WWVB Receiver...

[Gerhard Hoffmann]

Am 04.01.22 um 02:36 schrieb whit3rd:

On Monday, January 3, 2022 at 4:18:53 PM UTC-8, Gerhard Hoffmann wrote:
Am 03.01.22 um 23:58 schrieb Clifford Heath:

You seem to disregard multipath shortwave propagation and
effects of the ionosphere on WWV.

The longwave WWVB ( 60 kHz) is dominated by ionospheric
conduction, so multipath is a non-issue for that

Ok, we have DCF77 on 77.5 KHz instead

> source (other WWV stations at 5, 10, 15 MHz are less

I remember hearing them now & then in Europe

predictable). GPS ought to be less dependent on atmosphere
effects, of course \'cuz in the UP direction there\'s only a few miles of atmosphere.

... and you can remove the Faraday completely if you can carrier
lock the (lower) military frequency, too. Don\'t need no secret
polynom for that.

For best results, you might just want to make an atomic fountain of your own.

https://www.atlasobscura.com/places/nist-f1-cesium-fountain-atomic-clock

Don\'t need that; I\'ve made a comparator for a hydrogen maser and a
Cesium in space. In zero gravity you don\'t need a fountain;
a fountain is a sorry replacement for zero gravity.
The active hydrogen maser has the better phase noise, but cesium
is the law, at least in the long run.

Cheers, Gerhard

 

[Jan Frank]

whit3rd <whit3rd@gmail.com> wrote:

On Monday, January 3, 2022 at 4:18:53 PM UTC-8, Gerhard Hoffmann wrote:
Am 03.01.22 um 23:58 schrieb Clifford Heath:

You seem to disregard multipath shortwave propagation and effects of
the ionosphere on WWV.

The longwave WWVB ( 60 kHz) is dominated by ionospheric
conduction, so multipath is a non-issue for that
source (other WWV stations at 5, 10, 15 MHz are less
predictable). GPS ought to be less dependent on atmosphere
effects, of course \'cuz in the UP direction there\'s only a few miles of
atmosphere.

There are many errors in WWVB phase, primarily the diurnal phase shift at
sunrise and sunset. Tom Van Baak of Time-Nuts has published thse errors
against a Hydrogen Maser, and they are quite substantial.

GPS has many, many more errors besides a diurnal phase shift. You can get
multipath errors from local reflections, errors from the slant range when
satellites are near the horizon, even the error from the temperature
coefficient of the GPS antenna and coax cable leading to the receiver.

However, with care and special receivers and antennas, GPS is unbeatable. You
can get cm level accuracy, which is great for surveying. Getting accurate
time is a bit more difficult, but you can get down to 1e-12 accuracy. This is
why GPSDO\'s are so popular for in-house standards.

 

[Phil Hobbs]

Jan Frank wrote:
<snip>

I have designed a zdncpfd: zero deadband noise cancelling phase frequency
detector that locks two clocks together but omits the jitter of either
clock, such as the jitter in the 1PPS. It bypasses the 20*log(N) of
conventional multipliers and dividers.

Hmm. Well inside the loop bandwidth of a perfectly noiseless PLL, the
instantaneous frequency of the VCO in a PLL N-tupler is N times the
reference frequency.

Phase is the time integral of frequency, so the output noise phase is
also N times the reference phase, up to an additive constant of
integration.

It takes about three lines of algebra to show that (in the high SNR
limit) phase noise PSD goes as <delta phi> ** 2, so the output of a
perfectly locked PLL has 20 log N decibels worse phase noise than its
reference.

We know that sin phi ~= phi for small angles, so weak phase noise
behaves like additive noise. Thus that 20 log N is independent of the
phase noise spectrum as long as the total

<delta phi> << 2 pi radians,

Seems pretty conclusive to me.

How does your magic box avoid this?

(I\'m familiar with various tricks that reduce the additive phase noise
of dividers and phase detectors, but that\'s beside the point here--we\'ve
assumed they\'re noiseless to begin with.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

Phil Hobbs wrote:

Jan Frank wrote: <snip
I have designed a zdncpfd: zero deadband noise cancelling phase
frequency detector that locks two clocks together but omits the
jitter of either clock, such as the jitter in the 1PPS. It bypasses
the 20*log(N) of conventional multipliers and dividers.

