Automated Testing of Instruments
Francis Bauer
Posts: 364
Today after JonnyMac's great presentation on working with Stepper Motors using the Propeller P2, the group continued talking about topics related to using Stepper Motors, 3D Printers and related technologies.
My contribution to the conversation was a feeble description of the automation system that was developed when I was working for HP's Microwave Instrument division. I thought I would post this to do a better job of describing of what I was talking about.
During my tenure at HP I helped create/maintain Test & Measurement Software used in Manufacturing of HP RF/MW Instruments. We created Test processes that would use automated testing/equipment to align and verify the operation of the manufactured instruments. A few of the local HP Engineers came up with the idea of developing a system that would take Automated Testing of Microwave Instruments (RF/MW Signal Generators, Network Analyzers, Spectrum Analyzers, etc...) to the next level.
The next level consisted of creating a way to physically move the instruments through the various test processes automatically. Besides just moving the instruments, the system could make the needed (Signal, Power) connections to the instruments and even have them tested at temperature (55c). This new automation system allowed the production line to load up a line of instrument to be tested and have the shop floor controller/software pull them into the overall test process without any further human interaction. The various tests were run (some which could take a few hours to complete), the performance data was collected, pass/fail determined as the instruments flowed 24x7, without any humans even being around.
The beginnings of the automated system was done frugally (unusual for most HP projects). The extruded aluminum supports they used in the super-structure did cost a bit, so they found other innovative ways to keep costs down. Like using Sledge-Hammer heads from the local hardware store when they needed counter-weights for the safety-gates and using roller-blade wheels from local sporting goods stores to be used for the Instrument transport carriages.
The system that was developed worked very well, so well in fact that it was turned into a product itself and was put to use in other HP Instrument divisions and for outside customers.
HP split off its Instrument Divisions to Agilent Technologies, who eventually split them off to Keysight Technologies. Checkout the current Keysight Video to see an example of the automation system in action. Keysight has developed an entire Industry solution for automated manufacturing.
Comments
Thank you for the kind comment, Francis. Later that evening I was speaking with my friend John (of JonyJib) and told him about my jokey idea of making an X/Y plotter with a laser on it to do simple art. He told me he has such a device with a Schmaschmino in it, and that I can rip that out and replace it with a Propeller. That's a project for down the road a bit, but should be fun.
I think it would be fun for the community and good for those of us who enjoy spreading the good word about the Propeller to create small-scale industrial type projects. I think success with small projects gives people the confidence to look ahead to the big projects.
X/Y plotting/positioning has been morphed to do a variety of things over the years. Now with the addition of the Z axis we are living in a true 3D world of creations.
Things on the small scale can certainly be applied/leveraged and inspire others to do things on a larger scale.
Checkout this house-building 3D printer.
And let us not forget that HP wrote an amazing software suite to control their instruments, called VEE who followed from them to Agilent to Keysight. Who now seems they want to scuttle the works, not realizing that VEE is the best thing for instruments since the GPIB (HPIB) and a BASIC dialect for the computers from HP.
But still and all, that's an amazing story Francis.
In fact despite the misgivings of the community Keysight is indeed following the HP way when it comes to T&M systems.
Buck Rogers,
Keysight requires you to register if you want to get a user manual for equipment that you have.
I had no trouble getting user, calibration, and service manuals for an old HP and Keithley (now part of Tektronix).
We used VEE a little in the RF/MW manufacturing processes at HP/Agilent/Keysight, it worked well, the only issue was there was a lag in getting instrument drivers written/available for all the various pieces of Test Equipment.
When I started at HP in 1980 working on the HP8566 Spectrum Analyzer Production line, we were using HP9825 Computer/Calculators (Custom 16bit HP microprocessor based) running Hpl to perform all the automated testing of the instruments/boards. In 1982 or so we took delivery of the new HP9826 Series 200 desktop computers (Motorola 68000 microprocessor based) which were capable of running both Hpl and HPBASIC. Over the years we developed a huge library of HPBASIC software (>5 million lines of code) for Instrument control/measurement, including numerous Test environments. As HPBASIC moved to HP-UX and Linux, we moved with it. I'm retired now, but as far as I know Keysight is still using a lot of this HPBASIC code running in Linux in conjunction with the Shop Floor controller running their automated test processes including the robotic instrument movement/testing environment.
