Emergency Power for Data Centers Full Transcription

Full notes on specifying the right links and emergency power for data centers.

By Consulting Specifying Engineer Staff November 28, 2007

Specifying the Right Links: Emergency Power for Data Centers
Sept. 20, 2007, Webcast

Moderator: Michael Ivanovich, Chief Editor CSE
Participants: Brian Renner, PE, LEED AP, A. Epstein and Sons International, Chicago. Dennis DeCoster, Executive Principal, Mission Critical West, Redondo Beach, Calif. Kfir Goodrich, chief technology officer for EYP Mission Critical Facilities.

MI: Hello, everyone and thank you everyone for joining us today for today’s webcast. Specifying the right links and emergency power for data centers, hosted by CSE and sponsored by Generac Eaton, Caterpillar, Eaton Data Centers and ASCO. I am you host Michael Ivanovich, chief editor for CSE and I will be moderating the webcast that will last for one hour and 15 minutes. We will have three speakers presenting for ten to 15 minutes each, followed by an interactive Q and A session lasting about 30 minutes or so. We will have commercial breaks from our sponsors along the way and an exit survey. Before we begin, allow me to briefly go over the viewer panel and the controls you will have access to during the webcast. First, disable all pop up blockers you may have operating with your browser. This will enable the interactivity tools to work as we go along. If you have technical problems, click on the help button on the lower right corner of the screen and tech. will provide assistance. To enlarge the slides, click the enlarge slides button and the slides with enlarge. The slides will change automatically in both windows. In addition, you can click on the download slides button and download a PDF file of today’s files. Also, note that the logos of CSE and the webcast sponsors are hyperlinked to their respective websites and clicking on it will open a different window so you can be free to browse during the webcast without interrupting the webcast.their presentations. Those of you who wish to earn CEUs we will provide a hyperlink to a ten-question exam following the webcast and a thank you letter we will send via email and the link will be placed on CSE’s website shortly after the exam.  In order to pass the exam, you need to get 8 outof 10 questions right and you can retake the exam if necessary and we will notify you if you passed or not via email.


Today we will be covering the reliability of data center power. For any facility that handles key financial data, 100% up time is a must. When there is a power event, this kind of reliability depends on a seamless system of transferring power from primary to back up systems. In this webcast, we will take a close look at the design and technical details of switching from utility to on site stand by power and coordinating all the equipment in the emergency power system to protect vital data. We will have three speakers from engineering firms today, Brian Renner, Dennis DeCoster and Kfir Godrich. Before introducing our first speaker, let’s hear from our first sponsor, Generac.



MI: Welcome back, our first speaker is Brian Renner a PE and LEED Accredited Professional. He is vice president and CEE for A. Epstein and Sons International here in Chicago. Mr. Renner is a frequent contributor to CSE, particularly on the subject of specifying and installing switchgear, switchboards and transfer switches for back up and emergency power systems. Mr. Renner please begin.

BR: Welcome everyone my name is Brian Renner, vice president and CEE at A. Epstein and Sons in Chicago. We’ll be talking today about considerations for switchgear and transfer switches and generators for data centers. Were going to hit the highlights today.


hip with your local representatives. It is important to review codes, as we know some local codes use older national codes or codes with amendments and always be sure to review the codes where your project is located. Another excellent source of information is the IEEE color series books, which contain a wealth of information on switchgear, switchboards, generators and transfer switches.


Were going to start with switchgears and switchboards.


It is a common confusion among many people who aren’t experienced electrical engineers and don’t know the differences between switchgear and switchboard and they have very important difference when you are considering using them for data centers. When we talk about switchboards, we are talking about devices designed and built to UL Standard 891. They have front access and sometimes rear access on the larger gear. The breakers themselves are protective devices primarily fixed mount. Switchgears are primarily built to ANSI C 37 Standards. Almost always it has rear access and primarily drawout construction breakers.


You’ll see here typical configurations of switchboards and switchgear and in the first picture you will see a switchboard with front connected access, where the cables are terminated in the front in the channels and are routed directly to breakers that are fixed mounted in the switchboard, although we will talk about later about a hybrid. There are also switchboards that are rear connected in which case the gear has access in the back for cable connections. These switchboards are primarily fixed mounted can use some drawout devices including insulated case breakers. The last picture you see there are of a switchgear which is a draw out construction circuit breakers.


