Can someone explain Multiplexers to me

[THATguy]

Are they just simple IC's made from other logic gates, or can they be a single logic gate on their own?

Are they just a way to change an IC's operation to make it more versatile?

How useful are they? Are they found in CPU's?

If I'm just making simple circuits, how much time should I invest in studying Multiplexers?

[ataradov]

There is not much to study, it is a pretty simple concept.

Yes, they are found everywhere, you can't make a CPU without one.

They can be built from logic gates, but inside the ICs they are made out of transistors.

[CatalinaWOW]

You are aware that there are two fundamental kinds of multiplexors - digital and analog.  They both are a way of sharing wires, by assigning a slice of time on the wire to a specific signal.  The digital multiplexors focus on assigning a logic value to that slice of time, while the analog multiplexors actually connect one of several inputs to one or more outputs during a time slice.  Both types make the assumption that the input signal is constant during the time slice.  This assumption can be enforced on the digital side, while on the analog side there are many complexities associated with making that assumption more or less true.

[ataradov]

I don't get that thing about time slices. Multiplexer just select one of two or more inputs and puts the value of that input on the output. The input signal may change all it wants. This is true for both analog and digital MUXes.

[strawberry]

Multiplexer is bidirectional MOSFET switch that can digitaly select RLC components and values for MCU internal clock or filters that other logic cant do.
Some old single beam crt oscillooscope used MUX to make it as dual chanel..

Similar for digital signal only: https://en.wikipedia.org/wiki/Three-state_logic

[hamster_nz]

FWIW, a digital multiplexer can be built from logic gates:

x = (a AND c)  OR (b AND NOT c)

The output is either the value of 'a' or 'b', depending on the value of 'c'.

[EEVblog]

I don't get that thing about time slices. Multiplexer just select one of two or more inputs and puts the value of that input on the output. The input signal may change all it wants. This is true for both analog and digital MUXes.

And then you can have latched multiplexers!

[ataradov]

And then you can have latched multiplexers!

Well, yes, there is obviously variety. But none of them are very complicated to ponder whether you should study them or not

[james_s]

In the most basic form a multiplexer can be consisted of a single pole, multiple position switch. For simplicity let's take the most basic form, a single pole, 2 position switch. Feed one input to each of the two poles, now when you flip the switch the selected input will appear on the single output.

Now to demultiplex, imagine another identical switch, connect the common contact to the output of our multiplexer we created above, and connect an LED or other output to each of the other two contacts. Now imagine that the switch levers are mechanically linked. If you want, you can rapidly toggle it and the result is that you'll transmit the state of two inputs across a single wire to be indicated on the two outputs.

Real world multiplexers are typically 4 or 8 bits wide, and are often used for selecting data from or feeding data to two different buses.

[THATguy]

Thanks James_s
Now I can see a use of them.
I'll invest significant study time, when I come to building something more complex.

[fubar.gr]

The electrical equivalent of a mux would be a changeover or transfer switch.

Eg, in a typical scenario, a house can be powered either by grid power or by a backup generator. The changeover switch chooses which of the inputs (grid or generator) will connect to the output (house).

[hamster_nz]

I don't get that thing about time slices. Multiplexer just select one of two or more inputs and puts the value of that input on the output. The input signal may change all it wants. This is true for both analog and digital MUXes.

A multiplexer in data comms speak is different to a multiplexer in electronics or digital design speak

Time division multiplexing, wavelength-division multiplexing, frequency-division multiplexing and so on... all are different ways of a single high-bandwidth channel carrying multiple lower bandwidth signals at the same time,

[MattHollands]

There is not much to study, it is a pretty simple concept.

Yes, they are found everywhere, you can't make a CPU without one.

They can be built from logic gates, but inside the ICs they are made out of transistors.

Seeing as you can make any logic gate (with arbitrary number of inputs and outputs) out of muxes I don't see why you say you couldn't make a CPU out of them.

[Brumby]

The basic idea for using a multiplexer is to use one wire to transfer multiple signals.  As hamster has indicated, there are a number of ways of doing this.

The simplest is a "time share" system - where the wire has one signal on it for a period and then another for a second period ... and so on.  You can do this through dedicated ICs or program it through a microcontroller.

My first foray into multiplexing was when I built a digital clock with an HH:MM:SS display.  It used 7400 series TTL and six seven segment LED displays.

To connect the counting output directly to the displays would require 7 x 6 = 42 individual wires.  Multiplexing can do this in 13 wires (or 10 if you take it another step).

So - how to do this in 13?

First you take 7 wires and connect each of them to the same segment on each of the displays.  You then take 6 wires and connect each one of those to the common of a digit - the 'digit select'.  There has to be a little switching control going on - but that's the basic setup.

To display the time, you set the segment signals for the first digit on the 7 segment wires and turn on the digit select to the first digit.  You then do the same for the second digit, then the third and so on.  Once you the to the last, you start over.

