Hi

A crystal filter at the front end of a 60 KHz receiver is a bit of a win / loose sort of thing.

The selection of crystals available at 60 KHz is a bit limited. The typical parts have temperature coefficient’s that are a bit high. This results in a filter with phase delay that varies noticeably over even modest temperature changes. That’s not ideal for a oscillator locking application. For a wall clock that is doing great if it’s only tens of milliseconds off, it’s not an issue.

These days a SDR based approach is the better choice.

Bob

The thing about SDR is always that they have zero selectivity prior to the RF amp and mixer.  If you have a lot of interference around it's good to have selectivity before you get into anything potentially non-linear, and it also makes the ADC's job easier.  Surely if the crystal has known characteristics it should be possible to compensate for the varying phase shift?  Or, given that it's going to discipline an OCXO anyway, put it in an oven too.

-----Original Message-----
From: Bob Camp via time-nuts time-nuts@lists.febo.com
Sent: 28 September 2025 12:54
To: Discussion of precise time and frequency measurement time-nuts@lists.febo.com
Cc: Bob Camp kb8tq@n1k.org
Subject: [time-nuts] Re: vlf-disciplined OCXO circuit

Hi

A crystal filter at the front end of a 60 KHz receiver is a bit of a win / loose sort of thing.

The selection of crystals available at 60 KHz is a bit limited. The typical parts have temperature coefficient’s that are a bit high. This results in a filter with phase delay that varies noticeably over even modest temperature changes. That’s not ideal for a oscillator locking application. For a wall clock that is doing great if it’s only tens of milliseconds off, it’s not an issue.

These days a SDR based approach is the better choice.

Bob

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

WHat's your plan for an 60kHz antenna ?
If a ferrite loaded loop, you can get far better selectivity by building
it as a double tuned circuit :

• with the loop antenna  being one tank, and  something like a potcore
  inductor being the 2nd coupled tank.
• this will give you a nice flattop-ish  response with steeper skirts
  than a single mag loop antenna

I think the SDR using a single $10 STM32 microcontroller with a 2 Msps
12 bit ADC would be ideal as a back end.. run the ADC at something like
240 kHz
You'l
l need some sort of aliasing filter ahead as AM radio stations may alias
into the passband, depending on choice of sampling rate.

-glen

On 29/09/2025 07:52, john.haine--- via time-nuts wrote:

Hi

At VLF, you can run straight into an ADC converter. No mixer, and not much of an RF amp. Until you get a noise spike that saturates the (maybe) 5V input range on the converter ….. it pretty much just works. If you put a narrowband filter in front of the ADC, you then should come up with a noise source to “dither” the input.  A 24 bit converter can give you a lot of dynamic range and really good IMD performance.

Do you want the local AM broadcast station(s) blasting in? Probably not. In most areas, that’s a pretty broad filter. That’s good because the narrower the filter, the more it’s going to mess up the carrier phase as temperature and humidity change. Build the filter (or tuned antenna) with ferrite based inductors and it might get really crazy.

Once you get into the ADC, you can implement a really narrow filter and have essentially no impact on carrier phase (in terms of stability).

Bob

A wonderful evening!

Quite a few interesting things have been written on this thread,
but a few of the original questions remain unaswered.

On Mon, 22 Sep 2025 10:19:56 +0100
"john.haine--- via time-nuts" time-nuts@lists.febo.com wrote:

Yes, there have been several. But most of these have gone out of
fashion with the arrival of cheap GPS modules. Especially in the
German amateur radio community, various DCF77 disciplined oscillator
projects were shared.

E.g.:

"Baumappe zum DCF77-gesteuerten 10MHz-Frequenznomral",
by Funkamateur, 2009.
https://www.box73.de/file_dl/bausaetze/BX-176.pdf

Which is based on the design by DL1SNG and DL1FAC that was published
in Funkamateur December 2008
https://web.archive.org/web/20240609113503if_/https://www.worldradiohistory.com/INTERNATIONAL/FunkAmateur/FunkAmateur-2008-12.pdf

There are a few others, that work kind of in a similar way.

A more modern design is
"Performance Analysis and Receiver Architectures of DCF77
Radio-Controlled Clocks", by Daniel Engeler, 2012
https://doi.org/10.1109/TUFFC.2012.2272

Daniel used an FPGA to do tracking of the signal including the
PSK signal. One of the goals of this was to use this as a demonstrator
on how to decode and track time signals. The paper contains quite
a bit of analysis of different receiver strategies and how they
perform. So it's definitely worth a read.

