Noise Floor
Hi
Sorry about dumping the text from the previous post !!! Obviously the mailer on that
machine needs a bit of attention.
So here’s what it should have said:
When we are doing typical measurements on our time / frequency standards, much of it involves measuring noise. In many ways it is closer to a noise figure measurement than to a lot of other lab work. The “noise floor” of the measurement setup is an important partameter. Resolution might be 10,00X better than the noise floor. If so, resolution not a key parameter (at least not to me).
When setting up a lab, one of the first things to do is to get a working measurement setup and to characterize that setup. This often involves a lot of yelling, screaming, and kicking the dog. More often than not things like bum connectors, faulty cables, and bad grounding are what takes the most time to sort out. Weeks / months of work to get a setup going properly are not at all unusual.
One of the nice things about devices that work like a DMTD is that measuring the floor is a matter of driving the two inputs through a power splitter. With a single mixer setup (which is much easier to build) the floor is not as simple to estimate. The same is true of some (but not all) counter based setups.
Bob
Hoi Bob,
I see you got your hands on a PhaseStation. Color me jealous! :-)
The noise floor data is impressive! For reference: Expensive DMTD
systems for metrological applications are usually at 1e-13 @ 1s
and a lot more expensive.
I see I have to pester John more on how he designed the PhaseStation.
One intersting thing to note is, that the noise floor does not have
an exactly 1/τ slope. Which suggests that some additional effect
of higher order is affecting the measurement. This can be seen from
the phase data, which shows a quite prominent kink around 50ks and
is (almost?) linear before and after. It would be interesting to
know what caused this.
On Tue, 24 Mar 2020 08:03:31 -0400
Bob kb8tq kb8tq@n1k.org wrote:
One of the nice things about devices that work like a DMTD is that measuring
the floor is a matter of driving the two inputs through a power splitter.
With a single mixer setup (which is much easier to build) the floor is not
as simple to estimate. The same is true of some (but not all) counter based
setups.
Be careful here. DMTD and DMTD-like systems have a dependence of the noise
floor on the relative phase of the input signals. With the lowest noise floor
being at when both signals have the same phase. To trully assess the noise
floor, you have to shift the relative phases through 2π, while making sure
that the phase shift, however you implement it, does not degrade the signal.
And because you are shifting the singal, that the short term noise on the
signal is lower than the noise floor of the measurement system (in laser
systems is called the correlation length).
Attila Kinali
--
<JaberWorky> The bad part of Zurich is where the degenerates
throw DARK chocolate at you.
Hi
On Mar 24, 2020, at 10:01 AM, Attila Kinali attila@kinali.ch wrote:
Hoi Bob,
I see you got your hands on a PhaseStation. Color me jealous! :-)
If you recall all my grumbles, back in a bit I mentioned my pile of
ever more broken test gear. Keeping it all in repair had become so
time consuming that nothing else was getting done. A secondary
issue was the sort of bench space things like a three corner setup
with HP 3048 era gear takes up …..
The noise floor data is impressive! For reference: Expensive DMTD
systems for metrological applications are usually at 1e-13 @ 1s
and a lot more expensive.
I see I have to pester John more on how he designed the PhaseStation.
One intersting thing to note is, that the noise floor does not have
an exactly 1/τ slope. Which suggests that some additional effect
of higher order is affecting the measurement. This can be seen from
the phase data, which shows a quite prominent kink around 50ks and
is (almost?) linear before and after. It would be interesting to
know what caused this.
(which is why I included the phase data ….)
The mixer was hooked to the inputs with some pretty good / short microwave
cables. They both are as identical as Pasternack Enterprises could make
them. The guys at Mini Circuits did their best to deliver a splitter (via eBay).
None the less, they are by no means perfect. My guess is that the bump is an
artifact of temperature change / mechanical relaxation on either the cables or
the splitter.
I ran similar “floor” data on a TimePod back when I had one at work. My
vague recollection is that it did not go quite as far before “ripples” set in.
Since that was with different cables / splitter / environment it’s not clear
how they would compare directly. On the TimePod data run I could fairly
easily map the ripples to the changes in room temperature.
