NI PXI equipment (or similar)

[forrestc]

It's time for me redo some automatic test equipment in the assembly line.... or probably more accurately stated:  make the test equipment more automatic.   Today, some of our products are tested with a semi-automatic process, but most are tested the old fashioned way with standard test equipment.   For various reasons, this needs to change - but the biggest one is that failures are being missed just because there are humans involved in making the pass/fail decision.

I keep coming back to using some used PXI (or similar) instruments.  However, I've never actually used any of these, so it's a bit daunting to even know where to start.   What I have figured out so far (in summary) is that these are essentially rackmount, card caged, pci/pxi bus computers.  There is a processor, and the 'i/o cards' basically plug into a PXI bus which is basically card-cage PCI.   I get the impression that the preferred software to run on the PC is some sort of windows plus labview.   This feels somewhat familiar for me as I've done process control on an old ISA bus industrial computer with a passive bus, and before that with a S-100 bus.   So, the base of the hardware seems reasonably simple.

Where I'm lost is knowing what pieces I really need.  And how these things usually fit together to make a usable test system.  What the licensing situation is.  Is labview the right application for this?  Is there something else easily usable?  And so on.

I can give a bit more details about the application if necessary, but let's just say we make full use of a couple variable power supplies, various continuity and voltage meters, and a low-end scope (Think TDS2014).   Eventually we'll need to be doing some more advanced testing (such as switching in a scalar network analyzer), but I'd rather start with some of the simpler measurements.

Any ideas as to where I should start with this?  I'm mainly hoping someone can point me toward a resource for this basic getting started piece that I didn't know existed.

[awallin]

your NI sales rep should be eager to tell you what you need.

afaik at least the older real-time PCs ran some version of vxWorks as the real-time OS. This is programmed using a separate "target-VI" lab-view program that can be developed on a normal PC and the deployed to the real-time target. If you want to log data and monitor what the real-time box is doing you write a separate "host-VI" that runs on some PC on the same network and talks over Ethernet to the real-time target.
You probably need more or less full LabView license and the real-time module to make use of this. For high speed stuff they have FPGA:s which ofcourse requires their FPGA-module (you then have 3 separate labview diagrams: PC, real-time target, fpga)

some of their newer products are using linux, some version of preempt-rt I think. Probably not much cheaper on the license side (for you at least!).

[Galenbo]

...Where I'm lost is knowing what pieces I really need.  And how these things usually fit together to make a usable test system.  What the licensing situation is.  Is labview the right application for this?  Is there something else easily usable?  And so on...

Another question you could pose: Is your organisation fitted for the NI-Labview closed "solutions" organisation?

The first step is to get the right Labview Expert/Consultant to you, and don't generalise, nobody is NI-Vision AND Labview-compactRio, Rt AND Labview OOP expert.

If this is your first step into their solutions, perhaps it's better to ask a complete solution to one of their certified companies, and pay extra to get everything documented and explained to you.

[HAL-42b]

What a delightful corporate lock-in. 

[han]

ouch PXI..it's a nightmare to repair it without card extender

[forrestc]

your NI sales rep should be eager to tell you what you need.

Which is probably why I can't find the information I'm needing.  I don't have a NI sales rep.  I don't *want* a NI sales rep.

It seems amazing that there isn't any path to use of this old PXI gear without it going through NI.

[forrestc]

Another question you could pose: Is your organisation fitted for the NI-Labview closed "solutions" organisation?

The answer is ... NO.

The hope is something like this:   There's lots of used PXI gear out there, most of it NI.   I would love to make use of it without involving national instruments or a consultant. 

Is this even possible?

[forrestc]

What a delightful corporate lock-in. 

No kidding.   "Q: Lots of used gear out there, how do I use it?"  "A: Call NI or a consultant".   NI has got to be loving life.

[dom0]

Weren't there GPIB/VXI-11 controllers for PXI crates?

[joesixpack]

I absolutely hate Lab View.

You don't need to be locked into NI.  I've built PXI test systems with custom coded software running under Linux.   You just buy a chassis, buy the cards and buy a MXI card.  The PXI chassis becomes an extension of the host PC and the MXI card becomes a PCI bridge.  Most vendors have SDKs and you code code up your own software.
 

[Lukas]

Do you really need the compactness of PXI? Why not use regular benchtop instruments and talk to them using USB/LAN/GPIB? All of these interfaces are standardized, so zero vendor lock-in.
You can virtually use any programming language you want, you only need VISA bindings for it.

[Galenbo]

Another question you could pose: Is your organisation fitted for the NI-Labview closed "solutions" organisation?

The answer is ... NO.

The hope is something like this:   There's lots of used PXI gear out there, most of it NI.   I would love to make use of it without involving national instruments or a consultant. 

Is this even possible?

Of course, it's possible. depending on your time, effort, experience,...
The big difference is: Do you take responsability for everything yourself, or do you have to document and defend choices to superiors, collegues?

I had success with many homebrew "solutions" partly grabbed from their forum etc, mostly works fine.

