CompoundLearn

Engineering Interview Glossary

Plain-language definitions of the technical terms that come up in wireless, RF, hardware, firmware, signal processing, and ML engineering interviews. Each entry explains what the term means and why interviewers ask about it.

Wireless & RF

5G NR Numerology
The set of subcarrier-spacing options in 5G New Radio, indexed by μ (15 kHz × 2^μ). Larger numerologies use wider subcarriers and shorter slots, trading frequency-domain granularity for lower latency and better tolerance of high Doppler.Practice 5G NR Explained: Numerology, BWP, HARQ, and Frame Structure
OFDM Cyclic Prefix
A copy of an OFDM symbol's tail prepended to its front. As long as it is longer than the channel's delay spread, it turns the channel's linear convolution into a circular one, so each subcarrier can be equalized with a single complex multiply.Practice OFDM Explained: Subcarriers, Cyclic Prefix, PAPR, and OFDMA
MIMO
Multiple-Input Multiple-Output — using several antennas at both ends of a link to send independent data streams in parallel (spatial multiplexing) or to send data redundantly for robustness (diversity).Practice MIMO Explained: Spatial Multiplexing, Rank, Massive MIMO
Beamforming
Applying per-element phase and amplitude weights across an antenna array so the combined signal forms a directional beam. Steering that beam toward a user raises received signal power and suppresses interference.Practice Beamforming Interview Guide: Analog vs Digital vs Hybrid
Channel Estimation
Recovering the channel's frequency response from known pilot or reference symbols, so the receiver can equalize the unknown data symbols. Estimation error directly limits achievable throughput.Practice OFDM Explained: Subcarriers, Cyclic Prefix, PAPR, and OFDMA
Noise Figure
A measure of how much a component degrades signal-to-noise ratio, defined as the ratio of its actual SNR degradation to that of an ideal noiseless component. The first stage — typically the LNA — dominates a receiver’s overall noise figure.
OFDMA
Orthogonal Frequency-Division Multiple Access — introduced to Wi-Fi in IEEE 802.11ax (Wi-Fi 6). It partitions each channel into Resource Units in the frequency domain so the access point can serve many stations in one transmission instead of one at a time.Practice OFDMA Resource Units Explained (IEEE 802.11ax / 802.11bn)
Resource Unit (RU)
The allocation granularity in 802.11ax / be / bn OFDMA. Defined sizes are 26, 52, 106, 242, 484, and 996 tones — each a contiguous block of subcarriers the access point assigns to one station for the duration of an OFDMA transmission.Practice OFDMA Resource Units Explained (IEEE 802.11ax / 802.11bn)
RRU (Regular Resource Unit)
The traditional contiguous-tone OFDMA Resource Unit defined since IEEE 802.11ax (Wi-Fi 6). Data subcarriers are grouped together in adjacent frequency slots. Contrasted with the new IEEE 802.11bn (Wi-Fi 8) Distributed Resource Unit (DRU).Practice OFDMA Resource Units Explained (IEEE 802.11ax / 802.11bn)
DRU (Distributed Resource Unit)
A new IEEE 802.11bn (Wi-Fi 8) tone allocation where the data subcarriers are spread across the entire transmission bandwidth instead of being contiguous. The spreading delivers a power-boost coverage gain for far stations at the cost of fewer simultaneous OFDMA users and tighter spatial-stream limits.Practice OFDMA Resource Units Explained (IEEE 802.11ax / 802.11bn)
MDRU (Multiple Distributed Resource Unit)
An IEEE 802.11bn (Wi-Fi 8) allocation that uses more than one Distributed Resource Unit in the same transmission, letting an access point serve multiple far stations together while keeping the distributed-tone coverage gain. Adds preamble and signaling overhead versus a single DRU or RRU.Practice OFDMA Resource Units Explained (IEEE 802.11ax / 802.11bn)
STR (Simultaneous Transmit and Receive)
The MLO mode in which an 802.11be Multi-Link Device transmits on one link while receiving on another at the same instant. Requires about 30 dB of combined frequency separation plus antenna isolation between the radio chains so transmit leakage does not desensitize the cross-link receiver.Practice Multi-Link Operation (MLO) Explained: Wi-Fi 7 and Wi-Fi 8
Non-STR
The MLO mode where the Multi-Link Device alternates between links — only one link is active at a time — used when the device cannot meet the RF isolation budget required for Simultaneous Transmit and Receive (STR), typically on small clients with 2.4 + 5 GHz radio pairs.Practice Multi-Link Operation (MLO) Explained: Wi-Fi 7 and Wi-Fi 8
NPCA (Non-Primary Channel Access)
An IEEE 802.11bn (Wi-Fi 8) MAC mechanism that lets a device transmit on a non-primary 20 MHz subchannel when the primary is busy. Uses a secondary EDCA procedure on the non-primary channel and is gated by an RSSI threshold to bound hidden-node risk.Practice 802.11bn UHR: Wi-Fi 8 Ultra High Reliability
DSC (Dynamic Sensitivity Control)
An IEEE 802.11bn (Wi-Fi 8) MAC mechanism that adapts the Clear Channel Assessment (CCA) threshold based on network density and interference. Raising the threshold expands spatial reuse but increases hidden-node risk — it must coordinate with NPCA gating.Practice 802.11bn UHR: Wi-Fi 8 Ultra High Reliability