At a slow speed, you will see the digits switched on in turn - and it's not very pleasant to try and read the time, but if you increase this speed so that it cycles through all the digits 200 times a second (or more) then human persistence of vision will not notice this switching.

This is what my clock did.

If I wanted to get the wire count down to 10, instead of 6 'digit select' wires, I could have used 3 wires to supply a binary 'digit address'.  3 wires = 3 bits giving 8 combinations, enough for an 8 digit display.  I didn't go that far.

[james_s]

There is not much to study, it is a pretty simple concept.

Yes, they are found everywhere, you can't make a CPU without one.

They can be built from logic gates, but inside the ICs they are made out of transistors.

Seeing as you can make any logic gate (with arbitrary number of inputs and outputs) out of muxes I don't see why you say you couldn't make a CPU out of them.

He said you can't make a CPU without them, in other words you need muxes to make a CPU.

[MattHollands]

There is not much to study, it is a pretty simple concept.

Yes, they are found everywhere, you can't make a CPU without one.

They can be built from logic gates, but inside the ICs they are made out of transistors.

Seeing as you can make any logic gate (with arbitrary number of inputs and outputs) out of muxes I don't see why you say you couldn't make a CPU out of them.

He said you can't make a CPU without them, in other words you need muxes to make a CPU.

That's why you shouldn't read these posts while eating your lunch. My apologies - I totally agree.

[Zero999]

Just one note about analogue multiplexers: they use MOSFETs for switching, so the signal voltage needs to fall within the supply voltage rails. If the signal exceeds either supply rail, then the signal will be distorted and if it's 0.5V outside either rail, then it will become short circuited to the supply, via the IC's static protection diode.

[David Hess]

Just one note about analogue multiplexers: they use MOSFETs for switching, so the signal voltage needs to fall within the supply voltage rails. If the signal exceeds either supply rail, then the signal will be distorted and if it's 0.5V outside either rail, then it will become short circuited to the supply, via the IC's static protection diode.

Analog multiplexers can also be made with bipolar transistors in a couple of different ways and they used to even have specially designed "chopper" transistors for this sort of application if bidirectionally was required.  MOSFET and bipolar transistor based analog multiplexers are also one of the lower performance options at least as far as bandwidth so there is another way to handle it which uses neither ...

Multiplexer is bidirectional MOSFET switch that can digitaly select RLC components and values for MCU internal clock or filters that other logic cant do.
Some old single beam crt oscillooscope used MUX to make it as dual chanel.

In old oscilloscopes, an analog multiplexer was usually referred to as a channel switch but it was very rare at least initially to use transistors for the switching function because excessive capacitance limits bandwidth.  For the same reason, transistors are not suitable for high speed sampling switches.  I know of some oscilloscope families which used JFETs for analog multiplexing but only up to 2 MHz or so.  Later modern designs with MOSFET based analog multiplexers are not much faster.

The solution was to use diodes.  Instead of switching variable voltages, diodes can be used to switch variable currents and this has the advantage that when a diode is off, it can have a very low and controlled capacitance.  The forward voltage drop is not a problem if currents are being switched and even that can be handled if pairs of matched diodes are used as seen in diode bridges which are used as very fast analog voltage or current switches.  Besides high potential bandwidth, another feature of switching currents instead of voltages is that multiple switches can be activated at once allowing the currents to add together at the output.

Later current based analog multiplexers used bipolar transistors in common base mode as switches which is more amendable to integrated circuit design but I know of some examples where this was done with discrete transistors as well.  Some modern bipolar based video multiplexers work this way.

The logic version of this is wired-or logic with open collector/drain outputs and there were also the old DTL (diode transistor logic) families and some of the faster TTL families also use diodes for switching.  The slower and original TTL families use bipolar transistors with multiple emitters instead of multiple diodes.  Bipolar ECL logic also uses current based transistor switching internally.

It took a long time for MOSFET based analog multiplexers to catch up as far as speed but I think bipolar transistor and diode based analog multiplexers still have an advantage there.

One odd thing I have found about CMOS *and* JFET analog multiplexers is that old ones tend to have poor reliability compared to other ICs.  I have no idea why this would be.

[CatalinaWOW]

I don't get that thing about time slices. Multiplexer just select one of two or more inputs and puts the value of that input on the output. The input signal may change all it wants. This is true for both analog and digital MUXes.

Inartful use of language.  In many systems the switching between selected inputs comes on a regular basis, making time slices.  In a static or asynchronous system those "slices" last as long as a given state lasts.  From infinity down.  And since the state doesn't necessarily change at regular intervals the "slices" are not necessarily of the same size.

The next confusion comes from A/D systems, which often select between multiple inputs for conversion, and almost as frequently have an either explicit or implicit assumption that the input does not vary during the selected interval.  Sample and hold circuits are one way to partially address this assumption.  Video multiplexers and other similar systems don't have the same issues, though even there can be transients in system timing if a switch doesn't occur nicely relative to synchronization signals embedded in the signal train.  May or may not be an issue in any particular application.