We’re going to talk about here about the key items when talking about switchgear vs. switchboard, which is the breaker types. We talked about fixed mounted, that is molded case breakers, those protective devices, those circuit breakers are sealed within an external molded case and not accessible for service and inspection or maintenance or any of that type of thing. Those molded case breakers are bolted or a fixed to the gear and not able to be drawn out. The second type of breaker is the insulated case and its is sort of like the molded case circuit breakers in that it has some features of power circuit breakers and switchgear in that some parts are accessible for servicing and the insulated case is more rugged and capable of high withstands. The last type of circuit breakers is the power circuit breakers, which is primarily used in switchgear and usually draw out construction and is definitely a higher grade of circuit breaker compared to the other breaker types. An important thing to talk about when we are talking about draw out breakers, it allows you to do servicing and maintenance and other types of operations without taking down the switchgear for servicing. This is an important issue for reliability in data centers, because you do not have to do deal with live switchboards and switchgears, which can be to access a breaker, replace a breaker or service a breaker, that draw out ability disconnects the breaker from the BUS. I think it provides a higher level of reliability and serviceability for a switchgear type of application instead of switchboard or molded case. Some switchboards are hybrids and contain mixtures of insulated case breakers, sometimes a power circuit breaker and molded case breakers. Cost is always an issue, not so much with data centers but with power circuit breakers and switchgear is more expensive than switchboards.


I’d like to hit some highlights on automatic transfer switches. These are built to UL 1008 standards and they can be done in a number of different ways. They are traditionally done with switches and most people are familiar with automatic transfer switches. It is possible to use motor operated breakers in a switchboard or switchgear arrangement, which has an advantage of saving space and giving operations built right into your gear. The third type of automatic transfer switch is solid-state transfer switches. Initially these were very small, but solid-state transfer switches have been getting larger and larger and offer some advantages of repeated switching capabilities but they are very expensive. We are going to be talking about later generator grounding schemes, its important to understand how you are going to be grounding your generator when you are using your automatic transfer switch. Three pole vs. four pole were going to talk a little bit about that it is something that you need to review with your manufacturer or equipment supplier when you are writing your specifications. You also have to know if you want bypass isolation, which allows you, again we talked about the draw out type construction in switchgear. Many of the automatic transfer switches also have the ability to rack up the primary switching device and put it in bypass and is a great way to service your transfer switch without having to take down your data center, which is not desirable in any case. Withstand ratings are something else that you need to be careful in selecting your automatic transfer switch to match your short circuit and withstand ratings of the rest of your system. We are going to talk about open to laid and closed transition, but these are items where they used to call them make before break and break before make. Whether or not you are going to parallel for a small period of time, your generator and normal sources, these are used a lot in hospitals and other applications and programmed transition or delayed transition, where you might want to put in some delay for large motor loads. Clearances and accessibility are another issue that you need to be careful about automatic transfer switches. These transfer switches can be large and they might require rear or side access and you need to be aware of that when you are laying your transfer switches in your electrical rooms. We will talk a little bit about UPS coordination with the generator.


OK, were going to review some basic diagrams and these are from the IEEE color series book. The first figure shows a solidly inter-connected neutral conductor. This is one where the generator is not grounded at the neutral it is grounded for safety reasons on the frame, but there is no ground to neutral. In this case, the generator is not a separately derived system, but is grounded back at your main service transformer and main switchboard and switchgear. There are some advantages and disadvantages to that, which we won’t be able to get to today; I just wanted to make you aware of it.


The next configuration I wanted to make you aware of is the transfer neutral conductor. This is common in 480 and 277 system, it is a very common system in data centers in where the neutral in the generator is grounded, and the transfer switch has a forced switching device. It could be overlapping contracts, could be a true fourth pole for transferring and breaking the neutral conducting, allowing the generator to be separately grounded and derived from your main source. Again there are some advantages to this system that is more commonly used, you should review this with your manufacturer and use the IEEE color books.


The next hot topic we want to talk about is generators. This is a typical photo of an interior generator application. One of the things that jump out at me from this photo is to realize how big the silencers can get depending on what silence level, classification you give those silencers.


