Reliability and availability (Part 3: Semiconductor redundancy)
Today we gonna talk about N+1 semiconductor redundancy. When is it needed? What are the benefits and drawbacks?
The main purpose of redundancy is to increase reliability and availaibility of a system. Therefore, to answer if N+1 semiconductor redundancy is needed, we shall twist the question and ask: What availability is required? Does a system without redundancy fulfill the expectation or not?
As we have already suggested earlier, the user shall specify the requested availability rather than giving the manufacturer instructions how to design the product. If you buy a computer, you have certain criteria for selection. You need to decide between desktop and laptop, select operating system, what size of RAM you need, what graphics etc. However, you don’t normally define how the printed circuit boards are designed or how the computer architecture shall look like. When buying a car we have some performance characteristics in mind, but do not specify how the engine shall be designed or what components shall be used.
Whether N+1 semiconductor redundancy is needed, depends mainly on the drive topology. In general, series connection of components reduces the reliability. The more components in series the higher risk that some of them fail. Therefore, medium voltage drives based on low voltage semiconductors (usually LV IGBT) might benefit from semiconductor redundancy, at least in critical applications. On the other hand, drives with medium voltage semiconductors usually achieve same or even better reliability without any redundancy due to much lower amount of components. Keep in mind that semiconductor redundancy also has some drawbacks:
1. The drive operates with lower efficiency as there are additional semiconductors in series that generate extra heat losses.
2. Redundant semiconductors likely increase the footprint of the drive.
3. Extra components mean higher cost.
N+1 semiconductor redundancy might be a marketing trick of a manufacturer how to position himself and eliminate some competitors. You may ask him what is the percentage of the products that are actually delivered with N+1 redundancy and how many are delivered non-redundant (and please share the information with us ;).
Let’s illustrate the N+1 debate on a great example: LCI is one of the oldest AC drive technologies being on the market since more than 45 years and known for its robustness and high reliability. LCI is based on thyristors that easily allow series connection (used to scale the product in power and voltage). Therefore, each and every LCI can be designed and manufactured with N+1 thyristor redundancy. Despite of that, vast majority of these drives delivered up to now do not have semiconductor redundancy! Apparently, there is no drawback on reliability and LCI enjoys a good reputation exactly because of its convincing reliability figures.
For some topologies, especially those with medium voltage semiconductors, the N+1 redundancy is not practical or not possible. It does not mean that these topologies are worse or less reliable. Even the manufacturers pushing N+1 semiconductor redundancy for certain industries have non redundant topologies used in other industries and they sell non-negligible amount of them!
Make your own picture if N+1 is a technical need or rather a selling strategy. Once more I emphasize that your goal is the availability of the drive. Why not leave it to the manufacturer and his vast experience to propose the best solution from his perspective?