Hmm. Well inside the loop bandwidth of a perfectly noiseless PLL,
the instantaneous frequency of the VCO in a PLL N-tupler is N times
the reference frequency.

Phase is the time integral of frequency, so the output

phase is also N times the reference phase, up to an additive constant
of integration.

It takes about three lines of algebra to show that (in the high SNR
limit) phase noise PSD goes as <delta phi> ** 2, so the output of a
perfectly locked PLL has 20 log N decibels worse phase noise than its
reference.

We know that sin phi ~= phi for small angles, so weak phase noise
behaves like additive noise. Thus that 20 log N is independent of
the phase noise spectrum as long as the total

delta phi> << 2 pi radians,

Seems pretty conclusive to me.

How does your magic box avoid this?

(I\'m familiar with various tricks that reduce the additive phase
noise of dividers and phase detectors, but that\'s beside the point
here--we\'ve assumed they\'re noiseless to begin with.)

Cheers

Phil Hobbs

(corrected an editing scar)

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Jan Frank]

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
snip
I have designed a zdncpfd: zero deadband noise cancelling phase
frequency detector that locks two clocks together but omits the jitter
of either clock, such as the jitter in the 1PPS. It bypasses the
20*log(N) of conventional multipliers and dividers.

Hmm. Well inside the loop bandwidth of a perfectly noiseless PLL, the
instantaneous frequency of the VCO in a PLL N-tupler is N times the
reference frequency.

Phase is the time integral of frequency, so the output noise phase is
also N times the reference phase, up to an additive constant of
integration.

It takes about three lines of algebra to show that (in the high SNR
limit) phase noise PSD goes as <delta phi> ** 2, so the output of a
perfectly locked PLL has 20 log N decibels worse phase noise than its
reference.

We know that sin phi ~= phi for small angles, so weak phase noise
behaves like additive noise. Thus that 20 log N is independent of the
phase noise spectrum as long as the total

delta phi> << 2 pi radians,

Seems pretty conclusive to me.

How does your magic box avoid this?

I have posted complete details of the noise rejecting sampler here. The
zdncpfd is a combination of the sampler plus a very simple extension of the
comventional pfd.

The end result is the output doesn\'t follow the linear response of a
conventional pfd, so the conventional 20*log(N) response no longer applies.
It also has no differential delay for + and - phase errors, so there is no
deadband. It is a full phase-frequency detector, so it guarantees lock if
the oscillator has enough range. It is unlike any phase detector you have
ever seen.

(I\'m familiar with various tricks that reduce the additive phase noise
of dividers and phase detectors, but that\'s beside the point here--we\'ve
assumed they\'re noiseless to begin with.)

Cheers

Phil Hobbs

 

[Phil Hobbs]

Jan Frank wrote:

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
snip
I have designed a zdncpfd: zero deadband noise cancelling phase
frequency detector that locks two clocks together but omits the jitter
of either clock, such as the jitter in the 1PPS. It bypasses the
20*log(N) of conventional multipliers and dividers.

Hmm. Well inside the loop bandwidth of a perfectly noiseless PLL, the
instantaneous frequency of the VCO in a PLL N-tupler is N times the
reference frequency.

Phase is the time integral of frequency, so the output noise phase is
also N times the reference phase, up to an additive constant of
integration.

It takes about three lines of algebra to show that (in the high SNR
limit) phase noise PSD goes as <delta phi> ** 2, so the output of a
perfectly locked PLL has 20 log N decibels worse phase noise than its
reference.

We know that sin phi ~= phi for small angles, so weak phase noise
behaves like additive noise. Thus that 20 log N is independent of the
phase noise spectrum as long as the total

delta phi> << 2 pi radians,

Seems pretty conclusive to me.

How does your magic box avoid this?

I have posted complete details of the noise rejecting sampler here. The
zdncpfd is a combination of the sampler plus a very simple extension of the
comventional pfd.

The end result is the output doesn\'t follow the linear response of a
conventional pfd, so the conventional 20*log(N) response no longer applies.

So say I make a times-N PLL multiplier using your magic gizmo, and then
divide its output down by N again. I assume we agree that the divider
_reduces_ the phase noise by 20 log N decibels, yes? (Apart from its
additional jitter, which is not proportional to N.)

So just by multiplying up and dividing down again, with no additional
phase reference, we can reduce the phase noise of the fundamental to any
desired degree--doesn\'t that follow?