Again were going to hit the highlights here and the key issues and there’s so much talk about generators. These are just some things that I think are important to watch when you are dealing with data centers, watch the term emergency. We are all familiar with the term emergency generators where generators are required by code for life safety, lighting, high rises, elevators. Typically when we are talking about a data center, we’re not talking about an emergency generator, we’re talking about a stand by generator. If you use the term, emergency generator in your documents or with local officials it triggers a whole other set of requirements, demand factors and all sorts of other requirements that you want to be careful to stay away from. If the generator itself is just supplying your data center and if that generator does feed emergency required systems, that’s another thing you need to be careful about when you’re looking at reviewing the codes. Just make sure when you’re looking at your emergency systems vs.s gone by, there used to be a concern of over sizing the generator to 2 or 3 times the size of the UPS. These days the UPS technology has got pretty advanced to the point where often, you can get close to one to one if not then you can get 1.2 to 1.5 size times the size of the UPS. You should be very careful when you specifying the UPS and your generators to coordinate with your vendors and equipment suppliers to make sure you got this size properly. If not all, then most of the representatives have sizing software you can use where you can plug in the UPS and other equipment and its recommended that you run that sizing program, to help you avoid problems between the generator and the UPS. Another common issue are fuel issues, type, storage and run time and in most cases we are dealing with diesel generators, although in smaller generators, it’s possible to use natural gas, although natural gas is considered sometimes not as reliable as onsite fuel storage. The run time, be aware that the amount of fuel that your generator is going to use is significant. I think the general rule of thumb is that 7 gallons per hour for every 100 KW. If your going to be storing the fuel onsite,be careful not to just say you need two or three days or a day in the belly tank, because those belly tanks under the generator can get quite large. Make sure you considered an external fuel tank and they make those in concrete types that are fire rated and can be located both inside and outside. Ventilation issues are a critical issue to work with mechanical and electrical architects. If these generators are located inside the building, your going to have to have see how air moves in and our of the generator space, there’s a lot of air in there for cooling and a small amount for combustion air. Environmental issues, one of these types of things that is confusing to us who work in data centers, there are environmental ratings called tiers, for generator applications. Those tiers are actually 1 through 4. Confusing those of us who are used to tier 1 tier 2 tier 3 or tier 4 data centers. Don’t confuse those. Those are environmental regulations talking about the type of generator and specifications that it has to meet for pollution. Be sure to review with your local authorities and the manufacturer and supplier of the generator to review what kind of environmental compliance you need. Paralleling is another issue. Its becoming easier and easier to parallel generators. In days gone by it required a lot of switchgear and protective devices a lot of the generators now are pretty now they allowing you to daisy chain  or network the generators together in a simple paralleling configuration and it makes it a lot easier to parallel than it did in the past.


In conclusion, I would to thank CSE and the sponsors for this webcast. We just covered some of the highlights today and I hope to hear from some of your questions and I hope this spurs you to consider the right issues when looking at switchgear, transfer switches and generators for data centers thank you.

MI: Thank you Brian, before moving on to Mr. DeCoster, let’s hear from Eaton Powerware and Caterpillar.



MI: Welcome back. Allow me to introduce Dennis DeCoster. Mr. DeCoster is an executive Principal at Mission Critical West Inc., Redondo Beach, Calif. A graduate of Rutgers University with post graduate studies done at Seton Hall, Princeton and Rutgers in law, marketing and engineering, Mr. DeCoster also has a substantial background in multi-discipline mission critical project management encompassing all areas of engineering, planning, technical writing, installation and 7 by 24 service management. Dennis the floor is yours.

DD: Thank you all for joining us today and thanks to CSE for hosting the event. We’re happy to be here with you. Before I get into the presentation itself, let me give you a background on Mission Critical West. We’re rather unique in terms of power quality consultants go because we have direct hands on experience, not just with specifying critical electrical facilities’ infrastructure, but also installing, commissioning and overseeing the servicing of that seem infrastructure. This experience extends to every type of UPS and CPS ever built. In today’s presentation, I’ll focus on real world issues derived from MC West client data base and supplemented by IEEE gold book data, IEEE prep data, IEEE transaction data as well as data from other industry sources. A few ground rules first. When we discuss reliability or availability figures, we’ll assume proper commissioning, training, scripting and security for the model site. That is a big assumption because the majority of data center load losses are the result of human error. Nevertheless, the procedures to eliminate or reduce these problems are easy install. Good commissioning eliminates failures caused by loose connections, grounding issues, UPS compatibility and other factors we see from other data base. It will be impossible to cover such a broad topic in our time today, but I hope it will give all of you something to think about. Feel free to visit our website to see white papers describing these issues in more detail later on.


