Need >= 30 Mhz 16x16

[johngardner58]

Hi I am still waiting on feedback from our engineer as to whether we can relax some requirements.  If not it is a moot point.

I cannot go into detail about the hookup as this is a public forum but we do have some test cases for bandwidth that if the device can meet we'll be happy assuming that the other requirements can be relaxed.

[Now_With_Underscores]

Hey John,

Feel free to send me a private message and we can discuss your design requirements.

[Icefan11]

Hello John,

I work with John on a new test system(s), and I have a few questions about the 2532B:

1) The 2532B spec stated that the maximum current = 0.5A  it is not clear to me that the spec is for average current, peak current, peak to peak current, or spike current, could you clarify it for me?

2) The 2532B spec stated that the BW >=30MHz, it is not clear to me that the BW spec is for reed switch or for the module.  Could you clarify it for me?

3) If the BW >=30MHz at -3dB point specification is for the module, then what is the base frequency for your calculation?

4) What is the test setup for your BW test?  (The 2532B uses two high density connectors in the front panel, that kind of breakout box do you use?)

5)  Do you have a frequency response plot of the 2532B?

Best regards,

Fong

[johngardner58]

Hi John S.

For clarification:

1.  When Fong said "I work with John"  he meant me not you.  And this is correct, we do work together.  He is the engineer who will actually use the card under discussion

2.  Fong has a much more extensive hardware background than I do.   I have had more experience working with NI PXI equipment in general.

3.  When he wrote "module" you should substitute "PXI-2532B". 

4.  As I mentioned before we tested a competing product and performed a frequency scan.  Not only did it not meet the published -3db specification for 15MHz at a single crosspoint but there was a significant "hole" in the plot even before the sub-par -3db point.

5.  Fong will provide more details as to what we need exactly.

[Now_With_Underscores]

Hey Fong,

1) Short answer: Do not exceed 500mA peak while switching or 500mA rms while static.

1) Long answer: 500mA is both the peak and rms current, depending on the switch state.  When switching, exceeding 500mA - even for a moment - will quickly decrade the relay contacts. The typical failure mechanism for overcurrenting relays while opening/closing is that they'll begin to 'stick', which means they'll fail to open when you tell them to, as the contact surfaces are literally welded together.  A slight tap will break the contacts apart (i.e. physical shock, e.g. striking the chassis with your hand).  Obviously, you don't want to have to pay someone to hit your chassis each time you open a relay, so I strongly encourage you not to exceed the current spec.  Of course, there are multiple other failure modes due to current inrush while operating the relay, but they're all exponentially likely based on the switching current.  500mA is also the rms limit (or 'carry current') when the relay is closed.  This current limit is purely a function of internal self heating.  Exceeding 500mA rms current could cause the reed relay contacts, internal traces, and/or wires associated with the 2532 melt.

2) The BW spec is for the complete module, but is only typical (e.g. not tested in manufacturing, so not guaranteed).  Also, the specification is ONLY when a single crosspoint connection is made, into a 50Ω load, so it represents the 'best' case you could achieve with this module.  [aside: this is the industry standard, which is why that 'other' company's switch module didn't meet their spec when y'all tested it].  From what I've heard so far, your application is not the ideal case (multiple simultaneous crosspoints), and thus I wouldn't expect to achieve anywhere near 30MHz BW.  I'll note here that we're fighting with transmission line theory here, not a 'subpar' design or anything like that.  I believe that what y'all are attempting to achieve - a true 16x16 matrix with multiple (all?) simultaneous crosspoints - is not physically possible.  There are pseudo-matrix topologies that would work, but I'll need to know much more about your setup before we can optimize the design.

3) I do not know the exact frequency, but it's very low (if not +DC).  Of course, the bandwidth is incredibly linear at low frequencies, so the absolute low frequency we started at for our tests isn't important.  Please let me know if you need further clarification here, as I'm not 100% sure I understand what you're hunting for.

