On Monday 17 April 2006 15:14, Victor Lazzarini wrote:

> However, as memory is not a constraint these days, you
> can use truncating oscillators with large tables which
> will give you similar SNRs to interpolating oscillators
> with smaller tables. The overall SNR will always depend
> on the size of your function table.

Although, with a non-interpolating oscillator, you may need
a table size like 1048576 to get the same SNR as that of an
interpolating one with only a few thousand samples.
If I recall correctly, without interpolation the SNR improves
by 6 dB on doubling the table size, while the same value is
12 dB with linear interpolation; I am not entirely sure about
this, though.