This represents a basic block diagram of the key critical electrical infrastructure elements employed at most financial data centers or any data center. In tier 3 or 4 data centers, the blocks shown are multiplied for redundancy and reliability improvements. Despite what you may have seen in other presentations, these blocks are very reliable with one single exception that will discuss in detail later on.


This graph represents a more publicized a survey of 24 facilities in the United States over a 2 year period. It’s not really so important to know exactly how many disturbances were recorded at each under voltage, but rather understand that there were only a few events per year that lasted more than a few seconds in duration. This data parallels European utility data and more recently US data as well. By comparing these types of data, we can better understand cost effective power quality.


Here we see the timelines associated with the typical GENSET load transfers following the power outages and note that for a single generator installation, 12 to 15 seconds of UPS energy reserve is more than enough time to bridge any power outage or disturbance. Even for parallel GENSET systems, more than 30 seconds are rarely needed due to load priority switching and starting GENSET. In fact, most diesel GENSETs will start in as little as six seconds. You might be able to see from this that reserve time is not the essential issue, reliability and components are the real issue.


Most common causes of data center load losses mentioned in industry papers are circuit breakers, automatic transfer switches, and diesel generators. UPS systems and UPS battery failure to start. As we mentioned earlier, they are reliable. The automatic transfer switch has single point availability index and a 4.9 to 5.9 range determined by the pipe and maintenance history. Further, the GENSET is only necessary when a power outage occurs, which exceeds the UPS system’s ability to provide condition to power the load. They all take place less than 10% of the time on a sole disturbance basis. The UPS itself is exceptionally reliable with the advent of digital controls, IGBT inversion, and reliance on the system static bypass switch. This switch used to save loads by bypassing utilities during UPS failure is routinely rated at the single point reliability levels at the 6.9 level when it’s fully rated.


ATS failures are few and far between and I am going to defer discussion on exactly what those failures are for the interest of time. It’s sufficient to say they are very reliable. That’s not the case the DC length in this system, meaning the UPS batteries themselves. GENSET start batteries these electrical chemical devices spill more often than any other in the chain. Battery reliability gets far worse as systems age and chemical degradation occurs. 90% of the batteries used with UPS systems used today are of the steel type which experience failure rates in the order of magnitude higher than flooded batteries. This is due to primarily to sell dry out and chemical action on relatively thin plate arrays. I show one example here where I talk about a two string 15 minute battery system that tends to give users a nice comfort feeling when they see 15 minute, but little do they know if either one of those string goes out, if any cell on the spring bails open, they end up with zero minutes and zero seconds. So of all the critical and electrical infrastructure blocks and the model data center, the one with the highest incident of failure and greatest contribution to data center load loss is batteries. These electrical generators varies much more than UPS and ATS components as far as reliability goes, but if we assume if the model has well designed and maintained generator system, than the single module start index it is in well excess of 99.7%. The redundant reliability index is between 3.9 and 4.9 depending on the N. This is good enough for a financial data center, the time the site must depend on that GENSET system operating is small due to low incidents of total system failure. The single most common cause of failure of GENSETs is batteries. This is usually the result of improper maintenance in that UPS batteries are always air conditioned and GENSET batteries are not. Lead acid batteries will haveead acid and other issues can be addressed with periodic fuel cleaning, switch position, security and maintenance.


Now for the UPS batteries, here I present a series of quotes from UPS manufacturers and these guys are more in touch with the parts pieces that break on their systems and as you can see


What can high criticality sites do to minimize their exposure to failure from batteries. Flywheels are gaining more acceptances and come in both high speed and low speed versions and both come with a 15-30 second energy reserve, which is more predictable by far and more reliable than steel lead acid batteries. In some cases, flywheel reliable can be almost as reliable as flooded batteries, which take up a lot more space. The pictures shown here is a flooded battery of equivalent capacity to the flywheel shown would take up a room ten times this size. The flywheel has a spot in the ever-growing world of tight data center real estate.