Co-BF (Coordinated Beamforming)
A multi-AP beamforming mechanism in IEEE 802.11bn (Wi-Fi 8) where neighboring access points sound the channel together — via joint or sequential sounding — and form transmit beams that null interference toward each other’s served clients.Practice 802.11bn UHR: Wi-Fi 8 Ultra High Reliability
UHR-LTF (Ultra High Reliability LTF)
The new Long Training Field generation in IEEE 802.11bn (Wi-Fi 8) that supports the Distributed Resource Unit (DRU) and Multiple DRU (MDRU) tone patterns. Designed to coexist with EHT-LTF preambles for backward compatibility with Wi-Fi 7 clients.Practice 802.11bn UHR: Wi-Fi 8 Ultra High Reliability
TXOP (Transmit Opportunity)
A bounded time interval during which a station that wins channel access is permitted to transmit one or more frames. EDCA defines per-access-category TXOP duration limits; 802.11ax introduced TXOP sharing so an initiating station can share its TXOP with others.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
MU-MIMO
Multi-User MIMO — using independent spatial streams to transmit to multiple stations on the same frequency at the same time. Downlink MU-MIMO was added in IEEE 802.11ac (Wi-Fi 5); uplink MU-MIMO arrived in 802.11ax (Wi-Fi 6). Complements OFDMA frequency-domain partitioning.Practice MIMO Explained: Spatial Multiplexing, Rank, Massive MIMO
BSS Coloring
A 6-bit identifier assigned to each Basic Service Set in IEEE 802.11ax (Wi-Fi 6) so a station can recognize frames from a neighboring BSS and apply an adjusted carrier-sense threshold for OBSS/PD spatial reuse. Reduces unnecessary deferral in dense deployments.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
TWT (Target Wake Time)
An IEEE 802.11ax (Wi-Fi 6) power-management mechanism where stations negotiate scheduled wake intervals with the access point. Individual TWT serves per-station scheduling; broadcast TWT serves IoT groups; extended for multi-link operation in 802.11be (Wi-Fi 7).Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
R-TWT (Restricted Target Wake Time)
A latency-targeted extension of TWT introduced in IEEE 802.11be (Wi-Fi 7) that reserves specific wake windows for time-sensitive traffic. Stations participating in R-TWT defer to the reserved windows so latency-critical frames see predictable contention behavior.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
Preamble Puncturing
An IEEE 802.11be (Wi-Fi 7) PHY mechanism that lets a wide channel (up to 320 MHz) operate while excluding one or more 20 MHz subchannels occupied by interference or radar. The puncturing pattern is signaled in the U-SIG and EHT-SIG preamble fields.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
4096-QAM
A 12-bits-per-symbol modulation order introduced in IEEE 802.11be (Wi-Fi 7). Reaches the highest peak rates among Wi-Fi modulations but requires roughly 35–40 dB SNR with LDPC coding, so it activates only at close range with strong line-of-sight and low transmitter EVM.Practice Wi-Fi 8 (802.11bn) vs Wi-Fi 7: What Changed
OBSS/PD (Overlapping BSS Packet Detect)
An IEEE 802.11ax (Wi-Fi 6) spatial-reuse mechanism. When a station detects an inter-BSS frame (identified via BSS coloring), it can apply a more permissive CCA threshold for that frame and transmit concurrently if transmit-power constraints permit.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
AFC (Automated Frequency Coordination)
The regulatory model required for Standard Power operation in the 6 GHz band in the United States. Devices must query an AFC server for permitted channels and EIRP limits before transmitting. Low Power Indoor and Very Low Power device classes skip AFC but carry stricter EIRP caps.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
EIRP (Effective Isotropic Radiated Power)
The hypothetical power an isotropic antenna would have to radiate to match the peak power delivered by a real directional antenna in its main beam. The metric regulators cap (e.g., per-band EIRP limits in 5 / 6 GHz) and the figure beamforming directly raises.Practice Wireless Foundations
NDP Sounding (Null Data Packet Sounding)
The explicit-beamforming channel-state acquisition protocol in IEEE 802.11ac and later. The transmitter sends a Null Data Packet whose Long Training Fields probe the channel; the receiver replies with a compressed beamforming feedback frame the transmitter uses to compute steering weights.Practice MIMO Explained: Spatial Multiplexing, Rank, Massive MIMO
OFDMA Trigger Frame
A control frame the access point sends to coordinate uplink OFDMA. The trigger synchronizes the transmit start time, assigns Resource Units, sets the MCS, and signals a transmit-power adjustment so all simultaneous uplink transmissions arrive at the AP aligned in time and frequency.Practice OFDMA Resource Units Explained (IEEE 802.11ax / 802.11bn)
CSMA/CA
Carrier-Sense Multiple Access with Collision Avoidance — the baseline 802.11 medium-access protocol. Stations listen before transmitting and use random back-off intervals after the medium goes idle to reduce the chance that two stations transmit simultaneously.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
