🚀 go-pugleaf

In article <C0JMov.J7A@uceng.uc.edu>, xhan@uceng.uc.edu (Xiaoping Han)
wrote:
> 
> Hi netter,
>   I am thinking about building a white-noise generator (audio band)
> with a pseudo-random series generator, followed by a LP filter
> (sort of 1-bit D/A). Any comment, suggestion, etc?
>   Does anyone has any information on one-chip solution to pseudo-random
> generation (>16 bit)?  Basically there is nothing in it but a
> clock, a shift registor and several XOR gates. 
> 
> Thanks in advance. email or post are both fine.
> 
> Han.                            xhan@uceng.uc.edu

I have done exactly what you describe, except for a different (wider)
bandwidth.  I needed a noise source which included the spectrum between the
voice band and higher band generators that were avaliable at that time
(1981).  I used a 20 bit register clocked at 10 Mhz followed by a 1Mhz
lowpass to generate a noise bandwidth that was virtually flat from DC to 1
Mhz (filter cutoff).

As I recall, the bandwidth is defined by approximately 1/2 the clock
frequency as the high end and 1/(pseudo random pattern length) at the low
end.  So in my case I had an unfiltered bandwidth of around 9Hz to 5Mhz. .
.   Well, the more I think about it, the upper range may not be right,
seems like it may have been more like 1/10 the clock frequency. Sorry, its
been a while!

Anyway I believe the low end is right:

    Pattern length = 2^20 = 1.049Mb
    Pattern cycle time = 1.049Mb/10Mb/s = .105s
    Lowest frequency component = 1/.105 = 9.5Hz

Anyway, your idea is definitely valid and will work.  Maybe a PAL for your
single chip logic.  For voice band you probably will not need so many
registers; depends on how low you want to go.

Hope this helps.

dam

==================================================================

Dana Merrill                                Bell Northern Research
DMerrill@BNR.CA                         Research Triangle Park, NC
==================================================================