NRZ encodes each bit as a sustained signal level, NRZI encodes it by whether the signal changes, and Manchester encodes it with a required transition halfway through every bit. That difference determines how easily a receiver can recover timing—and how many signal symbols the code uses per data bit.
How does each encoding represent a bit?
NRZ: the signal level represents the bit
In a level-oriented NRZ scheme, the signal stays at its selected level for the duration of a bit cell, and that level identifies the bit value. A change in data value can produce a signal transition; repeated bits mapped to the same level produce no transition. NRZ uses one signal symbol per data bit in the cited Microchip description.
NRZ describes a data-to-waveform rule, not one required electrical interface. Implementations can differ in polarity and signaling levels.
NRZI: a change or no change represents the bit
NRZI makes the interpretation depend on whether the signal changes state, rather than only on its absolute level. The mapping is convention-dependent: William G. Wong’s Electronic Design overview illustrates transitions associated with a 1, while USB 2.0 uses 1 for no change and 0 for a transition. Always check the convention used by the particular protocol or diagram.
#1 Best Overall
NRZI does not guarantee a transition in every bit cell. A run of bits mapped to “no change” leaves the signal steady, which can make timing recovery difficult.
Manchester: the mid-bit transition represents the bit
Manchester divides each bit into two halves and requires a transition between them. The direction of that transition identifies the bit, with the 0/1 direction mapping depending on the convention. IEEE Std 802.3-2015 describes Manchester on the AUI as combining data and clock into bit-symbols; the transition halfway through each bit provides timing information.
Rank #2
How do the codes compare?
| Property | NRZ | NRZI | Manchester |
|---|---|---|---|
| What encodes the bit | Signal level during the bit cell | Presence or absence of a signal change, under a specified convention | Direction of the required mid-bit transition |
| Transition guaranteed by the code? | No. Repeated bits at the same level can produce no transition. | No. Bits mapped to no change can run without transitions. | Yes. At least one transition occurs at the midpoint of every bit. |
| Timing recovery | Sparse transitions can complicate recovery unless another clocking or coding method is used. | Transition-free runs are possible; protocols may constrain them. | Mid-bit transitions provide timing information regardless of the data sequence. |
| Symbols per data bit | One in the cited Microchip description | One signal symbol per bit in the basic encoding | Two in the cited Microchip description |
| Main tradeoff | Simple level representation, but clock recovery must be handled. | Encodes transitions rather than absolute levels, but mapping and transition density need attention. | Provides embedded timing and is described as having no DC component in the cited vendor reference, at twice the data rate as symbol rate. |
These are conceptual line-code comparisons, not substitutes for a PHY specification. Actual voltages, polarity, differential signaling, framing, and other details depend on the implementation.
Why do transitions matter for clock recovery?
A receiver has to identify where bit cells begin and end. If the waveform stays at one level through a long run of bits, it provides few edges to help the receiver keep its sampling clock aligned. This is the potential issue with NRZ and with NRZI runs mapped to no change.
Manchester supplies a transition halfway through each bit, even when consecutive data bits have the same value. Those regular edges give the receiver timing information. Microchip also describes Manchester as having no DC component; that property and the frequent transitions can be useful, but the code has a rate cost.
What is Manchester’s rate cost?
Manchester represents each data bit with two signal symbols in the cited Microchip documentation, so its symbol rate is twice the data rate. NRZ and basic NRZI use one signal symbol per data bit in the descriptions here. The doubled symbol rate is the directly supported comparison; actual required channel bandwidth depends on the signaling and filtering assumptions, so it should not be inferred from symbol rate alone.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How do protocols handle NRZI’s transition-free runs?
USB 2.0 is a concrete example of a protocol adding a transition-control rule around NRZI. Its convention is 1 = no signal change and 0 = signal change. After six consecutive 1s, USB 2.0 inserts a 0 before NRZI encoding; that inserted bit forces a line transition and helps the receiver maintain data and clock lock. This bit-stuffing rule is specific to USB 2.0 packet encoding, not an inherent feature of NRZI.
The USB-IF lists a USB 2.0 specification package dated 2025-06-03. The detailed clauses 7.1.8–7.1.9 are available in a reproduced specification copy; consult the official package for authoritative clause text: USB-IF USB 2.0 Specification listing.
Recommended Free Tools
Best Value
Where do these examples apply?
- USB 2.0: NRZI with 1 = no change, 0 = transition, plus bit stuffing after six consecutive 1s.
- IEEE 802.3-2015 AUI: The cited edition’s section 7.3.1.1 describes Manchester encoding on the Attachment Unit Interface. This historical interface example should not be generalized to every Ethernet PHY or generation.
Block codes such as 4B/5B, 8B/10B, and 64B/66B are related to transmission coding but are not synonyms for NRZ, NRZI, or Manchester; they operate at a different coding layer. For an introductory overview of these distinctions, see Electronic Design’s comparison of NRZ, NRZI, and Manchester encoding.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