It also has no differential delay for + and - phase errors, so there is no
deadband. It is a full phase-frequency detector, so it guarantees lock if
the oscillator has enough range. It is unlike any phase detector you have
ever seen.

No doubt. I\'m having trouble finding where you posted this design,
though--you\'ve only been posting as \"Jan Frank\" since Christmas Eve.
Should I be looking under some other nym?

It would save time if you could repost it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Jan Frank]

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
[...]

Seems pretty conclusive to me.

How does your magic box avoid this?

I have posted complete details of the noise rejecting sampler here. The
zdncpfd is a combination of the sampler plus a very simple extension of
the comventional pfd.

The end result is the output doesn\'t follow the linear response of a
conventional pfd, so the conventional 20*log(N) response no longer
applies.

So say I make a times-N PLL multiplier using your magic gizmo, and then
divide its output down by N again. I assume we agree that the divider
_reduces_ the phase noise by 20 log N decibels, yes? (Apart from its
additional jitter, which is not proportional to N.)

No. It does not REDUCE the phase noise. Each oscillator has it\'s own phase
noise which is not affected by the zdncpfd. It merely decouples one from
the other.

So just by multiplying up and dividing down again, with no additional
phase reference, we can reduce the phase noise of the fundamental to any
desired degree--doesn\'t that follow?

No. See above.

It also has no differential delay for + and - phase errors, so there is
no deadband. It is a full phase-frequency detector, so it guarantees
lock if the oscillator has enough range. It is unlike any phase
detector you have ever seen.

No doubt. I\'m having trouble finding where you posted this design,
though--you\'ve only been posting as \"Jan Frank\" since Christmas Eve.
Should I be looking under some other nym?

I have to keep changing my nym since Youtube bans me for violating their
idiotic rules which you don\'t find out about until you break one of them.

I used Steve Wilson for a long time. I searched through my recent posts and
could not find the entry, so it was a while ago. I finally got a new web
site to upload the information, but I have been busy searching for a new
pcb cad program to ease layouts. After going thorough a number such as
Altium, EasyCad, and many others, I finally settled on Diptrace. I consider
this is the best program since PCAD of the DOS days. Now I have to put it
to work and complete some projects.

> It would save time if you could repost it.

I plan to as soon as I get some other things out of the way. I also need to
build a cross-coupled phase noise tester to prove the results. See Rubiola.
Thanks for your interest and comments.

Cheers

Phil Hobbs

 

[Phil Hobbs]

Jan Frank wrote:

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
[...]

Seems pretty conclusive to me.

How does your magic box avoid this?

I have posted complete details of the noise rejecting sampler here. The
zdncpfd is a combination of the sampler plus a very simple extension of
the comventional pfd.

The end result is the output doesn\'t follow the linear response of a
conventional pfd, so the conventional 20*log(N) response no longer
applies.

So say I make a times-N PLL multiplier using your magic gizmo, and then
divide its output down by N again. I assume we agree that the divider
_reduces_ the phase noise by 20 log N decibels, yes? (Apart from its
additional jitter, which is not proportional to N.)

No. It does not REDUCE the phase noise. Each oscillator has it\'s own phase
noise which is not affected by the zdncpfd. It merely decouples one from
the other.

It isn\'t your magic thing that does the reducing, it\'s the divider, silly.

So just by multiplying up and dividing down again, with no additional
phase reference, we can reduce the phase noise of the fundamental to any
desired degree--doesn\'t that follow?

No. See above.

It also has no differential delay for + and - phase errors, so there is
no deadband. It is a full phase-frequency detector, so it guarantees
lock if the oscillator has enough range. It is unlike any phase
detector you have ever seen.

No doubt. I\'m having trouble finding where you posted this design,
though--you\'ve only been posting as \"Jan Frank\" since Christmas Eve.
Should I be looking under some other nym?

I have to keep changing my nym since Youtube bans me for violating their
idiotic rules which you don\'t find out about until you break one of them.

I used Steve Wilson for a long time. I searched through my recent posts and
could not find the entry, so it was a while ago.

Ah, okay, it\'s Mike Mpnett. Hiya.

I finally got a new web
site to upload the information, but I have been busy searching for a new
pcb cad program to ease layouts. After going thorough a number such as
Altium, EasyCad, and many others, I finally settled on Diptrace. I consider
this is the best program since PCAD of the DOS days. Now I have to put it
to work and complete some projects.

