Hi

The spurs are an issue, but their cause seems a mystery to me. However, it occurred to me today that they may be odd harmonics (or "sub-harmonics" for overtone crystals over 20 MHz).

This seems a plausible explanation as there shouldn't be such high phase noise that far from the DUT frequency.

The spurs change with frequency as per the YouTube video.

I have emailed the PN2060C maker to see what he thinks.

The commercial analysers may be aware of the harmonics as spurs and have scripts to remove them.

Any thoughts?

Regards Drew VK4ZXI

https://www.youtube.com/watch?v=wKDBUkGKul8&ab_channel=heisenberg

Subject: [time-nuts] Re: Seems like the new version PN2060C is
available for sale US$680
To: Discussion of precise time and frequency measurement
time-nuts@lists.febo.com
Message-ID: 618AFACD-623D-487B-916B-60DA01A1D3DE@n1k.org
Content-Type: text/plain;      charset=utf-8

Hi

Looking at their plot of the 10 MHz OCXO into both inputs:

There are some pretty significant spurs on that plot. From the notes below it, they seem to still be digging into the source of those spurs and ways to reduce / eliminate them. I think there still is a bit more work to be done. They imply the fix could be hardware rather than software. If so, I’d wait for the next version ….

Bob

Hi

One way or the other, the spurs are from an internal source in the device. If you do the math, you can’t get “heterodyne” spurs directly from the inputs when both of them are from the same source. As they mention in their little footnote, there may be issues with the internal clocking approach.

Simply blanking them is not a good idea. There are enough of them you could easily blank out something else. Based on their very limited comment, I don’t believe they are headed in that direction.

Any design like this is full of grubby details that need to be addressed. There is always going to be a lot of testing, finding things, working out what they are, and coming up with a fix. We tend to assume that will be a software fix and there’s just an update. Whatever we bought “early on” just gets new code and it works fine. With low level signal processing (which this is), that’s not always the case. Sometimes it’s the hardware than needs a fix …..

Bob

Bob Camp via time-nuts writes:

That is why I was surprised to see that they retained the USB interface ?

As a first order approximation, anything related to USB is cost
optimized, and therefore EMI/EMC efforts terminate the moment the
spurs are 3dB under the template.

They do have a comment about stationary vs. laptop giving different
spurs, so I guess they are aware of it.

At that level of ambition I would have reached for batteries +
linear PSU + fiber-optics, even though it would be both more
cumbersome and costly.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Hi

The referenced picture appears to show multiple bundles of batteries driving various bits and pieces of their test rig. They do appear to have reduced the likely hood of the problem coming from their test setup. Of course one can never be sure (as I’ve proven on the bench a number of times).

Doing a fiber optic interface is an interesting approach. How you do a two way link to the computer, at an affordable price. is a bit unclear to me. Maybe there is an obvious answer I’ve missed …

Bob

Bob Camp writes:

The crew behind OZ7IGY found a non-obvious answer for low bandwidth applications: TOSLINK

TOSLINK is a cheap LED+plastic fiber-optic audio connection.  As
with anything cheap it comes with footnotes, but they are workable.

For high bandwidths the answer is almost always ethernet (hardware)
and if one uses SFP modules, it is somewhat workable.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Hi

Is TOSLINK bi-directional or do you just use two? (and away you go to your custom
“gizmo” on the other end)

Bob

Bob Camp writes:

It's unidirectional, but since the devices cost less than five dollars...

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Yeah, USB2.0 is awful with 12/24/48/480 MHz clocks locally. And then you’re bringing in the signal from the host as well. And people usually don’t have USB isolation either so there’s also ground loop potentially in play.

With an SDR using a 6.144/12.288/24.576 clock for Nx48KHz sample rates, you either end up with DC offsets or the 12MHz at Nyquist nulls depending on sample rates.

Some of the TI ADCs use a 250x decimation instead of 256x decimation to avoid that problem.
Intel HD audio codecs take a similar approach.

With USB3.0, we have a 5GHz clock, with 8B/10B line coding for a 4Gbit/sec payload, spread spectrum clocked with 33KHz modulation. The bit scrambler LFSR gets reset under specific conditions. Inevitably, the USB3.0 will have PLLs and/or synthesizers involved.

While I'd expect the PN2060A/B/C doesn't use sigma delta ADCs, it's worth pointing out that in multibit sigma delta systems, that correlation might be different than expected.

How exactly all of the possible sources of trash can be excluded is a sticky problem.

From: Poul-Henning Kamp via time-nuts time-nuts@lists.febo.com
Sent: Saturday, October 7, 2023 9:07:08 AM
To: Discussion of precise time and frequency measurement time-nuts@lists.febo.com
Cc: Poul-Henning Kamp phk@phk.freebsd.dk
Subject: [time-nuts] Re: Seems like the new version PN2060C is available for sale US$680

Bob Camp via time-nuts writes:

That is why I was surprised to see that they retained the USB interface ?

As a first order approximation, anything related to USB is cost
optimized, and therefore EMI/EMC efforts terminate the moment the
spurs are 3dB under the template.

They do have a comment about stationary vs. laptop giving different
spurs, so I guess they are aware of it.

At that level of ambition I would have reached for batteries +
linear PSU + fiber-optics, even though it would be both more
cumbersome and costly.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com

Hi

Just to be clear:

I’m not knocking the folks who designed this gizmo or their efforts to date. They have come a long way so far. They obviously are quite capable and able to deal with a lot of complex issues. There likely are a lot of clocks running around inside the box. That makes sorting this and that out “fun”. I’m sure they can make progress as they dig in further.

My point is that these appear to be early days for this design. If someone heads off and buys one, that’s fine. It’s up to them. They will get a device that does some cool stuff. I think there still is a bit of work to be done. If you want a “finished product”, this hardware may or may not be what the designers ultimately settle on.

Bob