[78-L] Inverting and reversing digital audion

[DAVID BURNHAM]

We had a session at ARSC in New Orleans which demonstrated that 
reversing recordings digitally must NOT be done.  They demonstrated how 
a square wave turns into curicules and regular sound waves become almost 
unrecognizable in comparison to the original.  They might sound ok but 
they aren't.  Sound can be reversed in analogue with minimal problems, 
but even there you can run into an absolute-phase issue if you are not 
careful. But NEVER EVER reverse audio in the digital domain. 

Mike Biel 

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

I attended that presentation at ARSC and I felt a fair amount of skepticism at 
this finding.  When I first started working with digital audio, I was using a 
SONY PCM F1 processor.  To keep this on topic, one of the first things I 
recorded with this unit was a 78, initially to hear if I could detect any 
difference in the sound of the surface noise between the original and the 
recorded 78.  Not only did they both sound the same to me but I was amazed at 
how accurately it captured a surface irregularity which resulted in a periodic 
bounce of the bass speakers at 78 rpm.  This impressed me because, of course, it 
meant that the system was reproducing frequencies down to 2 or 3 Hz or lower. 
 At the time I also ran a square wave through the system and witnessed the 
visual distortion on a scope which Mike describes above, but this, of course, 
was a square wave which was passed through analog circuits.  As an analog square 
wave passes through reactive circuits, all of the component frequencies are 
still going to be there but they will be delayed by different amounts depending 
on their frequency - hence the visual distortion of the square wave while not 
perceptively altering the actual sound of the wave.

However, once we get into an all-digital domain, I discovered that you can take 
a square wave, reverse its phase or direction or both without changing the wave 
form, (except, of course that the phase is reversed).  I did a couple of 
experiments when I got home from New Orleans.  I recorded a sample which 
contained a sawtooth wave, a bit of silence and then a square wave.  I digitally 
reversed the phase of this signal, then the direction of the signal, (played it 
backwards), and then both the phase and the direction at the same time.  I even 
ran it through 10 generations of phase reversal but in the end, the square wave 
was as square as it was when I started.

I have some screen shots illustrating this but, unfortunately, I have no way of 
posting them here.  I will be sending them to Mike Biel and if he has a way and 
an interest in posting them somewhere that they can be viewed, it may be of 
interest to some members.

db