Make Hardware Great Again
US weakness in 5G and the coming AI race stems from a deeper problem, hardware decline and lack of CPU innovation. Jeff Brower argues that the software-only narrative has hollowed out semiconductor leadership, leaving only a few chipmakers and blocking vital R&D. He calls for targeted government action, funding for neural-net chips, and an industrial Hardhattan Project to rebuild CPU and hardware capabilities.
A Fast Real-Time Trapezoidal Rule Integrator
Rick Lyons presents a compact, recursive real-time Trapezoidal Rule integrator that computes N-sample discrete integration using only four arithmetic operations per input sample. The proposed network yields a finite-length, linear-phase impulse response with constant group delay (N-1)/2 and cuts substantial computation compared with a tapped-delay implementation, making it useful for speeding Romberg-based digital filters.
Third-Order Distortion of a Digitally-Modulated Signal
Amplifier third-order distortion is a common limiter in RF and communications chains, and Neil Robertson walks through why it matters using hands-on MATLAB simulations. He shows how a cubic nonlinearity creates IMD3 tones, causes spectral regrowth and degrades QAM constellations, and gives practical notes on estimating k3, computing ACPR from PSDs, and sampling considerations.
A Narrow Bandpass Filter in Octave or Matlab
Building very narrow FIR bandpass filters at high sample rates often yields extremely long impulse responses. This post shows a practical Octave/Matlab implementation that uses complex downconversion to baseband plus a multistage Matrix IFIR and running-sum cascade to slash computation. With the provided example (48 kHz, 850 Hz center, 10 Hz passband) you get <1 dB ripple and >60 dB stopband while running 20x to 100x faster than a single-stage FIR.
Second Order Discrete-Time System Demonstration
Want a hands-on way to see how continuous second-order dynamics appear in discrete time? Neil Robertson converts a canonical H(s) to H(z), shows z-plane pole mapping for different damping ratios, and walks through impulse-invariance scaling and zero placement. The post includes a MATLAB function so_demo.m that computes numerator and denominator coefficients, plots poles, and compares impulse and frequency responses so you can experiment with sampling effects.
A Beginner's Guide To Cascaded Integrator-Comb (CIC) Filters
Rick Lyons breaks down cascaded integrator-comb filters into clear, practical terms, showing why they are the efficient choice for high-rate decimation and interpolation in hardware. The post explains CIC structure, its sinc-like frequency response, multistage tradeoffs, register-bit-width rules, and simple FIR compensation tricks. Hands-on tips and references make it easy for engineers to design and implement robust CIC-based decimators and interpolators.
Are DSPs Dead ?
Jeff Brower argues that the science of digital signal processing is far from dead, but commercial DSP chips lost momentum when Texas Instruments refused to embrace server-centric AI and 5G markets. He traces how TI's embedded-only culture, halted multicore CPU roadmaps, and lack of server-class products pushed customers to GPUs and FPGAs. A comeback would demand PCIe cards, VM and container support, open-source engagement, and bold leadership.
Digging into an Audio Signal and the DSP Process Pipeline
Zooming into an audio waveform can be misleading if you rely on only one view. This post compares Audacity with a simple C++ WAV reader and shows how the same samples can look like zero in a GUI, while the raw data reveals a small nonzero value. It is a practical reminder that multiple tools help you inspect and verify signal data more accurately.
A Simplified Matlab Function for Power Spectral Density
Neil Robertson provides a tiny Matlab wrapper around pwelch that simplifies PSD computation by preselecting a Kaiser window, default overlap, and converting units from W/Hz to dBW/bin. Call psd_simple(x,nfft,fs) to get PdB and a frequency vector, with nfft controlling whether DFT averaging is used. The post includes examples showing the effect of averaging and explains the Kaiser window processing loss.
Already 3000+ Attendees Registered for the Upcoming Embedded Online Conference
More than 3,000 engineers have already signed up for the Embedded Online Conference, and free registration closes at the end of February. Stephane Boucher highlights four practical tracks—DSP and machine learning, FPGA, embedded systems programming, and embedded systems security—and notes that every talk will be available to stream on demand from May 20. If you prefer no-travel learning or want flexible access to world-class talks, register now.
