Howto Reduce Costs of Expensive National Instruments Test Systems

Although part of the R&D department, I have been recently asked to help our validation department with the automation of some environmental tests. The target project is an ECU (Electronic Control Unit) that is intended for use in cars as a Parking Assistance System. The purpose of these environmental tests is to check the functioning of the units throughout some extreme and diverse ambient operating conditions (extreme temperatures, high humidity, salt spray, vibrations, thermal shock etc). Some of these tests can be pretty long (a couple of them last hundreds of hours) and the number of parameters to be monitored throughout these tests can be pretty high. As a result, employing some automated test equipment instead of a technician would make sense; it would also be an advantage from a repeatability point of view, the same test setup being capable to assist in validating other similar projects too.

Unfortunately, I was only assigned the task when some major decisions regarding the test equipment had already been taken. And the decision had been to go with test equipment from National Instruments. While I must admit that NI is the one which sets the standards in the field of testing, their products are prohibitively expensive.

Testing several units at a time dictated the need for the following items: relay cards, high frequency multiplexer card, video frame grabber, digital multimeter, programmable power supply, PCMCIA card holder. All these were to be centrally controlled from a PC, and due to the fact that NI was chosen as a supplier the software controlling them was to be developed using Labview and Teststand.

This equipment made use of the PXI platform from NI which allows for easy integration and transparent software control. PXI stands for PCI eXtension for Instrumentation. It was a platform developed back in the 1990’s with the intention of offering a complete solution for integrated test systems, measurement and data acquisition.
Basically, a PXI system consists of a rack (chassis) in which various standard cards may be plugged, all of them being controlled by a WindowsXP based PC which has a special slot in the same chassis.

The development and operation of Windows-based PXI systems is no different from that of a standard Windows-based PC. Additionally, because the PXI backplane uses the industry-standard PCI bus, writing software to communicate with PXI modules is, in most cases, identical to that of PCI boards. Therefore, you do not have to rewrite existing application software, example code, and programming techniques when moving software between PC-based and PXI-based systems.
Apart from the chassis itself, the master PC is the most expensive piece of equipment, being in fact a highly integrated embedded system built around an Intel processor.

Not that everything else would be cheap! In the end the total needs for this particular project had been established as being:

The cost of the test system went up to the amazing amount of 23499 euros! And I have not added here neither the PXI based video frame grabber nor the programmable power supply, which were purchased from a third party supplier (and which I choose to keep). And they had really spent this money for some programmable relays (the high frequency multiplexers turned out to be still relays), a programmable multimeter and a fancy PC with a nice support for all the cards and with an easy-to-use software environment (Labview and Teststand).
Now I do not want to deny all the benefits of such an integrated test system. There are plenty of them and thoroughly advertised on the NI website. Read through, and MAYBE you will be convinced of the need to spend so much money. But I was not!
And as a consequence I proceeded to build an alternative system to it.
The first thing I crossed off the list was the Labview software. I know most of the benefits Labview brings, but in this particular case, we would not have used more than a fraction of them. You really do not need a 4k Euro software to drive some relays and to communicate with some equipment via RS232 or GPIB. I chose to write the software modules in plain C, or in C# (for which you may get a free and open-source compiler here). You do not need to be proficient in either of these languages to implement such simple tasks and there are plenty of code examples out there. I decided to keep the Teststand software, as it is pretty good as a sequencer and it is well able to call C and C# modules developed separately. In addition, I felt it was too much work to develop a sequencer of my own, so I considered the 3.9k euros spent on it justifiable.
The next items I crossed off the list were the chassis and the embedded controller. Basically, what we needed were some controllable relays, multiplexers, multimeter and power supply. So what I chose to do, was to buy a plain laptop from Dell, a simple and cheap Latitude E6500.
It has WindowsXP preinstalled and it has PCMCIA slot, which we can use for our CAN cards from Vector. Thus, in a single stroke, I eliminated the need for the chassis, embedded controller and the PXI 8221 holder which was only meant to allow the PXI system to hold a CAN card (PCMCIA format). Of course, all these being changed, I could not possibly contemplate the use of any PXI card. I had to reorient and find replacements for all the relays, multiplexers and digital multimeter (DMM).
A quick search on Google revealed many relay cards or boards with various interfaces: parallel port, serial port, USB etc. Due to the ease of use, I chose a Relay board with serial interface, and a decent price.
It only costs about 55 euros… Maybe it is not the same quality like the one from NI, but with 1.3k euros you can get more than 20 replacements!
After a thorough analysis, we reached the conclusion that even the high frequency multiplexer may be replaced by two such relay boards, as the frequency we needed them for was not that high (about 6 MHz).
As for the DMM, I stopped for an “affordable” one from Agilent with a RS232 interface and a price of under 1000 euros.

Once the analysis was complete, I was able to add up all the costs incurred by the new system:

The difference between the two solutions is more than 15000 euros! While this would not look like a huge amount to a large corporation, for small and medium businesses it is an amount to be reckoned with. I admit there are disadvantages to the new solution like:

• the system is not compact and neat, but it spreads across an entire work table
• the software has to be developed in C and C#, not in Labview
• the components are purchased from several suppliers, not from a single one
• there is no warranty or support center for the entire system

But still, you have to put into balance these disadvantages against a 15000 euros amount. For this kind of money you can get even an expensive software developer to work for three months to set things up. You can also get a cheap software developer to do the entire software for about 3000 euros and then you get a net profit more than 10k!
As I said, while it might sound a trivial amount for a huge corporation, for small and medium businesses it is an amount that matters. And it might just make the difference between success and bankruptcy.

Read the Italian version: Come ridurre i costi degli strumenti di test della National Instruments