CMI Code

Introduction to coded mark inversion (CMI)

Coded Mark Inversion (CMI) is a two-level line code used in digital transmission systems. ITU-T G.703 specifies it for the 139 264 kbit/s E4 interface. The code defines the line waveform; framing and payload functions remain separate. Each input bit occupies one interval $T$, and the receiver evaluates the waveform across that interval. Its name refers to the alternating polarity of successive binary marks:

• Coded: It maps the input binary stream to a defined two-level waveform.
• Mark: In telecommunication terminology, a mark denotes logical 1.
• Inversion: The level used for successive 1 bits alternates in polarity.

The mapping provides a mid-bit transition for every 0 and alternates the polarity of successive 1 bits. The polarity state changes only when another 1 is transmitted, so intervening zeros do not reset it. These properties support timing recovery and transformer or capacitor coupling.

CMI encoding rules

CMI uses two signal levels, denoted +V and -V here. A 0 changes from -V to +V at the middle of its bit period, while a 1 occupies one level for the full period.

Analysis of CMI properties

CMI determines the timing transitions, DC behavior, bandwidth, and code-violation checks available to the receiver.

• Clock timing transitions: Every encoded 0 contains a mid-bit transition. Successive encoded 1 bits alternate polarity, so their transitions also carry timing information. The receiver can use these transitions for clock recovery.
• DC behavior: The 0 mapping spends equal half-periods at -V and +V. Successive 1 bits alternate polarity, suppressing the DC component over successive marks. This property supports systems coupled through transformers or capacitors.
• Bandwidth usage: The mid-bit transition in each 0 increases high-frequency content. In this spectral model, the main lobe extends approximately to twice the clock frequency ($2f_{clk}$ or $2/T$), so CMI uses more bandwidth than NRZ.
• Simple error detection: A receiver can flag some violations, such as two consecutive 1 bits with the same polarity or a 0 without the required mid-bit transition. The flag indicates that the waveform does not follow the CMI mapping; it does not identify the cause by itself.