Gabriel Suranyi, Distinguished Member of Technical Staff, General Electric Critical Power

Systems demand lower power levels at turn-ON than at full power capacity when they must provide power during peak traffic, full processing speeds, and peak operating modes.

A critical step in rectifier design is the fine-tuning of a set of parameters to achieve the desired function of the final product. Turn-ON performance is a fine tuned parameter that controls the magnitude and duration of permissible current surges during startup while concurrently shielding the power train from potentially damaging overloads. The ultimate goal is to ensure that the front-end rectifier power system properly starts into the end-system it powers under all conditions.

Verification of the turn-ON performance of the rectifier is often performed using electronic loads that simulate the end-use environment of the power system. The latest market trend for some of these electronic loads is to offer a constant-power mode capability. In distributed power systems, constant power emulates more precisely the power drawn by post dc-dc converters. Here are test summaries of two constant-power loads: one manufactured by Chroma and the other by TDI. The test data show that these electronic loads when set to the constant-power mode may inhibit rectifier startup under certain conditions. It is important to understand the prevailing control dynamics during startup and recognize that the constant power setting of the electronic loads may indeed prevent startup because of their inherent control mechanism.

Startup difficulties under certain set-up conditions were present only when the electronic loads were programmed and connected prior to startup of the power system. If the electronic loads engaged after the power system were already up, then no peculiarities were observed. Some of the startup dynamics between the power system and the electronics loads were unavoidable. Understanding the cause for these abnormalities clarifies the reason for these startup issues and assists in quantifying whether the power system performs properly.

First, the desired simulated performance characteristic was conducted. This theory follows with recording the startup dynamics when a power system turns on into a bank of dc-dc post converters. These results are then compared to the results of turn-ON into a constant resistance mode setting of the electronic loads. This comparison is important because many developers are still using the constant resistance mode setting of the electronic loads and with good reason; it works at simulating the post converter performance and achieves quick and accurate results. Lastly, the startup characteristic when the electronic loads were set to the constant power mode, were recorded and explained.

System characterization
In typical distributed systems, power is fed into a bank of capacitors followed by the post dc-dc converters. These capacitors either hold up the bus during fuse clearing transients or provide input filtering for stability considerations of the post dc-dc converters. Following these capacitors are the dc-dc converters that convert the bus voltage to voltage levels required by the board electronics. The front-end rectifier must be able to charge these capacitors and then provide the energy demands of the post dc-dc converters.

The turn-ON current surge of the post conversion stage is fairly benign because the turn-ON sequence commences only after the bus voltage reaches a predetermined threshold, followed by soft start that ensures a steady rise of demanded power rather than a sudden application of instantaneous power.

Although the output voltage of front-ends varies between 12 Vdc and 48 Vdc, the focus is on the behavior and testing of 48 V front-ends. However, the testing methodology and conclusions drawn from these tests are equally applicable to all families of front ends independent of their output voltage level.

Converter turn-ON characterization
The largest market segment of converters operates over an input range from 36 Vdc to 75 Vdc, covering the voltage swings of both 48 V and 60 V battery plants of telecom systems. The turn-ON voltage of these converters is around 34 Vdc. When the voltage exceeds the input voltage turn-on threshold, the converter initiates a soft-start.

Another set of converters targeting the wireless telecom market operates over an input range of 18 Vdc to 60 Vdc, covering the needs of 24 V or 48 V battery plants. Plant voltage selection is predominantly being driven in these systems by the power requirements of the power amplifiers, since they comprise over 2/3 of the load. The turn-ON voltage of these converters is around 16 Vdc.

Evaluations using the resistive mode
Until recently, turn-ON characterization of the power system was verified using the resistive mode setting of electronic loads and this mode yielded well defined and predictable waveforms. In this mode the electronic load does not stress the rectifier beyond its ratings unless specifically set to do so.

The conclusion from these tests is that the front-end rectifier turn-ON, as observed by the voltage rise of the bus, is similar between the voltage rise of the actual load bank and that which was obtained when the electronic loads were set to resistive mode (albeit, excluding the dips caused by insufficient bus load capacitance). Therefore, the resistive mode of operation of the electronic loads precisely simulated the actual environment even without resorting to any turn-ON voltage control.

Evaluations with constant power mode
The dynamics of startup are more difficult to control in constant-power mode. Manufacturers resort to different means to handle this. Imagine a small system of around 2 kW that begins to start up. If the output voltage is only 1 Vdc then the electronic load should demand over 2000 A, to provide 2 kW of power. This is a large amount of current for either the electronic load to sink, or for the power system to source. In practice, neither piece of equipment is designed to handle such large transients; therefore, limitations are used.

Practically, if the electronic load demands a higher current level than the rectifier is programmed to deliver; the rectifier will limit current to its output in order to protect its internal circuitry. If the electronic load would demand more than 53 A, the rectifier would go into current limit and lower its output voltage. The load resistance that the electronic load was adjusted to determine the output voltage. Needless to say, if the electronic load attempts to start at 1 Vdc, the resultant power level will be less than what the electronic load is asking. Unless this condition or state is somehow disturbed by the control circuitry, the system will not start up.