EDCA (Enhanced Distributed Channel Access)
The QoS extension to CSMA/CA defined in IEEE 802.11e. Stations sort traffic into four Access Categories (voice, video, best-effort, background); higher-priority traffic uses shorter Arbitration Inter-Frame Spaces and smaller contention windows so it wins channel access more often.Practice Wi-Fi Standards Interview Questions & Answers (Wi-Fi 6/7/8)
IEEE 802.11 Amendment
A scoped revision to the IEEE 802.11 standard, identified by a single or double lowercase letter (a, b, g, n, ac, ax, be, bn, …) assigned by the working group in task-group authorization order. The Wi-Fi Alliance separately assigns consumer marketing names (Wi-Fi 4 / 5 / 6 / 7 / 8).Practice 802.11ax/be/bn → Wi-Fi 6, 7, 8: Amendment Letters Explained
Wi-Fi Alliance
The industry consortium that certifies Wi-Fi product interoperability and assigns the sequential consumer marketing names (Wi-Fi 1 through Wi-Fi 8). Distinct from the IEEE 802.11 working group, which drafts and publishes the underlying technical standard.Practice 802.11ax/be/bn → Wi-Fi 6, 7, 8: Amendment Letters Explained
PHY HARQ (Wi-Fi 8)
A physical-layer retransmission scheme introduced in IEEE 802.11bn (Wi-Fi 8) where the receiver soft-combines log-likelihood ratios from a failed frame with those from a retransmission, decoding the combined codeword at a lower effective SNR than either attempt would need alone.Practice Wi-Fi 8 (802.11bn) vs Wi-Fi 7: What Changed
gNB
The 5G New Radio base station — the access-network node connecting UEs to the 5G Core (5GC). Equivalent of the LTE eNB in 4G. In modern deployments, the gNB is usually split into CU and DU components (see CU/DU split).Practice 5G NR Explained: Numerology, BWP, HARQ, and Frame Structure
gNB-CU
The Central Unit of a 5G NR base station, running the higher protocol layers: RRC and PDCP. Connects to one or more DUs via the F1 interface. Centralization enables mobility optimization, unified security, and resource pooling across many cells.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
gNB-DU
The Distributed Unit of a 5G NR base station, running the lower protocol layers: RLC, MAC, and (upper) PHY. Sits close to the radio for sub-millisecond latency on HARQ and scheduling. Connects to the CU via F1 and to RUs via the O-RAN Open Fronthaul (7-2x split).Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
CU-CP
The Control Plane portion of the gNB-CU. Handles RRC signaling and control-plane PDCP. Connected to the user-plane half (CU-UP) via the E1 interface. The CU-CP/CU-UP split lets operators scale control and user planes independently.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
CU-UP
The User Plane portion of the gNB-CU. Handles user-plane PDCP — packet encryption, header compression, reordering — for the data traffic of UEs served by the gNB. Supports per-slice user-plane deployment for network slicing.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
AMF
Access and Mobility Management Function — the 5GC node that handles UE registration, mobility, paging, and security context. The 5G analog of the LTE MME. Connects to gNBs via the N2 interface.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
SMF
Session Management Function — the 5GC node that manages PDU sessions, IP-address allocation, and the policy and routing for user-plane traffic. Coordinates with the UPF for data-plane forwarding decisions.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
UPF
User Plane Function — the 5GC data-plane node that routes user traffic between the UE (via the gNB) and external networks. UPF can be deployed near the cell edge (MEC) to minimize latency for URLLC services.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
C-RNTI
Cell Radio Network Temporary Identifier — the UE’s permanent unicast identifier within a serving cell. Used to scramble the CRC of PDCCH carrying DCI addressed to the UE. Assigned during RACH contention resolution and held until the UE leaves the cell.Practice PDCCH, PDSCH, PUCCH, PUSCH: The 5G NR Physical Channels
RA-RNTI
Random Access RNTI — a temporary identifier used during the RACH procedure to address the Random Access Response (MSG2). Derived from the RACH occasion the UE used to transmit its preamble, shared across all UEs that picked that same occasion.Practice NR Initial Access: SSB, MIB, SIB1, and the RACH Procedure
SI-RNTI
System Information RNTI — a fixed reserved value used to address PDCCH scheduling broadcast of System Information Blocks (SIB1 and others). Any UE in the cell can descramble PDCCH with SI-RNTI to find broadcast information.Practice NR Initial Access: SSB, MIB, SIB1, and the RACH Procedure
INT-RNTI
Interruption RNTI — used to scramble DCI format 2_1 (preemption indication). When the gNB preempts an ongoing eMBB transmission with an urgent URLLC mini-slot, the affected eMBB UEs receive a preemption-indication bitmap via DCI scrambled with their INT-RNTI.Practice URLLC in 5G NR: Mini-Slots, Preemption, and Sub-1ms Latency
MCS-C-RNTI
