Need >= 30 Mhz 16x16

[johngardner58]

Hi,

Looking for PXI   16x16 1-wire  with <= 0.5Ω and enough bandwidth to carry a true -3Db 10Mhz with multiple switch closures  say 3 on a row.  It would be nice if it fit in a normal slot (I think 1U).   I am aware of the 2532b but not only is it way more than what we need, it is huge 3U and we would have to have a separate rack also its switch impedance is higher than what we would like.

We really need a minimum of 10Mhz which is why I have been looking at 40Mhz and above so that when we use multiple switches we can still get the bandwidth.  

[johngardner58]

Oops 3U IS standard.

We think we can use the PXI-2532b but the bandwidth question is a biggie.   

1. It really needs to handle the >= 30MHz at least at one node.  We have a pickering card that says it handles 15MHz but at 3db it is much less than expected even with a single Xpt.
2. It would be nice to know in what we get when we start throwing relays.
3. Let's say our topology is 16x32 1-wire.  And I do r0:c0-c5  (6 switches) what kind of bandwidth will I get measured at each of c0-c5 separately.  Do certain paths have significantly less bandwidth than the others?  If we achieve >=10MHz with 6 connected relays we'll be happy.
4. We really need the density this card provides too.

[Now_With_Underscores]

Tell us more about your source and DUT.  Are they 50Ω?  I imagine that with a 500mΩ requirement, they're not.  Thus, we'll have incident reflections from your source into our matrix, through our matrix, and again as we leave our matrix.  Each additional source/load you connect at the same time will complicate matters further.

If you're attempting to pass a 10MHz sine wave through our switch (the analog definition of BW), then you should expect variable attenuation and (possibly severe) reflections, depending on how many DUTs you connect simultaneously.

When SI is important on faster signals, I typically recommend a blocking matrix design.  Blocking matrices allow NxN connections well into the GHz, but unfortunately, only allow one connection on any particular path.  Take a peak at this KnowledgeBase I wrote for an overview.

Of course, 10MHz isn't typically fast enough for me to recommend a blocking matrix, but when said matrix is 16x16, that warrants more consideration.  If you can use a smaller matrix, I'd recommend the PXI-2541, which is 8x12.  Note: You can expand the matrix to 8x24, 8x36, etc, with multiple modules, but this will also affect BW. 

How do we overcome all the stubs?  Glad you asked!  Inside the 2541, we've got a bunch of smaller 2x4 interleaved matrices, thus reducing the stub length.  However, all those additional relays certainly increase the ESR: 2.1Ω.

You might ask "But I don't need 300MHz."  Yes. You. Do.  Our 300MHz spec is routing a single column to a single row in a single module.  Once you start connecting multiple rows and columns at the same time, your bandwidth is going to drop off like the great recession.  Unfortunately, I can't comment further since I don't know about your test system and I haven't gathered this particular empirical data. 

[Now_With_Underscores]

Let us know more about your test system, too.  I can usually decrease matrix size requirements with a little bit of thought... unless you truly need to connect everything to everything all at once.

[johngardner58]

Hi 

Normally for bandwidth we are using a Scope and Function generator (PXI) so I don't think the 500mΩ is an issue.   The low ON resistance requirement is for other tests where a DMM might be used.   We really need the 16xN not 8xN configuration so 8x32 isn't good for us.  The 16x32 is nearly ideal with a few spares.  Test signals will be DC and AC signals.

I don't know what a stub is?

[johngardner58]

Also our DUT is typically loaded with 500Ω loads.   And the blocking matrix won't work we need multiple closures on a given row as we may drive a single signal to multiple inputs for say CMRR tests.   The DUT is a medical implant.

[fromm8]

Hello johngardner58

I was filtering the search of the PXI switches according to your requirements, and as you said the 2532B seems to be fine, I did not find others. Could you repeat me what is the main issue with the 2532B?

Regards

Frank R.

[johngardner58]

Hi,

There are a couple of things that are outside our requirements:

1.  Current carrying capacity is a bit lower than our requirements  600 mA

2.  Switch impedance is a higher than our requirement of < 0.5Ω

3.  Bandwidth may be an issue.   We need a minimum of 10MHz with multiple switches.

4.  Our absolute matrix minimum is 13x11.  This one is solvable by the 2532b.   8 by n expandable matrixes won't work.  Also the fact that it is actually 16x32 actually simplifies our testing because we can combine two diffent hookup scenarios into one.   We are required by regulatory agencies to connect to regulated interfaces of which we need two.  with a 13x11 matrix we would have to do our tests in two stages each with a different test interface.  With the 16x32 we can connect both interfaces.

Items 1 & 2 we might be able to change our requirements and limit the test coverage.   Some of the tests are regulatory tests and we have next to no wiggle room on the specs.

If we we can get by items 1 & 2 we have a specific series of bandwidth tests that if the device is able to complete would mean that it is suitable for our use.   

[fromm8]

John,

I repeated the search and the closer device according to your needs is the 2532B again. Point number 2 is hard to accomplish.

Regards

Frank R.

[Now_With_Underscores]

Definitely eager to know more about your system, so we can see how to optimize.  What instruments do you have connected to the rows?  Columns?  Can we subdivide the columns?  Let me know what you want connected where (maybe give a few examples) and I'll see what we can do.