But also had huge problems with THEIR Cognex camera drive after a mandatory Labview update, witch turned into some "exception" when I switched between image and settings, (documented). Try to explain the boss the only solution is a workaround with more hardware. Problem too: I wasn't in power to FORCE them to come it explain to my boss, because i wasn't the guy that was in the position to SIGN/NOT SIGN for the yearly +4K they want.

I also had problems with THEIR XP-Modbus driver to access a linear drive. Mostly this worked like a rocket, Sometimes communication was slower than 2sec/message.
And sometimes: The application didn't want to start. Like in: You add a blinking led, and even that doesn't blink.
So another workaround that first start another application, from there, start the "real" application, wait for some heartbeat and then close the first application.
No info, no help from the NI officials off course. But no problem to get them for a demo of their product line.

Try to explain the above to a collegue that doesn't know the difference between a PC and a PLC, or the difference between C and Java.
I had to. Imagine.

My observation is you get help in the beginning, and help to guide you to your purchases. Help if they cooperate in a University, Nasa or media project.
But if it really goes wrong, and they can't find the problem in an hour, you're alone.

My conclusion is I would never pay again for their license BEFORE everything works and is ready to sell to customers.

[forrestc]

Do you really need the compactness of PXI? Why not use regular benchtop instruments and talk to them using USB/LAN/GPIB? All of these interfaces are standardized, so zero vendor lock-in.
You can virtually use any programming language you want, you only need VISA bindings for it.

The compactness has sort of been the attraction.   Actually, more accurately the density of certain cards available for pxi.   As a specific example:  I have one product in particular with over 300 connector pins.  Somehow I have to multiplex these (in varying combinations) into the appropriate test equipment.   To do this, I can get a PXI-2530B mux/matrix switch module which has 128 reed relays, all the driver circuitry, all in a nice compact unit for somewhere around $100 on ebay, maybe a bit more if I'm impatient.  Assuming I can live with being able to switch any *two* of the wires to a external dmm or similar, I should be able to get away with 5 of these to build a 320x2 mux (any one of 320 channels into either of two outputs).  Adding another mux off of the end I can then connect those two outputs to any one of 64 test equipment.   All for $600 of used gear.   Plus the cost of the chassis and cpu.   Operate and release time of <600uS, and operating life of over 100 million cycles (minus prior life cycles) makes a very economical and fast mux.

Buying that same mux card (only one) through official channels is over $2K.   

What I have discovered recently though is that evidentally these all have a standard VISA driver available, so you can run any tool which will utilize a visa driver.  With that discovery, utilizing something like a python library for visa seems more likely.

[forrestc]

Of course, it's possible. depending on your time, effort, experience,...
The big difference is: Do you take responsability for everything yourself, or do you have to document and defend choices to superiors, collegues?

This is in an environment where I'm pretty much the first, last, and only person I have to answer to.  And I have pretty extensive knowledge in doing this type of stuff.   As I posted above, I finally found out that all of these instruments are supposed to have standard visa drivers.   This opens up many possibilities - the most likely one will be for me to be able to write this stuff in python, which is a good option for me (instead of also learning labview).

[Galenbo]

... I finally found out that all of these instruments are supposed to have standard visa drivers.   This opens up many possibilities - the most likely one will be for me to be able to write this stuff in python, which is a good option for me (instead of also learning labview).

Good. there's also the LabWindows/CVI possiblity.
But I have no experience with that.

[View[+]Finder]

Well here's some good news for all the 'pixies' who would like an easy way to control NI gear: Python is now (for a few years) being embraced by NI as an alternative to LabVIEW, etc. There are modules on Github for PXI DMM's, AWG's and other instruments that bridge the device API (hidden) to the Python script.  https://github.com/ni/nimi-python

Code: [Select]

import nidmm
import time

PXI4071 = nidmm.Session("PXI-4071") # set this in NI MAX (nidmm is the name I use for this Session)
PXI4071.configure_measurement_digits(nidmm.Function.DC_VOLTS, 10, 7.5) # This is for a 7.5 digit DMM, 10v range
#PXI4071.aperture_time_units.SECONDS # maybe NPLC too, I can't get it to work (yet)
#PXI4071.aperture_time = 1 # Aperture of 1 second; can do micro-seconds as a float
PXI4071.adc_calibration.ON # This couples the ADC to the internal temperature of the DMM to do 'tempco' compensation
for i in range(1, 5001): # for 5000 observations
   t = (str(PXI4071.get_dev_temp()))
   print(str(i) +', ' + t + ', ' + str(PXI4071.read()))
   time.sleep(3)


I run this code in the Thonny IDE (the kid's version--simple) with Python 3.7 embedded. The NI MAX interface sets up all the drivers that you can download easily from the NI website. I run on Windows 7 with a NI PXI-8105 controller in a PXI-1002 chassis.

The above is for NI gear; PXI is my choice and eBay is my source. So, you want to access your Keysight DMM? No problem, MAX does VISA over ethernet for the modern instruments. Got an old HP3458a? There are GPIB modules for the chassis or GPIB-ENET/1000 to connect via gigabit ethernet.