[time-nuts] "The Penultimate HP5065 A15"

[Poul-Henning Kamp]

--------
Bob kb8tq writes:

> > The "modern" approach to that is to modulate or dither with a
> > good long PRNG to whiten the noise, and while good in theory,
> > it is not _that_ easy to get right in practice.
>
> Of course one could simply cave in to the fact that for the sort of
> gain required here, a pure DC controller works quite well.
>
> I know - that takes all the fun out of it .....

It's not like PI(D) controllers were unknown when the 5065 was
designed, if anything people were better at "servo technology"
back then, than they are today, so I cannot imagine they would
not have considered them, and for some reason ruled them out.

One problem with PI(D) that if your DUT has impedance (thermal in
our case) on the timescale of the main oscillatory external
disturbances (air-cons in our case), then at best a PI(D) will
introduce a delay in the oscillation, at worst it will amplify it.

Today you can buy PID controllers where then processor will attempt
to phase-lock the primary disturbance and average it out.  That
works great until somebody props the door open for ten minutes
(=half the air-con period) to fix the hinges.  Dont ask me how I know.

This is why I suspect the wien-bridge approach may not be a just a
homage to the HP200, I think they deliberately wanted to shift the
frequencies well north of the 137 Hz.

Dithering a PI(D) with a PRNG gets you the same effect, but with
wideband noise instead of a single spectral peak.

-- 
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.