The Breadboard RF103

When [ik1xpv] sets out to build a software-defined radio (SDR), he doesn’t fool around. His Breadboard RF103 sports USB 3.0, and 16-bit A/D converter that can sample up to 105 Msps, and can receive from 0 to 1800 MHz. Not bad. Thanks to the USB 3.0 port, all the signal processing occurs in the PC without the limitations of feeding data through a common sound port. You can see the device in action in the video below.

The Cypress FX3 USB device is an ARM processor, but it is only streaming data, not processing it. You can find the slightly modified firmware, a driver for using PC software, and schematics and board layouts on GitHub.

We aren’t quite sure where the term breadboard factors into the name since there is a PC board involved and no solderless breadboard. The basic receiver covers 0 to 30 MHz, and a mixer downconverts higher frequencies to that range.

There’s enough information that you should be able to duplicate this project, although you’d need to do a little analysis of the material. This isn’t a step-by-step build tutorial. However, this should be much more sophisticated than the usual RTL-SDR dongle. This is more on par with the SDRPlay, although it has a  higher maximum frequency and some additional hardware filtering. It also is a 12-bit digitizer instead of the RF103’s 16-bit one which is a point in the RF103’s favor.

13 thoughts on “The Breadboard RF103

  1. I’ve read many comments about DIY/USB oscilloscopes being impractical, but could a semi-respectable USB o-scope be constructed using a derivative design? My biggest use case for a USB o-scope is portability (eg. easily fits into an EDC), thus bringing up yet another DIY/USB o-scope possibility.

    PS, the last two sentences of the Hackaday post need a little more clarification around what “it” each sentence is referencing.

    1. The biggest issue with most portable and/or low end scopes is the front-end design. Getting a DC-accurate several 100MHz flat bandwidth (you need at least 2x the target scope bandwidth for good fidelity) amplifier with high-impedance inputs built out is hard. Making it stable and equal bandwidth over all attenuation settings is harder.

      1. LOL, OK, so just because one might be able to get upt to 20-50 Mhz (from the 105 Msps bandwidth, right?) out of this setup doesn’t necessarily make it good enough for a scope? Darn.

        Guess I’m still on the lookout for a travel-ready, respectable, open-source hardware platform to function as an o-scope, a logic analyzer, and a function generator.

          1. But part of that usually requires some high quality, mass produced, cheap silicon that can be shoehorned to the required purpose. e.g. STB/automotive/2G/3G/4G/PC. It is never just the right people, it is also keeping the price low. Setting a selling price at say $100 (excluding shipping) and dialling in everything to turn a profit (resellers would want ~60% profit margin). You need ~30% profit of the remaining $40, and ~30% for fabrication, and that leaves ~$13 to ~$16 for the BOM including the PCB.

            If you look at the above parts:
            16-bit 105MSPS ADC used (LTC2217) costs ~$82 in lots of 2k
            EZ FX3 USB 3.0 bridge chip is ~$17.40 @ 2k units

            Both are great parts but blow the budget and are suboptimal for a “cheap” USB oscilloscope. An 8-bit ADC would be better, for that application, an 8-bit 200MSPS ADC ~$9.5 @ 2.5k units (ADC08200C). Maybe use two of them interleaved to double the sample rate. Instead of needing USB 3.0 it would probably be cheaper to add RAM and use USB 2.0 to trickle the data across to a PC. Most of the time you do not need to see every single sample (it is not like a 1920×1080 screen can display them all at once anyhow), and can transfer one sample for every 100 captured. Keep the price down with smarter software e.g. store all the samples in RAM for later retrieval, and only transfer across if zoomed in and if you are zoomed in then you don’t need to see the samples outside the zoomed in range, so they do not need to be transferred (at least not in realtime).

            And then you have a power budget of 500mA@ 5V for USB 2.0 (2.5W for High-power) of which the ADC has used up at least ~210mW double that if the price was increased and you decided to you two interleaved.

            Getting a product to market is seeing reuse of silicon, back of the envelope number crunching and balancing a large number of suboptimal decisions to keep the price as low as possible to ship to as many as possible. And then revision 2 of the hardware is to weed out all the really poor decisions.

          2. Looking at their website, the LTC2261-14 ($52) and LTC2255 ($49) would be my next choice.
            Dropping to 12 bits gets you down to about $30 each, in lots of 1k.

        1. AD9288 $4-10 dual 8bit 100Msps
          thats whats inside Rigol DS1052E, so definitely good enough for hobby use + can be pumped whole without any processing into the computer with minimal interfacing. As a bonus you could rip off whole DS1054Z analog frontend, its VERY cheap because it was build with discrete transistors, EEVBlog even reverse engineered the diagram

          1. I dont have a smartphone to use .io (just guessing it works on smartphones, because it sure doesnt not work on desktop browsers, usability is about 5%)

            btw there is also $140 Jinhan JDS2022A – handheld 20MHz BW with theoretical 2x200Msps
            2x AD9288-40 on board so they just overclock the shit out of them, Rigol at least used 80MHz parts at 100MHz
            + probably older lattice CPLD + crap chinese frontend without cans with hilarious 2 cm free air connections :)), but its all nicely build on a separate module, and could be adapted into a DIY project.


Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.