It would save time if you could repost it.

I plan to as soon as I get some other things out of the way. I also need to
build a cross-coupled phase noise tester to prove the results. See Rubiola.
Thanks for your interest and comments.

Yeah, sure, I\'ll be waiting with bated breath. Please work on perpetual
motion next.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Jan Frank]

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:

No. It does not REDUCE the phase noise. Each oscillator has it\'s own
phase noise which is not affected by the zdncpfd. It merely decouples
one from the other.

It isn\'t your magic thing that does the reducing, it\'s the divider,
silly.

It has nothing to do with the divider. It is the phase detector. You can lock
two 10MHz signals together, such as a Rubidium and an OCXO and it works the
same.

[...]

Yeah, sure, I\'ll be waiting with bated breath. Please work on perpetual
motion next.

Some people felt the same about the noise rejecting sampler.

Cheers

Phil Hobbs

 

[Phil Hobbs]

Jan Frank wrote:

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:

No. It does not REDUCE the phase noise. Each oscillator has it\'s own
phase noise which is not affected by the zdncpfd. It merely decouples
one from the other.

It isn\'t your magic thing that does the reducing, it\'s the divider,
silly.

It has nothing to do with the divider.

My thought experiment has everything to do with the divider, because it
shows that your scheme is impossible even in principle. If you can
multiply without incurring the 20 log N penalty, then by the well known
properties of dividers, you can divide it down and get something 20 log
N decibels quieter than the signal you started with.

It is the phase detector. You can lock
two 10MHz signals together, such as a Rubidium and an OCXO and it works the
same.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

Phil Hobbs wrote:

Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:

No. It does not REDUCE the phase noise. Each oscillator has
it\'s own phase noise which is not affected by the zdncpfd. It
merely decouples one from the other.

It isn\'t your magic thing that does the reducing, it\'s the
divider, silly.
It has nothing to do with the divider.

My thought experiment has everything to do with the divider, because
it shows that your scheme

(however good it may be at reducing the phase detector jitter) cannot
do as you claim.

If you can multiply without incurring the 20 log N penalty, then by
the well known properties of dividers, you can divide it down and get
something 20 log N decibels quieter than the signal you started
with.

It is the phase detector. You can lock two 10MHz signals together,
such as a Rubidium and an OCXO and it works the same.

Cheers

Phil Hobbs

(Added a bit of clarification--the phase detector gizmo probably works
great, but it can\'t avoid the 20 log N.)


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Jan Frank]

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:

No. It does not REDUCE the phase noise. Each oscillator has it\'s own
phase noise which is not affected by the zdncpfd. It merely decouples
one from the other.

It isn\'t your magic thing that does the reducing, it\'s the divider,
silly.

It has nothing to do with the divider.

My thought experiment has everything to do with the divider, because it
shows that your scheme is impossible even in principle. If you can
multiply without incurring the 20 log N penalty, then by the well known
properties of dividers, you can divide it down and get something 20 log
N decibels quieter than the signal you started with.

The output is not the conventional linear pdf + and - pulses. The 20*log(N)
does not apply.

It is the phase detector. You can lock
two 10MHz signals together, such as a Rubidium and an OCXO and it works
the same.

Cheers

Phil Hobbs

 

[Jan Frank]

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

[...]

(Added a bit of clarification--the phase detector gizmo probably works
great, but it can\'t avoid the 20 log N.)

It can and does. It applies the technique used in the noise rejecting sampler
amplitude domain to the time domain. I searched for years to find a way to do
this, and finally found it.

As I stated, it is unlike any other pfd you have seen before.

 

[piglet]

On 04/01/2022 9:04 pm, Jan Frank wrote:

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

[...]

(Added a bit of clarification--the phase detector gizmo probably works
great, but it can\'t avoid the 20 log N.)

It can and does. It applies the technique used in the noise rejecting sampler
amplitude domain to the time domain. I searched for years to find a way to do
this, and finally found it.

As I stated, it is unlike any other pfd you have seen before.

Mention of amplitude domain makes me wonder if your development is
related to Pettigrew\'s amplitude locked loop?

piglet (who knows nothing about amplitude locked loops)

 

[Phil Hobbs]

Jan Frank wrote:

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:

No. It does not REDUCE the phase noise. Each oscillator has it\'s own
phase noise which is not affected by the zdncpfd. It merely decouples
one from the other.

