23cm Transverter for Quansheng UV-K5

Introduction

Many SOTA operators will tell you that the Quansheng UV-K5 is the best $25 they've ever spent. Especially with the 2026 SOTA challenge, projects like NR7Y's hardware mod and custom firmware, and the Brass Knuckle Gang taking over summits across the country...it's never been a better time to get a hacked UV-K5 on the air and make unusual contacts!

I got interested in the fascinating world of microwave engineering after building an ADS-B antenna and microstrip filter. I was hooked and saw no better learning opportunity than finding a way to get a Quansheng on the 23cm band (1296MHz). 

The 23cm band is really fun - the antennas are small enough that you can have large gain in a small package, but large enough that the precision of basic hand tools still works. The scarcity and expense of the equipment makes it a difficult band to use, though. Even used equipment like an Alinco DJ-G7T regularly sells for >$500+ on eBay. 

I hope that this project can make the band a little bit more accessible to hams.

Design Overview

Schematics and layout for my design can be found here. 3d models of the enclosure and Yagi antenna support can be found here.

The UV-K5 with modified firmware is capable of transmitting on the 23cm band, but performance is not good - the power amplifier output filter rolls off well before 1GHz. I have not measured this myself, but some write-ups show power output as low as -60dBm on the 23cm band. It does appear that some people have worked on hardware modifications to get higher output power (link1, link2). The low output power or need to modify the radio are the reasons I opted to design a transverter.

I modeled my block diagram and design principles after W1GHZ's excellent designs ("Gain is cheap"). One of his 23cm transverter designs can be found here. My design got more complicated than W1GHZ's as a result of a few factors:

• I wanted a single antenna output, not separate TX and RX ports. Needed to add a T/R switch onboard. I chose a quarter-wavelength T/R circuit for its isolation.

• A 1W output power target meant that the ERA-5 MMIC in W1GHZ's design is not sufficient. Needed much more transmit gain.

• The UV-K5 outputs at least a watt, so attenuation is needed on the IF input in the TX direction, and then gain to overcome the attenuation in the RX direction.

• A hairpin filter would be too large to fit in the form factor that I wanted. I designed a two-element coupled resonator filter for the output BPF, and used a SAW filter on the output of the mixer. The microstrip filter was necessary for high power. I started with the Marki Microwave calculator, then designed these filters after seeing K5TRA's design. Lacking a good design reference I found the correct geometry through trial and error.

To generate the LO, I wanted to avoid using the familiar SI5351, partly to change things up, and partly to have a software-free project. Rather than use frequency multipliers and filters to accomplish a 1150MHz oscillator with +7dBm output, I tried using a 50MHz TCXO clipped-sinewave output as the base and repeatedly filtering/amplifying through a SAW filter. This works OK as long as the SAW filter passband is less than 50MHz, and has good attenuation in the stopband. I was able to find an 1150MHz SAW filter from TAI-SAW at the "Refuse Counterfeit goods, Only Brand New And Original Store" on AliExpress.

I limited the design to only using MMICs. This was my first project at these frequencies and using MMICs made it manageable. On a future design I'd like to learn how to impedance match to a higher power transistor.

I didn't want this project to be a mess of cables or require external batteries, in order to retain a compact handheld form-factor. I opted to steal power from the UV-K5 battery's charging contacts.

Development Process

Here's a few notes on how I arrived at the current design.

I focused on modularity in my design work. The above photo shows some of the smaller modules that I built with PCBs from OSHPark. Even with 4 layer boards, the small size of these modules meant it was extremely affordable to iterate on a weekly basis. I started with smaller modules integrated together with SMA connectors/cables, then moved to subsystems (e.g. IF section, TX amplifier chain, T/R switching), and then after these all worked well enough I started on fully integrated boards. My v2 integrated board was the first one that sort of worked, and my v3 board is the one I used to make contacts.

I began using Murata SWF connectors on v2 and v3 boards. These connectors break the signal when plugged in. This makes possible testing of individual circuit blocks and turned out to be a huge game-changer in debug.

I ended up on version 15 of my microstrip filter. I iterated a lot while trying filters with a sharper response, and getting my two-resonator filters to have the right center frequency. I ultimately determined that I need to get access to a microstrip filter simulation tool, and it's tough to do better than 2.5dB of insertion loss on this stack-up. For fun and bench debug I built some metal filters that have ~1dB loss...but these are larger than the transverter itself. It's possible that I need more attenuation on the 3rd harmonic, but my equipment is not rated up to these frequencies so I am not sure if these are real results. I plan to look into a low loss lowpass filter, and why the there is so much third harmonic content even before the PA.