Modulation and Coding Scheme C-RNTI — used to scramble PDCCH when an alternate MCS table is in use for the UE. Typically used for URLLC bearers configured to use MCS Table 3 (qLowSE, targeting 10⁻⁵ BLER) while the same UE may still receive normal MCS Table 1 or 2 transmissions on its standard C-RNTI.Practice NR Channel Coding: LDPC, Polar, MCS, and Link Adaptation
DCI Format 0_0 / 0_1
NR uplink DCI formats. 0_0 is the fallback DCI: small fixed payload, default scheduling parameters, used during reconfiguration transitions. 0_1 is the configurable DCI: variable payload based on UE capabilities, supports MIMO layers, frequency-hopping, and the full NR uplink feature set.Practice PDCCH, PDSCH, PUCCH, PUSCH: The 5G NR Physical Channels
DCI Format 1_0 / 1_1
NR downlink DCI formats. 1_0 is the fallback DCI for downlink, also used to carry paging and SI grants. 1_1 is the configurable DCI for normal PDSCH scheduling. DCI 1_2 (added in Rel-16) is the URLLC compact variant.Practice PDCCH, PDSCH, PUCCH, PUSCH: The 5G NR Physical Channels
DCI Format 2_1
Group-common DCI carrying preemption indication for URLLC coexistence with eMBB. Scrambled with INT-RNTI per affected UE; carries a 14-bit bitmap indicating which time-frequency REs in the monitoring window were preempted by an urgent URLLC mini-slot.Practice URLLC in 5G NR: Mini-Slots, Preemption, and Sub-1ms Latency
SDAP
Service Data Adaptation Protocol — the 5G NR layer that maps 5GC QoS flows (identified by QFI) to RAN Data Radio Bearers (DRBs). SDAP marks each user-plane PDU with the source QFI so the gNB scheduler can apply per-flow QoS handling. SDAP did not exist in LTE; it was added in NR to bridge the per-flow 5GC QoS model with the per-DRB RAN bearer model.Practice 5G QoS and 5QI: Service Requirements and Network Slices
MAC CE
MAC Control Element — a control message carried at the MAC layer, multiplexed alongside data in a MAC PDU and identified by an LCID (Logical Channel ID). Examples: BSR (buffer status), PHR (power headroom), TCI state activation, SCell activation, SRS-trigger. MAC CEs are the layer-2 control plane complementing DCI on PDCCH.Practice NR MAC Scheduling: PDCCH Allocation, BSR, DRX, Configured Grant
BSR
Buffer Status Report — a MAC CE carried in PUSCH reporting per-Logical-Channel-Group buffer occupancy at the UE. Short BSR covers one LCG with a 5-bit buffer-size index; Long BSR covers up to 8 LCGs. The gNB uses BSR to size subsequent uplink grants accurately.Practice NR MAC Scheduling: PDCCH Allocation, BSR, DRX, Configured Grant
RLC Modes (TM, UM, AM)
Three modes the NR Radio Link Control layer can operate in. TM (Transparent Mode): no header, no segmentation, used for short control messages. UM (Unacknowledged Mode): segmentation/reassembly with sequence numbers but no retransmission; used for latency-sensitive traffic like voice. AM (Acknowledged Mode): full ARQ with retransmission; used for most user data.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
NAS Layer
Non-Access Stratum — the protocol layer between UE and the 5GC AMF, terminated above the RAN. NAS handles UE registration, authentication, PDU session establishment, mobility management at the core-network level (Tracking Area Update), and security context establishment. Distinct from the AS (Access Stratum) layers that terminate at the gNB.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
F1 Interface
The interface between the gNB-CU and gNB-DU in the CU/DU split architecture. F1-AP is the control-plane protocol (over SCTP) carrying UE context setup, bearer configuration, and mobility signaling. F1-U is the user-plane protocol (over GTP-U) carrying PDCP PDUs between CU and DU.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
E1 Interface
The interface between gNB-CU-CP and gNB-CU-UP within a split CU. Carries control signaling to set up, modify, and release the user-plane bearer state on the CU-UP. Allows independent scaling and per-slice deployment of the user plane.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
Xn Interface
The gNB-to-gNB interface in 5G NR, used for direct signaling between source and target gNB during handover (avoiding the AMF on the critical path) and for dual connectivity coordination. The 5G analog of the LTE X2 interface.Practice 5G NR Handover: Xn vs N2, Conditional HO, Beam Failure Recovery
N2 Interface
The interface between the gNB and the AMF in the 5GC. Carries NAS signaling tunneled through the gNB plus RAN-level control messages (UE context setup at the AMF, mobility coordination). Used for N2-based handover when Xn is unavailable.Practice 5G NR Architecture: NSA vs SA, CU/DU Split, and CA
I-RNTI
Inactive RNTI — a temporary identifier assigned to a UE when it is suspended from RRC_CONNECTED to RRC_INACTIVE. The I-RNTI lets the UE later resume to RRC_CONNECTED with a single RRC Resume Request, with the network fetching the stored UE context from the last-serving gNB over Xn.Practice RRC States in 5G NR: RRC_IDLE, RRC_INACTIVE, RRC_CONNECTED
RNA (RAN Notification Area)
The set of cells over which an RRC_INACTIVE UE can move without informing the network. Managed by the last-serving gNB. When the UE leaves the RNA, it triggers an RNA Update so paging continues to find it. The RAN-layer analog of the 5GC Tracking Area.Practice RRC States in 5G NR: RRC_IDLE, RRC_INACTIVE, RRC_CONNECTED

