LMX2572: LO application and clocking

Jon,

A few general comments:

• As far as the IF and LO signals needing to be "in-phase", this is more or less correct. You do not want there to be a frequency difference between your IF signal and LO signal on any detectable scale, or else the output signal will have unrelated FM noise. The most common way to make this happen is by frequency-locking the reference source of the system generating the IF and the system generating the LO. In some cases, this means providing a common reference to both systems; in other cases, this could mean generating multiple frequency-locked references for multiple parts of the system, all from one common frequency reference (often this is done with a PLL).
  
  
• Almost all frequency-related test equipment such as an AWG will have a reference port for a 10MHz signal. You can use this 10MHz signal as a frequency reference to the LMX2572, which will act as a phase-locked loop and keep your LO signal in-phase with the signal generated by the AWG. There are some non-ideal considerations such as the error from the flicker noise of the AWG reference, or the change in effective propagation delay of the signal across temperature (this shows up as very low frequency wander, think << 1Hz), which could impact the phase alignment or the noise performance, but these are usually not a big deal - test equipment usually uses an ovenized crystal oscillator for a very stable reference with low flicker noise, and you can probably start with a test of the system at a consistent temperature to avoid the wander effects above. Under normal circumstances, this kind of PLL scheme for the LO will retain the same overall stability as the AWG (to the limits of the flicker noise inherent to the PLL architecture).
  
  
• It's probably better for LMX2572 performance if you provide a higher-frequency reference e.g. 100MHz, since the in-band performance of the PLL will be to some extent a function of the phase detector frequency (for a more detailed investigation, consult PLLatinum Sim). Sometimes you get lucky and the test equipment also provides higher frequency reference signals, but often you are left with just 10MHz available. You might then benefit from a multi-stage PLL architecture that can lock frequency to the 10MHz, but step it up to a higher-frequency clean reference from a tunable oscillator such as a VCXO. As an example, the first PLL could lock to 10MHz and use a 100MHz VCXO as the VCO. Then, the output of the VCXO could be split or duplicated such that it is used both as first PLL feedback and as second PLL reference. The second PLL (LMX2572) could use the 100MHz reference instead of the 10MHz reference to improve the phase noise of the synthesized LO within the PLL loop bandwidth. 
  
  This kind of cascaded PLL approach is not strictly necessary just to get started on a project like this, but as a long-term optimization for better noise performance it could be helpful to remember this option (in case you find unacceptably high phase noise close to the carrier, contributed by the LO signal). I'll leave the task of finding a suitable PLL for this purpose to you, if you find it necessary - the best device for this purpose may not be a TI device.
  
  
• LMX2572 outputs are inherently differential. There are not too many differential mixers on the market. Consider using a balun to convert the output of LMX2572 from a differential signal to a single-ended signal for better compatibility with most mixers.
  
  
• You could also power-combine the output of two LMX2572 outputs to boost the LO power. The LO port is typically the highest power port in the mixer for spur performance reasons; increased power from two combined outputs instead of just one may be useful in omitting the need for a low-noise amplifier before the LO.
  
  
• No matter what, you need a SPI controller for the LMX2572. This could be a simple microcontroller with a SPI peripheral, a SPI programming dongle like an Aardvark, or a simple SoC like a Raspberry Pi which exposes its SPI interface. We include a SPI programmer for use with the LMX2572EVM.
  
  
• We also provide a GUI-based tool called TICS Pro which offers a graphical method of deriving the register programming for the LMX2572. You can check in advance if the configuration you want to generate is supported, and the tool will suggest sensible settings for many use cases. In combination with PLLatinum Sim, you can replicate the register programming from a PLLatinum Sim model, or simulate the performance of a proposed configuration.

Hopefully this helps!

Regards,

Derek Payne