1. Home
  2. Electronics Design
  3. Multiplexing and switching

Multiplexing and switching

Hi,

I have 32 current sources and I have to connect them to 144 different resistors of same resistance value of 2kohm. One source should be able to connect to any one of the144 different resistor. But the total number of connection will always be 32 out of 144 connections. The rest of the resistors will be unconnected or floating.

I am looking for an easy way and elegnt way to do it. The maximum current is 2mA per connection. All 32 current sources are independent of each other and compliance voltage acoss each current source is 20 volts.

I would like to control the switch digitally using some data aquisition board. I might get 8 digital lines to control this switching. Can anyone suggest some mutiplexer, some kind of marix chip etc. I will really appreciate it.

Thanks

John

Reply to
john
Loading thread data ...

What exactly ware you trying to achieve here? Why not just have a single programmable current source for starters? Why is the current source just driving a resistor? Presumably you can parallel any number of these 2K resistors?, otherwise what's the point? (not that there seems like any point in only switching them to individual resistors to start with)

More info please...

Ground independant too?

The supply voltage might prove troublesome with an IC solution. Something like this AD crosspoint switch is worth looking at, but I haven't looked at the details:

formatting link

You didn't mention cost, space, or current consumption? Have you considered a common bus relay matrix?, perhaps something like this:

formatting link
(jump to around 7:30)

Dave.

--
---------------------------------------------
Check out my Electronics Engineering Video Blog & Podcast:
http://www.eevblog.com
Reply to
David L. Jones 

--
Looking at it from the point of view of the loads, if any one of them
has to be able to connect to any one of 32 sources, then the total
number of switched points is 144 loads * 32 sources = 4608.

Since you have 32 sources, let\'s say that we use 32 bit shift registers
to do the switching and write in a word which will set the bits where we
want the switches to be ON, and clear them where we want them to be OFF.

OK so far?
Reply to
John Fields

It's been too long ago and the ones I used are mostly history because medical ultrasound has gone digital. You need to look for "crosspoint multiplexer" chips. But brace yourself, they have become very expensive boutique parts.

Of course you could cobble it together with a boat load of chips of the CD4051 type but 20V will be a stretch. You might want to contact Supertext if they have a specialty chip that isn't advertised on their web site. They specialize in highrr voltage ICs.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Reply to
Joerg

A bit more info on the application and it's requirements would help. Sounds like a ladder network ?, but there may be a better way to design this to reduce the number of resistors and switches, possibly by using multilevel sets, or different resistor values to allow binary code drive. Also, what sort of accuracy are you looking for ?...

Regards,

Chris

Reply to
ChrisQ

--- Here's how I see the switching matrix itself: (View in Courier)

Where S1 through S32 are the constant current sources, L1 through L144 are the loads, and "-O O-" are the switches.

S1 S2 S3 S4 . . . .S29 S30 S31 S32 | | | | | | | | | | | | | | | | L1---+--O O--+ | | | | | | | | | | | | | | | | +--O O--|---+ | | | | | | | | | | | | | | | +--O O--|---|---+ | | | | | | | | | | | | | | +--O O--|---|---|---+ | | | | | | | | | | | | | . . . . . . . . . . . . . . . . . . | | | | | | | | | +--O O--|---|---|---|----------+ | | | | | | | | | | | | +--O O--|---|---|---|----------|---+ | | | | | | | | | | | +--O O--|---|---|---|----------|---|---+ | | | | | | | | | | +--O O--|---|---|---|----------|---|---|---+ | | | | | | | | | | | | | | | | . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . | | | | | | | | | | | | | | | | L144-+--O O--+ | | | | | | | | | | | | | | | +--O O------+ | | | | | | | | | | | | | +--O O----------+ | | | | | | | | | | | +--O O--------------+ | | | | | | | | | . . . . . . . . . . | | | | | +--O O-------------------------+ | | | | | | | +--O O-----------------------------+ | | | | | +--O O---------------------------------+ | | | +--O O-------------------------------------+

It works by selectively making the 32 connections desired from each 144 resistor load bank, and breaking all of the other connections in the entire matrix.

I have to go do some stuff, but I'll be back in a while and post how to do the switch selection with shift registers.

Reply to
John Fields

On the face of it, this will require 32 x 144 switches. The only elegant way is to ditch those '32 current sources' and instead use 144 current sources. The condition that only 32 are set at nonzero values doesn't lead to any simplifications I can see.

Reply to
whit3rd

A non-blocking spanning switch network with 8 4x4 ingress units and 4 8x36 middle units and 36 4x4 egress units would use 1856 switches vs the 4608 switches of a 32x144 matrix. See

formatting link

Surely the OP has a handful of undisclosed conditions that would rule out reasonable approaches like that. :)

Reply to
joe

able

rest

Thanks. I was looking for ways to factor the task but could not get below about 4500 relays. Kind of cost prohibitive at ~$2 a piece.

I wonder what switch count you get with a 5 layer switch?

Reply to
JosephKK

If you don't mind a blocking architecture (i.e. you may have to change existing connections in order to add one), you can do quiet a bit better than MxN. Haven't got a reference for the minimal blocking configuration, sorry.

I'd have thought that making 144 current sources was probably cheaper than the switch fabric--how about just programming them via some bus, and just changing the programmed current values?

Cheers

Phil Hobbs

Reply to
Phil Hobbs

different

able

rest

=20

Mucho agreed, 1856 (let alone 4608) switches gets expensive really fast.

Reply to
JosephKK 

--
Sorry; I lost track of the thread.

Reading back over it, though, it seems the 144 current source idea is
much better than the 144 X 28 relay scheme.

If you\'re still out there, how accurately will the current need to be
set?

JF
Reply to
John Fields

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.