Hello,

on the reference done by direct conversion to a square wave (some simple
digital circuit like a limiting amplifier) or are more advanced clock
recovery approaches used that look only at the fundamental and use all
information in the 10 MHz fundamental, just like the Phase Station is doing?

The phase detector in a PLL has strict requirements in terms of low latency
due to loop stability. PhaseStation can take all the time it needs to
extract the phase information. You might use it with a very ultrastable
reference source and very stable local oscillator, but in this case the
ultrastable source is already very clean (see next lines) so I don't see
any real improvement.

There's no more noise on the zero crossing than in any other point of the
waveform. The issue with using zero-crossing detectors is that you are
effectively using a sampled time system, and therefore limited by the
sampling frequency. If you're locking a PLL with a 10 MHz reference signal,
and the phase detector is running at 10 MHz, any phase noise on the
reference signal modulated above a frequency offset of 5 MHz will fall back
towards the "baseband" (i.e. the carrier frequency, 10 MHz).

Il giorno ven 24 mar 2023 alle ore 14:28 Erik Kaashoek via time-nuts <
time-nuts@lists.febo.com> ha scritto:

Hi

Assuming that “modern measurement device” = “typical lab instrument” ( and not something
more exotic …).

If you dig into the typical counter or signal generator today, it is not all that different than one
from a decade or three ago. They depend on some sort of internal reference for achieving
the noise levels at the required “goodness”.

Why?

They have no idea at all just how good or (or far more often) bad the external reference input
will be. Their spec sheet might define an input level range. Past that, it’s tough to find one
that digs into reference noise specs.

Since the assumption is that it’s a pretty poor source (noise wise), they tend to do some pretty
basic lock setups. About the only thing they do is to run a narrow-ish bandwidth loop. That
might be 10 Hz, It could be a bit wider. The parts used are typically pretty standard stuff.

You could indeed do a much narrower loop with digital techniques these days.  It would not
take a lot. It would assume that the external reference is still worse than the internal down to
whatever the new bandwidth is.  Just how they would make that determination is unclear.
Doing anything more than the basics doesn’t seem to fit in their budget …..

If you are talking about a phase noise test set that does A to B comparisons (= a PhaseStation
ref input ) then indeed “that’s different ….”…...

Bob

Hi Erik,

On 2023-03-24 14:24, Erik Kaashoek via time-nuts wrote:

So, back in the day, the 10 MHz clock was muxed in when existing, rather
than using the internal clock. This was either through a very manual
switching, or later through signal detection and steering of mux.

In most modern counters, a PLL is used. The benefit of a PLL is that it
will also act as a clean-up oscillator so high frequency noise is
effectively supressed. but lower frequency stuff will still affect
performance.

Then, for instruments like TimePod and PhaseStation, you just digitize
the reference and ends up looking at the fundamental as the overtones
gets filtered out, with similar benefits as the PLL.

A PLL will not ignore noise at zero crossings, but may provide
significant reduction of it through it's filtering actions.

What signal is best for a particular counter one should take a little
bit of effort to study as one attempts to push the noise floor of ones
abilities.

One should also ponder on the reference actually being properly
terminated, and phase stability of the cable. At some point, it can
become relevant things to consider. Ensuring termination to avoid reflex
waves comes fairly cheap, so that one should be done out of normal
RF-cleanness.

Cheers,
Magnus

A lot of stuff to unpack here.  I can make a few general comments:

IMHO, you should make use of the whole sine wave, assuming you have a
sine wave to start with.  Precision frequencies should always be
distributed as sine waves.  The problem of deriving a square wave
from a sine wave without degrading phase noise has been discussed
many times on this reflector.  Spoiler alert:  it is non trivial :-)
Only attempt it if absolutely necessary.  Like if you have to
generate a low jitter clock signal for an ADC.

In a former life, I worked for the HP Santa Clara Division in the
frequency counter section.  Frequency counters always started with
a high speed comparator that generated a square wave from the
incoming sine wave.  This royally messed up the phase noise, but
frequency counters don't claim to measure phase noise.  I remember
someone had a wet dream where they would use the counter's comparator
like a zero cross detector, and then they could do DSP on the zero
crossings and recover all the information that was in the original
signal, based on the Nyquist theorem.  Fortunately, this ridiculous
idea never went anywhere.

Regarding PLLs:  You can't go wrong using a double balanced mixer
as a phase detector, if you want the best performance.  That's what
I did in the HP 5071 Cs standard.  Of course, that kind of phase
detector is not self acquiring, so I added a circuit using a couple
of flip flops that detected (1) if the loop was out of lock and (2)
if so, which direction it was out of lock. and (3) added an offset
of the correct polarity to the phase detector to make the VCO slew
into lock and then (4) removed the offset once the loop was locked.
There are whole books written about PLL's that you can read.
Clock recovery from data is another huge topic.  I even hold a
patent on a clock recovery circuit for 40 Gb/s data.  Way too much
complexity to discuss here.

Rick N6RK

On 3/24/2023 6:24 AM, Erik Kaashoek via time-nuts wrote:

Rick: Just the guy to answer a q for me(?).  I've a batch of 80's Hp
boatanchors that perform well. They have 10 MHz reference inputs and
outputs. I have wondered for some time if daisy chaining these devices
from a reference would gradually keep adding phase shift down the chain
output to input as opposed to tees, that is more than just the cable
lengths.
Thanks
Don

On 2023-03-24 11:01, Richard (Rick) Karlquist via time-nuts wrote:

"It's always something."
Roseanne Rosannadanna
----------------------"
Dr. Don Latham  AJ7LL
PO Box 404, Frenchtown, MT, 59834
VOX: 406-626-4304