multiport switches or multiplexers?

[mstacho]

Hi all,

I need to do the opposite of what a multiplexer does (no, not a demux): I have a single input, but it has to be able to be routed to multiple outputs.  So it's like a multiport switch -- the current can only flow through one pathway at a time, but I need to be able to select from, say, 4 pathways.

Will a standard analog multiplexer do the job?  I'm just wondering if it'll require reverse biasing the gates and therefore not work.

Any thoughts?

MIKE

[waltr]

but I need to be able to select from, say, 4 pathways

Or do you mean 'to' 4 pathways?

You may also need to terminate the input of the circuits that the signal is not routed to. For that you may need two analog switches in a NO & NC configuration.

What is the voltage levels of the signal?
What is the frequency range of the signal?
What is the required impedance of the inputs to be driven?
What is the output impedance of the input signal?

There are a number of analog switches that may do what you need.
Check these part numbers:
ADG1211
ADG1421
DG419
MAX4622
DG411
are a few of the analog switches I have used.

[Soeren]

Hi,

I need to do the opposite of what a multiplexer does (no, not a demux): I have a single input, but it has to be able to be routed to multiple outputs.  So it's like a multiport switch -- the current can only flow through one pathway at a time, but I need to be able to select from, say, 4 pathways.

So, in what way is it not a demux?

Will a standard analog multiplexer do the job?  I'm just wondering if it'll require reverse biasing the gates and therefore not work.

You ask about an analog mux, but is the signal analog or dogotal?
I've seen analog mux'es used for digital signals ad nauseam, but if the signal is digital, a digital mux may be a better choice.

If analog, what are your demands on signal quality (say max. THD allowed and such)?

[mstacho]

Thanks for the replies.  IN fact, I think it is a demux and I just had one of those "over complicate the works" processes.  It's actually simpler than anything relying on signal strength: the lines have to carry a constant voltage that will be anywhere between 5 and 7.5V depending on the rest of the setup (hence why I considered analog mux's in the first place, although I guess the exact voltage doesn't really matter?).

Basically, I have a 4x4 array of sensors accessed by 2 sets of 4 pins.  You put a voltage onto one of the 4 pins in set 1, and attach the other pin in set 2 to a circuit (in my case, just a voltage divider).  By sequentially reading all the pins, you get to address all 16 sensors.  But you have to make sure you only power one sensor at a time, or else you'll get crosstalk and useless data. 

Set 2 is familiar to me -- it's just a multiplexer, so the 4 pins go into the 4 inputs of a mux, and the signal comes out the output once I set the address right.  The hang up was that in set 1, you attach what was once the output as the input (the 5 to 7.5V line), then attach what was the inputs to act as the 4 outputs that go into the pins in set 1.  This way, you can make sure that only one pin in set 1 is attached to the voltage source at any time. 

But I have a feeling that a simple mux will work just fine, but operated "backwards" so it's a demux, right?  In other words, am I right to say that the signal can flow from the output to the input just as easily as they flow from the inputs to the output?

[waltr]

mstacho,
 You really need to read the data sheets and also any app notes from the manufactures.
Some analog switches are fully bi-directional whereas others are not specs as such but may be.

I'm currently looking at the Data sheet for the ADG1208/1209 analog mux. In the Pin function descriptions the analog switch pins have this description:
"S1A   Source Terminal 1A. Can be an input or an output".
"DA    Drain Terminal A. Can be an input or an output".

So the answer is in the data sheet.

Does the analog switches resistance affect your circuit performance?
Did you know that the analog switch resistance is voltage dependent?
Did you look up and read the data sheets for any of the parts I listed above?

[mstacho]

Yeah, I guess I'll hit the datasheets (which admittedly I should have done before, but for the 4051B chips I have I didn't find any information about bidirectionality). 

Thanks for the info, at least now I know it's possible.  :-P This really was just one of those episodes where I overcomplicated the explanation in my head.

MIKE

[waltr]

This really was just one of those episodes where I overcomplicated the explanation in my head

Yep, that can be easy to do.

If you don't understand something in the data sheet post the question and we'll help.

I just looked up the Phillips data sheet for the 74HC4051.
http://ics.nxp.com/products/switches/4051/

and it states applications as: "Analog multiplexing and demultiplexing"
So this implies bi-directional.
Also in the pin description for the Yx and Z pins are: "input or output".

This is a pretty good data sheet as it shows internal circuits, the 'Typical Ron' as a function of input voltage, leakage, isolation, frequency response and lots of other specs.

Just don't assume that another manufacture's part has the same specs.

[Soeren]

Hi,

I think there are better versions with lower on-resistance nowadays, but with a 4052 (dual 1:4 mux) each of the two mux'es in a chip could be paralleled to half the on-resistance of ~470 Ohm at 5V.

Did you consider other methods?
If you were addressing 4-20mA "transmitters" with open collector output, at each line, they could all be connected to the same point and the address would determine which one was "talking".

[mstacho]

Well, I'm definitely open to other methods.  Using a mux seemed the simplest, but let me explain the setup a bit more:

there are 16 force sensing resistors per array with a range between 300Kohm and 300ohm depending on the force applied.  If you want to access the resistor at location (i,j), you put a voltage on column (i) and open up row (j) to be read into your circuit (in my case, it's a voltage divider, standard way to read FSRs as far as I know).  So, even though all 4 in a column are connected and all 4 in a row are connected, current will only flow to the resistors that are currently addressed.

Since the resistances are so large, and since getting the sensors down to 300ohm would require way more force than my robot can handle anyway, I figured that demux'ing is the best option.  It also allows me to address the resistors sequentially by setting up a counter circuit.  My idea was to use a binary counter to run through all my sensors (with the count acting as the addressing signal), read them, and store them in an array, probably using an Axon or a dedicated PIC or something that's sending UART data to the Axon.  

Deal is, I have 6 such sensor arrays working at any time (and more in the future), so setting up 18 address pins on the axon is not feasible or necessary, as long as I'm willing to live with a lower reading rate of, say, 50Hz, which is way more than I need anyway.  I suppose I could connect the addresses together so that I read in parrallel (that is, whenever I want to read sensor 1 from array 1 I'll actually be reading sensor 1 from all the arrays), but part of the research is to play around with adaptive data updating, with the eventual goal of dealing with much larger arrays with much smaller components, so it's best if the arrays are all separate.

So yeah, my best solution was Axon -> counter -> mux -> sensors, but any suggestions are welcome  

**EDIT: Got the number of address pins needed wrong.  It's 18, not 48.  Still a lot :-P

MIKE

[waltr]

Mike,
 It sounds like to be a good way to do what you need.
A MUX to select and energize a sensor then a De-MUX to read the analog voltage from the sensor.
It could be wired to select both the MUX and De-MUX with the same addressing lines.

There are quite a number of analog switches and muxes available with different specs. Since you will be measuring the voltage of a 'resistive divider' the Ron of the mux can have an adverse affect on the measurement. You may be able to 'calibrate out' this affect.

I would like to see a schematic of the circuit.

[mstacho]

At the moment the circuit is in the conceptual idea stage, since I don't yet have the sensors anyway.  I'll be sure to post it once I get things going a bit further.

Thanks!

MIKE