In an earlier version of the design, I selected a 70cm IF because a higher IF should make filtering easier and there are more readily available SAW filters available for an 864MHz LO (27th harmonic of 32MHz). This turned out to be a huge mistake due to mixer products, because the LO (864MHz) is close to an integer multiple of the IF (432MHz). Mixer products therefore tend to land near the IF, for example f_RF-2*f_LO=432MHz.

I fried a handful of GVA-91 amplifiers on earlier versions of the design. On v1 and v2 boards, I had a parasitic oscillation on the transmit chain that could be stopped by breaking the RF receive amplifier loop. The clear conclusion is that my T/R switch had insufficient isolation for the ~40-50dB of transmit chain gain I put in the design. An additional PIN diode in the T/R switches (D2) resolved this on v3.

Minimum discernable signal looked good at about -120dBm with the Quansheng in CW mode, but I am not confident that I have cables with enough shielding to make this measurement properly. In order to get a practical test of receiver performance, on a couple of occasions I set up my HackRF and some amplifiers as a 50mW beacon and confirmed I could receive on top of a hill a couple of miles away from home.

The GVA-91 power amplifier needs input/output matching to 50ohm. The reference designs show matching networks for 900MHz and 2.1GHz, so I interpolated values to 1.3GHz and then manually tuned component values to maximize power output. 

I built boards by purchasing stencils for solder paste, hand-placing parts, and reflowing on a hot plate. Once organized it took me about 5 hours to place and reflow. 

Below are some plots of performance (left to right, row by row). All measurements with 8.0V supplied from external supply.

1. Spectrum at RF output (J6 connector)

2. Spectrum before microstrip BPF (J8 connector)

3. Spectrum before final PA (J9 connector)

4. IF output with no RF signal

5. IF output with -65dBm input at 1290MHz

6. Mixer output (J4 connector, installed backwards

7. U13 output (between first and second GALI-52 transmit amplifier)

8. UV-K5 transmit spectrum on Low power with CEC_0.3Q firmware

Field Test

My first field test of the radio was on South Sister, W7O/CM-001. I tried calling CQ after alerting, knowing there was a critical mass of hams in town, but this was an exercise in futility. Microwave contacts need to be scheduled...

The next day at the W7O campout, NR7Y and K7AHR were kind enough to spend their time helping me test the rig. My first microwave contacts were with them at camp over about 6ft distance, and later in the day we had a S2S contact on 23cm (both FM and CW) from Mt Bachelor to Lookout Mountain, about 12 miles. The CW tone was somewhat drifty but still copyable on both ends. Success!

Update: v4

The above write-up corresponds to "v3" of this design. These are the major changes I made:

• Removed one GALI-52 in the RF TX chain. I found +60dB RF gain to be unmanageable in this design. I had to put a fair amount of effort into maintaining isolation to avoid oscillation, and also created a lot of harmonics from overdriving one of the GALI-52's. I lowered my goal from 1W to 500mW as part of this.

• Removed the GALI-S66 used as an IF RX amplifier. I found that the RF RX amplifiers were both necessary for MDS performance, but the IF RX amplifier was just adding noise.

• Separated RF RX and TX filters, changing from the microstrip coupled resonator to a SAW filter on RX and a LTCC lowpass filter on TX. I really liked the microstrip filter but it did not provide as much rejection and was a couple dB lossier.

• Added an undervoltage lockout IC.

I was hoping for closer to +27dBm output, but excessive mixer conversion loss put the output closer to +23dBm. I believe this is a result of compression on the mixer.

The RF output on v4 is much cleaner.

Update: v5

Major changes for v5:

• Swapped the LNA from GALI-S66 to PGA-103. At low battery voltage, GALI-S66 had insufficient bias current and noise figure fell off a cliff. Powered from a 3.3V regulator, would be better from 5V to reduce loss in LDO and increase gain.

• Dropped corner frequency of IF filter to 160MHz from 200MHz. It was not attenuating 2*F_IF well.

• Increased resistances in PIN T/R switches. Lower bias current, but less power absorbed in the biasing resistors.

I plan to experiment with PHA-23 instead of GVA-91, since there are pads to support both on this board. This might get +1dB of output. It also would eliminate the narrowband input/output matching, making the design one step closer to supporting other bands with changes to the clock and SAW filters. 

What's Next / Lessons Learned

Next up: more field testing, and building more units to get additional tester feedback.

In hindsight:

• Better thermal isolation on the LO may improve drift. TCXO instability results in drifty but usable CW tones

• The LO is not as clean as I'd like, and a PLL might perform better. Unwanted 50MHz harmonics in the LO output are present at both the RF and IF ports.