Tiny Ten DSP-Based HF Transceiver | Hacker News Reader

Tiny Ten DSP-Based HF Transceiver

janbob.com

71 pointswglb19 comments

I loved the article! Even though I only understood about 1/3 of the technical details, I was honestly quite amazed at how “simple” was producing RF! The guy is using “sound” at 48kHz to produce the signals needed and then a Si5351A (signal generator for up to 200MHz or so controlled via I2C) to produce RF. I’m not sure, or don’t understand yet how the RF is being modulated, but conceptually is understandable! DSP is a field I want to dig deeper into, specifically DSP as it pertains RF.

The reference signal generated by Si5351A is used to clock two flip flops which I believe are used to generate a 90° quadrature local oscillator (LO) which enters a dual 1-4 multiplexer acting as a Tayloe quadrature mixer. And as you know, the nonlinear switching action introduces frequency products and allows frequency conversion.

This is a pretty clever system. The RF is low enough frequency that faster logic chips can be used effectively. I recommend getting yourself a cheap receive-only SDR for around 15 bucks and play with GNURadio, if you want to get more familiar with DSP for telecommunications/radio detection

Thanks! I think the limitation is in my knowledge and not in the article or schematics!

I have a rtl-sdr and I’ve played with gnu radio, but the signals that I had to explore were very limited! Apart from few remotes for fans and stuff around the house I only had fm radio and little more.

What I find very interesting from this is the possibility of taking a Si5351A breakout board, configuring it with I2C (I can do that), modulate it with something like audio and a mixer (I can code that… maybe) and plug that directly, likely through and attenuator, to gnu radio!

It would certainly be a lot easier to get a sdr transmitter, but I wasn’t aware of the possibility of modulating a Si5351A with a “simple” signal like 48kHz audio!

I have been wondering if/when we could get a direct synthesis SDR transmitter. Obviously the amplifier and filter stages will have to be discrete & analog, but to me it seems more elegant that you can software define any radio signal below 1Ghz or so and pass it to an amplifier.

That doesn't detract from this project, which is quite an achievement for a single hobbier.

After working at a company using SDRs, I realized that a well performing SDR is a lot more than an antenna, a DAC/ADC, amplifiers and a processor. You still have analog paths you need to worry about and for many applications those aren't going to cover your 1GHz range. You need to worry about noise, amplification, signal fidelity, etc. You still need a good portion of what makes up a traditional radio. Sure, there are narrowband, low power SDR xcvrs and xmitters but that's not the same thing.

That said, a long time ago I made a test fixture for a 6-bit, 12GHz DAC and it was cool to run it as an arbitrary waveform generator.

The popular IC-7300 transceiver from Icom does that. Receiver samples the whole band and does downconversion in the digital domain (DDC) and same for the transmitter (DUC).

On the hobbyist side there is the TRX Wolf by UA3REO.

Sure, just do the Zeta SDR schematic in reverse, but like others have said you'd have to figure out the rest of the chain as well as a tone of filtering. I've transmitted with direct digital synth chips as well as rpitx which is literally bit-banging RF off of a pin on a raspberry pi with NTP sync even. Again, still had to do a lot of filtering, but with wsprrypi I was able to be heard one time in New Zealand from Ohio, USA without any amplification.

How much RF power is actually coming out from the RPi transmitter? A few mW?

Yes maybe a couple of hundred milliwatts at most.

Naturally there needs to be filtering and some sort of analog frontend. My main point was that mainstream rigs now do direct sampling, instead of zero-if or superhet with dsp IF.

And WSPR sure is magic, people do thousands of km with it off an rp2040 gpio pin.

Yes sorry for the double reply I didn't know they were using the rp2040! Thanks for mentioning that !!

I'm not sure this is what you have in mind as far as "mainstream" rigs or architecture, but the FTDX-101D main receiver is a "narrowband SDR" system, as Yaesu describes it. Signal is downconverted through an analog chain including a narrow crystal filter before it reaches the 18 bit A/D+FPGA. And yes, it does perform very well, currently at the top of the "Sherwood" Third-Order Dynamic Range ranking, where it's been for about 5 years now.

If I'm interpreting your view correctly, then I share it: there is a lot of performance to be had with a well designed analog front end, and the specs of the digital parts can be quite modest and still perform extremely well.

Quite a few radios are direct-- no intermediate frequency, just a set of band selection filters. The direct route does not quite hit the absolute state of the art in dynamic range, but it's close. IC-7300, IC-7610, Elecraft K4, etc. all have this design. I wouldn't be too surprised if one more generation of ADC improvements take it over the line. Certainly there is a lot more engineering might being put into better ADC designs than there is for superhet radios.

Beyond being simpler on the analog side the direct conversion route makes spectrum/waterfall views of the whole band trivial. In theory radios could use this wide bandwidth for improved impulse blanker performance, but I'm not sure which if any do.

Some (like the flex radio, and Elecraft) will do multiple in-one-band simultaneous receive with on each ADC using it... but (other than flex perhaps) the advantages of direct converting a whole band are currently under-utilized. Once they are I doubt there will be much interest in superhet even if the dynamic range isn't quite matched.

I think the future of radio designs though will be finding ways to move the RF portion closer to the antenna(s), away from RF-noisy buildings, and avoiding expensive, lossy, and annoying coax runs.

Sure, just do the Zeta SDR schematic in reverse, but like others have said you'd have to figure out the rest of the chain as well as a ton of filtering. I've transmitted with direct digital synth chips as well as rpitx which is literally bit-banging RF off of a pin on a raspberry pi with NTP sync even. Again, still had to do a lot of filtering, but with wsprrypi I was able to be heard one time in New Zealand from Ohio, USA without any amplification.

> Again, still had to do a lot of filtering, but with wsprrypi I was able to be heard one time in New Zealand from Ohio, USA without any amplification.

Hope you got a QSL card out of that kind of DX effort. What antenna did you use?

I would imagine he was talking about running WSPR.

Yes WSPR ... The acknowledgement of contact was online. I had an MFJ antenna tuner but it was just a long strand of wire lying over the roof maybe 14 gauge? Slim. Something like 100mw output power? 200mw Maybe? It ran for a couple years unattended and was regularly heard east of the Mississippi and sometimes with some regularity in Europe.

We've been in the age of direct RF for a while. The new cellular base stations that Nokia makes are literally eight RFSoC chips connected to 64 power amplifiers. You can get one for yourself for pretty cheap: https://www.realdigital.org/hardware/rfsoc-4x2

Here's another: https://www.4sqrp.com/T41main.php - I just bought one that someone else had assembled. It's 100% open so you can buy boards and make your own, but the kits themselves are no longer available. I am thinking of ordering a set to take the design further, as a learning exercise.

Reply on news.ycombinator.com