Generally it should generate signal in frequency range 4 to 7GHz. Some assumptions:
- Range 4-7GHz, output level about 10-13dBm (to drive a mixer), not necessarily low noise/jitter (should just work)
- The design is not necessarily optimal, it could be simplified by using some more advanced parts
- Using planar components wherever possible - to shorten BOM and for fun
- Cheap materials and technology - 1mm thick FR4, double sided board
- Mostly simple components (all but PLL IC)
- Design with open source software: Qucs, OpenEMS, KiCad and more (Linsmith, Qucs-RFlayout,...)
- The bandwidth consists of three ranges - many circuits are in three versions for the ranges, that's because low bandwidth circuits are easier to design
I've made a new revision of the project. See LO1rev2_design_pack.zip for KiCad design, various simulations, design description. It's mostly redesigned, a little simplified although still I didn't try to avoid silly design decisions so don't expect high performance or usefulness of the design.
Short summary of changes:
In general, more components are wideband, previously there were mostly three signal paths for each sub-band (4-5, 5-6, 6-7GHz)
VCOs - Changed to a normal Clapp oscillator, I had some problems with this topology in 1st revision but it seems to be working OK if parasitics are reduced (tight layout, only single tuning diode). It works (that's all I can tell about it's quality).
Attenuator for gain control - Simple 4 diodes attenuator IC. It's simpler and works better because it's wideband (previously three circuits with distributed elements).
Amplifiers - 2-3.5GHz and 4-7GHz amplifiers are similar. I used some resistive matching to avoid bad SWR which was a problem in previous design. In my previous design lossless components were used for matching so in order to reduce gain at lower frequency (to get flat gain over frequency) a lot of power had to be reflected. In the rev2 design the power at lower frequencies is mostly dissipated in resistors (kind of a simple non-reflection filter at the input). I had too much gain before frequency doubler so I added attenuator and not enough gain after the doubler so I've cut a hole in PCB and inserted 4th gain block.
Frequency doubler - It's a wideband doubler, based on marchand balun. It was quite a fun to make an useful balun on poor man's substrate, this part mostly forced substrate thickness for the whole design. Details in documentation. Insertion loss is about 10-15dB which is sufficiently good for me, considering simplicity and price.
Switches for filters - The first design in rev2 was very good in simulations but totally useless in real life (parasitics...). I made a simpler switches to reduce insertion losses (cut the old from PCB and replaced with a new one on self made PCB). The best solution would be to use a real, integrated circuit switch instead of building it from diodes (7GHz is too much).
Filters - Only small adjustments, mostly the same as rev1.
Added low pass filter at the output - not sure if it helps or not (may be even harmful because of high losses @7GHz, @FR-4
PLL and ALC similar to rev1.
Unfortunately I've not much possibilities of measurements above 6GHz (up to 6GHz I use arinst-ssa, although it has undefined performance above 3GHz), also measurements in the circuit (without cutting out a measured block or making a new copy) are inaccurate (SWR of connections matters a lot) so I'm highly uncertain about performance of particular circuits. The whole synthesizer provides about 7dBm max over the 4-7GHz range, if I can trust a 20$ power meter (I can't actually...).
I closed the level-control loop and the PLL. After a few more small changes it works.
I skipped the second part of output amplifier (was unstable) and the first part seems to work good enough. I also skipped amplifiers after VCOs and added attenuator after 2-2.5GHz pin attenuator (too much gain). Variable attenuators were too sensitive to tuning voltage (at certain frequency and tuning voltage) so I introduced some imbalance as workaround. Other changes are listed in changes_rev1 text file in the project zip.
Recently I had chance to access a good spectrum analyzer for a while so I checked the results. The output signal looks bad (spurs, jitter), partially (mostly?) because I didn't even shield anything. Also I'm not 100% sure if everything is stable and not overdriven. The main tone is about 12dBm, it drops to about 8dBm close to 7GHz but in the remaining part of frequency range the output level is surprisingly flat (maybe about 2dB flatness below 6.7GHz). See the video in "files" (note 10dB attenuator was installed).
1. VCOs - they work (after some tuning, and replacing one transistor)
2. Amplifiers (2-3.5GHz in three ranges) - they are ok, VCOs output signal is higher than expected so maybe I need one instead of two amplifiers or I must add some attenuators
3. Atternuators - they work, attenuation range about 10dB (not much but close enough to simulations)
4. Doublers - they work (2nd harmonic much above the input), no accurate measurements done yet
5. PIN switch - it does something but I can't test it separately with the current PCB layout (maybe I'll attempt it later)
6. Filters - the lowest band looks fine, it's hard to test the others (measurements errors can be bad at so high frequency and currently I can't go above 6GHz)
7. Output amplifier first part (3 stage) - not good, hard to test, seems to be unstable but possibly the problem is caused by test setup (depending of arrangement of coax cables connected to PCB it oscillates or not); possibly it can be useable after some tuning
8. Output amplifier the second part (2 way) - part1+part2 output amplifier is much more unstable than the first part only so I'll skip the two-way amplifier for now; it's bad because of huge out of band gain, especially about 2GHz where it usually oscillates.
So the plan for now is to put all together, without the final amplifier.
The PCB has arrived, now the prototype is mostly assembled.
The fist impressions:
- I checked one VCO and seems that it works surprisingly well
- I also checked one 3-3.5GHz amplifier - it's fine
- The wideband 4-7GHz seems to be fail. First of all it's unstable. Also it's quite difficult to make any meaningful measurements, even with a good VNA (I had chance to use it for a few days) - U.FL connectors are quite bad even at their specified 6GHz and currently I can't test the two stages of the amplifier separately because I don't have connector between them. The problems are likely caused by radial stubs which may couple to each other and also diverge from simulation.
I hope to make more measurements soon.
The main challenge will be to fix the 4-7GHz amplifier (at least up to 4-6GHz where I can currently measure it). The worst case will be to replace it with integrated gain blocks but they are noisy, expensive and boring.
The design of the first version is finished and I'm waiting for PCBs.
I expect that not all the circuits will work as intended and most of them will require some tuning. On the other hand I'm pretty sure I'll be able to make them to work because I did a lot of simulations and also some experiments. I tested some of the used (or similar) circuits, mostly using cheap Arinst SSA-TG spectrum analyzer (up to ~6GHz) and it looks like the simulations really help to roughly estimate how the real world hardware will work. Not all the parts were tested so problems may arise, especially at the highest frequencies.
To be continued...
Design and reports are attached (as zip file) in "Files" section.