tinyPFA feature request
I have been using a tinyPFA for several weeks now, and it does exactly what
it's intended to do. Thanks and congratulations to Eric for creating this!
I have one request for an additional feature. Many systems of interest to
time-nuts have a 10 MHz output, but some important ones only have 5 MHz.
I can't always arrange doublers and dividers to put signals from different
gear on the same frequency. However, for signals that are very close in
frequency except for a factor of two, as most time-nuts signals are, it
seems like the tinyPFA could measure them directly without much additional
effort.
I don't know exactly how the internal software works, but it seems like
there is a software PLL which includes a software NCO. It seems that
shifting the NCO phase output word by one bit left or right in software
before applying it to the B channel signal would have the effect of
changing the target frequency by a factor of two. This seems like a fairly
simple software change to get a significant additional capability.
Possible side effects of this mode might include lower resolution or
additional noise (due to lost LSBs) and a narrower window of operation.
Currently both signal frequencies need to be within (200 Hz)/Tau of each
other, and that window would probably shrink by the ratio of the
frequencies.
It might be useful to allow two or three bits of shifting. There are old
OCXOs at 2.5 MHz, and I have gear on my bench at 20 and 40 MHz that I would
like to compare to a 10 MHz or 5 MHz reference.
I would try to implement this myself as a proof-of-concept, but the
measurement source code does not seem to be available.
Cheers!
--Stu
On 17/08/2023 20:51, Stewart Cobb via time-nuts wrote:
I would try to implement this myself as a proof-of-concept, but the
measurement source code does not seem to be available.
Cheers!
--Stu
Stu,
You might like to ask on the TinySA group:
There is (or was) a TinyPFA sub-group but I can't find the details now.
Cheers,
David
SatSignal Software - Quality software for you
Web: https://www.satsignal.eu
Email: david-taylor@blueyonder.co.uk
Twitter: @gm8arv
Here you can find the tinyPFA support forum
https://tinydevices.org/forum/index.php
Erik.
On 18-8-2023 9:03, David Taylor via time-nuts wrote:
On 17/08/2023 20:51, Stewart Cobb via time-nuts wrote:
I would try to implement this myself as a proof-of-concept, but the
measurement source code does not seem to be available.
Cheers!
--Stu
Stu,
You might like to ask on the TinySA group:
There is (or was) a TinyPFA sub-group but I can't find the details now.
Cheers,
David
I did some testing
Input A gets 10 MHz
Input B gets either 2.5 MHz or 5 MHz.
This works well if the input to B is a pulse instead of a sine wave or
symmetrical square wave as the pulse has enough second or fourth order
harmonics for the tinyPFA to work
As both tinyPFA inputs use the same NCO for down conversion to fit in
the sample rate of the ADC, your idea can not be implemented.
Erik.
On 17-8-2023 21:51, Stewart Cobb via time-nuts wrote:
I have been using a tinyPFA for several weeks now, and it does exactly what
it's intended to do. Thanks and congratulations to Eric for creating this!
I have one request for an additional feature. Many systems of interest to
time-nuts have a 10 MHz output, but some important ones only have 5 MHz.
I can't always arrange doublers and dividers to put signals from different
gear on the same frequency. However, for signals that are very close in
frequency except for a factor of two, as most time-nuts signals are, it
seems like the tinyPFA could measure them directly without much additional
effort.
I don't know exactly how the internal software works, but it seems like
there is a software PLL which includes a software NCO. It seems that
shifting the NCO phase output word by one bit left or right in software
before applying it to the B channel signal would have the effect of
changing the target frequency by a factor of two. This seems like a fairly
simple software change to get a significant additional capability.
Possible side effects of this mode might include lower resolution or
additional noise (due to lost LSBs) and a narrower window of operation.
Currently both signal frequencies need to be within (200 Hz)/Tau of each
other, and that window would probably shrink by the ratio of the
frequencies.
It might be useful to allow two or three bits of shifting. There are old
OCXOs at 2.5 MHz, and I have gear on my bench at 20 and 40 MHz that I would
like to compare to a 10 MHz or 5 MHz reference.
I would try to implement this myself as a proof-of-concept, but the
measurement source code does not seem to be available.
Cheers!