When I'm not playing with Parallax Propellers, I still play with some of my HP computer systems (HP9836,HP382, C3000, etc...) running Hpl and HPBASIC interacting with my collection of vintage HP Test & Measurement instruments. HPBASIC even though it is old, is still very powerful since it has Pascal like structures/commands, Extensive Technical/Mathematical libraries (including Matrices, Real/Imaginary) and Instrument control/measurement abilities.
Keysight still has elements (more or less) of the HP way flowing in its veins , probably more than what the current HP named companies have... Long Live the legacy of Bill & Dave.
It is a thought. However the Tek scope who lives here does have his manuals. I was commenting on the incredible explanation that Francis gave us. And naturally we've reached the level of discussion that we've reached now.
However..... This does not explain why there are fifty million ping-pong balls bouncing along towards the place of one of the group someplace.
Mascot on vacation.
That was really interesting, thanks Francis. Its interesting what they do with that extruded AL profile, and I see the test instruments are already hooked into the assembly, does this mean there is a mating microwave connection somehow?
We need to automate some P2 board testing and curve tracing. I've had a look at the SCPI protocol and BenchVue, but because the P2 is also going to be involved it feels like the P2 might end up being both the DUT, and the testing master, which is somewhat unusual. I say this because at the end of the day it makes most sense to store the calibration data in the flash on the same board as the calibrated P2
The alternative approach is more along the HP solution with higher up servers managing things, and perhaps the P2 could emulate a SCPI node and be made to respond to SCPI commands to it can fit in with a compatible 'standard' test and logging methodology, and allow statistical analysis across batches etc
The focus on 'solutions' is interesting too, many of the big companies are racing to establish their industry 4 solution credentials. It'll be interesting how this all plays out over the next decade. There's an interesting eevblog interview with John Kenny talking about this kind of thing too.
Hello!
@Tubular , don't use BenchVue for your plans. It has a learning curve that makes Labview look easy. (Which it is not.) Look instead at the VEE product. The company has not said they will, or won't be discontinuing it. So there are even some new users out there.
I've been using it on-and-off for the past 20 years (the entire time I've been based here.), to automate some of the more delicate, and difficult, activities and frankly the system is perfect for it. There's even a user group that's still running constantly out there.
Ok I'll take a look. Thanks!
I believe they use a DUT (Device-Under-Test) specific universal adapter assembly on each Instrument's moving carriage to mate it to the various Test stations. These consist of various RF/MW connectors both input and output and other stimulus depending on the DUT and Test measurement station. They would have the Metrology group do analysis/characterization of the various connections and instruments to ensure that Test measurements are accurate and NIST traceable.
I believe in the latest incarnation of Keysight's Automated Testing environment they have designed the various Test Stations to target specific types of Tests. So they have specific Test stations that can make Phase Noise or Power or Frequency or other performance measurements for many different types of DUTs (Signal Sources, Spectrum Analyzers, Network Analyzers, etc...) . Their goal was to create an overall Test process that could test many different types of Instruments on the same set of equipment. It has been close to 10 years since I've looked at what they are up to, so my information is probably out-of-date.
The DUTs are connected through HP-IB(GP-IB) and probably other communication methods by now (USB, LAN, etc...). Depending on the Instruments involved they may even be doing Wireless communication for Instrument control. Wireless communication is problematic though since a lot of the Instruments being tested are used to develop/measure Wireless communication and consistent performance of RF/MW measurements could be affected by too many spurious signals floating around.
You certainly could design your testing process to use the P2 to be both the running target DUT and be the testing controller in different cogs, depending on how complex the boards/DUTs are.
SCPI was and I assume still is heavily supported by the various Test & Measurement Instrument manufacturers.
I've never used BenchVue, so I don't know how well it works. I have used VEE and it was/is capable of doing a lot, but I don't know how much it is still being supported.
You can also look into National Instruments LabView. It is very capable, but also does cost some big $$$
In the past at HP/Agilent/Keysight each Test station had its own dedicated controller (usually a workstation running Linux or HP-UX or MS Windows). These controllers would run the Test environment software and interact with the Shop Floor controller indicating the overall Pass/Fail status of the DUT so the Shop Floor controller knows how to route the DUT, either to the next Test station in the process or back to a rework process to look into the failure(s).
>
Let me look around and see what I can find on published white papers/articles concerning HP/Agilent/Keysight efforts into their fully automated testing processes. There may be some gems of knowledge to help your efforts...