Things We Should Not Do In Future Radios, (Future Designs Should Not Include Past Mistakes)

Wireless technology is a shining example of a disruptive innovation that has changed society in remarkable ways. The innovation has altered how people communicate, how people access information, how people are entertained, and how people conduct and schedule their social lives. Every human activity advances and grows through a number of influences. One is experience, one is market forces, another is effective education, and yet another is common wisdom. Common wisdom is entrenched perspectives and levels of understanding accepted by the community as guide posts of the process. In fact there are many examples to be found in the wireless community of common wisdom being faulty. Samuel Clemens’ comment “It ain’t what you don’t know that gets you in trouble, it’s what you know for sure that just ain’t so” The wireless community is not free of entrenched faulty common wisdom which is passed on to successive practitioners of the art. Universities are just as liable as industry for not examining and questioning common wisdom. In this presentation we examine the evolution of wireless technology from the early days through now and show how a number of wisdoms can be shown to not be wise but never-the-less have become entrenched in the fabric of our wireless technology

What this presentation is about and why it matters

Fred Harris surveys “things we should not do” when building next‑generation radios. The talk is a practical critique of entrenched design habits—from analog I/Q downconversion and poor pulse‑shaping choices to windowing, filter design, finite‑precision effects and FFT implementations—and it shows how those habits cost you dynamic range, spectral cleanliness, synchronization robustness and battery life. If you design or implement radios, baseband DSP, spectrum analyzers, or DSP hardware, these are not just academic points: they directly affect EVM, adjacent‑channel leakage, PLL lock range, ADC effective bits, and the real-world performance of commercial wireless systems.

Who will benefit the most from this presentation

• RF and baseband DSP engineers implementing receivers and transmitters (SDR designers).
• System architects who choose standards, filters, and tradeoffs for products.
• Firmware and FPGA developers concerned with finite‑precision effects and accumulator growth.
• Graduate students and educators who want practical, historically informed lessons beyond textbook idealizations.

What you need to know

To get the most from the talk, be comfortable with these core ideas and the basic vocabulary of radio DSP:

• I/Q downconversion: how multiplying with cosine/sine creates complex baseband (I + jQ) and why gain/phase imbalance leaks energy between positive and negative frequencies (images).
• Image rejection and correction: imbalance causes mirror images to fold into your band; you can detect and cancel them with correlators/adaptive cancellers, or avoid them by digitizing earlier.
• Pulse shaping and Nyquist filters: the usual raised‑cosine (cosine‑taper) has bad sidelobes and in‑band ripple once you window it; Harris taper (a modern alternative) can reduce EVM and sidelobes dramatically for the same length and transition width.
• Excess bandwidth vs. synchronization and PAPR: narrower excess bandwidth reduces spectral waste but increases filter length, PAPR and weakens the energy feeding PLL/time recovery (“the genie”). Wider excess bandwidth often improves lock performance and reduces required PA backoff.
• Finite‑support windows and time–bandwidth: the Gaussian minimizes time–bandwidth only for infinite support. For finite windows use prolate spheroidal (Slepian), Kaiser or optimized Remez windows for better main‑lobe width versus sidelobe tradeoffs.
• DFT/FFT, finite arithmetic and scaling: accumulation and scaling introduce noise; don’t scale inputs or coefficients to hide filter gain—use extended accumulators and keep coefficients at full scale to preserve dynamic range (truncation increases noise floor ≈5 dB/bit in practice).
• Overlap and polyphase spectral analysis: good windows with low sidelobes require higher overlap (up to 75% for very low sidelobes). Implementing folded polyphase (N:1 folding) gives the time resolution of a short transform and the spectral resolution of a much longer window efficiently.
• FFT algorithms and prime‑factor/Winograd methods: Cooley–Tukey is standard, but prime‑factor (Good–Thomas) and Winograd short‑convolvers can eliminate twiddle tables and greatly reduce multiplies—important for low‑power devices and very large transforms.

If you want quick math reminders: the RMS time–bandwidth product obeys \(\sigma_t\sigma_f\ge 1/2\) (equality for an infinite Gaussian). And small phase errors satisfy \(\sin\theta\approx\theta\) for correction approximations in I/Q imbalance analysis.

Glossary

• I/Q imbalance — gain and/or phase mismatch between the in‑phase and quadrature analog paths that causes images and constellation tilting.
• Image frequency — an unwanted mirror component (negative frequency) folded into baseband by imperfect downconversion.
• Square‑root Nyquist filter — transmit/receive half of a Nyquist pulse that together yield zero ISI when convolved.
• Raised‑cosine (cosine taper) — a traditional pulse shape whose straightforward windowing often produces undesirable time‑domain echoes and in‑band ripple.
• Harris taper — an alternative pulse/window design Harris describes that reduces in‑band ripple and sidelobes for the same filter length.
• Excess bandwidth — the extra spectral width beyond the symbol rate used to shape pulses; trades spectral efficiency for synchronization robustness.
• PAPR (peak‑to‑average power ratio) — the ratio of instantaneous peak power to average power; larger PAPR forces PA backoff and reduces range.
• Matched filter — a filter matched to the transmitted pulse that maximizes SNR for white noise and is the optimal energy collector.
• Window function — the time‑domain taper applied to finite data blocks (Hann, Blackman‑Harris, Kaiser, prolate spheroidal, etc.) that controls sidelobes and main‑lobe width.
• Remez / Parks–McClellan — an FIR design algorithm that produces equiripple filters; can be modified (1/f penalty) to produce decaying sidelobes that avoid aliasing harm.

Final note

Fred Harris brings decades of hands‑on experience and entertaining clarity to practical DSP pitfalls. The talk mixes concise theory, concrete experiments, useful heuristics and historical perspective—plus a healthy dose of humor—so you’ll come away with actionable rules (and counterintuitive warnings) you can apply the next time you design a receiver, pick a pulse shape, or implement an FFT on hardware. If you care about real‑world performance rather than idealized textbook answers, this lecture is well worth your time.

This overview is AI-generated from the session transcript. Spot an issue? Let us know.