multiplexing five analog signals

[legacy]

so, I have to multiplex five analog signals in the range of -2V..2V from { Source0, Source1 } to Target0

Target0 <= mux( Source0, Source1 ) on select

What do you suggest to physically implement the multiplexer? Micro-rele'?

I have never done something similar, but I know it's similar to the technology used to implement KVM for computers.

A KVM (that doesn't stand for Kernel virtual machine or similar staff) is a device that multiplex two or more computer's keyboard's and video's sources (from computers) to a physical input/output devices (the physical keyboard and mouse, and the CRT/LCD).

Now, for the VGA, these signals, concerning the RGB, are analogic, while the synchronism is digital, as well as the PS/2 signals are completely digital.

In my case, I have 5 analog signals... I can't use a KVM, but I can "copy" its idea 

[Marco]

What's wrong with some SPDT analogue switch ICs?

[JPortici]

depending on the signal bandwidth it could be as simple as a CD4051

[PA0PBZ]

DC/AC-Frequency?
Accuracy?
Current?
Source&Destination Impedance?
Distance?

[Marco]

CD4051

AFAICS he wants 5PDT, the 4051 (SP8T) is not very efficient to implement that.

[legacy]

Source and Destination impedance is 50 Ohm
There are 400-600 mm of distance between source and destination.
The bandwidth is something like 100Mhz
The signal range is -2V .. 2V, with a current of 20mA max
and good accuracy is mandatory.


Not so easy (for me)

[JPortici]

you are right, i misread.

i though 2x SP5T, not 5x SPDT. So, depending on the other parameters besides voltage range 2x 4053 may suffice

-- Given the bandwidth i would probably use 5x SPDT relays
Otherwise I know that there are analogue switch ICs with 50 ohm paths for high bandwidth applications..

[legacy]

the time to switch from source0 to source1 is also a constraint: must be inferior to 10u sec

[legacy]

What do you suggest to physically implement the multiplexer? Micro-rele'?

When I wrote Micro-rele', I meant SPDT relays, sorry for my lapses of memory 

(edit:
I have to multi-task too much job, switching from field programmable Gatorades (fpga-s), to MPUs, to analog-fish&chips, and back again to digital-stuff .... and why on the why? I also have to write all the boring documentation that needs to be written? ... can't it be done by pretty secretaries?!? ... for sure my customers haven't yet understood this is the best way to see the label out of my office, "out for vacation"   )

[Ian.M]

More questions that need to be answered by the O.P. for us to make useful suggestions:

• Do the signals need to be DC coupled?
  .
• Does your device have both positive and negative supply rails?
  .

You certainly aren't going to get 10us switching times from mechanical relays, and unbuffered analog switches are problematic when you need 100MHz bandwidth and matched 50R impedance.

[GeorgeOfTheJungle]

I've done that for VGA signals with 4053s and it worked wonderfully. Are you sure you need 100 MHz bandwidth?

[legacy]

> Do the signals need to be DC coupled?

there is no DC component, and the min frequency is 20Hz
Say, the bandwidth is 20hz ... 100Mhz

> Does your device have both positive and negative supply rails?

yes

You certainly aren't going to get 10us switching times from mechanical relays, and unbuffered analog switches are problematic when you need 100MHz bandwidth and matched 50R impedance.

Yup. It's a damn problem, apparently (and conceptually) it sounds not tricky, but the more you look at constraints the more you understand it's complex. Anyway, I don't know WHY I have to handle even this task, I am not the right guy,  I was looking at KVMs as examples, definitively they are not the good thing to look at, I mean I know someone solved this problem in some way for the company called "Discreet Logic" that provides solutions for SGI in the 2000s. If you name SGI and Discreet Logic in the same sentence ... you are sure you are not talking about a piece of cake 

[legacy]

I've done that for VGA with 4053s and it worked wonderfully. Are you sure you need 100 MHz bandwidth?

Not sure, I have never seen a single wave crest for any frequencies above 50Mhz by attaching the probe of my spectrum analyzer in what the customer sent to me, and the spectrum of their testing equipment seems condensed around 30Mhz, but it's written in the customer's letter as a constraint "needed" - they wrote -  100Mhz of bandwidth.

I have to phone them to have a confirm

[PA0PBZ]

The DC or 20Hz makes it expensive, normal RF switches start at 300KHz - 10MHz and can be had for under $1. However, because you need DC you need something like 5 of those:

https://nl.mouser.com/datasheet/2/609/HMC545A_545AE-760918.pdf

That's about $3 per piece...

[GeorgeOfTheJungle]

Good, if 30 MHz might do: http://www.ti.com/product/CD4053B these work fine I can say for sure.

[Marco]

Tons of ICs which can do it, how about 2xFSA2367. The ability to handle negative signals without a negative supply is convenient.

[GeorgeOfTheJungle]

If signals are AC coupled you don't need a negative supply.

[David Hess]

Source and Destination impedance is 50 Ohm
There are 400-600 mm of distance between source and destination.
The bandwidth is something like 100Mhz
The signal range is -2V .. 2V, with a current of 20mA max
and good accuracy is mandatory.

Whoa, 50 ohm environment and 100 MHz bandwidth?  That rules out practically all analog switch type multiplexers.  5 inputs is a rather inconvenient number.

Relays are the easiest to understand solution.

Linear Technology made some current feedback amplifiers which can multiplex their outputs that would be perfect for this but they are now hidden and unfindable in Analog Device's selection guides.

[SiliconWizard]

Something like this may be close to what you need: http://www.analog.com/en/products/switches-multiplexers/buffered-analog-multiplexers/adv3222.html#product-overview

[joeqsmith]

I have  no idea what country you are in but you could look at the vintage 59307A off eBay.  Simple to program if you are already setup for GPIB.  Even if you don't, you may be able to just use the relays from them. 

[Gibson486]

I used the SPI version of this before to make a multplexing AVR porgammer.

http://www.analog.com/media/en/technical-documentation/data-sheets/ADG726_732.pdf

[MT]

There used to be analog RF multiplexers around. Harris

[splin]

Tons of ICs which can do it, how about 2xFSA2367. The ability to handle negative signals without a negative supply is convenient.

Omly 10dB of off isolation at 100MHz. Worse still 55pF on capacitance = 58MHz BW when driven from 50 ohm source.

How about OPA3875 triple 2:1 video mux?

Or TS3USB221? 1.1GHz BW, 6pF Con. Approx -55dB crosstalk and off isolation at 100MHz. Vcc max is 3.6V but can switch 5V signals. Ron rather high at 6 ohms, but with only 6pF you could afford to parallel several.

[obiwanjacobi]

I thought that perhaps Video switch matrix ICs would do.
Searching for that I ran into this:
http://www.datasheetlib.com/datasheet/99336/sky13299-321lf-evb_skyworks-solutions.html

[2c]

[Doctorandus_P]

My first go would also be an analog switch aka 4051 / 4051 / 4053. There are lots of variants of these, and some variants go down to 5 ohm series resistance. Just saw a bunch of them in the EEVblog Teardown of the Rigol 1054Z ( EEVblog #650 or so from 2015), where these are used in  the frontend.

For signal integrity you probably want to add some buffer or line driver.