Hi,
The main issue with the W3PM based design is the usage of fractional-N mode for the correction of the Si5351a output frequencies.
There have been several discussions showing that a clean output cannot be realized when using an Si5351 in fractional-N mode (as required for the frequency corrections with this design).
An Si5351running with integer mode is able to generate 25MHz and 40Mhz with a signal quality that is good enough for QO-100 operation.
73 Jean DJ0VL
Hi all,
I do not own nor intend to buy a DX-Patrol groundstation. This said, I would be disappointed seing an offset of 200Hz between RX and TX with a setup claiming to be GPS locked.
In principle, the ADF4351 can be programmed to output the exact frequency for generating the LNB LO of 10057,0 MHz as required for an IF of 432,5 MHz. The trick is to calculate using fractions and not trying to solve the fraction.
According to the datasheet, the output frequency of the ADF4351 calculates to RFOUT = fPFD × (INT + (FRAC/MOD))
Assuming fPDF as 10Mhz, 1650.379487xxx MHz can be generated exactly (meaning exactly 64 * 10057/390) with e.g. INT=165, FRAC = 37 and MOD = 975.
73 Jean DJ0VL
Hi Peter,
On the NB-Transponder, the signal seemed to move in from lower frequencies, see the noise level shape. A few seconds after taking the screenshot, a huge signal (20 or more carriers spread around 100 kHz) swept in from the lower band end. Unfortunately, the signal was shut down before I could make another screenshot.
73 Jean DJ0VL
Hi Peter,
The incident was also visible on the NB-Transponder where all signals - including the beacons - faided away for some 30 sec.
73 Jean DJ0VL
Hello Rene
Thank you for the details on your nice setup with a Raspi for the beacon generation. For my test (I wanted to double-check the stability of a recently build homebrew GPSDO in addition to some ADEV measurements), I used my Pluto with just the SPF5189 driver into a 75cm dish with helix feed.
The reports given by Edson nicely reflect the output level difference.
73 Jean DJ0VL
Hi all,
My beacon transmission on Tuesday was just a short test for the stability of a new homebrew GPSDO. So far I do not intend to operate the beacon on a regular base.
Nevertheless nice to read that he signal could be received with 1-2 db higher SNR levels as I experienced with my 75cm dish.
73 Jean DJ0VL
.... Btw the Brazilian WebSDR is about 250 Hz higher than the PSK-beacon, see my previous message with screenshot.
Hello Thomas,
Somewhat off-topic and just for fun: the accuracy of the Brazil WebSDR depends on the time of day. At night-time in Brazil, it goes off by more than 1kHz 
I am sure he knows. I will ask next time. The reasons for these errors are probably interesting and subtle.
Thank you for this highly interesting paper. Figure 4 confirms my observations regarding the short term stability. I commented on this as the observed stability may be normal behaviour for the type of GPDSO used, however it seems somewhat higher than the stated and measured stability data for this GPSDO.
I'm fully aware that we are discussing deviations on a very high level, nevertheless I'm interested in knowing the reason for the errors. We do not know whether its all due to the stability of the GPSDO, but imho a decent GPSDO has a significantly higher stability.
Allan deviation is stability.. 130Hz in 10 GHz is accuracy. Check here for a good explanation of the subject.
That said, there are numerous explanations for this offset including hw failures. Maybe telling the WebSDR team directly?
Thanks for the link and of course you are right. I'm familiar with the difference between stability and accuracy, my comment referred to the noticeable jitter. Sorry for the unclear wording.
73 Jean
PS: I would like to tell the team directly, but finding a contact address on the webpage is not easy.
130Hz in 10 GHz? Shocking. Don't think anyone said it was a standard frequency reference, but don't forget to account for Doppler and relativity.
Not shocking, just wondering as it apparently uses a GPSDO reference that states an Allan Deviation of 5E-11 @Tau=1s.
73 Jean DJ0VL
Hi all,
Half a year later, the issue remains unchanged. The Goonhilly WebSDR now seems to be off by around 130Hz. Apart from the offset, there is a visible jitter of quite a few Hz.
The screenshot below shows the same audio signal as received simultaneously by the Brazil WebSDR (trace at the left side), the Goonhilly WebSDR (trace at the right side) and by my own receiver (trace at 762Hz).
The Brazil WebSDR (http://appr.org.br:8902/) is shown just for comparison as Thomas mentioned it above and I guess it uses a simple TCXO as a reference. Both offset and drift are heavily dependent on the time of day.
73 Jean DJ0VL
Trevor,
C123 and C124 form a capacitive voltage divider to adapt the ~1.8Vpp output level of the original Pluto oscillator to the 1.3Vpp required by the AD9363.
The XTALN input of the AD9363 has an input resistance of about 10kOhm in parallel with 10pF, so that the capacitance of the input pin is sufficient to form the voltage divider with C123. Therefore C124 is not placed on the PCB.
73 Jean
Trevor,
You can find videos showing how to connect PGND to GND on Youtube, e.g. this one: Pluto ground mod (near the end of the video).
Jean
Hello Mike,
'ip' must be entered as lowercase letters, the string is not accepted when you enter uppercase letters.
73 Jean
Rumen,
The Si5351 is able to generate a quite clean 25MHz signal if you consider some basic requirements:
- you need a clean 10MHz input signal.
- the Si5351 is sensitive to supply voltage noise. Use linear voltage regulators and low esr decoupling caps.
- most important: use even integer divider ratios, as well for the pll stage as for the multisynth divider.
Usage of even integer divider ratios is essential for a clean output signal, a fractional-n divider leads to a messy signal with lots of spurs.
The output signal is a square wave, so you may want to add some low-pass filtering.
73 Jean
