K8TRC Digital SDR

This is the block diagram for the receiver portion of the radio. I've filled in some parts I'm going to try and some hopeful gain numbers. I'm going to try a direct conversion receiver first and see if I get usable data for the low signal modes.

First, the antenna goes right into a bad pre-selector. This will likely be a changeable component. It'll be pretty tight for the initial digital mode implementation, just the CW portion of the expected band. This should drop the noise figure significantly. It will also help with image rejection if I end up having to change over to superhet.

The RF pre-amp will use the monolithic RF amp and should provide on the order of 19 dB of gain which will split to the two mixer sections, one for in-phase and one for quadrature. I'm factoring in 3 dB loss here which should be a good enough for ball park estimates.

I'm going to try an SI5351 a shot. It's a clock generator so there will be some noise introduced because of the square wave harmonics but a lot of people have had good results. It's easily controllable with a microcontroller and it's possible to set the phase of the output signals allowing for direct generation of the I and Q LOs. I'm not sure what the phase distortion will be so this will be an experimental point.

I'm estimating the power at this point will be between -46 and -67 dBm given a 50 to 100 µV signal from the antenna. I sampled a general purpose op-amp from TI. The TSH82 is billed as a high bandwidth video op-amp. It's got a minimum large-signal voltage gain in the 70+ dB range so it should be able to give the 60 dB I think I'll need to reach 1.1 V at the ADCs. This will need to have some trim resistors to balance the I-Q signals but otherwise it should work pretty much as-is. These are only $3 or so in single quantities and they are dual devices so only on IC should be needed.

Initially I plan to use a STM32F103 breakout board, sometimes known as the Black Pill. It's very inexpensive, about $2 a piece from Chinese suppliers. The ADCs are 12-bit and can push 2 Msps, although a more realistic number for this application is probably about 500 ksps which is still about 5 times the 96 kHz I'm going to try and sample. If I lower the sample rate and pull a much smaller chunk of the spectrum around the digital channels, I can oversample and clean up the signal significantly. The F103 has some ADC modes that support DMA and synchronized sampling so I can ensure the I and Q are properly aligned. I'm planning on using the USB audio profile to publish a 96 kHz stereo signal for processing in SDR software. The middleware from ST doesn't support the newer USB audio profile with higher sample rates but I don't think it will be a problem. The plan is to eventually take the PC out of the equation and do all the processing on the micro.

The F103 doesn't really have the processing power to handle the DSP functions, but there's a STM32F373 device with an M4F core with ARM DSP extensions. It's also got 16-bit ADCs for increased sampling resolution. I found a breakout board on Oshpark for $10 but it's just the board. I grabbed a couple of samples of the micro as well (I think it's about a $5 or $6 part normally in single quantity). This probably puts the full board in the $20-$25 range. Another option is to get a Teensy 3.6 for about $10 more. It's a proven SDR board with a similar Cortex M4F. It has an easy to use Arduino DSP library and it's already assembled. It'll push the cost about $50 for the radio but might be worth it anyway.

I'm just waiting on parts to arrive to do some prototyping. If it looks like I'm getting decent amplification and noise statistics, I'll put together a proto PCB.

I've been wanting a small, low power, portable radio for working digital modes. Given the vast array of low power MMICs and other ICs available these days there must be some gold to be mined digging through manufacturer's product lists. My idea is to use low cost, modern parts, many of which are available for free in small quantities via sample requests.

As a starting point for requirements I've come up with the following list. The radio will likely support a much wider array of modes and frequencies, but this is the core requirements.

1. Support at a minimum, FT8 and WSPR modes for transmit and receive. The heavy lifting will initially be on a PC but I'd like to support fully autonomous use.
2. Support at least 40m, 30m, and 20m HAM bands but likely expand to 17m and 15m initially. Full HF, VHF, and 70cm UHF support would be great, likely changing out the front end band filters. I.E. not supporting all simultaneously, but allowing for a component change.
3. Low enough power use to run off a reasonably sized battery pack for a reasonably amount of time. A couple of amp-hours for 4+ hours should be a workable starting point.
4. Small enough to be portable for day hikes. I'd like to keep the size and weight as low as possible for multi-day backpacking in rugged terrain for more difficult to reach SOTA activations.
5. Price should be as low as possible, utilizing sampled parts when available. If I make this available as a full kit, the price should be under $100 for 40/30/20m.

I started with a Google for some mixers and stumbled on an interesting NXP part. Since I was already requesting samples I grabbed a couple more parts (NXP only allows 3 parts at a time, but multiple counts of varying maximums).

The SA612A is a low-power VHF monolithic double-balanced mixer with on-board oscillator and voltage regulator. It is intended for low cost, low-power communication systems with signal frequencies to 500 MHz and local oscillator frequencies as high as 200 MHz. The mixer is a ‘Gilbert cell’ multiplier configuration that provides gain of 14 dB or more at 45 MHz.

This is a Gilbert cell double balanced mixer that operates up to 500 MHz in a very manageable SO8 package. It's got a built-in oscillator and is available for $2.36 in single quantities from Digi-Key. I was able to sample 20 chips which is probably a lifetime buy if these work out.

The MMG3H21NT1 is a general purpose amplifier that is internally input
matched and internally output matched. It is designed for a broad range of
Class A, small--signal, high linearity, general purpose applications. It is
suitable for applications with frequencies from 0 to 6000 MHz such as
cellular, PCS, BWA, WLL, PHS, CATV, VHF, UHF, UMTS and general
small--signal RF.

This is an RF amplifier that'll go DC to 6 GHz. Small signal gain is on the order of 19 dB. It's 50Ω matched in a SOT package and is available in single quantities from Digi-Key for $2.57. I figure this will make a good RF pre-amp if it's needed given the 14 dB gain from the mixer and will also work as a pre-amp for the transmitter's power amp. I got 5 of them sampled.

The NX3DV2567 is a four-pole double-throw analog switch (4PDT) optimized for
switching WLAN-SIM supply, data and control signals. It has one digital select input (S)
and four switches each with two independent input/outputs (nY0 and nY1) and a common
input/output (nZ). Schmitt trigger action at S makes the circuit tolerant to slower input rise
and fall times across the entire VCC range from 1.4 V to 4.3 V.

Finally, an analog switch that's 4PDT. To save power and cost I'm going to avoid relays and this seemed to fit the bill. I'm not sure if I'll need 4 poles since I haven't sketched out the transceiver yet but it's the highest pole count they had available for sampling. I got 7 of them and it's available at Digi-Key for $0.62 in single quantity. It's a 16 pin QFN package which should be manageable by hand or certainly by hand placed reflow.

Stay tuned for a post with an initial design.