Another alternative to lead batteries, an emerging technology, are nickel alkaline batteries. These NiCAD batteries’ reliability is established and rated almost ten times better the flooded lead acid batteries. Nickel metal hydride is similar electrochemically, but has much less memory effect problem available in the steel version and it takes up little space and requires no maintenance. Either of these batteries can be designed for reserve times for as low as 15 seconds. Summarizing on the UPS side, these flooded batteries of space weight and budget, never use steel batteries without one redundant string, in moist cases we’ve seen 2 or 3 strings are preferred. Consider fly wheels when redundant GENSETs exist.


Finally a few considerations facility wide, for highest reliability in your facility infrastructure we recommend design for reliability not reserve time on your weakest link, which is the DC link. Spend the money exhaustively and make maintenance a design priority and be vigilant about training, scripting and security to minimize the human error aspect. Thank you for your time and feel free our resources page at www.mcwestinc.com for more information and a collection of white papers on this topic.

MI: Thank you Dennis, we have one for presentation, but before we begin let’s hear from our sponsors Eaton Data Systems.



MI: Welcome back, Kfir Goodrich, chief technology officer for EYP Mission Critical Facilities has expertise in powering critical infrastructure and complex system applications. He received his education at Ben-Gurion University and is a PhD candidate at the Polytechnic University in New York. He is a member of ANSI and homeland security standards panel task force, where he identifies reviews and makes suggestions on key guidelines in the power security and continuity area for national homeland security. Mr. Godrich please begin.

KG: Our presentation is selecting the right emergency power for financial data centers. It’s going to be tough to discuss the whole idea of emergency power for financial data centers in a short presentation like this. We’re going to discuss several items that this webcast is all about. ATS systems and their implementation in data center their availability and reliability in financial data centers were going to discuss compatibility of generators and UPSs the old stuff that it still looks like it exists. We’re going to discuss all kind of issues of bypassing and what it is all about how much is too much.


First off all before we start, we need to understand what the new problems facing financial data centers now. One of the biggest problems we need to design a facility for say 10 years and we have technology refresh of 11 months. How can we bridge that? This is the paradox of the facilities because this is the mot problematic thing in the design of data centers. The other thing that we are seeing that is a new problem has to do with high density, how do I implant this high density? If we are building a new data center, we have quite a lot of solutions for doing that; in an existing data center it can be much more challenging. Besides these issues, we have requested a high reliability we are talking about having the highest reliability having the right cost and we have been looking at the high density and how can we plug this in. Lately, we have come out with a new issue of energy efficiency and people are paying about 30 to 50% of the operational costs on utilities. You have to come with something more innovative, the issue of high efficiency in LEED data centers is a hot topic.


Additional new problem is the way the load is behaving in the data center, for forty years we had to measure the load and we used to have to have a constant load in the data center. We used to look at only the difference between the loads vs. what we used to have before and we kind of had a constant curve. In the last two years, the load is not constant and we can see that the load is changing based on the way the servers are using their application. The last topic that is hot, has to do with cooling. It is very difficult in dealing with all kinds of high density like were dealing right now. The financial data centers are more conservative275 sq-ft. with standard cooling solution before going into water and stuff and this is one of the topics that will bother us for several years because we are at the edge of power that the CPUs can take before going from air cooling to a kind of coolant over the CPUs.


If we are trying to understand the story of the ATS, how many problems there are with cooling the data center? What we want to understand about the incompatibility between the generators and UPSs how many problems is they facing in the data center? When you are looking at such as equipment failure is about 28% we don’t have too much information about how this is getting splintering for ATS incompatibility. Stories of system design deal with incompatibility and not so much with the ATSs because the system design, you can take another change over system or another ATS system and plug it into a different part of your data center, if you find its right. This has nothing to do with the fact that you can stay out of this curve that the ATSs are X amount or X % out of the whole problem in the data center and the compatibilities are the same. You can say how many times these bypasses have to do with the times they actually face the data center. This is going to be tough to do in any scenario. The bypasses are automatic like the UPS situation or an automatic design.