On Tue, 24 Mar 2020 08:03:31 -0400
Bob kb8tq kb8tq@n1k.org wrote:
One of the nice things about devices that work like a DMTD is that measuring
the floor is a matter of driving the two inputs through a power splitter.
With a single mixer setup (which is much easier to build) the floor is not
as simple to estimate. The same is true of some (but not all) counter based
setups.
Be careful here. DMTD and DMTD-like systems have a dependence of the noise
floor on the relative phase of the input signals. With the lowest noise floor
being at when both signals have the same phase. To trully assess the noise
floor, you have to shift the relative phases through 2π, while making sure
that the phase shift, however you implement it, does not degrade the signal.
And because you are shifting the singal, that the short term noise on the
signal is lower than the noise floor of the measurement system (in laser
systems is called the correlation length).
Yup, but at least you have a way to do it…..
Bob
Attila Kinali
--
<JaberWorky> The bad part of Zurich is where the degenerates
throw DARK chocolate at you.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
Thanks for posting these!
The user-manual provides some details on the internals:
http://www.miles.io/PhaseStation_53100A_user_manual.pdf
Figure 6, page 35, would indicate that there is no analog mixing stage,
just direct ADC.
Maybe it also shows the ADCs clocked at (max) 125 MHz, with some HPF and
under-sampling going on in order to measure a 200 MHz carrier!?
It would be interesting to know what ADC was used and if there's an
SDR-board out there that uses the same ADC.
Our experience with the Ettus radios is that although they have 2 channels
there is a lot of cross-talk between rx1 and rx2 channels, which makes a
diy SDR-based phase-meter challenging.
Another paper to look at is the R&S one
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/pdfs/FSWP_Paper_EFTF2016_V02.pdf
I can try to dig out my old comparison-data [1] and include this new data,
if Bob makes the results (or raw data) available as numbers somewhere.
On Tue, 24 Mar 2020 18:26:28 +0200
Anders Wallin anders.e.e.wallin@gmail.com wrote:
The user-manual provides some details on the internals:
http://www.miles.io/PhaseStation_53100A_user_manual.pdf
Figure 6, page 35, would indicate that there is no analog mixing stage,
just direct ADC.
I was in the lucky position that John gave me a quick tour of the
hardware last year at IFCS. The design is basically the same
as in the TimePod (see the schematics in the back of the TimePod manual),
but with everything that John had learned from the TimePod applied.
There are a few details and these details make up quite a bit of
the better performance.
Maybe it also shows the ADCs clocked at (max) 125 MHz, with some HPF and
under-sampling going on in order to measure a 200 MHz carrier!?
One of the big changes is that the sampling clock can be changed.
This allows more flexibility in measurements, especially not having
any frequencies at which measurement performance would drop, because
of being too close to integer fraction of the sampling frequency.
Also the mandatory low-pass filter that the TimePod had has been
replaced by a set of filters (or no filter) to make it even more flexible.
It would be interesting to know what ADC was used and if there's an
SDR-board out there that uses the same ADC.
Uh.. I remember John telling me what ADC it was, but I forgot, sorry.
Our experience with the Ettus radios is that although they have 2 channels
there is a lot of cross-talk between rx1 and rx2 channels, which makes a
diy SDR-based phase-meter challenging.
Yes. This is the key point here. The Ettus SDRs are just that: radios.
They don't need to have high isolation, because it doesn't matter for
what they are doing. Neither do they need to keep the two signal paths
closely matched. The PhaseStation and the TimePod are measurement
instruments and John designed them as such. When you see the inards
of the PhaseStation, you can tell that.
I am pretty sure I could design something like the PhaseStation as well.
The working principle is easy and can be explained on a napkin in 5 minutes.
But getting it to this remarkable perfomrance? Not without a lot of
trial and error. And even then, I wouldn't be sure.
Jupp! I need one of those! And I need to learn how exactly they work!
For SCIENCE! :-D
Attila Kinali
--
Science is made up of so many things that appear obvious
after they are explained. -- Pardot Kynes
On 3/24/20 9:26 AM, Anders Wallin wrote:
Our experience with the Ettus radios is that although they have 2 channels
there is a lot of cross-talk between rx1 and rx2 channels, which makes a
diy SDR-based phase-meter challenging.
