Do time nuts only use and measure 10MHz clocks?
In a DMTD one can use one clock send to both mixer for measuring very
accurately the frequency/phase difference of two clocks at roughly the
same frequency.
One can also imagine having two different clocks derived from the same
reference, one for each mixer, enabling the measurement of two very
difference clocks (such as 10MHz and 1, 5 or 100MHz) with hopefully
equal accuracy and without having to use external dividers.
Is this for time nuts a relevant use case?
If yes, what other frequencies besides 10MHz (and 1PPS) are used?
Erik.
Am 2022-10-06 17:10, schrieb Erik Kaashoek via time-nuts:
In a DMTD one can use one clock send to both mixer for measuring very
accurately the frequency/phase difference of two clocks at roughly
the same frequency.
One can also imagine having two different clocks derived from the same
reference, one for each mixer, enabling the measurement of two very
difference clocks (such as 10MHz and 1, 5 or 100MHz) with hopefully
equal accuracy and without having to use external dividers.
Is this for time nuts a relevant use case?
If yes, what other frequencies besides 10MHz (and 1PPS) are used?
Erik.
I did 100 MHz Cesium (it's the law) vs 100 MHz hydrogen maser (phase
noise);
at least the first, analogue stage.
DMTD has its drawback already in its name: TIME DIFFERENCE. That means
that the conversion oscillator does not cancel out completely since its
mixing results are used at different points of time in the two channels.
BTW, regarding my isolation channel for 100+ MHz:
I got new transistors from Digikey; Central Semi CXT3904 and
Diodes, Inc DXT2222, both in SOT-89. Both good for 5 MHz, perhaps. :-(
The surprise was in the wrong direction.
There is simply a big hole between BFQ19S (too hot) and the others.
Nothing available, at least for my voltage to current converter.
And LTspice and the VNA have completely different points of view.
I tend to believe the VNA.
It will take some time to digest Peter Staric°/Eric Margan's
"Wideband Amplifiers". Gurus from the time when Tektronix was king.
Cheers, Gerhard
That would be the case when using a TIC or a timestamping counter I believe.
How about when the DMTD actually measures phase using ADC's and DSP and
looks at the whole sinus?
Would that cancel out?
Erik.
On 6-10-2022 23:48, Gerhard Hoffmann via time-nuts wrote:
DMTD has its drawback already in its name: TIME DIFFERENCE. That means
that the conversion oscillator does not cancel out completely since its
mixing results are used at different points of time in the two channels.
Hi,
5 MHz is also very common. For some uses 100 MHz have become popular too.
I will build up a 5 MHz setup here for more continuous monitoring, as it
is the common demoninator among my sources, not 10 MHz.
Cheers,
Magnus
On 2022-10-06 17:10, Erik Kaashoek via time-nuts wrote:
In a DMTD one can use one clock send to both mixer for measuring very
accurately the frequency/phase difference of two clocks at roughly
the same frequency.
One can also imagine having two different clocks derived from the same
reference, one for each mixer, enabling the measurement of two very
difference clocks (such as 10MHz and 1, 5 or 100MHz) with hopefully
equal accuracy and without having to use external dividers.
Is this for time nuts a relevant use case?
If yes, what other frequencies besides 10MHz (and 1PPS) are used?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Do you know Andrew Holme's phase noise project:
http://www.aholme.co.uk/PhaseNoise/Main.htm
There it is possible that the DUT and references have different
frequencies up to a factor 4. Even differnt frequencies for the two
references would be possible, but that is not yet implemented (it would
need a second NCO in the FPGA - I did ask for that in a future version
of the software).
I measured clocks for audio puposes (24.576MHz, resp 22.5792Mhz and
multiples). My OCXO with best close-in phase noise have 16.384MHz. So
this freedom in choice of frequencies is very helpfull.
Am 06.10.22 um 17:10 schrieb Erik Kaashoek via time-nuts:
In a DMTD one can use one clock send to both mixer for measuring very
accurately the frequency/phase difference of two clocks at roughly the
same frequency.
One can also imagine having two different clocks derived from the same
reference, one for each mixer, enabling the measurement of two very
difference clocks (such as 10MHz and 1, 5 or 100MHz) with hopefully
equal accuracy and without having to use external dividers.