Another interesting one is
"Software-Defined Radio Decoding of DCF77: Time and Frequency
Dissemination with a Sound Card" by Jean-Michel Friedt,
Clément Eustache, Emile Carry and Enrico Rubiola, 2018
https://doi.org/10.1002/2017RS006420
where they use a soundcard and GnuRadio to decode the signal.

While this is true, the disadvantage of VLF transmissions is that
they have a very high variability during the day and from day to day.
Not only does the signal strength vary a lot (up to several 10s of dB)
but also the timing varies by 100s of µs to 10s of ms, depending on
where you are, relative to the station. So, if you need time accurate
to a few µs (which is a pretty normal requirement for data centers
these days), then VLF isn't going to cut it. In these cases it is
easier to just wire up an antenna to the roof and get <100ns from
GPS, then distribute time inside the building, either as a dedicated
pulse signal (often used for wall clocks) or over the network via
NTP or PTP.

While this is true, most of the jaming in the US is actually accidental
jamming from faulty RF equipment (like active antennas for TV reception).
Though, GPS jammers of disgruntled drivers who don't want their boss
to know they are going for a coffee break, are on the rise.
But, while jamming is getting more prevalent, it's by far not common.
A lot of our infrastructure does depend on GNSS reception. And most
of it works pretty well.

In eastern Europe, things are a little bit different at the moment.
(see, e.g. https://www.flightradar24.com/data/gps-jamming )
I am not fully aware what the different users of time do to cope
with this situation, but given that there has been very few
articles covering the situation, they seem to cope quite fine.

For all practical purposes, Loran is dead. While e-Loran does try
to make a come-back, it has failed to achieve that in the past 10 years.
Given that the reliability of GNSS is good enough for most applications
that use it, an additional expensive service isn't going to get
investors. And the states do not seem to be interested in it.

There is an European iniciative for reliable time distribution
on a continental scale, but as far as I know, VLF transmissions
are not being considered.

I have thought of building a multi-standard receiver myself
a couple of years ago. But quite honestly, the availability,
reliability and accuracy of GNSS makes the other time signals
just not worthwhile. At least not as a product. Ok, that's not
100% correct. Companies like Meinberg still sell DCF77/MSF/WWVB
systems. I am not quite sure what their purpose is, but apparently
people are paying for them. But I wouldn't develop such a new
product for VLF reception these days, as the demand is rather low
and does not justify the development cost. Though, it could be a
lot of fun as a hobby project.

These modules are mostly built based on chips used for alarm clocks.
These are fine as long as they aren't more than a few ms off. Hence
they don't use the carrier. And that's also the reason why they are
so cheap: the chips they use are produced in the millions and a few
of those end up on boards used as time references for computers and
such.

If you want to build your own VLF receiver, I would go the way PHK
described: Use a losely tuned antenna, add some pre-amplifier.
A simple JFET stage as a high impedance first stage for the
antenna and a low noise opamp should do the trick. Maybe add an
intermediate BJT based amplifier for additional low noise gain.
If you want to receive only one or two specific transmitters, you
might want to add some filtering too. Amplify the signal up to the
point where you can feed it to an ADC of sorts. Anything above 4bits
is enough, unless you are dealing with a very weak signal that you
need to dig out of the noise, then go for as many bits as possible.
Feed the ADC samples into a µC that does all the magic DSP stuff.
How you want to do it, depends a bit on how complicated you want
it to be and which signal you are tracking. Average the signal
for at least 24h and steer a good OCXO with it.

		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

Bob Camp via time-nuts writes:

So this gets to choice of antenna.

The best, flattest wide-band antennas are indisputably "E-field
probes", basically a piece of conductor connected to an amplifier
with as high input impedance as you can make it..

It is almost trivial to build an E-field probe which is flat from
DC to north of a GHz.  I'm personally partial to Chris Trask's
designs ("Complementary Push-Pull Amplifiers for Active Antennas:
A Critical Review") but there are many others.

But because they are wideband, they also pick up "static" and in
particular the insanely wide spectrum[1] of nearby lightning strikes,
which are the major cause of the big transients you talk about.

Below a MHz one can also use "M-field probes", which is a coil
attached to an amplifier with a balanced input, commonly known as
a loop-antenna.

The kind of noise spikes you talk about only happen in loop-antenna
if somebody quenches the superconducting magnet in the MR-scanner
next door.

A major difference between loop-antennas and e-field probes is
that loop-antennas have a figure-of-eight sensitivity pattern.

This is great if, like me, you have a hundreds of kW LF transmitter
in the next town over, but less great if you want to receive several
signals from all over at the same time.

