[LAU] Cleaning RAM

[Fons Adriaensen]

On Sun, Apr 01, 2018 at 12:50:07AM +0100, Will Godfrey wrote:

> ...
> This is the problem of memory slowing down as it ages. Every time a bit is
> flipped the underlying structures are stressed and an infinitesimal change
> takes place that makes it slightly more difficult for the next change. We are
> of course talking incredibly tiny amounts here so it's hardly surprising that
> it's not really been noticed up to now.

It has often been claimed that this 'RAM ageing' is why wavetable
oscillators tend to detune over time, but this is not true. The
real reason for this is a phenomenon called 'sample rot'. 

While RAM ageing is a slow, continuous process, sample rot is
periodic. It seems to be related to the first derivative of the
amount of solar energy impact, and therefore reaches a maximum
shortly after the spring equinox. The delay has been measured as
eleven days. It seems to be fairly constant over time, but there
is (disputed) evidence that it may change due to global warming.

Anyway, users noticing deviations in the pitch of their synths
are advised to refresh their wavetables shortly after the yearly
peak. Note that algorithmic oscillators are not affected, if
these show any detuning their must be another cause (usually
sloppy programming).

Ciao,

-- 
FA