The early USRPs in particular (I don't have as much experience with the
new little ones) are really intended for use as an educational tool -
they don't have particularly high performance in many ways - spurious
radiated emissions are a big problem (the internal LO leaks and gets
into the input of your receiver) - cross talk - lack of an easy way to
time tag data precisely or to have deterministic timing between Tx and Rx.
Hi
Since the unit is still in warranty, I’m a bit hesitant to crack it open and start
poking around :).
While it’s not 100% obvious, the ADC clock comes out the back and then
returns to the unit via a SMA jumper. At least in theory you could supply
your own clock to the unit. How well that would work …. no idea.
The internal ADC clock also is under program control !! You can move it around
from TimeLab if you think you might have a spur issue. Pretty neat feature.
I always wondered what I’d do if I had to test something that had a harmonic
at / near the clock in the TimePod. I never had a problem so it’s very much
a “who knows?” sort of thing to me.
ADC wise, there are at least 4 channels into the device. You can do a four
device cross comparison. How well that works … no idea. Fiddling with it
is on the list.
Isolation channel to channel is very good. I had issues early on with (phrase
edited out) cables and questionable fixture wring. Channel to channel isolation
issues inside the box… not so much. At 10 MHz it was below the noise (at -130
dbc) when I last looked.
The three corner hat math works, but with the “normal” limitations. Comparing
one ultra stable device (say a 5065a) to two not so stable parts (say a pair of
10811’s at 10K seconds) can result in strange plots. Not the fault of the device
(what ever you are using). It’s the limitation of the math involved. When one
is “decades” more stable it falls apart.
Just as with the good old 3048 ( and the TimePod), you can edit the FFT
segment table. If you are after spurs in a specific region or have other application
specific needs, that’s a nice feature. It’s one thing that kept the 3048’s going long
after newer gear had mostly replaced them.
Lots of fun !!!!
Bob
On Mar 24, 2020, at 12:26 PM, Anders Wallin anders.e.e.wallin@gmail.com wrote:
Thanks for posting these!
The user-manual provides some details on the internals:
http://www.miles.io/PhaseStation_53100A_user_manual.pdf
Figure 6, page 35, would indicate that there is no analog mixing stage,
just direct ADC.
Maybe it also shows the ADCs clocked at (max) 125 MHz, with some HPF and
under-sampling going on in order to measure a 200 MHz carrier!?
It would be interesting to know what ADC was used and if there's an
SDR-board out there that uses the same ADC.
Our experience with the Ettus radios is that although they have 2 channels
there is a lot of cross-talk between rx1 and rx2 channels, which makes a
diy SDR-based phase-meter challenging.
Another paper to look at is the R&S one
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_application/pdfs/FSWP_Paper_EFTF2016_V02.pdf
I can try to dig out my old comparison-data [1] and include this new data,
if Bob makes the results (or raw data) available as numbers somewhere.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.
On Tue, 24 Mar 2020 18:26:28 +0200
Anders Wallin anders.e.e.wallin@gmail.com wrote:
It would be interesting to know what ADC was used and if there's an
SDR-board out there that uses the same ADC.
Our experience with the Ettus radios is that although they have 2 channels
there is a lot of cross-talk between rx1 and rx2 channels, which makes a
diy SDR-based phase-meter challenging.
BTW: I forgot to mention:
If you want an SDR like system to play with, try the ADC/FPGA board
that Nicholas designed as part of his master thesis at CERN (in
the group that brought us white rabbit) [1]. Design files are all
online, including his thesis itself that should contain all the
design decisions (if not just ask). To the best of my knowledge
the design should be as low noise as the ADC can get to. You will
not get the isolation the Timepod or PhaseStation has, due to using
a 4-channel ADC, but it should be good for quite a few things.
Cost for the small run (5 boards) they did was IIRC 600CHF per board.
I don't remember whether that included the MicroZedBoard or not.
