[time-nuts] EFC tracking

Poul-Henning Kamp phk at phk.freebsd.dk
Sun Jun 27 07:07:57 UTC 2010

In message <HMEBKHFEACMNLMHIAFDNEEGPONAC.jmiles at pop.net>, "John Miles" writes:

>> That is actually a very debatable proposition, a lot of them are
>> tracking types that conveniently cover up any lack of linearity
>> on the analog side of the fence.
>Can you elaborate on that?  Linearity is linearity - wouldn't it ultimately
>show up in the card's IMD specs?

That would sort of assume there were relevant specs provided.

The Audio ADC's are provided with audio relevant specs, which are
not the specs you would care about in the suggested application.

Broadly speaking, the important ability for a audio ADC is to be
able to capture the top 16 significant bits of a continuous signal
in an approx 24 bit space, sort of like a 16b*2^8b floating point

Concerns like long term stability top out at "30 seconds", whereas
woodoo-concerns like clock-jitter has gotten all the snakeoil

And then there are the features that really will totally get
you in this application: autozero features, that are built with
the assumption that any DC on the ADC should be compensated out.

These ADC's are great for sound and they are almost perfect for Ham's
SDR experiements.

But they are not and will never be 7½ digit 200 ksps instrumentation
grade ADCs.

You can get a brand new china-style DVM with USB port for approx
150USD on ebay, and as other said, you can get a HP3456A for as
little as 50USD, but would need a GPIB-USB converter.

By the time you have fixed a long term reference for your soundcard,
which likely entails a buried zener chip, a totally new input
section and a thermally controlled enclosure, you will have spent
many times that, and still get a performance much worse than the

So if the objective is to learn about analog/digital converters
and their ailments: go for it.

If the objective is to measure the EFC voltage over the long term,
hit ebay, find instrument built for such purposes and move on.


Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

More information about the time-nuts mailing list