Re: [OctDev] wavwrite() with integer arguments

[Mike Miller]

On Wed, Oct 24, 2012 at 5:21 AM, Froebe, Matej wrote:
> Hi,
> Thank you for your answer. The modified function is in the attachment. The 
> difference is just the added if statement around the multiplications:

Hi, I have a few comments about this:

>   if isa(yi, 'float')

isfloat (yi) seems better suited.

>     ## scale samples
>     switch (bits_per_sample)
>       case 8
>         yi = round (yi*128 + 128);
>       case 16
>         yi = round (yi*32768);
>       case 32
>         yi = round (yi*2147483648);
>     endswitch
>   endif

I read the matlab documentation and as usual it's not fully specified,
so can you tell us:

1. If you pass in an integer vector and do *not* specify N, does the
integer size select the default value of N?
2. If you pass in an integer vector but a mismatched value for N, are
the samples clipped or does wavwrite report an error? For example if y
is an int16 or int32 vector and N=8?

-- 
mike