Attila Kinali
[1] https://www.ohwr.org/project/r19-tdc-del-a/wikis/home
--
<JaberWorky> The bad part of Zurich is where the degenerates
throw DARK chocolate at you.
It would be interesting to know what ADC was used and if there's an
SDR-board out there that uses the same ADC.
Uh.. I remember John telling me what ADC it was, but I forgot, sorry.
It uses four AD9265s. The TimePod used four LTC2216s, but the AD9265s
support higher clock rates with less power consumption, and both of those
attributes were important this time around.
I am pretty sure I could design something like the PhaseStation as well.
The working principle is easy and can be explained on a napkin in 5
minutes.
But getting it to this remarkable perfomrance? Not without a lot of
trial and error. And even then, I wouldn't be sure.
I don't think there's any way to avoid the trial-and-error part unless you
have the luxury of an unlimited ceiling for both the R&D budget and the
target retail price, and maybe not even then. One reason it took longer
than expected to ship the 53100A was that a lot of lessons that I thought
had been adequately learned on the 5330A/3120A project didn't pay off when
different ADCs were used, and when the carrier and offset frequency
requirements grew by 6x and 10x respectively.
With the TimePod, for example, noise and spur performance weren't strongly
influenced by ADC clock distribution. On PhaseStation, that particular
"unlearned lesson" cost me a respin.
Keeping parts and manufacturing costs under control was also more difficult
than anticipated. Another lesson that wasn't learned soon enough was that a
design with four or five internal PCBs ends up being much more expensive
than one that uses only two, even if the total board area is similar. We
had to increase the price twice to maintain standard T&M industry margins,
and (having just come back from visiting Said and Giovanni at Jackson Labs)
that's about to happen again. The original vision of a four-figure price
tag was unrealistic, and that's definitely a lesson for next time.
-- john, KE5FX
Miles Design LLC / Jackson Labs Technologies, Inc.
Hi
Another way to check what your measurement system is doing is to
compare two similar devices. Your guess is then that the floor must be
well below the combination of the two devices. As the difference in the
ADEV and Mod ADEV plots shows, you also need to stick with one xDEV
as you look at this or that.
So, what does the ADEV plot actually show?
It's comparing device Rb 1 (the famous 5065C) to a random eBay 5065
that has taken some significant effort to get running. (So much for
guaranteed
good / working ....).
The plot shows that at a given point the noise contribution of Rb 1 plus
the noise of Rb 2 plus the noise floor of the system is some magic number.
Assuming these are all random noise, then normal noise power math
applies. (two equal signals bump the total by 3db).
When would normal "noise math" not apply? If the noise was something like
a correlated spur (think 60Hz in the US) it could easily add as a voltage
rather
than a power. Knowing what you are measuring impacts how you interpret
the results.
The normal setup for a testing bench normally includes a lot of A to B
comparisons between a lot of sources to find "the best of the best of the
best". This is true even if one will ultimately use three corner hat
methods.
That stuff goes much better with sources that are fairly close to each
other.
Lots of grubby details do get into this. A fixture that "only" has 110 db
isolation
gives you a nasty spur at the offset frequency between your OCXO's. It
messes
up the phase noise and ADEV. Figuring out what this or that is and where it
comes
from is not a trivial task. Despite not being a lot of fun, it is part of
setting this
stuff up.
Bob
-----Original Message-----
From: time-nuts time-nuts-bounces@lists.febo.com On Behalf Of jimlux
Sent: Tuesday, March 24, 2020 1:21 PM
To: time-nuts@lists.febo.com
Subject: Re: [time-nuts] Noise Floor
On 3/24/20 9:26 AM, Anders Wallin wrote:
Our experience with the Ettus radios is that although they have 2 channels
there is a lot of cross-talk between rx1 and rx2 channels, which makes a
diy SDR-based phase-meter challenging.
The early USRPs in particular (I don't have as much experience with the
new little ones) are really intended for use as an educational tool -
they don't have particularly high performance in many ways - spurious
radiated emissions are a big problem (the internal LO leaks and gets
into the input of your receiver) - cross talk - lack of an easy way to
time tag data precisely or to have deterministic timing between Tx and Rx.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe, go to
http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
and follow the instructions there.