Loop-antennas can also be tuned to a particular frequency band
by adding a capacitor in parallel to the coil, and you can get
amazing "amplification" by using a high impedance input amplifier
because it operates near-resonance.  The downside is that it
takes forever for the resonance to die out again, which is
why it is almost only used in the "run forever on an AAA battery"
radio-controlled clocks, which only need a ~3Hz bandwidth.

I have experimented with both E-field and M-field probes in the VLF
band and I far prefer (untuned) M-field probes.

Poul-Henning

[1] Zero to many kA in less than 5 microsecond, you do the FFT.

--
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.

 
Yes, but you DO need a bunch o'bits in the ADC (depending on that input filter).   However, you could have a fairly wide filter, chosen for small tempco of phase shift, and that would help a lot.  But, for instance, if you try to do it with the mighty $20 RTL-SDR, which is 8 bits I/Q and 2 MHz BW, AM/MW stations will be a problem.

On Sun, 28 Sep 2025 07:54:01 -0400, Bob Camp via time-nuts time-nuts@lists.febo.com wrote:

Hi

A crystal filter at the front end of a 60 KHz receiver is a bit of a win / loose sort of thing.

The selection of crystals available at 60 KHz is a bit limited. The typical parts have temperature coefficient’s that are a bit high. This results in a filter with phase delay that varies noticeably over even modest temperature changes. That’s not ideal for a oscillator locking application. For a wall clock that is doing great if it’s only tens of milliseconds off, it’s not an issue.

These days a SDR based approach is the better choice.

Bob

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

Hi

Assuming we have the bits and pieces worked out (yes that’s a big assumption): Next up is how this fits together.

A disciplined oscillator can be looked at in many ways. The most common is to look at it as a frequency reference. It is not unusual to use ADEV to describe what it’s doing. There are a bunch of other objectives you might have and ways to look at things. Let’s keep it somewhat simple for now. This already is going to be a long post.

Your OCXO has an ADEV curve. Let’s pick a fairly common one and use it.

2x10^-12 at 1 second
1x10^-12 at 10 seconds
1x10^-12 at 100 seconds
5x10^-12 at 1K seconds (maybe)
keeps going up past 1K seconds

That’s what an HP 10811 that has been on power for a few months in a very stable environment might be doing. In a not quite so stable environment, you might be above 1x10^-11 at 1K seconds. Net is, you get the famous “bathtub” shaped curve.

Your disciplining source “as received”  has an ADEV plot. Typically there is a number at 1 second and it proceeds down a curve from there. It may go as 1/tau or as 1/squrt(tau). This eventually hits a limit and the slope changes.

The typical objective is to combine the “best” of the OCXO curve with the curve from the disciplining source. The hope is that you get a resulting curve that is quite good. Ideally, the performance of the disciplining source needs to be better than the OCXO at a reasonable tau. If the tau gets to long, the OCXO is heading up the right hand side of the bathtub curve.

A GNSS module might give you a 1 second ADEV anywhere from 1x10^-8 to 1x10^-10 depending on the specific part you are using. It is likely that it goes down at 1/tau out past 10K seconds. If you start at 1x10^-9, you will hit 1x10^-11 at 100 seconds. You will cross the OCXO ADEV curve shown above before you get to 1,000 seconds.

A VLF carrier is a much less stable source. It also likely drops off as 1/sqrt(tau). If it’s 1x10^-6 at 1 second, that’s way better than anything you would need for a wall clock. If it’s 1x10^-7 and drops with a power law (square root), you would be at 1x10^-9 at 10,000 seconds.

The cross over between the OCXO and disciplining source curves is done by your control loop. The time constants are very long. Doing this with anything other than (custom) code running in an MCU of some sort is not a good way to go.

You very much need to do the best job you can on your receiver if you are going to get any reasonable sort of result. A delta of 1x10-7 at 1 second would come out to 2.2 degrees at 60KHz.

Bob

Hi

Since this is 60 KHz, you can get a whole lot of bits on the ADC and still not be spending an insane amount of money. There are devices that do quite well. Start moving to parts that have clock rates above about a MHz and the cost does indeed rise at an alarming rate :)

One “side benefit” of all the man made crud down at VLF: Assuming you are in a normal urban environment, it’s there 24/7/365. Your setup does not need to get to anywhere close to the “thermal noise floor”. Part of that impacts how you hook to the antenna. The other part impacts how you rig up the ADC.

As you go looking at SDR’s showing their reception on the web, you do need to pay attention to just what you have tapped into. The combination of antenna plus SDR design will have an impact on what you do or don’t see on their plots. It’s not unusual for them show coverage down to frequencies that they really do not handle very well.

Bob