Thank you KE5FX for insightful suggestions
We are not getting -125dBc/Hz PN out of ADF4356. We used ADF4356 + LTC6945 combination as the main PLL plus HMC830 as the tunable reference generator to get -120 dBc/Hz in our current revision. At that time, LMX2594 wasn't around (we started this project in August 2016). Just before the launch of our campaign, TI released LMX2594, and it was scary good (*). We were going to do one final revision anyway, so we decided to use LMX2594 as the main PLL as well as the tunable reference generator. We purchased two copies of the LMX25945 the eval board and tested to see everything was inline with the datasheet, and it was. Here is the block diagram of both current revision and the final revision of ERASynth:
http://erainstruments.com/erasynth/erasynth-current-block.pnghttp://erainstruments.com/erasynth/erasynth-final-block.pngThe phase noise graphs shown on the campaign page were taken with our current revision. Hence, tunable ref is HMC830, VCO is the VCO of ADF4356 and Int-N PLL is LTC6945. You may see the PN graphs here:
https://www.crowdsupply.com/img/35ff/erasynth-pn-10ghz.pnghttps://www.crowdsupply.com/img/f040/erasynth-pn-1ghz.pngThese are respectable numbers to squeeze out of ADF4356. Yet, we are targeting another 5 dB improvement (in the pedestal region) over these numbers in the final revision.
We will be using a DDS to go down as much as possible in the plus version in the final revision, we are planning on using AD9913. We will consider adding an LF output. It will require an extra SMA connector, which may not be possible. We will see what we can do when we do the final layout.
There is a model on the campaign without the OCXO, but it only goes up to 6 GHz. However, both of the ERASynth models also accept external 10 MHz reference signal. If you have your own more stable 10MHz signal you may use it. Default inclusion of the OCXO just increased the cost a little bit, but people who doesn't have their own ultrastable timebase will appreciate it.
As for ALC, it is really beyond the scope of this project. Open loop calibration doesn't mean that the amplitude accuracy will be all over the place. At the moment, we are doing tests to see how our open loop calibration performs. We will publish the results as an upcoming campaign update. Please follow
our campaign page for updates.
(*) LMX2594 has the best flat phase noise FOM -236dBc/Hz in the industry at the moment. However, LTC6945 is still better when it comes to flicker FOM (Flicker FOMs: LTC6945=-274dBc/Hz vs LMX2594=-269dBc/Hz).
Hi everyone,
We are running our crowdfund campaign ERASynth on CrowdSupply. We would like to see your valuable feedback and recommendations. Your comments are very important since we are still in the process of taping out last revision of the board.
Thanks in advance.
edensrock
Hey, this really does look cool. -125 dBc/Hz at 10 kHz from a 1 GHz carrier out of an ADF4356 is pretty impressive. It's good to see so much technical detail up front. Good luck with the Kickstarter!
Some brief thoughts:
- If you use a DDS anywhere in the design, consider adding an LF output path for it. It'd be nice to offer 9 kHz-15 GHz coverage, or even lower. Not a huge deal but occasionally useful.
- Offer a version on your Kickstarter without the OCXO. Many of your customers will have their own stable 10 MHz references available, so it's a waste of money to include OCXOs at the instrument level.
- Lack of ALC is probably the biggest Achilles' heel I see. Calibrated and leveled output power is pretty important. I'd suggest prioritizing this in the next spin.
- LMX2594 just came out. Could be worth looking at, but it sounds like you're getting similar performance from the ADF4356 already, so maybe not.