On Fri, 21 Feb 2025 12:46:11 -0700
Skip Withrow via time-nuts time-nuts@lists.febo.com wrote:

I had the chance to have a short look at it at IFCS 2023.
It is a neat device, but I say only very little more than that.
They are slightly taller than the CNT-90/CNT-91 and a bit
heavier, but otherwise feel about the same. The big display
and the possibilty to use a tablet for controls make a
huge difference, though. And that's after taking into
account that the CNT-90/91 has a much nice user interface
than any of the other counters I've used over the years.

Indeed. That they went from kind-of start-stop counters to
fully time-stamping counters is a big thing, IMHO. The four
inputs are also very nice. The better single-shot resolution
makes these things even better.

They sell at 10k€ here in Germany [1], plus the various options
you'd want (e.g. OCXO, GHz input, etc). That's roughly twice
the price of a CNT-91 [2]. Is it worth the price? IMHO yes.

If you intend to buy a CNT-104 (or a CNT-91), please be aware
that the OCXO and GHz input options cannot be retrofitted.
They are soldered onto the board itself.

		Attila Kinali

[1] https://www.datatec.eu/de/en/pendulum-cnt-104s
[2] https://www.datatec.eu/de/en/pendulum-cnt-91

--
The driving force behind research is the question: "Why?"
There are things we don't understand and things we always
wonder about. And that's why we do research.
-- Kobayashi Makoto

On Fri, 21 Feb 2025 23:48:16 +0000
Tom Knox via time-nuts time-nuts@lists.febo.com wrote:

In terms of time-frequency counters, the CNT-104 is a pretty big
step. Just compare what happened the last 20 years. After the
reciprocal counter was introduced in the 70s, the first real
change was the continious time-stamping, which came in the early
2000s by Pendulum with the CNT-90.

Well, that's where the money is. I talked with quite a few
people who design and sell time-stamping systems and frequency
counters over the past decade. The big market for them are
post-production test. Be it electronics manufacturing or mechanical
testing systems (one intersting application of time-nuttery is
the testing of turbine blade alignment/behaviour where the passing
of each blade creates a pulse). Labs, while they need measurement
equipment, are not that big of a market. Old lab equipment will
be generally used until it doesn't work anymore. Only when they
have to do a measurement that is really beyond the capabilities
of the equipment they have, buying a new instrument is considered.

"A bit more accurate" is a slight understatement.
Getting to a single-shot resolution of 100ps is easy with today's
electronics, 50ps requires some thought. Going down to 7ps is
pretty hard. It's not that getting the counter resolution down
to a few ps is hard, while not trivial, that's the easy part
(there is this one time-stamping system manufacturer, we do not
name, that claims to get to 0.5ps). The hard part is to get from
an input that can accept a wide range of signals with different
voltage, different trigger points, different slopes, etc pp.,
to the counter circuitry with little noise added. It's the input
conditioning electronics and the comparator that are the limitation
in single-shot resolution, not the counter itself. Long ago, I
did a back-of-the-envelope calculation of what would be the limit
of a time-to-amplitude converter and came to the conclusion it
would be around 0.5-1ps. Further investigation then revealed that
I only could achieve this resolution if I could feed my input signal
directly to the T-t-A converter, without the need of comparator
or zero-crossing detector. The comparator and the level selection
circuitry add a lot of noise to the whole system. A lot.

There is a reason, why the SR620 has been a staple of labs
for such a long time. It was one of the few counters that
got down to 25ps. There were a few others over the years,
but they were all specialized devices meant for a specific
use cases rather than general lab equipment.

I do agree that an ADEV window would have been a very nice additon.
Especially given that the display has a high enough resolution
that it could easily display it. Though, for me myself, it's not
such a big deal as I do these kind of analysis offline anyways,
after checking the data for problems.

Cross correlation is a harder topic. There are many traps and
pittfalls there. I can understand that they were reluctant
to add this feature, as it requires proper data handling and
conditioning.

Well, what limits the resolution of a counter? For single-shot
time measurement it is the input stage, as mentioned above.
But for frequency measurmenent, it's the frequency reference.
They could add more digits, but those would be just random
numbers, because of the noise/instability of the reference.
But, as you know, the Pendulum counters do add more digits once
they have averaged over a certain amount of time and brought
the reference noise down.

For example, I currently am doing some precision measurements.
Therefor I calibrate the CNT-90 (option 30) that I have sitting
behind me every morning against a GPSDO. It drifts about 2-3e-10
a day (specified: <5e-10). So, the best I can hope for, with this
setup, are 10 digits of accuracy. I do get 13 digits of reslution
displayed (measuring over 100s), but the last 2-3 digits are
are pretty much random. So my precision is 10-11 digits.