Hardware & Digital Design

Setup and Hold Time
Setup time is how long a data input must be stable before a clock edge; hold time is how long it must stay stable after. Violating either can drive a flip-flop into metastability.Practice Digital Logic
Metastability
The unstable, undefined state a flip-flop can enter when its input changes too close to the clock edge. It eventually resolves to 0 or 1, but only after an unbounded settling time — the core hazard when crossing clock domains.Practice Digital Logic
Pipelining
Splitting a long combinational path into stages separated by registers, so the clock period only has to cover the slowest stage. It raises throughput and clock frequency at the cost of added latency.Practice Computer Architecture
FPGA Timing Closure
The iterative process of getting a design to meet every setup and hold constraint after place-and-route, so it runs reliably at the target clock frequency. It typically drives constraint tuning, pipelining, and floorplanning changes.Practice FPGA Design

Firmware & Embedded

RTOS Preemption
A scheduler’s ability to immediately suspend a running lower-priority task when a higher-priority task becomes ready, rather than waiting for the current task to yield. It is what keeps worst-case response time bounded.Practice RTOS
Interrupt Latency
The delay between an interrupt being asserted and the first instruction of its handler executing. A key real-time metric, driven by interrupt masking, higher-priority handlers, and context-save overhead.Practice RTOS
Priority Inversion
When a high-priority task is blocked on a resource held by a low-priority task — and a medium-priority task can run in between, delaying the high-priority task indefinitely. Priority inheritance is the standard fix.Practice RTOS
Direct Memory Access (DMA)
A hardware mechanism that transfers data between memory and peripherals without CPU involvement, freeing the processor during bulk transfers and cutting per-byte interrupt overhead.

Signal Processing

Nyquist Rate
Twice the highest frequency present in a signal — the minimum sampling rate that preserves it without aliasing. Sampling below it folds high frequencies down onto lower ones, corrupting the signal irreversibly.Practice Signals and Systems
FFT (Fast Fourier Transform)
An O(N log N) algorithm for computing the Discrete Fourier Transform, converting a sampled time-domain signal into its frequency-domain representation. It underpins spectral analysis, OFDM, and most practical DSP.Practice DSP Fundamentals
FIR vs IIR Filters
FIR (finite impulse response) filters use only past inputs — always stable and easy to give linear phase. IIR (infinite impulse response) filters also feed back past outputs — sharper responses for far fewer taps, but they can be unstable and have nonlinear phase.Practice DSP Fundamentals

Machine Learning

Self-Attention
The mechanism in a transformer where each token computes a weighted combination of all tokens in the sequence, with weights derived from query–key similarity. It lets the model relate distant elements without recurrence.Practice Transformer Architecture
Quantization (ML)
Reducing the numeric precision of a model’s weights and activations — for example FP32 to INT8 — to cut memory footprint and speed up inference, usually at a small, controllable accuracy cost.
Inference Latency
The time a trained model takes to produce a single prediction. Distinct from throughput, it is the binding constraint for real-time ML systems such as on-device or interactive applications.