A number of transient states could offset this steady-state, stale-mate condition. One is the inadvertent delay of the output current limit circuit of the rectifier that may satisfy the initial surge demand of the load. Another is either a delay in turn-ON, or a slow soft start of the electronic load limiting the instantaneous current demand. Nonetheless, something must trigger an alteration of the steady-state condition.

Other helpful triggers are to program the electronic loads so that they would not start until the bus voltage is above the point where the modules would start. Simultaneously, it reduces the initial power being applied to around 20 – 30% of the rectifier capacity. These are all reasonable initial conditions to better simulate what the modules would do. But in the end, these are all changes to the steady-state condition that may be sufficient to provide the kick required to start up the system.

Response of the Chroma model 63203A electronic load
One trigger is the power limit turn-ON. The Chroma load executes a turn-OFF, followed by a delay and a subsequent turn-ON. The current drain calculation gets recomputed with every turn-ON. Whether a successful turn-ON is yielded depends on when the voltage measurement is taken for a specific power computation, and the amount of current programmed to be drained to meet the power capacity demand requirement.

The second set of tests of the front-end consisted of a shelf of four CP2725AC54 rectifiers to provide power into a Chroma model 63203A electronic load, with a power capacity of 6000 W or 600 A. A number of peculiarities using this load were observed. One peculiarity was the startup behavior not repeating for consecutive turn-ON events. The start voltage was factory set at 0 Vdc for the first set of tests. Out of 10 attempts, approximately 8 different startup characteristics were captured. In the following two waveforms the load was set to 4 kW, about 40% of the capacity of the power system. The output voltage of the rectifier shelf is at 20 V/div, and the output current captured by a 100 A rated current probe is at 50 A/div with the period at 50 ms/div. The power system goes into current limit during the startup on two different occasions. The electronic load sees the collapsing voltage and turns off for a pre-set duration. This off time causes recovery from current limit and permits the power system to continue to climb its soft start ladder. By the third attempt, sufficient voltage exists for the load to limit its current demand below the current limit of the rectifiers.

Another observation showed the response to the same identical setup as before. The electronic load demands significantly higher current levels than in the first waveform as can be deduced from the rapid fall time of the output voltage of the power system. The behavior of the load is unexplainable. The power system voltage is higher than the first waveform. Yet, the demanded instantaneous current is also significantly higher than before. This rapid ON/OFF cycle continued for 20 sec after which, the power system executed a hiccup; turned off, timed out for 9 sec, and re-started without soft start. Rapid rise, instead of soft-start, triggered a turn-ON. But, it took 30 sec for an eventual startup. Vout – 20 V/div, Iout – 50 A/div, period – 50 ms/div.

TDI Dynaload RBL488 electronic load
TDI offers an electronic load with constant power capability. Testing TDI’s load revealed a number of differences when compared to the Chroma load. An important observation was the turn-ON waveform showed a similar signature with each successive turn-ON into the same set-up. This was an important observation because it substantiated that the dissimilar startup signatures when the Chroma load was connected is a function of the Chroma load control circuitry and not the power system. A second observation was during the turn-ON event. Two instability areas were seen: one at a low output voltage, and the other around 52 Vdc. The voltage spikes captured on screen were assumed to originate from the inductively coupled point of measurement, because the rectifiers did not trip for overvoltage. The system always turned on even though the waveform above the 90% point was not consistent with previously recorded data, either using real modules or the Chroma load. This type of instability was not observed with the Chroma load. Every ON/OFF sequence, for the Chroma, was precisely explained. Therefore, it was believed that peculiarity captured the behavior of the TDI load and not the power system.

Testing the TDI load was rather limited due to the inability to control the turn-ON voltage level. It was not possible to simulate turn-ON characteristics of the post dc-dc converters.

Testing the turn-ON performance of rectifiers is not trivial when the power system is mated to commercially available electronic loads in the constant-power setting. The Chroma and TDI loads exhibited totally different startup characteristics. The TDI Dynaload exhibited some form of a controlled soft-start, while the Chroma load was more likely to demand significant power instantaneously from the front-end power system.

Performance of the Chroma load improved when a start up voltage was introduced. But even the selection of this feature had its peculiarities. The load tended to turn ON and OFF in rapid cycles even at the slight deviation of the output voltage. An additional feature that would have kept the load ON, once it started, did not work well and made the response much less predictable. The only form of meaningful recovery from these instances consisted of powering OFF the load, waiting for about a minute and then repowering the load. This step cleared the internal registers of the electronic load. This form of restart worked every time and the load reset into a predictable state. The power system started in every instance, except under some conditions it took as long as 30 seconds.

It was also demonstrated that placing electronic loads into their resistive mode more precisely represented the startup characteristics of a real system with post dc-dc converters. Our testing concluded that these loads do not behave flawlessly during startup when programmed into their constant-power mode of operation. Intricacies that were concerns happened during startup of the power system-electronic load combination. If the load was connected after the power system was already powered, there were no peculiarities observed.

See graphs/charts for this article online at http://dwo.me/11r52m2

Reprint Info >>

General Electric Critical Power
www.geindustrial.com

---

Filed Under: Electronics • electrical
Tagged With: generalelectriccriticalpower
 

---