Date   : Fri, 21 May 2004 17:36:49 +0100
From   : Richard Gellman <splodge@...>
Subject: Re: Sound wave representation / Goldie Lookin' Chain

Charles Ripperton wrote:

>>>Yep, thats correct. You will need to set the pitch to its highest
>>>value, which is so ridiculously high that nobody can actually hear
>>>it. This causes a "half position" on the speaker cone, which can
>>>then be varied by adjusting the attentuation values (the chip
>>>works inversely, having 0 as "full volume" and 15 as
>>>"infinite attentuation").
>>>      
>>>
>>In terms of nybbles :-
>>
>>0000="full volume"
>>.
>>.
>>.
>>1111="infinite attentuation"
>>
>>
>>I'm not sure I follow, since in my mind there are two 'full volume'
>>values (i.e. maximum positive, and maximum negative amplitudes)?
>>
>>Am I asking a silly question? :/ :)
>>    
>>
>
>
>Hi all,
>
>Could someone clear up that question, please? How are 'two' volumes
>(the negative and the positive) described in nybbles?
>
>Thanks,
>Charles
>
>  
>
*cranks handle on projector*

Ok, the sound chip has a backwards volume system for each channel. 
Rather than "how loud is this?" it refers to "How much quieter than 
normal is this?".

A nybble (four bits) of the amplitude register is used to set this 
attenuation value. This is why a humble beeb makes that low rapsing 
sound on start-up. Since all registers are set to 0, thats a pitch of 0 
(low frequency) with no attenuation (full volume). The noise channel 
isnt heard as a pitch value of 0 equates to "no noise".

The confusion is arising because of the suggestion of superimposing a 
waveform onto this system. The values used in the attentuation are 
stepped evenly, so if you use the values 0 to 14 (i.e. excluding 15)  
you get 15 evenly spaced amplitudes.

(Still with me?)

Now, a wave form requires a "zero-point" in order to have negative 
amplitude. We are forcing the speaker cone to a half point by settings 
it pitch impossily high. The human ear can't hear this so ignores it 
(the same as if you put a DC voltage into a speaker). By varying the 
amplitude of the channel, we can move the distance from the speaker of 
the half-point. This allows us to generate the waveform.

If we now assume our waveform has a low resolution, with peak to peak 
being from -7 to +7, then we can superimpose that onto our 
"postive-only" values by subtracting 7 and then getting the absolute 
(negative -> positive) result.. Thus +7 less 7 becomes 0,  i.e. full 
volume. +6 becomes -1, absolute is 1, becomes attentuation level 1, i.e. 
slightly quieter than full volume.

At the opposite end of the scale, -7 subtract 7 becomes -14. Absolute is 
+14, i.e. quietest level (one short of "channel off").

This puts the "zero point" as being a chip attentuation value of 
abs(0-7) = 7.

With all this knowledge to hand one can put together a speedy piece of 
machine code to vary the amplitudes of the half-point and produce 
digital audio.

*switches off projector*

Clear? :)

-- Richard