It isn\'t your magic thing that does the reducing, it\'s the divider,
silly.

It has nothing to do with the divider.

My thought experiment has everything to do with the divider, because it
shows that your scheme is impossible even in principle. If you can
multiply without incurring the 20 log N penalty, then by the well known
properties of dividers, you can divide it down and get something 20 log
N decibels quieter than the signal you started with.

The output is not the conventional linear pdf + and - pulses. The 20*log(N)
does not apply.

No sort of phase detector--linear mixer, sampling, bang/bang, xor, you
name it--can get rid of the 20 log N, because it ain\'t the phase
detector that produces it. It\'s inherent to the multiplication process.

It\'s inherent in the frequency multiplication process.

It is the phase detector. You can lock
two 10MHz signals together, such as a Rubidium and an OCXO and it works
the same.

Sure, because then 20 log N = 0.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Clifford Heath]

On 4/1/22 1:58 pm, Jan Frank wrote:

Clifford Heath <no.spam@please.net> wrote:

On 20/12/21 9:33 am, Jan Frank wrote:
jlarkin@highlandsniptechnology.com wrote:
1 PPS doesn\'t provide much information

The 10 MHz GPS clock is derived from the 1 PPS. I will use the phase of
the WWVB carrier to compare against the GPS signal.

The 1PPS is also derived by a PLL locked to the GPS signals, so you\'re
still dependent on the stability of the underlying VCO for low phase
noise. The GPS signal is well below the noise floor, so the PLL loop
filter is necessarily slow - meaning the VCO phase noise will probably
extend to sub-Hz frequencies before the GPS locking signal starts to
pull it in. The 10MHz output is no better, of course.

Folk who are enthusiastic about their GPS disciplined 10MHz reference
source seem to mostly disregard these issues.

Clifford Heath.

You have got it backwards. The 1PPS is from the GPS signal. The 10MHz is
from an OCXO locked to the 1PPS.

Ohai, Mike M aka Steve aka Jan. You don\'t need to change your USENET nym
to defeat Youtube, and I wish you wouldn\'t.

This is where most people go wrong. There is 20ns to 50ns jitter in the
1PPS signal. This is from the internal clock in the GPS receiver that
decodes the GPS signal and issues the 1PPS pulse. It knows how much error
there is between the GPS time and when it is going to issue the 1PPS pulse.

\"it knows how much error... and when it is going to issue\" means it\'s
making a prediction. The prediction must be continually corrected, and
that requires some kind of LPF doesn\'t it? So the 1PPS is effectively a
PLL locked to the 1PPS from the GPS signal, as I said. I didn\'t mention
the 10MHz signal.

Cheers,

Clifford Heath

 

[Jan Panteltje]

On a sunny day (Tue, 4 Jan 2022 15:25:06 -0500) it happened Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote in
<bbcf2d8f-003e-d901-5b25-f7115f9f6677@electrooptical.net>:

Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

Jan Frank wrote:

No. It does not REDUCE the phase noise. Each oscillator has it\'s own
phase noise which is not affected by the zdncpfd. It merely decouples
one from the other.

It isn\'t your magic thing that does the reducing, it\'s the divider,
silly.

It has nothing to do with the divider.

My thought experiment has everything to do with the divider, because it
shows that your scheme is impossible even in principle. If you can
multiply without incurring the 20 log N penalty, then by the well known
properties of dividers, you can divide it down and get something 20 log
N decibels quieter than the signal you started with.

I cannot follow you there
sure the timeconstant in the PLL loop matters,
your oscillator can be super super quiet, low phase noise, but slowly steered to the exact frequency?

Or are we talking about different things?

 

[Jan Frank]

piglet <erichpwagner@hotmail.com> wrote:

On 04/01/2022 9:04 pm, Jan Frank wrote:
Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

[...]

(Added a bit of clarification--the phase detector gizmo probably works
great, but it can\'t avoid the 20 log N.)

It can and does. It applies the technique used in the noise rejecting
sampler amplitude domain to the time domain. I searched for years to
find a way to do this, and finally found it.

As I stated, it is unlike any other pfd you have seen before.



