[sdiy] Harmonic bandwidth

[Tom Wiltshire]

On 8 Jan 2008, at 16:09, Tim Daugard wrote:

>> There is one source of possible differences between the waveforms,
>> which is that they've all been normalized to their maximum level.
>> Given that the waveshapes won't have the same maximum peak level,
>> this is likely to produce minor level differences between them.
>>
>> However, they've all got exactly the same 20 harmonics in exactly the
>> same proportions.
>
> Measured with?

Not measured, generated. The samples have been created artificially  
from a script and then saved as AIFF files. The script simply adds  
together 20 harmonics in the required amounts for a ramp wave ( 1,  
1/2, 1/3, etc). Thus the generated sound is as accurate as the  
floating-point internal maths and the 16-bit AIFF output. They  
definitely do have the same proportions of each harmonic.

The only possible source of error is the normalization of the result  
to the full 16-bit range. I could remove that too, if anyone _really_  
thinks it affects the experiment.

> You would probably have to measure the frequency of each harmonic  
> within a
> few hertz and the amplitudes within 1db or less. I was surprised,  
> after many years as
> being a technician, how little the change was required to make a  
> huge difference in sound.
> No magic wire theories, but there are changes that fall in the  
> tolerance ranges of test
> equipment that can be heard by the ear. It's amazing what the ear  
> and brain can interpret
> (and misinterpret).

I think the key there is 'misinterpret'. There's lots of evidence  
that people hear differences where they think there are some - like  
when they plug in their super-duper audiophile hifi using a new  
platinum-plated mains cable.The improvement is subtle, but  
_definitely_ there, ask anyone who just paid $400 for a mains cable! ;)

T.