In an ideal world, if we are looking at networks, you can barely get 99.9% availability out of the utility, so the 5 and 6 9th requested by the financial market you need to do something there. So you are coming to solve this solution and you are applying all kinds of GENSETS and UPS systems and you creating a dual part distribution, calculating all kinds of reliability. And then you are looking into the equipment you need to achieve this, in order to implement the GENSETS you need the ATS, so you are looking at various technologies and you creating complexities based on the e experience you have in the in design of such systems. There are also natural complexities that you can’t avoid. So theoretically you have a serial system introduced between the critical load and the sources and therefore you have a bottleneck. This is something that you have to take care of somehow. In the ideal world, we again want to do this source to critical load, but you know people are looking at the ATS and saying UPS system. UPS systems we need to bypass, we have a bypass we are synchronizing to this bypass to the output of the UPS. ATS doesn’t need to synchronize to his bypass or something like this.that you have this process of volume engineering in the data centers process. And this by some of the bypasses might fail.


What are the financial fundamentals of design of the data center? You have to segregate the systems and enter them into independent logs, you have to eliminate the bottlenecks and you have to try and push the single point of failure as close as possible to the load. This is one of the items that when you design a bypass you are trying to bring the source as low as possible to the load. Some people are trying to do that to the input of the UPS systems, some people are trying to do it with switchgears, but you always create two different sources of power for the critical load. You are looking into to interdependency and interaction, how will the A work with the B, can they work together or do they need to work separately. You need to look at control circuits; you have to keep these things running automatically. If you do any reliability calculations that do not include automatic modes, that’s a different calculation. You are not going to achieve what you were looking for. And you want to keep the human intervention to a minimum.


So you are starting with the dual pass system, which most of people will call it the Tier 3 plus or tier 4 system. You will see in this kind of typology you don’t have any kind of bypasses and you are not seeing too many things, because most of things should do work and you have A and B side and why bother it. The financial do bother, they want to solve all kinds of issues, they have a lot of experience in their systems, they want to improve every point that is a weak point that they have seen.


You are contemplating all kinds of bypass lines over the medium voltage level and most of the ATSs in the financial world for medium to large scale data centers use around 15 kV level. You are trying to bring the sources as close as possible to the critical UPS station. So you create that and take it out and bring it down to the UPS input sub station and one of the inputs into the sub station. Some of the people they will take the utilities bypass lines and they are also some of examples of people taking the generators and bypassing it. This has to do with the very basic philosophy if I have my sources and introduce new things with my sources if I am not able while I have one of the problems in this thing that I introduced, if I can’t bypass them than I am the one who created the problem. I have the problem. This is the logic working behind designing bypass systems you have to understand what you have to bypass and what you want to achieve with that. So basically, if you are looking down and looking into the sub station, how much is too much? You have dual pass line, then you are starting plugging all these bypasses and you can do all these bypasses from the utilities, the generator, so you have the automatic process and the bypass is beautiful, but how many of these bypasses are too much to the point you are creating too much complexity. Well, this is a tough question and difficult to respond to, in an analytical way of actually implementing reliability analysis.


And of course every great idea of all kinds of bypasses, they are tied to the budget. Because you cannot afford yourself to create all these lines and circuit breakers that in addition that you actually need and already implemented and they say dual distribution power supply.


When looking at this case study, we are seeing why these automatic power processes and how to implement them. Here is a mid level type of data center where we are looking at the system that the reliability and shown on the upper figure, and you can see there a system with the reliability of 78 or 20 % of reliability or failure of 21.8%. In the system B, all we did was add an automatic transfer switch, before basically between the sources and the system downstream. We got the reliability to 90.51% and the vulnerability failure was improved by a factor of 2. This is very serious, basically if you have a medium data center and you have only one ATS system, only by adding the second ATS system and creating a dual disability of basically bypassing the ATS with another ATS system or creating a dual disability you are going to get a probability of failure improvement by a factor of 2. This is a major thing to understand.


The next thing I would like to talk about is the story of coordinating UPS system with generators. The problem is with large UPS system they have an input filter, if the input filter system knows to solve all kinds of morning problems, but the UPS especially the static double conversion UPS systems are creating. So you introduce the field tests, then you are running low in the load and these filters are not even connected, so the harmonics are going up to the generators and creating all kinds of problems with voltage regulation that the generators are suffering from. What can you do there? Some people are trying to delay and doing all kinds of ratio of generators and UPS up to three times, other solutions people are going and using all kinds of switching devices as I mentioned. These switching devices, they might create all kind of notching when used with contactors or used with all kinds of morning systems, they are very expensive. Another solution is to introduce the most important element in this situation is serial chalk together with the filter so you are trying a low kilowatt filter. One manufacturer suggests designing this kind of compatibility by being always into this green curve of kW, this will take c are of any kind of behavior of the UPS with the generator upstream when the designed in the right ratio. Lately, the UPS has started to have the same problems as generators used to have with a new power supply creating a problem of limited power factor and additional problems we didn’t see before in the data centers. And this is required upstream and sometimes downstream. In the generator side, a lot of things have been done. The fact that we can deal with these harmonics in all kinds of ways, not only by design of the generator, but also by the design of the hardware working around the generator and the way we are going to automation.