RF & Analog Circuits

Phase-Locked Loop (PLL)
A feedback loop that locks a tunable oscillator's phase to a reference by comparing the two in a phase detector and steering a VCO through a loop filter. The basis of frequency synthesis and clock generation.Practice Phase-Locked Loops (PLLs) & Frequency Synthesis
Fractional-N Synthesis
A PLL technique that produces output frequencies at non-integer multiples of the reference by dithering the feedback divider (typically with a delta-sigma modulator), giving fine resolution at the cost of fractional spurs.Practice Phase-Locked Loops (PLLs) & Frequency Synthesis
Reference Spur
A discrete spurious tone offset from a PLL's carrier by the comparison (reference) frequency, caused by periodic leakage from the phase detector and charge pump modulating the VCO.Practice Phase-Locked Loops (PLLs) & Frequency Synthesis
Phase Noise
An oscillator's short-term random phase fluctuation, specified as single-sideband power relative to the carrier per hertz (dBc/Hz) at a given offset. The frequency-domain dual of timing jitter.Practice Phase Noise & Jitter
Reciprocal Mixing
A receiver impairment in which local-oscillator phase noise mixes a strong adjacent-channel signal onto the wanted signal's IF, raising the noise floor even when the wanted signal is weak.Practice Phase Noise & Jitter
Low-Noise Amplifier (LNA)
A receiver's first active stage, designed to add minimal noise while providing gain. Because it sets the front-end noise figure (per the Friis formula), its design dominates receiver sensitivity.Practice Low-Noise Amplifier (LNA) Design
Friis Noise Formula
The cascade equation for noise figure: each stage's noise contribution is divided by the total gain ahead of it, so the first stage dominates — the reason a low-noise, high-gain LNA goes first.Practice Low-Noise Amplifier (LNA) Design
Third-Order Intercept (IP3)
The extrapolated power at which a circuit's third-order intermodulation products would equal the fundamental. Higher IP3 means better large-signal linearity; it trades against noise and power consumption.Practice Low-Noise Amplifier (LNA) Design
1-dB Compression Point (P1dB)
The input or output power at which an amplifier's gain has fallen 1 dB below its small-signal value, marking the onset of saturation and the practical top of its linear range.Practice Power Amplifier Design & Linearization
Power Amplifier Back-Off
Operating a PA below its saturated output power to keep distortion low enough for high-PAPR signals such as OFDM. Back-off improves linearity but lowers efficiency.Practice Power Amplifier Design & Linearization
Doherty Amplifier
A PA architecture using a main and a peaking amplifier with active load modulation to keep efficiency high at power back-off, widely used in cellular base-station transmitters.Practice Power Amplifier Design & Linearization
Adjacent Channel Leakage Ratio (ACLR)
The ratio of in-channel power to the power that PA nonlinearity spills into an adjacent channel (spectral regrowth). A key transmitter conformance metric.Practice Power Amplifier Design & Linearization
AM-AM / AM-PM Distortion
The two ways a PA distorts a signal: AM-AM is gain compression versus input amplitude; AM-PM is input-amplitude-dependent phase shift. Both are what predistortion corrects.Practice Power Amplifier Design & Linearization
Load-Pull
A characterization technique that sweeps the impedance presented to a PA's output to map contours of output power, efficiency, and linearity, guiding the design of the output matching network.Practice Power Amplifier Design & Linearization
Image Frequency
An unwanted input frequency on the opposite side of the local oscillator from the wanted signal that down-converts to the same IF. It must be filtered or cancelled by an image-reject mixer.Practice Mixers & Frequency Conversion
Conversion Gain
The ratio of a mixer's output (IF) level to its input (RF) level. Active mixers provide gain; passive mixers have conversion loss but typically better linearity.Practice Mixers & Frequency Conversion
Effective Number of Bits (ENOB)
A data converter’s real resolution after noise and distortion, derived from measured SINAD via ENOB = (SINAD − 1.76) / 6.02. It is always below the nominal bit count.Practice Data Converters: ADC & DAC Fundamentals
Spurious-Free Dynamic Range (SFDR)
The ratio of a converter's full-scale signal to its largest spurious tone, setting how small a signal can be resolved next to a large one.Practice Data Converters: ADC & DAC Fundamentals
INL / DNL
Integral and Differential Non-Linearity: static converter errors. DNL is each code step's deviation from 1 LSB (DNL ≤ −1 means a missing code); INL is the accumulated deviation of the transfer curve.Practice Data Converters: ADC & DAC Fundamentals
Aperture Jitter
Random variation in an ADC's sampling instant. It caps achievable SNR for high-input-frequency signals, since SNR ≈ −20·log₁₀(2π·f_in·t_jitter).Practice Data Converters: ADC & DAC Fundamentals
Sigma-Delta Modulator
An oversampling converter that puts a coarse quantizer in a feedback loop to push quantization noise out of the signal band (noise shaping), then removes it with a digital filter to reach high resolution.Practice Data Converters: ADC & DAC Fundamentals

