A two ADCs HF/50MHz direct sampling SDR transceiver with OpenHPSDR v2 compatible protocol.

Some time ago I designed an SDR module for my other projects. The module has two RX inputs (with DVGA and 14bits AD9255 ADC) and one TX output (with 14bits AD9744 DAC). There are also Cyclone 4 FPGA EP4CE22E22, configuration memory and all necessary components for clocking and power it up. The ADCs are clocked at 77.76MHz and the DAC is clocked at 155.52MHz by the low noise ABLJO-155.52MHz VCXO.

The goal of this project was to build a base for the OpenHPSDR compatible SDR transceiver based on that SDR module. The FPGA firmware is based on the OpenHPSDR Angelia code, the NCO code is from the HermesLite2 project. There were many changes in the code to fit 4 DDCs into the relatively small and low pin count EP4CE22E22 FPGA, some changes were requied because of the different ADC/DAC sample rates.

The second board contains Eternet PHY, standard ALEX, OC, keyer and PTT interfaces, four analog inputs, diagnostic LEDs, switching regulator (so it can be powered from the single 12V supply). There is also STM32F072 MCU on the secons board.

You need to add some RX filtering and TX amplifier with LPFs to get a full featured transceiver.

There are some limitations:

• The maximun supported output samplerate is 192kSPS
• There is no audio CODEC on the boards
• The Ethernet connection has 100Mbit/s speed
• The ADC opaerates on the second Nyquist zone on 50MHz band. The board has LPF filter with 65MHz cutoff frequency, so additional selectivity is needed to avoid images reception. It can be as simple as switchable 30MHz LPF and 50MHz bandpass filter.
• There is no TX power amplifier at the DAC output - just an LPF filter-diplexer. So, you will need some amplfication/filtering in the TX path.
• Altera USBBlaster or similar JTAG adapter is needed for FPGA firmware update

The SDR operates with the SDR Console v3 and Thetis.

The receiver BW was 500Hz during the measurements.

The usual method of determining IMD3 receiver performance does not give useful data when testing direct sampling receiver (because of IMD products does not follow cubical law). So, the IMD3 performance data presented in graphical form showing IM3 levels depending of the test tones levels for the different attenuator settings.

AngeliaLite uses different ADC/DAC sampling frequencies, so all decimation stages and filters were redesigned for the new decimation ratios.

LPF filter at the SDR module input rejects ADC aliases. Only signals in the 1st and 2nd Nyquist zones passed. As you can see some additional filtering is required to separate signals in the first and second zones and improove rejection in the third and higher zones. Here is the LPF response:

When the signal is decimated aliasing occurs, so some filtering is needed. CIC filter provides such filtering for the first 9x decimation. The worst case alias rejection is 116dB. Here is the filter response:

The second decimation stage and one more CIC filter. The worst case alial rejection is 121dB (for 5x decimation 96kHz bandwidth). Here is filter response:

The aliases of the last decimation by 9 are suppressed by the FIR filter. AngeliaLite uses relatively small and inexpensive FPGA, so last stage FIR filter is optimized for the minimum logic/memory FPGA resources use. The performance is limited by the 18bits FIR filter coefficients, but rejection is still at a respectable 100dB. Here are the calculated ideal filter response (blue line) and the real one with quantized coefficients (red line):

SDR module

AngeliaLite mainboard

73! Oleg UR3IQO