EEVblog Electronics Community Forum
Electronics => RF, Microwave, Ham Radio => Topic started by: ezalys on February 25, 2018, 12:01:05 am
-
I recently snagged one of these ADF4351 boards from ebay https://www.ebay.com/itm/35M-4-4GHz-PLL-RF-Signal-Source-Frequency-Synthesizer-ADF4351-Development-Board/282385472127?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649 (https://www.ebay.com/itm/35M-4-4GHz-PLL-RF-Signal-Source-Frequency-Synthesizer-ADF4351-Development-Board/282385472127?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649) . Does anyone have any experience with these? I managed to get it to lock to 2500 MHz in INT-N mode with a PFD frequency of 25 MHz, which is also the reference clock. I'm seeing reference spurs (2500 + n*25 MHz) at 60 dBc! Pretty crappy. Has anyone had similar problems with this board? How would you go about tracking down the issue?
-
I recently snagged one of these ADF4351 boards from ebay https://www.ebay.com/itm/35M-4-4GHz-PLL-RF-Signal-Source-Frequency-Synthesizer-ADF4351-Development-Board/282385472127?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649 (https://www.ebay.com/itm/35M-4-4GHz-PLL-RF-Signal-Source-Frequency-Synthesizer-ADF4351-Development-Board/282385472127?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649) . Does anyone have any experience with these? I managed to get it to lock to 2500 MHz in INT-N mode with a PFD frequency of 25 MHz, which is also the reference clock. I'm seeing reference spurs (2500 + n*25 MHz) at 60 dBc! Pretty crappy. Has anyone had similar problems with this board? How would you go about tracking down the issue?
I'm going to assume you mean -60dBc. And to me, that sounds pretty okay-ish? I'm not sure. I think for sure -60 dBc would be excellent for the harmonics of the fundamental.
Reading the datasheet (http://www.analog.com/media/en/technical-documentation/data-sheets/ADF4351.pdf), it looks like "Charge Cancelleation" might be of use. Later, in "Reference Spurs" section it seems like the reference spur levels even in Fractional-N mode can be as high as -80dBc (though it sounds like this is in badly designed boards, which one would assume the reference board or a copy of that board, are not).
I have the ADF4350 (https://sigrok.org/wiki/BG7TBL)-based board which can probably demonstrate this problem (if indeed it's using integer-N to generate that signal). I have never driven it directly.
I have a specan that should easily be able to see such spurs, so I'll post back later.
-
I'm not referring to harmonics of the fundamental. I'm referring to reference spurs... which in my case appear at 2525, 2550, etc. and likewise 2475, 2450, etc.
-
https://imgur.com/a/l6Xik
Here's a screenshot from the specan.
-
https://imgur.com/a/l6Xik
Here's a screenshot from the specan.
if you are using a reasonably stable 25 MHz reference you should be able to use a much lower PFD frequeyncy?
The charge-pump filter must be tailored to the PFD frequency - here it looks like your filter doesn't attenuate the 25 MHs PFD very much and it feeds through to the output (?).
Try a PFD that is much much lower, 1 kHz (?), and check what filter you have between charge_pump_out and vco_tune_in pins.
-
https://imgur.com/a/l6Xik
Here's a screenshot from the specan.
if you are using a reasonably stable 25 MHz reference you should be able to use a much lower PFD frequeyncy?
The charge-pump filter must be tailored to the PFD frequency - here it looks like your filter doesn't attenuate the 25 MHs PFD very much and it feeds through to the output (?).
Try a PFD that is much much lower, 1 kHz (?), and check what filter you have between charge_pump_out and vco_tune_in pins.
I am following this thread because I may learn something about solving such spur issues.
But I dont quite understand why would anyone want lower PFD frequency if fractional-N PLL is available? Would not in this case it will be even harder to filter PFD/reference components, because PLL loop filters are usually low-pass?
-
Yes this is correct. The loop filter on this board is designed for 25 MHz pfd. I suspect bad supply bypassing. Additionally if I set the divider, I do indeed see the pfd spurs go down, but the reference spurs persist.
-
Though honestly it looks like the bypassing is very similar to the reference dev board. I'm beginning to suspect the routing.
-
I cut the supply for the oscillator and drove the reference with an external source. The spurs are still there. Seems like the synthesizer itself needs more bypassing or something.
Unfortunately I have no rf probes... so I hacked it and I took a BNC bulkhead connector and soldered wires from the center conductor to the 3.3v rail and the ground to ground. I then put a DC block on this and hooked it into my specan. I realize this isn't impedance matched but I figured maybe I'd learn something. Anyhow, this showed some 50 MHz content at the level of -70 dBm. I guess this is 0.2 mVp-p into 50 ohms or 0.07 mVrms into 50 ohms... On a 3.3 volt rail, I don't see how this would induce -50 dBc sidebands.
-
I took some measurements with my ADF4530 board, which apprently has a 10.5 MHz reference in it. I see the same spurs, but they are -60dBc.
I suspect the fact that you see them at 50dBc has to do with your relatively high RBW.
-
https://imgur.com/a/l6Xik
Here's a screenshot from the specan.
for a 1V signal, there will be 4mV (-48dB) harmonics, spurs, noise or whatever you want to call it. at $30, what do you expect? -150dB noise?
-
I made an RF sniffer probe to try and track down the issue so if I make my own version I don't repeat the mistake. The board layout I think is just crappy. You can see a ton of 50 MHz trash floating around near the loop filter. The board appears to be two layers... so I think a 4 layer board with much more via stitching is likely needed to get better performance. Splitting the power rails might help, too, I'd imagine.
Mechatrommer:
No, I don't expect it to be great, but I also want to understand exactly why it's not great.
-
No, I don't expect it to be great, but I also want to understand exactly why it's not great.
maybe you can investigate into the filtering used after the synthesized signal output. or maybe less via count beside the tracks making it less ideal ground plane. (i'm waiting for my AF4355 generator its been more than a month |O)
-
Sorry to say, but those boards are crap. They use a single power rail from a single and noisy LM1117 LDO instead of isolating VCO and CP supply from digital and ref oscillator supply.
Two layer PCB doesnt allow you to make a good low noise synthesizer implementation based on ADF4351 with continuous ground plane, separated supply rails, good decoupling very near the IC, and a quiet loop filter. Its very easy to couple oscillator noise to loop filter and VCO supply, leading to FM modulating it and creating those sidebands.
I've analyzed some of those eBay boards and found some having 1117-5.0 regulators mistakenly mounted instead of the 3.3V version. As well as 25MHz CMOS oscillators, with high phase noise and harmonics at the output and supply ripple, instead of the more suitable clipped sinewave oscillator.
You can change the LDO and ref oscillator, and add additional capacitors. It would help, but you cant fix all the layout errors, so you will never get the ADF435x IC promised performance, as you get (or nearly) from the eval board or similar good design.