bandlimited pseudo noise

[Tim Ressel]

(deep magic mode on)
Here is how it worked: We were using 461 lines out the 512 lines of a 1024
sample FFT (don't ask). So you take a buffer with 1024 complex samples. You fill
the first 461 slots with unity amplitude, random phase data. Then for each of
those samples you fill the mirror bucket (1024-i) with the complex conjugate of
the sample. Then you perform an inverse FFT and presto! you have your 1024
samples.
(deep magic mode off)

I can make a file with the eprom data, what format I wonder? S-records? Intel
hex? Other?


Tim Ressel--Compliance Engineer
Hewlett-Packard
Verifone Division
3755 Atherton Rd.
Rocklin, Cal
916-630-2541  
timothy_ressel at hp.com



-----Original Message-----
From: Magnus Danielson [mailto:cfmd at swipnet.se]
Sent: Friday, August 11, 2000 2:10 PM
To: Tim_R1 at verifone.com
Cc: synth-diy at node12b53.a2000.nl
Subject: RE: bandlimited pseudo noise


From: Tim Ressel <Tim_R1 at verifone.com>
Subject: RE: bandlimited pseudo noise
Date: Fri, 11 Aug 2000 12:33:40 -0700

> Martin,
> 
> Back when I did vibration analysis using FFT's, we had a calibration signal
for
> measureing phase & amplitude through the filters. The signal was a PRN
> synthesized from an inverse FFT and burned into an eprom. The eprom was then
> 'played' with the same sample clock as the a/d.  Would this work in your case?

> 
> By the by, i think I still remember how to do the inverse FFT trick. Something
> about a phase conjugate reflection....

That would be the IRFFT, right?

If you want to do a FFT on a real-valued vector of the size of n then you can
do this correctly with an n/2 FFT and some pre and post processing. This is
being called RFFT in some places. Naturally you can inverse this process and
then get a IRFFT.

If you use the normal (complex-valued) FFT on a real-valued signal (a normal
set of samples) you will get mirror frequencies.

> By the by part 2, I do still have some eproms from that era. Its only 12-bit
> data, tho'.

Please ;D

Cheers,
Magnus