So we can put together a coiled ditch design in the generator and with that we can eliminate some of the harmonics created by the UPS system. And from the other hand, with the latest new technologies you can actually have a lot of control at the level of the activity so floating with generators they basically this control is so fast and so accurate, that a lot of problems that we used to have in the past, they are disappearing. But it is also important that not everything that is very fast and not all the digital fast voltage regulators, are great and they can work fantastic sometimes and the speed is creating different type of problems in the systems that we are designing for financial data centers.


In the conclusion, I would like to basically say that ATS industry has a reliable and adequate solution for the financial data center market. We feel that a lot has been done here. All of the companies that have automatic transfer switches are very advanced and there are a lot of solutions out there in the market for all types of applications, especially for financial data centers. There is nothing perfect out there, this is why the design should be taken into consideration for all the problems, regarding the bypassing as long as you are not creating a more complex system, some bypass systems they are ok in the financial world, they are more than the required and the financial clients they are requiring to be able to doing live maintenance down to the level of almost direct. Not only, line of bypasses is required, the most bypasses are required and remote PDU bypasses are required with other PDU and we creating such solutions for in creased availability. People are not ready to lose availability in any kind of maintenance. IN the end we can say a lot has been done with UPS and generator compatibility and the functionality is in the hands of the designer. I would like to thank all the companies mentioned and everyone who put this presentation together.

MI: Thank you Kfir, I appreciate you taking the time to speak with us today, and before we get to this question and answer section, let’s hear from our final sponsor, ASCO.


CSE Webcast: Specifying the Right Links: Emergency Power for Data Centers
Answers to questions not answered during the live Webcast
Note: some unanswered questions have been removed because they were not answered by the panelists.
Responses by Brian Renner, PE, LEED AP, A. Epstein and Sons International, Chicago
Responses by Dennis DeCoster, Executive Principal, Mission Critical West, Redondo Beach, Calif.
Do generators still need to be sized at 1.5x the total UPS capacity?
As a rule of thumb this is a conservative starting point – but consult with your UPS and Generator Sales Rep
Many new generation UPS systems use active front ends employing IGBT rectification. This eliminates filters, THD & capacitance issues on gensets so you can close-size at 1.2 or less.
Does UL 1008 also apply to Medium Voltage Transfer Switches?
No – only to 600 Volts and Below
Low voltage only.
Do you have any experience with High Speed Transfer Switches? We are installing an 11MW switch fed by different substations on different utility feeds. Are these effective in place of a UPS system?
In my experience – no, this is not an effective substitute for UPS or other backup systems.
Historically, there have been problems with large freestanding 1/4 cycle STS. Many of these had to do with applications involving motor loads which are not suited for these due to field hold-up. Other issues involve implied independence of dual feeds only to find they in fact share paths at some point.
If generator alternator is 3 phase 4 wire, but the interconnect to the main switchgear does not use the neutral, should I bond the neutral and ground at the generator?
Are generators for data centers normally specified as standby or continuous rated, vs. prime?
Standby is my experience
Since prime ratings are lower, many specify prime as an additional derating factor but standby ratings should be used if proper UPS & load compatibility has been insured.
How often should stand-by generators and ATS be load-bank tested?
Monthly at a minimum in accordance with NFPA 110
Make sure you run at load long enough to reach recommended stack temp (700-750 degrees typically – check with genset provider)
How are phase relationships maintained when paralleling generators?
Synchronizing relays in the gear or in the controllers of the generators
Fuel sizing equation again?
Rule of thumb 7 gal/hr for every 100kw.
Allow for catastrophic event logistics when planning refuel options and vendor delivery guarantees.
How does an open cell in a parallel battery remove all reliability?
It occurs when a 2 string, non-redundant lead acid battery is used to meet capacity at a 5 to 15 minute reserve. With one string open, the other will not have enough energy to provide any reserve time, not even 15 seconds.