Add the Hilbert Transformer to Your DSP Toolkit, Part 2
This post shows a simple practical route to a Hilbert transformer by starting from a half-band FIR filter and tweaking its symmetry. It walks through a 19-tap example synthesized with Matlab's firpm (Parks-McClellan), explains the required frequency scaling, and shows how even-numbered taps become (or can be forced) zero through symmetry and coefficient quantization. Useful design rules are summarized for choosing ntaps.
Second Order Discrete-Time System Demonstration
Want a hands-on way to see how continuous second-order dynamics appear in discrete time? Neil Robertson converts a canonical H(s) to H(z), shows z-plane pole mapping for different damping ratios, and walks through impulse-invariance scaling and zero placement. The post includes a MATLAB function so_demo.m that computes numerator and denominator coefficients, plots poles, and compares impulse and frequency responses so you can experiment with sampling effects.
Sum of Two Equal-Frequency Sinusoids
Rick Lyons exposes a frequent trig mistake and delivers complete closed-form expressions for collapsing two equal-frequency sinusoids into a single sinusoid. Using complex-exponential phasor addition and equating real and imaginary parts, he compiles easy-to-use tables for cosine+cosine, sine+sine, and cosine+sine cases and shows how to derive each form. Engineers get corrected identities and compact derivations useful for analysis and communications.
Simplest Calculation of Half-band Filter Coefficients
Half-band FIR filters put the cutoff at one-quarter of the sampling rate, and nearly half their coefficients are exactly zero, which makes them highly efficient for decimation-by-2 and interpolation-by-2. This post shows the straightforward window-method derivation of half-band coefficients from the ideal sinc impulse response, providing a clear, hands-on explanation for engineers learning filter design. It also points to equiripple options such as Matlab's firhalfband and a later Parks-McClellan implementation.
Round Round Get Around: Why Fixed-Point Right-Shifts Are Just Fine
Jason Sachs explains why, in most embedded systems, simple bitwise right-shifts are an acceptable way to do fixed-point division rather than paying the runtime cost to round. He shows the cheap trick of adding 2^(N-1) to implement round-to-nearest, explains unbiased "round-to-even" issues, and compares arithmetic error to much larger ADC and sensor errors. The takeaway: save cycles unless your algorithm or inputs require extra precision.
Pulse Shaping in Single-Carrier Communication Systems
Eric Jacobsen clears up common confusion around pulse shaping in single-carrier communications, focusing on matched filtering, Nyquist filtering, and related terminology. He uses the NRZ rectangular pulse as a concrete example to show how the transmit spectrum becomes a sinc envelope when the bitstream has enough randomness, and he highlights how bit patterns and context-sensitive terms can change the observed behavior.
A Simple Complex Down-conversion Scheme
Rick Lyons shows a compact way to turn a real bandpass signal centered at ±fs/4 into a complex, zero-centered analytic signal. The trick uses a delay, a Hilbert transform filter, and a 4:1 downsample, with a small compensation filter to widen the usable passband. He also points out a no-multiplier implementation using shift-and-add coefficients, or a higher-attenuation version with two multiplies per output sample.
Digital Envelope Detection: The Good, the Bad, and the Ugly
Envelope detection sounds simple, but implementation choices change everything. Rick Lyons gathers common digital detectors, including half-wave, full-wave, square-law, Hilbert-based complex, and synchronous coherent designs, and explains how harmonics, filtering, and carrier recovery change results. He ranks detectors by output SNR from a representative simulation and offers practical tips on filter cutoff, Hilbert transformer bandwidth, and when a simple detector is good enough.
Multiplierless Half-band Filters and Hilbert Transformers
This article provides coefficients of multiplierless Finite Impulse Response 7-tap, 11-tap, and 15-tap half-band filters and Hilbert Transformers. Since Hilbert transformer coefficients are simply related to half-band coefficients, multiplierless Hilbert transformers are easily derived from multiplierless half-bands.
Candan's Tweaks of Jacobsen's Frequency Approximation
Cedron Dawg shows how small tweaks to Jacobsen's three-bin frequency estimator turn a popular approximation into an exact formula, and how a modest cubic correction yields a near-exact, low-cost alternative. The article derives an arctan/tan exact recovery, relates it to Candan's 2011/2013 tweaks, and supplies reference C code and numerical tables so engineers can see when each formula is sufficient.
