Cut the Copper
Snubber components inside transformer enclosures
Placing snubber components inside transformer enclosures creates a dilemma for dry-type transformer manufacturers.
Some larger consulting firms have resorted to the practice of “defensive engineering,” revising their standard specifications to require snubbers on the primary windings of every distribution transformer to be installed in a data center, including even liquid transformers - without ever performing a systems study. (My apologies here - I guess that in order to be more politically correct I should say instead “some engineers have recently taken a very conservative approach.”)
I think this approach is actually a disservice to the clients of those firms. It adds unnecessary costs to be borne by their clients, adds physical space requirements inside data centers that have a very high-cost per square-foot, and, most importantly, it adds new and unnecessary points of potential failures.
In the case of liquid transformers, I think that the practice of adding snubbers actually makes the installation more likely to fail. Connecting snubbers to liquid transformers requires deliberately compromising their very well-insulated, dead-front elbow connections, and adding components that will be insulated only in open air. Everything inside the transformer tank remains well insulated under dielectric fluid, but the snubber components that are being added outside the tank have only air-insulated live parts that can too easily flashed over to each other or to ground and fail.
Effectively, this is adding large costs (anywhere from $10,000 to $40,000 per transformer), and is adding additional physical space requirements and future maintenance headaches - all in order to make the installation less reliable.
Moreover, the failure of an RC snubber is not an event to be taken lightly. It can be reasonably expected that at some point in time, the capacitor will fail, and it will always fail “shorted.” When that happens, the resistors would be destroyed in a matter of seconds, unless there are fuses installed between the line terminals and the resistors. So, now you have three fuses, three resistors, and three capacitors, all with un-insulated live parts at 15 kV, 25 kV, or 35 kV, all in close proximity to each other, and all in close proximity to the core and coil of the transformer.
Then, the data center owner will ask, “How would I know if a fuse has blown, and my snubber has become disabled?” So now, layered on top of everything else, come the blown fuse detectors and current transformers and other monitoring system components, all again un-insulated live parts, and all increasing the likelihood that if any of these components of the snubber fails violently enough, that failure could trigger a failure of the very transformer that it’s trying to protect.
I have concerns about this very problem. Most data centers owners who use static UPS systems will tell you about the maintenance chores and rate of failures involved with the filter capacitors in their UPS. A surge capacitor as part of a transformer’s RC snubber network has a relatively easy service duty - under steady state conditions with a smooth 60 Hz waveform, it normally would conduct less than an amp of current. However, when mounted inside the enclosure of a dry-type transformer, right next to the core and coil - which might be operating at around 250 F – I have concerns that the ambient heating will accelerate the aging of the capacitor, and cause it to fail prematurely. A violent failure involving a case rupture could spew liquid and shoot metal shrapnel into the core and coil.
Even metal oxide surge arresters (MOSAs), which are frequently mounted inside the enclosures of dry-type transformers, don’t like that kind of ambient heat, and care must be taken to mount them near the intake air vents, and away from the core and coil. These arrestors are typically rated for application in an ambient air temperature of 40 C, which is easily exceeded inside a dry-type transformer case, and the failure of an arrester can also be a violent event.
In contrast to dry-type transformers, liquid transformers generally use dead-front elbow-type MOSAs, which are located outside the tank, so that even a violent failure can’t touch the core and coil.
Moreover, those arresters are generally rated for application in an ambient temperature of 85 C.
Placing all of these snubber components inside transformer enclosures is creating a dilemma for dry-type transformer manufacturers, who are now between a rock and a hard place. By and large, transformer manufacturers are very concerned about the possibility of failures of all of these snubber components they are being asked to install in their enclosures, and they would prefer not to install them. Yet, they realize that if they don’t install them, then the transformer itself can become susceptible to catastrophic failure from switching-induced transients. (Again, this is not a transformer problem - it’s a systems problem.
So far, I’ve heard of only two incidents of snubber component failures. But, the widespread application of RC snubbers inside transformer enclosures is a relatively new thing, a recent trend that began just 5-10 years ago. Only more time and more experience will tell what their longevity is and what modes of failure will occur.
Helping Joe on these blogs posts is Brian Steinbrecher, an electrical engineer focused on medium-voltage power distribution systems. His 30 year career includes work with an end-user (IOU), a manufacturer of power systems equipment, and as a system designer/consultant. Brian has a wide breadth of experience within the utility segment from systems design to equipment specifications and from system studies to construction and start-up. He has written many technical documents, papers, and reports and holds over a dozen active patents.
A good portion of Brian’s career was with Cooper Power Systems where he performed engineering and marketing work in behalf of their major product groups. Prior to moving into his current role, Brian was the Director of Engineering for a product group at Cooper. Brian is currently the Owner and Principal Engineer at Galt Engineering Solutions located in Brookfield, Wis.
>> Return to the Cut the Copper blog listings.