[sdiy] PRN Noise Gennie Question

[Harry Bissell Jr]

If you took the DAC outputs from widely
spaced shift register points (assuming that you could
as in a hardware implementation)... wouldn't that
also remove the lowpass characteristic.

Far enough apart, those bits would be random, yes ?

(ok pseudo-random but who could tell the difference in
the short term ?)

H^) harry



--- "James J. Clark" <clark at cim.mcgill.ca> wrote:
> 
> 
> 
> >   >
>
http://www.oldcrows.net/~patchell/synthmodulesII/200-1021.pdf
> >   >
> >   > In the above schematic, I use a 32 bit PRN
> source, and tap off 8 bits to
> >   > run to a DAC...on this application, I did it
> this way so I could use the
> >   > DAC output got generate control voltages in
> predictable sized steps....
> >
> >Jim,
> >
> >Um, it's no longer white noise then, right? 
> Running the shift
> >register outputs to the DAC effectively creates a
> digital filter, so
> >the result is filtered white noise.  In this case a
> single pulse
> >through the shift register will show up as an
> exponentially decaying
> >impulse response.
> 
> Or as an exponentially increasing one, depending on
> how you arrange the
> LSB/MSB order of the DAC.
> 
> >
> >I have not done the math so I don't know what the
> effect of this
> >filtering is.  (Anybody?)
> 
> This is a good point. The output of the DAC, if run
> at the same rate as
> the PRN source, will have a lowpass spectrum. But,
> if you have an 8-bit
> DAC with a very long PRN source, you can get a
> (nearly) white signal by
> updating the DAC at 1/8th the clock rate of the PRN.
> This is because the
> input to successive DAC conversions will not contain
> any bits in common,
> they will all be decorrelated. If you have an N-bit
> DAC you would update
> the DAC at 1/Nth of the PRN clock rate.
> 
> Note that this type of "filter", while linear and
> having a lowpass
> characteristic, can be un-done (deconvolved) in a
> perfect noise-free
> fashion by applying an ADC giving the individual
> input bitstream again.
> Most filters cannot be reliably de-convolved. What
> that is good for, I
> can't say!
> 
> Jim
> 
> 
>