Designer’s Guide Community

vivkr

Community Fellow

Offline

Posts: 780

pseudo-random sequence generator
Nov 02^nd, 2005, 8:10am

Hi,

I am looking for a simple way to generate PRBS sequences. I should like to have one with a fairly long period (2^20 say), but the bigger problem is that I would
like to use this to get 3 states which are equally (or almost equally) likely, but the period in which these states occur is fairly long.

Typically, it is easy to get 2 such states out of a PRBS (any of the outputs) but trying to get 3 states out of it has me stumped.

I would appreciate any help. I am not much of a digital designer but if there are references describing design of such a block, then I think I can pick it up.

Regards
Vivek

Back to top

IP Logged

Paul

Community Fellow

Offline

Posts: 351
Switzerland

Re: pseudo-random sequence generator
Reply #1 - Nov 2^nd, 2005, 11:16am

Hi Vivek,

PRBS sequences are generated using linear feedback shift registers (LFSRs). You can find sample code here (doulos.com):
http://www.doulos.com/knowhow/vhdl_designers_guide/models/bist_circuits_part_one...

and here (generic implementation):
http://vnsnes.freeshell.org/lfsr/LFSR.v

In order to make sure each sequence only occurs once each 2^N-1, you must chose the taps correctly. More on this and an introduction to Galois field math are presented here:
http://www.newwaveinstruments.com/resources/articles/m_sequence_linear_feedback_...

The two implementations of LFSRs are shown in this reference, but you should be aware that the tap definition to obtain PRBS sequences is opposite.

I hope this answers your question.

Paul

Back to top

IP Logged

jbdavid

Community Fellow

Offline

Posts: 378
Silicon Valley

Re: pseudo-random sequence generator
Reply #2 - Nov 3^rd, 2005, 9:03am

The first thing that occurs to me for a 3 level output, would be to alternate the "1" value between +1 and -1.. so every even 0-1 transition would go positive and the odd ones would go negative. you'd maintain the same direction until the next 0..
---
another idea.. use the state change of the prbs sequence to drive a 3 state counter/modulator.. this has the disadvantage of either a 1 to -1 or -1 to 1 transition on every third state change.
--
a third idea .. bring out 2 bits from the prbs sequence..
then assign two patterns to the same value.. ie
00 = 11 = 0 , 10 = -1 , 01 = +1
this should, in effect, be the differential of the PRBS pattern..
Hope this helps

Back to top

jbdavid
Mixed Signal Design Verification

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: pseudo-random sequence generator
Reply #3 - Nov 4^th, 2005, 4:33am

Hi Paul & David,

Thanks for the feedback and the links. I think it should provide useful reference material.

Regarding the suggestion of using a PRBS to drive a 3-state machine, I already thought of that but if you look carefully, you have 3 states with probabilities of
1/2, 1/4, 1/4. This would be undesirable. Perhaps, I need to insert a random scrambler in between before I map 2 bits from the PRBS to a 3-state machine. This ought to even out the probabilities a bit more.

I need to try out these suggestions first. Thanks again.

Vivek

Back to top

IP Logged