Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!pacbell!att-ih!alberta!auvax!rwa
From: rwa@auvax.UUCP (Ross Alexander)
Newsgroups: comp.arch
Subject: Re: Computer noises (was Re: Cray architecture)
Message-ID: <583@auvax.UUCP>
Date: 31 Mar 88 19:59:00 GMT
References: <7762@alice.UUCP> <418@ole.UUCP> <3216@phri.UUCP> <1574@osiris.UUCP> <1503@vaxb.calgary.UUCP>
Organization: Athabasca U., Alberta, Canada
Lines: 70
Summary: it _can_ be done

In article <1503@vaxb.calgary.UUCP>, radford@calgary.UUCP (Radford Neal) writes:
> I have a vague recollection back from when I was a first-year student
> (1974) of someone demonstrating a program for the PDP-8 that played a quite
> good rendition of some piece of music on a radio placed beside the 
> CPU. In retrospect, this seems moderately unlikely, though not impossible.

and continues by observing that it would be difficult to fit enough
information into a 4k pdp8 to allow more than 10 seconds of music,
assuming 6k samples/second.  True enough; but that's not how it's
done ;-).  The basic code is something like


.=0010

MUSICP,	MUSIC		/ autoincrement pointer to list of notes

.=0200			/ code lives on page 1

	CLA		/ 0 --> accumulator

PLAY,	TLS		/ essentially, start a 1/10 second timer :-)

	TAD I	MUSIC	/ get next note code
	TAD	DELAYL	/ convert to index into delay list
	DCA	DELAYP	/ save address of delay count

PAUSE,	TAD I	DELAYP	/ refresh delay counter
	DCA	STALL

	ISZ	STALL	/ sleep for a while
	JMP	.-1

	TSF		/ has timer expired yet ? ******
	JMP	PAUSE	/ no, go stall some more

	JMP	PLAY	/ yes, get next note

STALL,	0		/ sleep counter
DELAYP,	0		/ pointer into delay list
DELAYL,	DELAYS		/ constant to form above pointer

DELAYS,	0100		/ these values are clearly incorrect
	0200
	0300
	.
	.
	.

MUSIC=0400		/ list of notes starts on page 2

	END

The reason this works (aside from coding errors; it's been _years_!)
is that the TSF (marked with the asterisks above) induces a spike of
RF energy as a side effect.  If you sleep 1/4000 second, do a TSF,
and loop back to the sleep, a radio will produce a 4 kilohertz tone.
After 1/10 second, the TSF will skip over the JMP PAUSE and we will
advance to the next note.

This basic technique can be extended ( there is room for both a
duration and a note code in each 12-bit word, for example ) get
better encoding and also more than one voice; three voices is doable
but tricky.  You can also do a light show in the accumulator display
while you're at it - this is really impressive.

I sure wish we had an 8 in the machine room nowadays - it would add
a nice traditional touch that's lacking (the vaxen are visually
boring).

Ross Alexander, Athabasca University
alberta!auvax!rwa