Recently, while exchanging ideas with some industry clients, I heard a rather "novel" claim: a certain manufacturer's modular UPS has a remarkable feature—each power module can be used independently as a complete UPS, which is touted as a major selling point. It seems marketing has to keep up with the times. Isn't this just the old "distributed bypass" modular UPS? Technology that hasn't changed in over a decade is now being packaged as a new selling point. Truly, "knowledge is power!"
Those familiar with the development of modular UPS systems should know that there have been two different technical approaches to modular UPS system architecture from the outset: distributed bypass and centralized bypass. Here, I would like to compare and contrast these two approaches in terms of their technological origins and performance reliability, hoping to provide readers with some insights and assistance.
I. Architecture Definitions and Origins of Two Bypass Schemes
Modular UPS, as the name suggests, divides a high-power UPS system into multiple sub-modules connected in parallel. Through optimized system control, it enables online expansion, upgrades, and maintenance of the system, significantly improving system reliability, availability, and energy efficiency, while reducing customer maintenance costs. In recent years, it has gradually become the preferred choice for mainstream customers. The following analysis uses a typical 300kVA system based on 10 30kVA power modules as an example.
(1) Distributed bypass architecture
A distributed bypass architecture means that each power module contains rectification, inversion, and battery conversion components, as well as a static bypass with a capacity equal to that of the power module. It can be considered a UPS without LCD monitoring. Multiple modules are connected in parallel in a cabinet to form a system, and the interrelationships between the modules are similar to those of a traditional multi-parallel UPS system. When the system switches to bypass power, the load is powered in parallel by the distributed bypasses within all power modules. The system architecture diagram is shown in Figure 1.
Figure 1. Distributed bypass architecture diagram
(2) Centralized bypass architecture
The centralized bypass architecture features a single centralized bypass module with a capacity equal to the system's capacity. Each power module contains only rectification, inversion, and battery conversion circuits, each controlled by an independent controller. Parallel connections between modules are no longer traditional UPS parallel systems, but rather incorporate complex logic for inverter current sharing, bypass control, and monitoring. The system architecture diagram is shown in Figure 2.
Figure 2. Centralized Bypass Architecture Diagram
(3) The origins of the development of the two technical solutions
The concept of modular UPS originated from customers' need for simplified system maintenance, hoping to restore the system without affecting critical business operations in the event of a failure, through simple replacement operations. Manufacturers naturally conceived of designing UPS parallel systems with a modular structure, which is the origin of distributed bypass solutions.
The advantages of distributed bypass solutions are: simple control, low development difficulty, only the original UPS parallel system needs to be transplanted and the monitoring part optimized; low cabinet cost; the bypass devices are relatively inexpensive due to their small capacity; and static bypass has multiple redundancies.
Centralized bypass is a technical approach that has emerged after distributed bypass. Compared with traditional parallel UPS systems, it has undergone significant changes in terms of parallel current sharing control, system logic coordination, and fault tolerance. It can be said to be a completely new technical field with great development difficulty.
The following sections will describe the differences in performance and reliability between the two technical approaches.
II. Performance differences between the two schemes
Static bypass, as the last line of defense in UPS power supply, is undeniably important. Common bypass power supply scenarios include: inverter failure, inverter overload or overtemperature, and output short circuit. As can be seen, bypass power supply conditions are mostly extreme, requiring even more stringent testing of the components.
(1) Steady-state operating conditions
When using bypass power supply, the centralized bypass scheme is easy to understand. Only one bypass provides the full current, and the bypass capacity is designed according to the maximum capacity of the system. It is unrelated to the number of modules configured and there are no problems.
The distributed bypass scheme uses multiple low-power static bypasses to bear the load. Since the bypass loops are low-impedance loops, the current sharing of multiple loops cannot be controlled by software. The current sharing between modules depends entirely on the following factors:
1) The differences between individual devices are mainly due to differences in on-state voltage drop, and the dispersion of device manufacturers is unavoidable;
2) The difference in loop impedance is mainly due to the inability to guarantee that the length of the cables in each loop is consistent, and the impedance of the cable connection points cannot be controlled due to process control and other reasons.
