RFQ/Demo
Reference handbook · BNC-HB-002 · Rev A

SDR Product Handbook

Choosing a software-defined receiver or a real-time spectrum analyzer: why the market looks the way it does, what separates one class of device from the next, and how to specify without being told what to think.

Download the handbook (PDF, 57 pages) Read it in the browser Talk to an RF engineer

There are more software-defined radios on the market than any engineer can evaluate, and from a distance they look interchangeable: similar frequency ranges, similar bandwidths, and prices that cluster by class. They are not interchangeable, and the reason they are not is rarely the number on the front of the datasheet.

This handbook explains why the market clusters the way it does, starting from the silicon underneath it. It separates the three receiver architectures and the two families of converter that define them, teaches the sampling, dynamic-range, phase-noise and noise-cascade arithmetic that decides what a receiver can actually see, walks the software stack from abstraction layer to application, and ends with a vendor-neutral selection method, a specification you can send to several suppliers, and a reference section built to be photocopied.

57 pagesFive parts, 66 sections, 18 figures and 25 tables, every one referenced from the text.
Vendor-neutral methodA seven-question selection order, two worked selections, ten jobs and what actually decides each one.
Every relation evaluatedTen numbered equations, each worked to a number you can check on a calculator, with the condition that makes it true.
Current to 2026Direct RF-sampling silicon, FR3 and WRC-27, satellite direct-to-device, O-RAN fronthaul, and the EU Cyber Resilience Act.

What is in it

Part I. Why the landscape looks the way it does

  • 2. What software-defined actually moved
  • 3. How the market got here, and where it is going
  • 4. Three architectures, and what each one gives up
  • 5. The silicon underneath, and why platforms cluster
  • 6. The second cluster, and where the digitizer moved
  • 7. Where the programmable boundary falls, drawn
  • 8. Sampling, and the three numbers people confuse
  • 9. Dynamic range, and why the bit count misleads
  • 10. The two ends of the range, and the span between them
  • 11. Phase noise, and the signal you cannot see next to a strong one
  • 12. What zero IF costs, and where it shows
  • 13. The five settings that decide what a screen shows
  • 14. Processing gain, and what it can and cannot recover
  • 15. Noise, gain and the front end you attach

Part II. The product landscape

  • 16. Reading an SDR datasheet
  • 17. The classes of device
  • 18. Transmit, duplex and multiple channels
  • 19. Antennas, preselection and the part nobody budgets for
  • 20. What each class actually costs you to own
  • 21. Timing, reference and synchronization
  • 22. Form factor, environment and the things that break
  • 23. Interfaces, transport and the standards that matter

Part III. The software world

  • 24. Frameworks, abstractions and applications
  • 25. Abstraction layers, and what they deliberately omit
  • 26. The arithmetic that decides whether a system is buildable
  • 27. How a processing framework actually runs
  • 28. Samples, scaling and time: the three that break integrations
  • 29. Platforms, and where RF data goes after the disk
  • 30. Diagnosing an overrun, and testing a pipeline
  • 31. What a software lead should ask before anyone commits

Part IV. From samples to measurements

  • 32. What separates a reading from a measurement
  • 33. Real-time capture, and what gap-free guarantees
  • 34. Triggering, and how a finite buffer sees a long day
  • 35. Distributed systems, and what changes when there is more than one
  • 36. How a swept analyzer works, and the relation that governs it
  • 37. The evidence a swept trace cannot give you
  • 38. What the open world does with a calibrated front end
  • 39. Calibration, and what traceability actually costs someone
  • 40. Machine learning on spectrum, and what it inherits
  • 41. Where the ICX-FieldHawk sits
  • 42. What an instrument does that a receiver does not
  • 43. Four more measurements, each checkable from its own screen
  • 44. One engine, three form factors
  • 45. Where this is not the answer
  • 46. An uncertainty budget, worked

Part V. Choosing, and reference

  • 47. Bandwidth, and the four numbers a span can mean
  • 48. Nyquist zones, and deliberate undersampling
  • 49. Aliasing, and the spurs that are not there
  • 50. Selection, by the job rather than by the datasheet
  • 51. Ten jobs, and what decides each one
  • 52. Troubleshooting, by symptom
  • 53. Buy, build, or integrate
  • 54. Compliance, export and supply chain
  • 55. Provenance, and the questions procurement will ask
  • 56. Reading somebody else's datasheet honestly
  • 57. Measurement practice that applies whatever you bought
  • 58. Two worked selections
  • 59. What lives where: a working band reference
  • 60. Where the bands are going
  • 61. Two more movements a receiver will meet
  • 62. Quick reference
  • 63. A specification you can send to several vendors
  • 64. Standards worth knowing by number
  • 65. A reproducibility checklist
  • 66. Definitions, symbols and further reading
  • 67. What to take away

Who it is for

Engineers specifying or buying a receiver who need a method rather than a recommendation. Students and newcomers meeting the field for the first time, who need the vocabulary and the arithmetic before the catalogue. Product architects weighing buy, build and integrate. Software leads handed an integration and a deadline. It is written to be useful to somebody who never buys anything from us.

Download the handbook (PDF, 57 pages) Read the companion: SDR Integration Handbook