High-Speed Digital & Signal Integrity

SerDes
Serializer/Deserializer: a circuit pair that converts parallel data to a high-rate serial stream and back, with the clock recovered from the data. The backbone of PCIe, Ethernet, and chip-to-chip links.Practice SerDes Design Fundamentals
PAM4
Four-level Pulse-Amplitude Modulation, carrying two bits per symbol so a link doubles throughput at the same symbol rate as NRZ — at the cost of roughly one-third the voltage margin and tougher equalization.Practice SerDes Design Fundamentals
Clock Data Recovery (CDR)
The receiver function that extracts a sampling clock from the transitions of an embedded-clock serial stream, so no separate clock lane is required.Practice SerDes Design Fundamentals
Decision-Feedback Equalization (DFE)
A receiver equalizer that subtracts the trailing intersymbol interference of already-decided bits. Because it feeds back decisions, it cancels ISI without amplifying noise, unlike a linear equalizer.Practice SerDes Design Fundamentals
CTLE
Continuous-Time Linear Equalizer: an analog high-frequency-boost filter at a SerDes receiver that flattens channel loss before sampling, usually paired with a DFE.Practice SerDes Design Fundamentals
Eye Diagram
An overlay of many received bit periods whose open 'eye' reveals timing and voltage margin; its height and width quantify ISI, jitter, and noise on a high-speed link.Practice SerDes Design Fundamentals
PCIe FLIT Mode
Flow-control-unit mode introduced in PCIe 6.0: data travels in fixed-size FLITs with integrated forward error correction and CRC, replacing per-packet framing to handle the higher error rate of PAM4 signaling.Practice PCIe Gen5/Gen6 Explained
PCIe Lane
A single full-duplex differential SerDes pair in a PCIe link. Lanes aggregate into x1/x4/x8/x16 links whose bandwidth scales with lane count and per-generation data rate.Practice PCIe Gen5/Gen6 Explained
Data Strobe (DQS)
The source-synchronous strobe a DDR interface sends alongside data so the receiver can capture DQ bits without a shared clock; it is center-aligned on writes and edge-aligned on reads.Practice DDR Memory Interfaces
Write Leveling
A DDR training step that adjusts each DRAM's strobe launch timing to compensate for the deliberate clock skew of fly-by command/address routing, realigning strobe and clock at every device.Practice DDR Memory Interfaces
On-Die Termination (ODT)
Termination resistors integrated inside the DRAM or controller, switched on dynamically during reads and writes to absorb reflections on the high-speed memory bus.Practice DDR Memory Interfaces
Memory Prefetch
The number of bits a DDR device fetches per column access (e.g. 8n in DDR4, 16n in DDR5/LPDDR5), letting a slow internal array feed a fast I/O bus by widening the internal transfer.Practice DDR Memory Interfaces
Insertion Loss
The frequency-dependent attenuation a signal suffers traversing an interconnect (S21). Rising loss at high frequency closes the eye and is the main impairment equalizers fight on high-speed links.Practice Signal Integrity Fundamentals
Crosstalk (NEXT / FEXT)
Unwanted coupling from an aggressor trace onto a victim, measured as near-end (NEXT) at the driver side and far-end (FEXT) at the receiver side. A dominant noise source on dense high-speed buses.Practice Signal Integrity Fundamentals
Intersymbol Interference (ISI)
Smearing of one symbol's energy into adjacent symbols caused by a band-limited or reflective channel — the primary cause of eye closure that equalization corrects.Practice Signal Integrity Fundamentals
Characteristic Impedance
The instantaneous voltage-to-current ratio a signal sees on a uniform transmission line. Discontinuities in it cause reflections, so controlled-impedance routing (e.g. 50 Ω, 100 Ω differential) is essential.Practice Signal Integrity Fundamentals
Power Delivery Network (PDN)
The full path delivering current from regulator to on-chip transistors — planes, vias, package, and decoupling capacitors — engineered to keep impedance below a target across frequency.Practice Power Integrity & PDN Design
Target Impedance
The maximum PDN impedance that keeps supply ripple within budget for a given transient current (Z_target ≈ ΔV / ΔI). PDN design is largely the art of staying under it across frequency.Practice Power Integrity & PDN Design
Simultaneous Switching Noise (SSN)
Power/ground-rail bounce caused by many outputs switching at once drawing current through supply inductance (also called ground bounce). A key power-integrity limit on I/O timing.Practice Power Integrity & PDN Design
Clock Domain Crossing (CDC)
Passing a signal between asynchronous or unrelated clock domains. Done naively it causes metastability and data-coherency bugs, so it requires synchronizers, gray coding, or handshakes.Practice Clock Domain Crossing (CDC)
Two-Flop Synchronizer
Two back-to-back flip-flops in the destination clock domain that let a single-bit signal's metastability resolve before it is used — the standard fix for a one-bit clock-domain crossing.Practice Clock Domain Crossing (CDC)
Gray Code
A binary encoding where consecutive values differ in exactly one bit. Used for multi-bit clock-domain crossings (e.g. FIFO pointers) so at most one bit can be mid-transition when sampled.Practice Clock Domain Crossing (CDC)