Generally speaking, even with the most optimistic estimates, the current sharing difference is unlikely to be less than 20%, which means there is a risk that some modules will have excessive current, which is very dangerous in harsh applications.
Due to this uncontrollable current sharing capability, some manufacturers have proposed a "solution"—bypass current sharing inductors. The principle is simple and crude: each bypass circuit is connected in series with an inductor (as shown in Figure 3), using the inductor's impedance to balance the current in each branch (which is also the method used in conventional parallel systems). Aside from the fact that the 10% individual variation in inductance leads to greater system losses, this solution also has the following insurmountable gap in transient performance.
Figure 3. Bypass current sharing inductor of a certain manufacturer
(2) Transient operating conditions
Switching from inverter to bypass is essentially an emergency situation, requiring extremely precise switching timing; otherwise, critical load interruptions can easily occur. Switching under heavy load or fault current conditions can result in instantaneous operating currents several times the system's rated current, which is why static bypass designs require a larger margin.
The key parameter for static bypass devices to withstand transient current surges is I²t, which is the integral of the current over a short period (generally less than 10ms). If I²t is too large, the device is likely to burn out. Commonly specified bypass overload capacity in UPS performance parameters is 1000% for 10ms, meaning the bypass needs to provide at least 10 times the rated current within the 10ms protection time of the distribution switch. The following analysis uses a 300kVA system as an example to examine the differences in surge resistance among different devices.
Due to current technological limitations, the maximum current rating of a single distributed static bypass device is 70A. According to the device datasheet of a well-known manufacturer, the maximum I²t is 7200A²s (<10ms). A 300k system can be considered as 10 devices operating in parallel.
Centralized static bypass uses SCR modules, with the most mainstream manufacturer being German SEMIKRON. Let's look at the I2t parameters of one of the models, SKKT323/16E. Under the same 10ms condition, it is 450,000 A2S, a difference of more than 60 times!
Let's calculate the I2T requirement for a typical 1000% overload with 10ms latency, for a 300kVA system.
In other words, a single SCR module with centralized bypass can provide 10ms protection capability of more than 10 times the rated current, while static bypass based on discrete components is far from sufficient, even without considering uneven current distribution among components!
Transient switching current sharing control is not only related to the devices and the impedance of each loop, but also to the control mechanism. Since each module has its own controller, factors such as processor speed, communication delay, and module differences affect the actual switching action of each module, resulting in varying delays. This means the first module to switch to bypass may be subjected to a current 100 times its rated capacity! Due to the transient high current, even a series bypass current sharing inductor will not provide any current limiting effect. This is an impossible task for any device; such a switch is tantamount to an explosion. A schematic diagram of the short-circuit fault current is shown in Figure 4.
Figure 4. Schematic diagram of short-circuit fault current
Of course, manufacturers of distributed bypass systems are well aware of this principle and have provided corresponding "solutions": in the event of a short circuit, only the inverter is maintained for 200ms, and then the bypass is not switched off, and the system is shut down directly!
Let me explain. The 10 times rated current operating condition is common in output short-circuit conditions. When the inverter cannot provide enough current to interrupt the fault (usually 3 times the rated current for 200ms), the system will switch to bypass power supply. The low impedance and high current of the bypass will be used to break the protection device (switch or fuse) at the short circuit point. This must be considered in power distribution design. If the power distribution system is designed correctly, the protection design of each branch should not produce over-level protection, that is, the downstream fault should not cause the upstream switch to operate. The worst case of the system is to switch to bypass and then use the strong overload capacity of the bypass to break the downstream protection device. This is the source of the bypass surge resistance requirement.
