A simple setup for measurements using the TinyPFA side channel

[Jarl]

The TinyPFA is a highly useful tool for measuring small phase/frequency differences due to the instruments low noise floor and its high resolution. And as shown in
 
https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.Examples   (last example)

using the PFA's side channel facility significantly reduce phase errors during long term measurement sessions.

The practical side of the side channel setup was briefly touched upon in the TinyPFA Forum in january 2023 under the heading tinyPFA hardware improvements? .

Out of curiosity I wondered if a straight forward solution just consisting of  power dividers/splitters and attenuators could do the job.

Using surplus building blocks bought at hamfests I had fine results with the first version -  but this solution would be rather expensive to replicate due to the high cost of the components involved – even second hand.

I later rebuilt the circuit using 3 cheap ADP-2-1 Mini Circuits combiners mounted in ”Ugly Bug” fashion with ordinary metal film resistors in the T-attenuator sections just to try out if it could be done for a few Euros on a piece of scrap printed circuit board.

Just for the record: This solution is meant for measurements in the range of 1 – 10 MHz.

I attach a set of Noise Floor measurements at 10 MHz of the two versions for information. They are obviously very close to each other and also very close to the example shown in the TinyPFA Measurements Examples. In my measurements I have used 1 s sample interval and decimation = 1.

I also attach a sketch of the circuit diagram and pictures of the two versions. The signal levels in the diagram are measured values from the printed circuit board version which demonstrates that the reqiured level of isolation between channels is obtained.

To produce the 10,008 MHz signal for the side channel I use a Si 5351A based generator with a Kuhne QH40A Crystal Heater attached to the Si-chip and to the crystal in order to stabilize the output frequency.

The Si5351A generator is visible as the grey box in the background on the picture of the first setup. It deliver a  fixed + 7 dBm output signal. This fairly high output level is the reason for including the 6 dB attenuator ahead of the first power splitter in the setup in order to produce the -25 dBm side channel levels at the A and B outputs.

The Si5351 generator together with the printed side channel circuit board is now a permanent part of the TinyPFA setup in my workshop. When in use, the TinyPFA and the side channel circuit board is firmly fixed with plastic clamps. Connecting coax cables to and from the setup must be left untouched during measurement sessions in order to avoid influencing the output from the TinyPFA.

I hope the description above demonstrate that a fairly cheap and simple setup for TinyPFA measurements using the side channel facility is indeed possible.

[bnz]

What is the idea behind the side channel approach? Is it just to have a common signal for the DUT-DAC and REF-DAC to cancel differences between these two DACs, or is there more?

[ch_scr]

I found it surprising that the relatively high phase-noise signal from SI5351 works. But I guess since it's the same for both channels, it's ok?

[Jarl]

To bnz and ch_scr:

I invite you to consult the information provided by the originator of the TinyPFA, Erik Kaashoek, in the TinyPFA Forum :

https://tinydevices.org/forum/viewtopic.php?t=31

General information about the TinyPFA can be found in https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.Homepage

I am - unfortunately - not aware of any more detailed description of the details behind the Tiny PFA, at least in its present form, but I am sure you can ask for further information from Erik Kaashoek at the TinyPFA Forum.

I can just confirm from my own observations that by using the side channel as described during long time measurements you can obtain several decades of lower noise level from the TinyPFA compared to measurements without using the side channel feature and that the frequency stability and level of phase noise from the Si5351 generator is sufficient for the requirements of the side channel

- Jarl

[WatchfulEye]

There is some non-ideality in the tinyPFA hardware which results in temperature dependent drift of the phase offset between the two channels.

By injecting a reference oscillator which is know to have identical phase on both channels, inter-channel phase offset can be measured and subtracted.

[thinkfat]

I found it surprising that the relatively high phase-noise signal from SI5351 works. But I guess since it's the same for both channels, it's ok?

For one, it's the same on both channels so the phase noise is identical. Also, it's 8 kHz away from the frequency of the DUT. Phase noise and frequency drift don't matter because the software just looks for the difference between the two channels and since the injected offset signal is from the same source, there is a fixed phase/frequency relation. Whatever difference between the offset signal is detected between the channels, must be due to inequalities of the channels themselves. The channel output at the DUT frequency can then be adjusted accordingly.

[Solder_Junkie]

I am puzzled by this side channel business. I have a PFA that has been used extensively for stability measurements with a rubidium standard in one port and a GPSDO in the other. I use TimeLab for the actual measurements.

I find it necessary to let the TinyPFA warm up/settle for several minutes before use, perhaps 20 mins or so.

I can easily compare GPSDO settings and also compare the four different GPSDOs that I have, with results are what I would expect. The “best” produces frequency differences of around +/- 2 parts in 10^11 over 60 seconds. The poorest shows significant short term jitter.

Considering the low cost of a TinyPFA, the results and repeatability are pretty impressive.

SJ

[Jarl]

Hi Solder_Junkie

Thanks for bringing this question up. It inspired me to make some noise floor tests on my own setup, see the attachment below.

The red and black curve refer to noise floor tests made with the Tiny PFA immediately after it is switched on after a long break. The other two curves are tests made more than one hour after switch-on of the Tiny PFA. All tests are made with tau=1 and decimation=1 and the side channel engaged.

As can be seen from the tests of the Tiny PFA immediately after switch-on, the curves deviate from the curves from the warmed-up PFA in the region below 1E - 15.

If your Tiny PFA test setup is used for testing GPSDO's using a rubidium standard as reference my guess is, that your measurements ought to be independent of the cold or warm status of the Tiny PFA since you are operating in the 1E - 11 region. But just to be sure: I suggest you make your own noise floor tests to see if other factors influence your measurements.

By way of information I also attach a picture of my present mechanical lay-out when using my Tiny PFA together with its side-channel circuit board. The base plate is made from perspex, offering some thermal decoupling from the surroundings. The Tiny PFA itself is held in place by self adhesive plastic feet mounted upside down on the base plate.

[Solder_Junkie]

Hi Jarl

The noise floor Allan Deviation I measured with my TinyPFA is in the region of 1E-14 at 10 seconds or greater. My Efratom FRS-C rubidium is specified at +/- 3 parts in 10^11 over 10 seconds, and 1 part in 10^11 over 100 seconds. There are higher stability rubidium oscillators, but that is all I have.

As the highest accuracy I realistically need is 1 part in 10^9 (1 Hz in 1 GHz) and even then 10 Hz at 1 GHz would be OK.

It is not easy to make meaningful comparison measurements at these stability levels as none of the equipment I have is officially calibrated, although for my needs any of my GPSDOs are plenty good enough, as is the rubidium.

SJ