[time-nuts] To improve a Tbolt?

[John Miles]

> What power supply did the TAPR Thunderbolt use?

The green, orange, and red traces used the supply module that came with the
TAPR order.  You can see the characteristic hump around 3 kHz that appears
in many of Tom's plots with this supply as well.

The blue trace was the old-style packaged Thunderbolt with the integrated
switching supply.

> Have you considered using a variant of the Wenzel supply filters to
> cleanup the spurs in the 1-10kHz region.

It couldn't hurt.  However, I don't tend to take spurs below -120 dBc very
seriously, because they can come from so many different sources outside my
control.  The HP 3048A system components were scattered all over the room at
the time these graphs were taken, and I had a strong cellular/pager
transmitter on the building immediately next door.  In any event the 3048A's
own spur-level specification is about -115 dBc.

I try not to lose sleep over low-level spurs unless/until they interfere
with an actual application, then I look into the specific one(s) that are
causing trouble.  Sloppy, perhaps, but ars longa, vita brevis and all that
good stuff...

> This is trivial with the low current -12V supply but would require some
> design effort for the +12V and +5V supplies.
> Using a more modern opamp (simplifies biasing) to augment the
> performance of such a filter for the higher current supplies would be
> worthwhile.
> Even dropping the spur level by 20dB is probably worthwhile.

I think the best way to deal with wideband noise and spurs is at the
architectural level.  Wall off the GPS clock and its supplies entirely, and
use its 10 MHz output to phase lock the clock you're actually using to drive
critical stuff.  That clock would be fed with a clean supply and tightly
integrated with its application, as opposed to mounted 20 feet away on
another rack.

With that philosophy, the GPS clock's performance at 10 Hz and below is
worth sweating out, but phase noise and spurs at wider offsets may not be.

-- john, KE5FX