An Efficient Linear Interpolation Scheme
A simple trick slashes the cost of linear interpolation to at most one multiply per output sample, and often to none. The post shows a zero-order-hold based network that preserves input samples, has a short L-1 transient, and lets 1/L scaling be implemented as a binary shift when L is a power of two. It also gives a fixed-point layout that moves scaling to the end to reduce quantization distortion.
Spectral Flipping Around Signal Center Frequency
Most DSP engineers know that multiplying a real signal by (-1)^n inverts its spectrum about fs/4, but that trick fails when you need to flip around a specific carrier. Rick Lyons presents two practical techniques: a multirate upsample-by-two solution using paired lowpass filters and cosine mixing, and a computationally heavier complex-multiply plus real-part method attributed to Dirk Bell, both yielding the desired fcntr-centered flip.
Accurate Measurement of a Sinusoid's Peak Amplitude Based on FFT Data
Measuring a sinewave's peak from FFT data can be severely biased by scalloping loss, producing errors up to 36.3 percent. Rick Lyons demonstrates how to apply a flat-top window via frequency-domain convolution to the FFT bins, cutting maximum amplitude error to about 0.02 dB compared with 3.9 dB for rectangular windows. The post includes Matlab code and practical caveats for reliable use.
IIR Bandpass Filters Using Cascaded Biquads
This post provides a Matlab function that builds Butterworth bandpass IIR filters by cascading second-order biquad sections. The biquad approach, implemented in Direct Form II, reduces sensitivity to coefficient quantization, which matters most for narrowband filters. The included biquad_bp function computes each section's feedforward and feedback coefficients plus gains from a lowpass prototype order, center frequency, bandwidth, and sampling rate.
Generating Complex Baseband and Analytic Bandpass Signals
Rick Lyons gathers and compares practical methods for creating complex baseband and analytic bandpass signals in one compact reference. The post clarifies definitions, lists time and frequency domain techniques from quadrature sampling to FFT-based analytic generation, and notes implementation tradeoffs such as sample-rate constraints, Hilbert transformer use, and phase linearity concerns. Engineers get a quick Hit Parade of options and pointers to deeper references.
Design IIR Highpass Filters
Neil Robertson walks through a compact, six-step procedure to synthesize IIR Butterworth highpass filters using pre-warping and the bilinear transform. The post gives the pole transformations, the placement of N zeros at z=1, the scaling to unity gain at fs/2, and a ready-to-run MATLAB hp_synth implementation that reproduces MATLAB's butter results.
Phase and Amplitude Calculation for a Pure Real Tone in a DFT: Method 1
Cedron Dawg shows how to get exact amplitude and phase for a real sinusoid whose frequency does not land on an integer DFT bin. The method treats a small neighborhood of DFT bins as a complex vector, builds two basis vectors from the cosine and sine transforms, and solves a 2x2 system using conjugate dot products to recover real coefficients that give amplitude and phase. A C++ example and sample output verify the formulas.
A Simplified Matlab Function for Power Spectral Density
Neil Robertson provides a tiny Matlab wrapper around pwelch that simplifies PSD computation by preselecting a Kaiser window, default overlap, and converting units from W/Hz to dBW/bin. Call psd_simple(x,nfft,fs) to get PdB and a frequency vector, with nfft controlling whether DFT averaging is used. The post includes examples showing the effect of averaging and explains the Kaiser window processing loss.
Peak to Average Power Ratio and CCDF
Setting digital modulator levels depends on peak-to-average power ratio, because random signals produce occasional high peaks that cause clipping. This post shows how to compute the CCDF of PAPR from a signal vector, with MATLAB code and examples for a sine wave and Gaussian noise. The examples reveal the fixed 3.01 dB PAPR of a sine and the need for large sample counts to capture rare AWGN peaks.
Setting the 3-dB Cutoff Frequency of an Exponential Averager
Many engineers use a simple exponential averager but need the correct α to achieve a specified 3-dB cutoff. Rick Lyons compares a common approximation with the exact closed-form solution, shows when the approximation is valid, and derives the exact α in the appendix. The approximation works well for fc < 0.1fs, but it becomes noticeably inaccurate as the normalized cutoff increases.
