtain a specific modulus)   
   assuming a Fibonacci configuration.   
      
   The number of levels of logic (as in number of shift register   
   stages) can often be rejiggered using a Galois configuration.   
      
   Motogorilla (?) had an excellent app note in the 70's on creating   
   LFSRs of various moduli.  They also see use as pseudo-random number   
   generators where lengths of 1000 bits are not uncommon (as you   
   don't want an observer to be able to predict state from any   
   rational number of observations!)   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)