DSP & Signal Processing

Digital Predistortion (DPD)
Applying the inverse of a power amplifier's distortion to the baseband signal so the PA output is linear, letting it run closer to saturation for efficiency while meeting ACLR and EVM limits.Practice Digital Predistortion (DPD)
Memory Polynomial
A DPD model that captures a PA's memory effects by combining current and past signal samples with nonlinear terms — a tractable simplification of the full Volterra series.Practice Digital Predistortion (DPD)
Spectral Leakage
The spreading of an FFT bin's energy into neighbors when a signal's frequency is not an exact bin multiple, because the finite observation window is not periodic in the signal. Reduced by windowing.Practice FFT & Spectral Analysis
Window Function
A taper (Hann, Hamming, Blackman, etc.) applied before an FFT to reduce spectral leakage, trading a wider main lobe (coarser resolution) for lower side lobes (better dynamic range).Practice FFT & Spectral Analysis
Zero-Padding
Appending zeros before an FFT to interpolate the spectrum onto a finer grid. It smooths the displayed shape but adds no true resolution, which is set by the observation time.Practice FFT & Spectral Analysis
Group Delay
The negative derivative of a filter's phase versus frequency — the delay each frequency component sees. Constant group delay (linear phase, as in symmetric FIR) preserves waveform shape.Practice FIR vs IIR Filter Design
Automatic Gain Control (AGC)
A feedback loop that adjusts receiver gain to keep signal level within the ADC's range despite a wide range of input power, balancing settling speed against added distortion.Practice Automatic Gain Control (AGC)
Variable-Gain Amplifier (VGA)
An amplifier whose gain is set by a control voltage or code — the actuator in an AGC loop, often built with decibel-linear (exponential) gain control for constant loop dynamics.Practice Automatic Gain Control (AGC)

Cellular IoT, V2X & Deterministic Networking

RedCap (Reduced Capability)
Reduced Capability NR (3GPP Rel-17, 'NR-Light'): a mid-tier device class between eMBB and LPWA, with reduced bandwidth (up to 20 MHz in FR1), fewer receive branches, and relaxed MIMO for wearables and sensors.Practice RedCap (Reduced Capability) in 5G
eRedCap
Enhanced RedCap (3GPP Rel-18): a further-reduced device tier capping peak throughput near 10 Mbps while keeping a 20 MHz RF channel, for the lowest-cost cellular IoT endpoints.Practice RedCap (Reduced Capability) in 5G
NB-IoT
Narrowband IoT: a 180 kHz LPWA cellular technology for stationary, low-rate, deep-coverage devices, using coverage-enhancement repetitions and power-saving modes; it omits handover and voice.Practice NB-IoT vs LTE-M
LTE-M (eMTC)
LTE-M (eMTC): a 1.4 MHz LPWA cellular technology supporting mobility, VoLTE, and higher data rates than NB-IoT, suited to moving or voice-capable IoT devices.Practice NB-IoT vs LTE-M
Power Saving Mode (PSM)
A cellular IoT low-power state where a device stays registered but unreachable for long periods, drawing near-zero current to extend battery life to years, at the cost of downlink latency.Practice NB-IoT vs LTE-M
eDRX (Extended DRX)
Extended Discontinuous Reception: lengthening the idle-mode paging cycle so an IoT device wakes its receiver far less often, trading downlink reachability latency for battery life.Practice NB-IoT vs LTE-M
C-V2X
Cellular Vehicle-to-Everything: 3GPP sidelink-based communication for vehicles (LTE Rel-14 onward, NR Rel-16 onward), supporting safety use cases such as collision avoidance and platooning over PC5.Practice 5G Sidelink & C-V2X
Time-Sensitive Networking (TSN)
A set of IEEE 802.1 standards adding bounded latency and jitter to standard Ethernet via time synchronization, scheduled traffic, and frame preemption — for industrial, automotive, and fronthaul use.Practice Time-Sensitive Networking (TSN)
Time-Aware Shaper (802.1Qbv)
The IEEE 802.1Qbv mechanism that opens and closes per-queue transmission gates on a synchronized schedule, reserving protected windows for time-critical traffic on a shared Ethernet link.Practice Time-Sensitive Networking (TSN)