Is this for time nuts a relevant use case?
If yes, what other frequencies besides 10MHz (and 1PPS) are used?
Erik.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Yes, I took inspiration from Andrew's design, although using substantial
less HW power.
After implementing the ability to have two different input clocks I
discovered (yes, I could have known before implementing) that as soon as
you use two different frequencies for the two LO's, used for down mixing
to the ADC input , you loose coherence, even if they are derived from
the same clock, because they mix different frequencies into a common IF
frequency so the difference in LO frequencies causes different impact of
the phase noise into the IF frequency.
So, different input frequencies for ref and DUT are possible but the
resolution is destroyed by the phase noise of the internal reference
TCXO. Maybe I could use the ref input as reference for the internal PLL
to eliminate the TCXO, something to test.
Erik.
On 7-10-2022 12:00, zfe via time-nuts wrote:
Do you know Andrew Holme's phase noise project:
http://www.aholme.co.uk/PhaseNoise/Main.htm
There it is possible that the DUT and references have different
frequencies up to a factor 4. Even differnt frequencies for the two
references would be possible, but that is not yet implemented (it would
need a second NCO in the FPGA - I did ask for that in a future version
of the software).
I measured clocks for audio puposes (24.576MHz, resp 22.5792Mhz and
multiples). My OCXO with best close-in phase noise have 16.384MHz. So
this freedom in choice of frequencies is very helpfull.
I did, up to now, not experience a loss in measurement resolution using
different frequencies for DUT and REF.
The first picture is a 10MHz DUT vs two 16.384MHz REFs. The picture
using 10MHz Refs looks alike (not attached).
The second is a 22.5792MHz NZ2520SDA as DUT vs two 10MHz REFs (with PN
of -150dBc for higher frequencies). It looks quite close data in the
data sheet. The differences are surely more due my not very perfect
measurement setup.
The sampling clock for the ADC comes from a free running SI5345
evaluation board. Its phase noise is decent for high frequencies, but
not for the frequencies measured with Andrew's setup (<1MHz).
Am 07.10.22 um 18:26 schrieb Erik Kaashoek via time-nuts:
Yes, I took inspiration from Andrew's design, although using substantial
less HW power.
After implementing the ability to have two different input clocks I
discovered (yes, I could have known before implementing) that as soon as
you use two different frequencies for the two LO's, used for down mixing
to the ADC input , you loose coherence, even if they are derived from
the same clock, because they mix different frequencies into a common IF
frequency so the difference in LO frequencies causes different impact of
the phase noise into the IF frequency.
So, different input frequencies for ref and DUT are possible but the
resolution is destroyed by the phase noise of the internal reference
TCXO. Maybe I could use the ref input as reference for the internal PLL
to eliminate the TCXO, something to test.
Erik.
On 7-10-2022 12:00, zfe via time-nuts wrote:
Do you know Andrew Holme's phase noise project:
http://www.aholme.co.uk/PhaseNoise/Main.htm
There it is possible that the DUT and references have different
frequencies up to a factor 4. Even differnt frequencies for the two
references would be possible, but that is not yet implemented (it would
need a second NCO in the FPGA - I did ask for that in a future version
of the software).
I measured clocks for audio puposes (24.576MHz, resp 22.5792Mhz and
multiples). My OCXO with best close-in phase noise have 16.384MHz. So
this freedom in choice of frequencies is very helpfull.
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Thanks for the input.
After some more testing the coherence turned out to be lost even when the
inputs use the same mixing frequency. The current assumption is that the
two fractional dividers used to generate the two mixer input frequencies,
although using same settings, already cause the loss of coherence. Probably
because they are not started at the same moment so the fractional shifts
happen at different moments. Next test is to us only integer dividers to
generate the two mixing frequencies but this will lead to two IF
frequencies and it is not sure if this can be implemented on the not so
powerful MCU used.
Erik
On Sun, Oct 9, 2022, 11:30 zfe via time-nuts time-nuts@lists.febo.com
wrote:
I did, up to now, not experience a loss in measurement resolution using
different frequencies for DUT and REF.