[time-nuts] Re: Accuracy of a sound card

Magnus Danielson cfmd at bredband.net
Tue Aug 23 16:19:52 EDT 2005

From: Chuck Harris <cfharris at erols.com>
Subject: Re: [time-nuts] Re: Accuracy of a sound card
Date: Tue, 23 Aug 2005 15:43:54 -0400
Message-ID: <430B7C7A.6040408 at erols.com>

> Hi Poul,

Hi Chuck,

> I am quite aware that making a soundcard more accurate is gilding the
> lilly.
> But if you return to the middle of this thread, where I offered powerline
> noise up as a reasonably accurate, ubiquitious timing reference, and had
> my suggestion refuted by a gentleman with a soundcard based spectrum
> analyzer, you would understand my point.
> If you are going to use a soundcard as the basis for a spectrum analyzer,
> and you are going to let your software readout 5 or 6 significant digits,
> you are going to have to also realize that the oscillator in the sound card
> is not very good, and your data is suspicious.
> In an effort to illustrate this point, TVB made a graph of the characteristics
> of his high quality sound card, and low and behold, it behaves just like it has an
> uncompensated crystal oscillator...imagine!

Like expected, indeed. Frequency accuracy and stability does not magically
apear from thin air (but you can get a fair sense of it from space with
magicless GPS for instance) in an el-cheapo solution.

> There are no adjustments of any kind on these boards, let alone any for
> frequency.  There is no reason to believe that they will be any more accurate
> than musicality requires.

If you don't have a sampling clock input of any kind, modding one in should be
considered unless a audio-card upgrade is over the horizon. Having it lock to a
10 MHz (or other similar frequency) source of sufficient longterm stability and
frequency accuracy should be of interest.


More information about the time-nuts mailing list