4) I'd need to go review V&V data to determine the exact test setup, but I'll note that each individual terminal block's specification sheet notes it's overall contribution to BW.  For example, the TB-2642 - the module you'd likely use for your matrix - lists a -3dB BW of just 15MHz, into a 50Ω load, which again is with just a single crosspoint.

5) We don't have any published frequency response plot.  Even if we did, it would be typical and not warranted.

[johngardner58]

Thanks for your feedback John,

I would like to jump in to say that according to Fong's testing we did not see much BW difference between one and two crosspoints (connected together). Perhaps if we had 6 xpts (again connected together) it would still allow 10MHz BW.   We probably will not use the terminal except for prototyping.  We plan on building our own interface where we can carefully control the layout.  We may even avoid the cable by pluggin in this interface directly into the PXI card.  The one big disadvantage (for us) with this card is that because it is so configurable it needs to expose more pins and two connectors.   A different pickering unit (than the one that failed our tests) which is very similar in specs to the 2532b just exposes the matrix edges and so this makes for an easier connection.  However we haven't found working with them as easy as with NI (My rack is all NI.  Fong had to pick some other cards for his rack to get the specs he needed)

Assuming that we can't get an evaluation card for testing,  if we gave you some specific tests, would you be willing to run them?  We don't want to plunk down 6 grand and find the card won't work.  Or is an eval a possibility?

[Icefan11]

Hello John S,

Thanks for the very good answers.

2) I used a network analyzer to double check my measurement results, and found the switch matrix resonate at less than 4MHz with on single cross point.  I war surprise by the network analyzer test results.

3) I expected that your based frequency is between 100 Hz to 10 kHz. 

4) We plan to design our own interface board/card to the switch matrix to minimize all the cabling. But the 2532B use two connectors, which may cause some difficulty of designing the interface board for the 2532B.

5) I want to see the flatness of amplitude response and the phase response.  I do not like surprises.

Regards,

Fong

[Now_With_Underscores]

Hey John G,

For terminal block inspiration, take a peak at the PXI-2531's terminal blocks, which utilize a similar front panel connector, but which break out into SH-68-68 connectors instead of the ribbon cables used on the PXI-2532/2532B.  Unfortunately, the 2531 doesn't natively support the 16x16 matrix you're looking for, so I can't recommend that module.  Of course, I also can't in good conscience recommend the 2532B y'all are looking at either, due to the points I've raised previously, particularly as they relate to your application.

FYI: To achieve the 2532's industry-leading crosspoint density, we were not able to configure the topology on the module, hence the need to configure the topology on each specific terminal block.  Thus, the front panel connector requires connections for each of the 2532's native 2x16 matrices.  Typically, we configure each specific topology with dedicated configuration relays, but we simply didn't have room on this module.  Of course, we could have released multiple '2532' modules - each with a dedicated topology - but that would have required substantially more R&D effort than the 2532's terminal block solution.

Please contact your local NI representative to discuss evaluation options.  For the record, I'm not convinved the 2532B will work for your application, and will reiterate that I'd be more than happy to look over your project deliverables to see if we can develop a better quasi-matrix solution.

[Now_With_Underscores]

3) The low end frequency sounds reasonable.

4) To account for mechanical uncertainty, our board uses a flexible PCB strip.  Mounting a PCB orthogonally to the two connectors could result in alignment issues, which will place strain on the connectors.

5) See my previous post to John about obtaining evaluation units.

[johngardner58]

Jumping in here.

@John S.  We've already analyzed for a quasi-matrix solution and could not find one.  We've looked at increasing the number of instrument channels to reduce the switching need. We've considered doing some stuff in an adapter unit.  Nothing really works out.  I've even looked at LXI and SCXI.  Also we have very little control over the interfaces that we have to connect to they are specified for us.

Unless of course we simply hard wire the exact set of switches which is what we may have to do.