Switch matrix card

[switch78]

hi

i am looking for switching matrix card one of my application. my requirement is input is 100 signals can be switched to 320 output lines. can you able to suggest switching mechanism using NI modules

[Now_With_Underscores]

Hey Switch78,

How do you need to connect the signals?  Can all of the signals that you want connected be connected together?  If so, that's going to be the cheapest solution, and will require two muxes: the first mux is 100:1, and plugs into the second mux, which is 1:320.  Knowing nothing else of your system (voltage, current, frequency), I'll propose the PXI-2575.  Theoretically, this is only 420 relays.

If you need to connect multiple channels simultaneously without shorting them together, what is the most number of simultaneous connections you would need to make?  Depending on that answer, I'd recommend using multiple matrix modules in the NI-2800 switchblock carrier.

Of course, if you want the capability to simultaneously connect any number of signals to any other signals, then you're going to need a VERY large matrix: 320x100 is 32,000 relays and would be very difficult to actually manufacture. 

Let us know more specifics and I'll hone my recommendation.  Happy clicking!

[switch78]

Hi,

My system requirement is to simultaneously connect any number of signals(i.e.. Input 100) to any other signals(i.e.. Output 320),I think its VERY large matrix: 320x100.

Signal type is TTL with frequency 100MHz and analog signals of 100Khz. and current is less than 1A.

Regards,

balakrishna

[crossrulz]

I have never heard of a 100 row XPT.  I usually work with a 4 or 8 row matrix.  Are you actually hooking up 100 different connections all at the same time?

[Now_With_Underscores]

Hey Balakrishna,

You could easily spend $2,000,000 building such a large matrix.  A traditional matrix is out of the question due to the several hundred MHz of analog bandwidth your requirements mandate; the stubs on such a large array would destroy your high frequency performance.

To get the bandwidth you're looking for, we could theoretically build a blocking matrix, but that would require 420 1:420 multiplexers (~$8,800,000), and each multiplexer would need to consist of several smaller multiplexers to keep the bandwidth up.

Tell us more about your application so we can determine a simpler routing solution.  There's usually a much easier method once we know all the requirements. 

[switch78]

Dear sir,

in one of my system i am using NI PCI-6225,16-Bit, 250 kS/s, 80 Analog Inputs, module.

now i am facing some problem in reading voltage i.e when i read voltage by using API, we could not able to get constant voltage, their is difference in one reading to next reading even though input is constant(upto module). difference is in terms of 100mV i.e(5.110354543, next reading is 5.205435436),

since we are using 16bit card, voltage should read in terms of micro volt but here we are getting in terms of milivolt.

kindly suggest regarding this, 

thanks & Regards,

balakrishna

[crossrulz]

Are you reading a single channel or multiple?  Did you try measuring your voltage with a scope?  How long of cables are you using?  Do you have any big machines (motors) around?

There's a ton of ways you could be getting extra noise in there.

[Now_With_Underscores]

Because your analog input card multiplexes all channels to one ADC, if you sample too quickly, then you'll see aliasing from your other channels.  Try measuring just the one channel and see if the issue goes away.  If it doesn't, scope the signal and see what's happening.  If if does go away, then sample the previous channel and the channel under test... does the issue come back?  If so, the internal capacitance on the ADC is holding charge from one signal to the next.  You can improve the settling time many ways... the easiest method is to reduce the resistance on each input channel, if possible.  Another method is to slow the sample rate down, etc.  Let us know what you find with these simple tests and we can elaborate further.