23cm / 13cm Upconverter

2024

Parts

RF Parts

Mixer

This mixer has a LO buffer. It requires output matching the examples for which are not explicitly given for either band, but s-parameter files are available which allow for band specific matches to be obtained. I have used QUCS (Quite Universal Circuit Simulator) in the past, but I had trouble setting it up in Linux. It turns out that the original QUCS project is kind of dead but that people are still maintaining QUCS-S and QucsStudio. I ran into abstruse errors that seemed to be dependency related with QUCS-S, but I was able to get QucsStudio running in Wine on my first try. I did have to play with the wine configuration settings to get transmission lines to simulate: http://qucsstudio.de/forums/topic/linux-support/

I tried a couple broadband matching ideas, and I tried a dual matching idea, but a 1GHz wide match isn't a trivial problem. I ultimately decided to go for a single band match to be determined by component stuffing options.

LO

For initial testing, I can use a tinySA for the LO source. ADF4350 is a good frequency synthesizer for more precise work involving a highly stable LO. The tinySA puts out a -27 to -18dB signal depending on the mode, so it needs to be amplified before being applied to the mixer. MAX2671 does have an LO buffer, but it requires an input in the range of -10 to +5dBm. The BGA2802,115 should do the trick here. -27+32=5dBm. -18+32=14dBm. Absolute max LO input power is +10dBm, so I'll throw a 6dB attenuator in front of the PA to make the setup safe against operator error.

IF

I'd like to use this mixer with SX1262 based transmitters which means that I will need to attenuate and limit the input to protect the mixer. The P1dB compression point of the mixer is -6.4dB, and it has a conversion gain of as much as 11.9dB, so with a 10dB backoff from that I probably want to drive it with -30dBm (which happens to be the datasheet typical IF power value). Since I plan to use this with SX1262 based transmitters with +20dBm maximum outputs, I will want to both attenuate the input considerably. I don't strictly need a limiter since I will have to attenuate by 30dB of to get to -30dBm at the input to the mixer (assuming a nominal 0dBm input). Since the absolute maximum rating of the mixer is +10dBm, this means that I would be able to apply +40dBm (10W) to the input before causing damage. I decided to add a limiter anyway since I haven't had a chance to play with them before. The lineup is: 10dB resistive attenuator -> limiter -> 20dB resistive attenuator.

Amplifier

Since the 1dB compression point of the mixer is -8.3dBm and because I will be operating 10dB backed off from that, an amplifier will be needed to produce a reasonable RF output power. I opted for a two stage solution for simplicity. Assuming an operating point of -18dBm at the output of the mixer and 32dB of gain from the first stage amplifier (BGA2802,115), I would have +14dBm which is well over the P1dB of that amplifier (+8dBm). To resolve this, I will add a 10dB attenuator before the first stage amplifier. I could further attenuate IF, but that would result in 10dB more LO leakage power at the input to the first stage amplifier. I plan to filter after the first stage amplifier which will throw in 1.5dB additional loss for a total of -18 -10 +32 -1.5 = 2.5dBm at the output of the first stage amplifier. The second stage amplifier (GVA-84+) adds enough gain to drive it to its P1dB at a nominal IF input power to the board of 0dBm which is a good place to be since there will likely be unaccounted for losses in the design. I can always further attenuate the input.

Both of these amplifiers are wideband internally matched devices because I don't feel like matching non-linear devices for a personal project and the only real impact of using them is const and power consumption. I can always add an external amplifier if 100mW isn't good enough for me.

I also listed SST12CP33SST12CP33 as an amplifier option for the 13cm band. I may play with it in the future and I suspect it will work fine at 1.35GHz even though that is a bit below what it is characterized at. It provides 10dB more power than the wideband solutions.

Power

Everything can run off of 5V, so I made the board USB powered for simplicity. I designed it into a black extruded aluminum case since its easy to do that even though this is just a test board.

Two Iterations of JLC PCB and OSHPARK SMA Transition Test Boards