An inverse multiplexer (often abbreviated to "inverse mux", "imux" or "demux") allows a data stream to be broken into multiple lower data rate communications links. An inverse multiplexor differs from a demultiplexer in that each of the low rate links coming from it is related to the other ones and they all work together to carry the same data. By contrast, the output streams from a demultiplexer may each be completely independent from eachother and the demultiplexer does not have to understand them in any way.

Note that this is the opposite of a multiplexer in that it creates one high speed link from multiple low speed ones whilst a multiplexer creates multiple slow speed links from one high speed link.

example:

 |--low rate link #1--| DTE ---high rate data---inverse mux |--low rate link #2--|(de)inverse mux--DCE |--low rate link #3--| 

 DTE = Data terminal equipment DCE = Data circuit terminating equipment 

This provides an end to end connection of 3 x the data rate available on each of the low rate data links. (Strictly, slightly less than 3 x, because the inverse mux will use some of the available bandwidth to communicate with the inverse mux at the far end of the link, i.e. checksum or other error checking codes). Note that, as with multiplexers, links are almost always bi-directional and an inverse mux will practically always be combined with its reverse and still be called an inverse mux. This means that the "de-inverse mux" will actually be an inverse mux.

Inverse muxes are used, for example, to combine a number of ISDN channels together into one high rate circuit, where the DTE needs a higher rate connection than is available from a single ISDN connection. This is typically useful in areas where higher rate circuits are not available.

An alternative to an inverse mux is to use three separate links and load sharing of data between them. In the case of IP network packets could be sent in round robin mode between each separate link. Advantages of using an inverse mux over separate links include

• lower link latency (one single packet can be spread across all links)
• fairer load sharing
• network simplicity (no router needed between boxes with high speed interfaces)

See also MDM