Indeed. The CNT-104 is on my list of devices to aquire,....
as soon as I can justify getting another counter ;-)

Have a nice weekend!

		Attila Kinali

--
The driving force behind research is the question: "Why?"
There are things we don't understand and things we always
wonder about. And that's why we do research.
-- Kobayashi Makoto

Hi,

On 2/22/25 10:54, Attila Kinali via time-nuts wrote:

I agree fully. The CNT102/104 builds on the continuous improvements from
CNT80/81 (ASIC based), to CNT90/91 (First FPGA based, ADC interpolator)
to now CNT102/104 (FPGA based, FPGA interpolator).

I actually think I recall that the CNT80/81 could do continious
timestamping too, but it was more a traditional counter in that regard.

What has developed in how much that was accessable, and while the
documentation was always there, it has gone more into the thinking.

I proposed a more multichannel counter for them back in the day, as it
would enable monitoring multiple sources at the same time, either for
testing, for more elaborate measurement setups or for thinks like small
ensamble clocks. The CNT104 can do exactly that. Back in the day of
CNT90/91 just released focus was on improved performance. CNT81 had a 25
ps precision, CNT90 has 100 ps, CNT91 pushed that to 50 ps (minor tweaks
on essentially the same hardware). Also, the separate line was the
Wandermeters with WM10, WM11 and then the large screen platform now
available as Calnex Senteniel.

In this regard, CNT104 is just a step forward in performance and
capabilities. This is to be expected for such general instrument.

In the end of the day, these instrument makes needs to get their return
on investment, and there is only so much interest of certain instruments
available as a market. This general counter market needs to be updated
with an instrument that can live long and maintain or even gain market
share. For that it makes good sense.

Can you build much a much finer dedicated instrument? Sure you can. That
is however another market and another investment decision.

Actually HP5370A/B and CNT81 was there. Certain HP5328A/B could do that
resolution if allowed to interpolate over multiple pulses, so not "PPS".
SR620 is still commonly seen and used.

The Wavecrest counters of DTS-series for instance was at 2 ps. They
however had more focus on signal integrity use. An oscilloscope of that
vintage also had about 2 ps trigger jitter.

For Pendulum and others, they look at that as a separate software tool,
which they are happy to sell you. Pendulum did include the TDEV measure
into their wanter meters, and later, as that fit a particular market.

Cross correlation is a different signal path. Cross correlation is a
different way of thinking altogether. Doing a straight DMTD has lots of
problems associated with it, so the oversampling approach with modern
ADC creates a much better environment, but a very different signal path
altogether. The trigger noise you can then filter adapted to the signal
in digital rather than analog.

I'd like to point out that the class of generic counters needs inputs
having analog bandwidth for the full frequency range supported. This
bandwidth will scale the noise that the input comparator sees. Through
the slewrate S of the signal, the trigger jitter tj = en / S. If you
make a dedicated signal input, you can adapt the bandwidth such that the
noise reduces and also optimize other parts of that path. This is why
you never can compare dedicated instruments from general instruments, as
the later always suffer from them needing to support a much wider range
of signals in frequency and amplitude, and thus does not allow them to
be optimized.

Yes, indeed.

I hope I can repair the CNT81. Should be possible. The PSU have failed.

Cheers,
Magnus

Hej Magnus

On Sat, 22 Feb 2025 16:26:14 +0100
Magnus Danielson via time-nuts time-nuts@lists.febo.com wrote:

Interesting. I was not aware that FPGA based interpolators could reach
this low resolution. The lowest I know of, was in the order of 100ps.
While not bad, it's a far cry from the 7ps of the CNT-104.

I would have thought that they either managed to get the time-to-amplitude
approach to a very low noise level, or used a ring oscillator based ASIC
(see e.g. the design done at CERN by Perktold and Christiansen [1,2])

Are you able to share more details of what Pendulum did?

That's what I meant. The Wavecrest are designed for one specialized use-case.
That shows itself in the limited input volage range of just 1V. Not something
you would want in a general purpose instrument.

		Attila Kinali

[1] "Fine-Time Resolution Measurements for High Energy Physics Experiments",
by Lukas Perktold and Jorgen Christiansen, in "High-Performance AD and DA
Converters, IC Design in Scaled Technologies, and Time-Domain Signal Processing,"
2015.
https://dx.doi.org/10.1007/978-3-319-07938-7_14

[2] "A multichannel Time-To-Digital Converter ASIC with better than 3ps RMS
Time Resolution", by Lukas Perktold and Jorgen Christiansen, 2013
https://indico.cern.ch/event/228972/contributions/1539621/attachments/378552/526492/TDC_TWEPP_2013.pdf

--
Science is made up of so many things that appear obvious
after they are explained. -- Pardot Kynes