Mention of amplitude domain makes me wonder if your development is
related to Pettigrew\'s amplitude locked loop?

piglet (who knows nothing about amplitude locked loops)

I found some links in google, but the articles were unintelligible. It has no
relation to my work.

 

[Jan Frank]

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:

> Jan Frank wrote:

[...]

The output is not the conventional linear pdf + and - pulses. The
20*log(N) does not apply.

No sort of phase detector--linear mixer, sampling, bang/bang, xor, you
name it--can get rid of the 20 log N, because it ain\'t the phase
detector that produces it. It\'s inherent to the multiplication process.

I keep telling you it is not linear and does not follow the same
limitations.

It\'s inherent in the frequency multiplication process.

It is the phase detector. You can lock
two 10MHz signals together, such as a Rubidium and an OCXO and it
works the same.

Sure, because then 20 log N = 0.

If you use a conventional pfd, the rubidium noise will transfer to the OCXO
since the pfd is linear. The 20*log(1) is meaningless here.

My detector is not linear and will not transfer the rubidium noise to the
OCXO. It will lock on frequency and phase, but the rubidium jitter will not
affect the OCXO.

Every case of locking one oscillator to another shows a shelf in the phase
noise where the jitter is inside the loop bandwidth. My detector eliminates
this shelf.

For example, you could lock a SAW oscillator at 1 GHZ to a crystal
oscillator at 1 MHz. Normally, the 20*log(N) would produce 60 dB of
degredation in the SAW phase noise.

My detector decouples the SAW from the crystal, and keeps the original SAW
phase noise.

similarly, you could lock a GUNN at 10GHz to a 1 MHz crystal. This would
normally produce 80 dB degredation in the GUNN phase noise, which is why
ultra pure oscillators running at as high a frequency are sought after. See
Wenzel 100Mhz ULN oscillators.

With my detector, this is no longer true. You can use any stable crystal to
achieve the same frequency stability without the additional noise due to
frequency multiplication.

Thanks for your comments. You are giving me excellent ammunition when it
comes to selling my detector to major companies in the low noise business.

Cheers

Phil Hobbs

 

[Jan Frank]

Clifford Heath <no.spam@please.net> wrote:

On 4/1/22 1:58 pm, Jan Frank wrote:
Clifford Heath <no.spam@please.net> wrote:

On 20/12/21 9:33 am, Jan Frank wrote:
jlarkin@highlandsniptechnology.com wrote:
1 PPS doesn\'t provide much information

The 10 MHz GPS clock is derived from the 1 PPS. I will use the phase
of the WWVB carrier to compare against the GPS signal.

The 1PPS is also derived by a PLL locked to the GPS signals, so you\'re
still dependent on the stability of the underlying VCO for low phase
noise. The GPS signal is well below the noise floor, so the PLL loop
filter is necessarily slow - meaning the VCO phase noise will probably
extend to sub-Hz frequencies before the GPS locking signal starts to
pull it in. The 10MHz output is no better, of course.

Folk who are enthusiastic about their GPS disciplined 10MHz reference
source seem to mostly disregard these issues.

Clifford Heath.

You have got it backwards. The 1PPS is from the GPS signal. The 10MHz
is from an OCXO locked to the 1PPS.

Ohai, Mike M aka Steve aka Jan. You don\'t need to change your USENET nym
to defeat Youtube, and I wish you wouldn\'t.

Too bad.

This is where most people go wrong. There is 20ns to 50ns jitter in the
1PPS signal. This is from the internal clock in the GPS receiver that
decodes the GPS signal and issues the 1PPS pulse. It knows how much
error there is between the GPS time and when it is going to issue the
1PPS pulse.

\"it knows how much error... and when it is going to issue\" means it\'s
making a prediction. The prediction must be continually corrected, and
that requires some kind of LPF doesn\'t it? So the 1PPS is effectively a
PLL locked to the 1PPS from the GPS signal, as I said. I didn\'t mention
the 10MHz signal.

No, you still don\'t understand. There is no locking of the 1PPS to GPS. You
can have multiple phase reversals as the receiver clock drifts high and
low. You can have hanging bridges, where the internal clock is temporarily
at the correct frequency, but the wrong phase.

See Tom Van Baak, Sawtooth and hanging bridges, timing stability at
http://www.leapsecond.com/pages/MG1613S/

Cheers,

Clifford Heath