parasitic phase modulation in the signal from the ADF4351 board

[pulsarvictor]

I have found parasitic phase modulation in the signal from the chinese ADF4351 board:
https://www.aliexpress.com/item/1005005340663396.html?spm=a2g0o.order_list.order_list_main.117.21ef180290Ym5P
Please suggest how to fix this problem ?

[radiolistener]

I also found parasitic phase/frequency modulation on my LTDZ board with ADF4351.
I investigated it more deep and found that the root of cause is onboard oscillator, which has some phase/frequency fluctuations...

So it needs to be replaced with more clean VCXO. Unfortunately I don't have replacement to test if it helps to resolve issue.

It looks that your board supports external clock source, so you can use it to feed ADF4351 with more clean clock from external signal generator.

[pulsarvictor]

Tnak you for your reply.
I used external clock source from 10 MHz out of Agilent E4422B.
Is it clock enought clean from your point of view ?

[radiolistener]

at a glance it should be clean. How you determined that ADF4351 has parasitic modulation on the output? Did you checked power supply?

[Bud]

Reverse engineer the board and check if the IC is properly wired according to the datasheet. Those Ali "engineers" often have no clue what they are doing. See an example here:

https://www.eevblog.com/forum/rf-microwave/all-you-need-to-know-about-the-chinese-ad9959-ebay-board/#msg2205405

[radiogeek381]

How much phase modulation?

In the end, phase noise or phase modulation or close-in FM are all part of what makes the difference between an inexpensive synthesizer/oscillator and something that costs big-time money.

There are lots and lots of factors here, among them:

• power supply ripple/noise
• big divider ratios in the feedback loop
• fractional-N synthesis (spurs more than noise)
• reference oscillator feed-through (spurs and noise)
• reference oscillator drift/phase-noise/phase-modulation/FM
• substrate (on chip) noise near/under the VCO

I've seen each of these at various times.

How about someone who has designed one of these hopping in?

[Bud]

Oh, it has a loop filter. High chances are it was designed incorrectly.

[pulsarvictor]

Thank you for all.
I am preparing the answers.
Regards.

[pulsarvictor]

I do not have a phase noise analyzer, so analysis of the signal phase behavior was carried out as follows:
the signal under study with the frequency Fin was applied to the ADC input, and
coherent signal with frequency Fs was applied to ADC external clock.
Frequencies ratio Fin/Fs=5/4.
The same phase points (for example, 1, 5, 9, 13, ... etc) of the investigated signal discrete implementation in the absence of modulation
in both signals must fall on the same phase points of the studied periodical signal.
The graphs of the same phase points amplitudes should look like horizontal lines, and the histogram of the amplitudes dispersion
should seems like the law of the random variables distribution (if there are no regular components).
As an example, the results of such an experiment can be given:

Signal under study:
--------------------
Agilent's E4422B output (Fout=87.5MHz; Urms=707mV) ->
-> ADC (ADS5541) input;

ADC external clock signal:
---------------------------------
External output 10MHz of Agilent's E4422B TXCO -> Freq. Multiplyer 7x10=70MHz -> Amplifier -> Quartz Filter 70MHz ->
-> external input of ADC (ADS5541) synchronization;

Please look results in the attached files:
Figure 1.png
shows the behavior of similar phase points amplitudes in the absence of modulation of two coherent signals.
Total points in the discrete implementation of the signal - 16M.

Figure 2.png -
the histogram of the distribution for 4 same phase points is shown (because with a frequency ratio of 5/4 we get 4 points for "period")

Figure 3.png -
the same histogram for one point on an enlarged scale.
This graph looks like a normal distribution law of a random variable.

Figure 4.png
shows the behavior of similar phase points amplitudes in the case when the signal under study was taken
from a Chinese synthesizer board based on the ADF4351 chip (during the experiment, it was assumed that the chip was really from ADI)
In the absence of amplitude modulation, this result can be interpreted as a presence of phase/frequency parasitic modulation.
   
Unfortunately, a mistake was made during this experiment, because low-frequency amplitude modulation was not previously detected
in the output signal of synthesizer. The period of this modulating signal = 0.4 miliSek.
Thus, my interpretation of the experimental result as a presence of parasitic phase modulation in the signal was wrong.
From my point of view, the ADF4351 chip should not have amplitude modulation, but it will be better to move the discussion for this problem to another post
so the title of the current will not be match to new topic.

Thanks to everyone who tried to help me.

[prutser]

First thing to look is probably the power supply. The VCO sensitivity of the ADF4351 is 40MHz/V, which means that with 0.5mV noise you already get 20kHz deviation !
For some other (similar) PLL synthesizer IC's Analog devices advices the ADM7170 voltage regulator (Not the cheapest, but 'Low noise: 5 μV rms independent of output voltage at 100 Hz to 100 kHz') and high PSRR which makes sense if you consider the VCO sensitivity.