In systems using distributed bypasses, forcibly switching to bypass will undoubtedly lead to device damage and system downtime due to insufficient shock resistance and asynchronous switching. Therefore, manufacturers can only design systems that prohibit switching to bypass. Imagine a complex computer room or factory where a short circuit in just one branch would render the entire system unusable! This is unacceptable in practical applications, but it is also an inherent problem that distributed bypasses cannot solve.
III. System Reliability Analysis
The only claimable advantage of distributed bypasses is bypass redundancy, while centralized bypasses are considered to have a single point of failure. Let's analyze this further.
(1) Analysis from the perspective of device selection
From a component selection perspective, the reliability of a single high-power SCR is far higher than that of a system composed of numerous small components. Centralized bypass modules are simple in function, requiring consideration only of the components and a small number of external drive circuits. Distributed bypasses, however, are located within the power module and are affected by numerous components within the module. Experienced engineers know that faults in the rectifier and inverter circuits can cause faults in other parts of the circuit due to sparks or other reasons. In other words, static bypasses face more uncertainties and risks. If centralized bypasses represent a single fault, distributed bypasses might be considered "multi-point faults."
(2) Analysis from the perspective of system capacity
From a system capacity perspective, the capacity of a centralized bypass is determined by the rack design and is independent of the number of modules configured. However, the static bypass capacity of a distributed bypass is determined by the module capacity. This means that when a module fails, the system loses its corresponding static bypass capacity. In an extreme example, when a rack has two power modules and the load rate is around 55%, if one module fails, the remaining module will be under 110% overload, ultimately resulting in a system power outage. This same condition is not a problem for a centralized bypass. Due to the advantage of device capacity, some manufacturers even offer centralized bypass modules with a long-term overload capacity of 125%, providing absolute assurance for system reliability.
(3) Reliability design analysis of centralized bypass
For centralized bypass reliability design, many mainstream manufacturers have proposed many solutions to improve reliability, such as redundant backup control loop solutions, communication bus redundancy solutions, power module and bypass module control decoupling solutions, and power module participation in bypass control solutions. Each manufacturer's solution has its own characteristics. After years of market verification, they can greatly improve system availability. In addition, the bypass module's common hot-swappable design makes maintenance and upgrades as simple as those of the power module.
IV. Summary
Through the above analysis, we hope to help you further understand the differences in the overall system performance and product reliability between the two solutions.
Debates and choices regarding technological approaches are normal phenomena in product development. For users, correctly understanding the advantages and disadvantages of each approach is crucial; listening to all sides leads to clarity and avoids falling into marketing misconceptions. However, for manufacturers, the choice of technological approach is significant. Once the approach is determined, product development cannot be changed midway, and subsequent product lines must continue in that direction. This is why, regardless of industry developments, manufacturers using decentralized bypass systems cannot switch to another camp. Currently, the most mainstream modular UPS manufacturers, such as Emerson, Eaton, APC, Invt, and Huawei, all adopt centralized bypass solutions. Savvy customers should understand the reasons behind this.
For more details, please visit our official website www.invt.com.cn or follow our WeChat service account.
About Invt:
Founded in 2002, INVT is committed to becoming a leading and respected global provider of products and services in industrial automation and energy. It was listed on the Shenzhen Stock Exchange A-share market in 2010 (stock code: 002334). INVT is a key high-tech enterprise under the National Torch Program, currently possessing 16 holding subsidiaries and 12 R&D centers across China. It has applied for over 800 patents and relies on key technologies in power electronics, electrical drives, automatic control, and information technology. Its main products cover high, medium, and low voltage frequency converters, intelligent elevator control systems, servo systems, PLCs, HMIs, SVG, UPS, motors and electric spindles, photovoltaic inverters, energy-saving and emission-reduction online management systems, rail transit traction systems, and new energy vehicle electronic control systems. INVT currently has over 2,500 employees, three large-scale production bases, and a marketing network covering more than 60 countries and regions both domestically and internationally.
