[time-nuts] Modern signal generators

[jimlux]

On 12/12/18 9:40 AM, Russ Ramirez wrote:
> Don, you mention digital attenuators. What parts do you have experience
> with? I am in the midst of a signal generator design for a club build
> project. Looking at digital attenuator parts now, but many are spec'd in
> the GHz range. Looking to get down to -140 dBm from +20, or as close to
> this range as possible. TIA
> 

You're looking for 160 dB of attenuation - that's hard - if something 
leaks at 100 dB down, that's 60 dB more than what you want.

If you're interested in precision phase measurements, it's even tougher, 
because a leaking signal that's "on frequency" wont necessarily make 
your level wrong, but will perturb the phase (e.g. 20 dB down spurious 
signal will make the phase wrong by 25 degrees (arccos(0.9))

It's awfully easy to get leakage that's -100 or -120dBc - backwards 
through the power supply is a favorite one. Through gaps or holes in 
enclosures. Through the braid of the coax.

At JPL we regularly do measurements of signals at <-150 dBm into 
sensitive deep space receivers, and we're obsessive about phase (since 
that's how we measure the distance to spacecraft to a gnat's eyelash). 
Historically, we found, for instance that most signal generators leak 
more through the chassis than come out the output jack at minimum level. 
   We test the receivers in a screen room, with the generators outside, 
so you get about 100dB from the screen room to help. Waveguide helps too 
(WG below cutoff is a very effective filter) and it's not too tough to 
make waveguide joints that have better than 150dB isolation to the 
outside (lap them smooth, use pins, wrap in foil tape)

Coax is always a challenge because the shields leak.


What we used to do is take an 8663 and low pass filter it real well, 
then do a low level multiply by 7 inside the screen room to get it up to 
the 7 or 8 GHz we need.  That way we can make sure that spurs and 
leakage is out of band.