I have constantly heard complaints about GPIB both from a cost standpoint and a throughput standpoint. Yet it still lives on 35 years later. Why is that? First, GPIB is the first and still one of the best buses designed as an industrial communications standard. Both LAN and USB are attempts to modify computer standards to meet industrial needs. As a starting point look at the cabling and connectors. GPIB is double (sometimes triple) shielded. The line drivers are all designed and specified to work in high noise environments. The connector has been thoroughly tested to withstand repeated abuse and is also heavily shielded. The three wire handshake was absolutely brilliant and allows devices with radically different performance characteristics to work flawlessly together. There is little to no minimum timing requirements to co-exist on the bus. Configuration is a simple 1 of 31 address switch configuration. Not only that but manufacturers have worked for years on predicting and avoiding collisions for the default addresses so that 90% of the time, you don't even need to worry about that. The existence of 488.2 and SCPI has standardized communication to the point where instruments can be auto discovered and configured. Further since GPIB is exclusively an instrumentation bus, programmers don't need to worry about conflicting device drivers and protocols interfering. Short answer, it just works and works well. The two major criticisms are cost and performance. Let's look at cost first. Yes, the cables are expensive. They are also robust. We already talked about shielding. USB has minimal shielding, LAN has none, just twisted pairs. We used to test instruments and computers by wrapping 10 meters of GPIB cable around an industrial arc welder. The connector is also tough. It solidly connects to the instrument. USB is strictly an interference fit. LAN has a simple plastic tab that I see broken off more often than not. How often have you gone out on the factory floor and found that the problem was a disconnected USB cable? You won't see that with GPIB. I once hung from a GPIB cable connected to an instrument and swung on it for 30 seconds. What would happen if I tried that with a LAN cable? Next, the speed issue. People have been complaining about the 1 Mbyte/sec limit for as long as I can remember, yet when I analyze programs, I find that they are not bothering to do some of the simple tricks to utilize the available bandwidth and are generally transferring at under 50 Kbytes/sec. There is a factor of 20 available to them. I am talking simple things like using binary instead of ASCII, buffering up readings for burst transfer to avoid addressing overhead, and lowering resolution to what they really need to increase transfer speed. Finally there is the issue of stability. GPIB has been around forever. Computer interface standards have a habit of changing every three to five years. Test systems last a long time, and it may be difficult to service that USB system in a few years. Tried to find an RS-232 equipped computer lately? How about traditional ISA (or even PCI) slots? These were the 'new' way of doing interfaces just ten years ago. How much faith do we have that there will be USB available in another 10 years? How about 802.3 LAN? Yes, GPIB will eventually die. Everything does, but I don't see it happening anytime soon. I predict a life for another 10-20 years as a primary instrument interface. The killer will likely be obsolescence of GPIB computer interfaces, not those on the instruments. Jay Nemeth-Johannes Smart Sensor Systems

I hear what Bus Death and Martin are saying. There is no logical reason to throw away GPIB. We have too much legacy and investment to just toss it. But going forward, LAN and especially LXI address applications that GPIB, and in some case USB, could never adequately address. Remote test applications, such as central office monitoring, need the capability of LAN. And analysis of complex waveforms needs a high speed datalink for efficient test that GPIB will never address. Cable robustness aside, LXI can be implemented for so much less hardware investment. As the bottom line is scrutinized more than ever, I appreciate this. As to simplicity, no new bus architecture will ever match the ease of GPIB addressing. But software tols will evolve to make our efforts seem that simple. In summary, GPIB will not go away tomorrow, so hybrid systms will remian the norm. But LXI and USB developments will make the transisition inevitable.

I received an e-mail from a reader who said "the death of GPIB has to be the most anticipated event never to occur in technology."