--Stu
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
After finding and analyzing the schematic for the NanoVNA-H4, which is the
underlying hardware for the tinyPFA, I now realize that my requested
feature is impossible. The hardware is extremely narrowband. It only
digitizes signals within a few kHz of the selected frequency. Therefore, it
can't simultaneously receive (say) 5 and 10 MHz. To do that, it would need
wideband digitizers, such as are found in the (much more expensive) TimePod
and PhaseStation designed by John Miles. The tinyPFA is a very clever reuse
of existing inexpensive hardware, but it does have limitations.
Cheers!
--Stu
On Thu, Aug 17, 2023, 14:51 Stewart Cobb stewart.cobb@gmail.com wrote:
I have been using a tinyPFA for several weeks now, and it does exactly
what it's intended to do. Thanks and congratulations to Eric for creating
this!
I have one request for an additional feature. Many systems of interest to
time-nuts have a 10 MHz output, but some important ones only have 5 MHz.
I can't always arrange doublers and dividers to put signals from different
gear on the same frequency. However, for signals that are very close in
frequency except for a factor of two, as most time-nuts signals are, it
seems like the tinyPFA could measure them directly without much additional
effort.
I don't know exactly how the internal software works, but it seems like
there is a software PLL which includes a software NCO. It seems that
shifting the NCO phase output word by one bit left or right in software
before applying it to the B channel signal would have the effect of
changing the target frequency by a factor of two. This seems like a fairly
simple software change to get a significant additional capability.
Possible side effects of this mode might include lower resolution or
additional noise (due to lost LSBs) and a narrower window of operation.
Currently both signal frequencies need to be within (200 Hz)/Tau of each
other, and that window would probably shrink by the ratio of the
frequencies.
It might be useful to allow two or three bits of shifting. There are old
OCXOs at 2.5 MHz, and I have gear on my bench at 20 and 40 MHz that I would
like to compare to a 10 MHz or 5 MHz reference.
I would try to implement this myself as a proof-of-concept, but the
measurement source code does not seem to be available.
Cheers!
--Stu
All,
I’ve been experimenting with the Tiny PFA and it’s an excellent piece of engineering. Thanks Erik!
Not completely understanding the statistics of ADEV, I’ve searched the forum for any information on the decimation that is selectable in the tPFA software. Are there any threads or references someone can point me to on this topic?
Thanks
Sent from my iPhone
Adam MacDonald
calvinf15@sbcglobal.net
850-460-5535
On Aug 20, 2023, at 10:01, Erik Kaashoek via time-nuts time-nuts@lists.febo.com wrote:
Here you can find the tinyPFA support forum
https://tinydevices.org/forum/index.php
Erik.
On 18-8-2023 9:03, David Taylor via time-nuts wrote:
On 17/08/2023 20:51, Stewart Cobb via time-nuts wrote:
I would try to implement this myself as a proof-of-concept, but the
measurement source code does not seem to be available.
Cheers!
--Stu
Stu,
You might like to ask on the TinySA group:
There is (or was) a TinyPFA sub-group but I can't find the details now.
Cheers,
David
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
Adam,
Both the Timepod and the tinyPFA normally would use all samples over the
set measurement interval to calculate the phase.
If the phase rotates quickly versus the measurement interval this leads
to some kind of averaging of the phase over the measurement interval
resulting in a reduction of the detection of any phase changes with a
period close to the measurement interval.
Both the Timepod and the tinyPFA solve this by not using the the data
from the whole interval but only a part of it. Say only the first 10%.
Any phase change with a period close to the measurement interval will
now no longer be attenuated.
This reduction in measurement time is similar to decimation used in
digital signal processing, hence the word decimation is used to describe
the amount of data that is not used from a measurement interval.
Setting the decimation to 5 in a tinyPFA delivers similar results to a
Timepod as the Timepod is using similar amount of decimation
Erik.
On 22-8-2023 1:13, Adam MacDonald wrote:
All,
I’ve been experimenting with the Tiny PFA and it’s an excellent piece of engineering. Thanks Erik!
Not completely understanding the statistics of ADEV, I’ve searched the forum for any information on the decimation that is selectable in the tPFA software. Are there any threads or references someone can point me to on this topic?
Thanks
Sent from my iPhone
Adam MacDonald
calvinf15@sbcglobal.net
850-460-5535