Author Topic: PLL (ADF4351) + 2-ch DDS (AD9912) board comments?  (Read 312 times)

0 Members and 1 Guest are viewing this topic.

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 676
PLL (ADF4351) + 2-ch DDS (AD9912) board comments?
« on: June 03, 2020, 04:53:58 pm »
Hi all you gurus that are used to laying out 4-layer boards with multi-GHz mixed signal ICs..  8)

I want to build a 2-channel AD9912 DDS-board that takes either 10 MHz or 100 MHz input reference frequency and converts that (ADF4351 PLL+VCO) to a 1 GHz SYSCLK for the DDS.
Here is a first design draft:

  • any general comments?
  • The power-supply side is quite extensive here, with a separate low-noise LDO everywhere that AD suggest separate LDOs - I wonder if this is useful or just overkill?
  • Although the board is 4-layer the supply-layer (3rd) and the SPI control-lines end up not being routed that nicely - I am hoping this will not matter much as the proper RF-lines are only on the top layer with a GND-plane as layer 2.
  • I don't understand the positioning of ADF4351 pins CP_OUT and VTUNE - I would have thought these should be close together so the passive filter can be nicely between the two pins. Not so - now I have a passive filter close to CP_OUT and then a via and a long-ish trace to the VTUNE pin..

thanks for comments,

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1206
  • Country: cn
  • RF Engineer @ OwOComm. Discord: スメグマ#2236
Re: PLL (ADF4351) + 2-ch DDS (AD9912) board comments?
« Reply #1 on: June 04, 2020, 05:48:18 am »
1. I would suggest using a buck converter to take the supply voltage down to 3 - 5V before the linear regulators to avoid power dissipation, but be careful about minimum load requirements - most buck ICs will enter "pulse skipping mode" at low load generating a crap ton of noise at LF. You will have to experiment with various buck ICs to find one that enters continuous mode at your load current, but this is easy as long as you stick to the "standard" pinout SOT-23 buck ICs.

2. There are a lot of what looks like 0603 parts on the ADF4351 output signal path. The big pads will add a capacitive discontinuity, so I would switch to 0402 there.

3. The ferrite beads followed by 0.1uF capacitors will give you a resonance. Make sure there is at least 10uF of capacitance on the load side of each ferrite bead.

4. R303, R302, R301 - don't terminate a differential RF signal like that, just use 2 50ohm resistors to ground. The idea being you want to preferably shunt out common mode signals. The ADF4351 does NOT have very good differential balance, and terminating with 3 resistors will weaken the differential mode signal compared to the common mode one. That is one big mistake made in the original nanovna's coupler bridge that worsened performance for no benefit.

You could probably get away with using fewer LDOs or even using buck converters for everything, but keep this in mind: the purpose of LDOs are for rejecting *low frequency* noise. High frequency noise is very easily filtered out by ferrite beads. If you have a circuit that you believe is going to introduce HF noise on its supply rail, a ferrite bead L-C filter or two will deal with it, no need for a separate LDO. Similarly, buck ICs are problematic not because of the switching action (1MHz and its harmonics are easily filtered out), but because of the control loop that introduces LF noise. If you have a circuit that is insensitive to LF noise (for example you have two synthesizers and you only care about relative phase noise between the two), you can get away with using ordinary buck ICs.
Discord: スメグマ#2236
GitHub: gabriel-tenma-white

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 676
Re: PLL (ADF4351) + 2-ch DDS (AD9912) board comments?
« Reply #2 on: June 05, 2020, 04:57:04 am »
1. I will probably feed the LDOs with about +5VDC from an AC/DC power-brick. I don't want to make this board more complex with a buck IC at this moment..

2. yes, in fact they are mostly 0805-sized and smaller parts might be better. I will think about going to 0603 at least!

3. yes, this was shown correctly on the AD dev-kit board but I had not copied the component values correctly - now fixed.

4. I guess I can just not populate the 'across'-resistor between the differential pair.

thanks for the comments!

Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo