By Jesse E. Hayes
Research Specialist
University of Connecticut, Connecticut Global Fuel Cell Center

10-May-2004

ABSTRACT
Fuel cell technology has the potential to alter the course of electrical power generation. With this new direct current technology comes the need to reevaluate power regulation and distribution. In most cases this renewable and clean source of energy can be generated at the point of load. The most practical methods of converting the fuel cell output to a useable level are DC/AC inverters and DC/DC converters. This paper presents a DC/DC converter designed for a 10-cell fuel cell stack and the issues associated with voltage regulation of low cell count stacks.

INTRODUCTION
Often, a fuel cell stack is designed to have enough cells in series to provide an output voltage that matches that of commercial off-the-shelf inverters or DC/DC regulators. The most important drawback is that the stack reliability is equal to the reliability of a single cell to the nth power, where n is the number of cells in series. It is analogous to a battery pack with a large number of cells in series. If one cell fails, the battery pack is rendered useless. There is also a need to optimize the number of cells for reliability and the costs of power supply regulation. The results of our design project have yielded a solid basis for further investigation in this area.

PROCEDURE
Selection of a voltage regulator to match our fuel cell output was quite a challenge. Since our estimated stack output power was 240 Watts at 6.5 Volts, we were unable to find a commercially available single module solution. We chose a 5 Volt nominal input, 40-Watt module from Beta Dyne, Inc. Using six of these modules operating in parallel, we expected to deliver an output of 15 Volts at 240 Watts.

In our fuel cell application, we were concerned with high frequency switching noise that might come from the DC/DC converters and adversely affect the electrochemical reaction from our energy source. Intuition guided us to employ a common mode choke (1mH 12A) to present a high impedance to noise from the input pins of each of the DC/DC modules. In addition, we chose input capacitors of different types to present a low impedance path between the positive and negative input pins for a broad range of frequencies. The output noise was not of as much concern because we did not have a specific load in mind. Although no calculations were performed to predict our noise reduction requirements, we obtained satisfactory results and leave any further necessary work for a more specific investigation.

The parallel operation of our six DC/DC modules was accomplished using a master-slave scheme. The printed circuit boards were manufactured identically but a master board was equipped with a 312 kHz synchronization pulse generator circuit. This circuit was designed to force the modules to synchronize at the master board's sync frequency. This helped to reduce switching noise from the overall regulator. Beta Dyne also recommended that we place a small resistance at the positive output terminal. This served to help the modules share the load current. When the module with the highest output voltage (V_O) delivers current, a voltage drop (IR) lowers the bus voltage and allows modules with a slightly lower V_O to deliver more current.

Component selection was done with some research, guidelines and past experience. The fuse was selected to be 12 Amps because the input current to each module was expected to be approximately 8 Amps. A typical 150% margin was the guideline. It would be possible in the future to use a low voltage automotive fuse to save space. The common mode choke was selected because of its low profile and high current rating. It was commercially available from Vicor Corporation for $14. The input capacitors were a general purpose 3,300uF and a high frequency electrolytic 1,800uF capacitor. In addition a 100uF tantalum and 4.7uF multilayer ceramic capacitor were used because of their low equivalent series resistance (ESR) at high frequencies.

The series resistor at the positive output terminal of each regulator was selected to minimize power loss and to drop an adequate voltage for load sharing. We selected a 0.300-ohm, 1%, 5-Watt wirewound resistor (Ohmite: 45FR30) based on calculations of a 2% module output accuracy and full-load power dissipation. The upper and lower limits for the output voltage were calculated using equation 1. The voltage drop across the series resistor under full load current (I_FL) was calculated using equation 2.

In the worst-case scenario where the output voltage of two modules are + and - 2% accurate, the lower voltage module will begin to supply load current before the higher voltage module has reached full load. The module output voltage is de-rated for currents greater than I_FL.

The power dissipation across the series resistor under full load current was calculated using equation 3. The 5-Watt resistor selected gives greater than a 150% margin.

The output capacitors were chosen to handle part of the dynamic load conditions and ripple currents of each module. In the interest of space we selected low profile (20mm) capacitors of 1,000uF at 25VDC. Two of these capacitors and one 33uF tantalum were employed at each module.

The implementation of master module of our DC/DC power supply is shown in figure 1. For the slave modules, the sync circuit is absent. Otherwise, the schematic is identical.

Figure 1. Master DC/DC Module Schematic
(Click to enlarge)

The power supply was assembled using the 6-modular boards (figure 2) connected to aluminum bus rails in parallel. Care was taken to minimize contact resistances since we were expecting currents in the 50-Ampere range. The sync pins were connected on each board in parallel. A partial fiberglass enclosure was constructed for ease of handling. The result is shown in figure 3.

Figure 2. Master DC/DC Converter Module

Figure 3. Power Supply Assembly

RESULTS
We were not able to extensively test our DC/DC power supply due to the lack of proper test equipment. To properly test our design, we needed a linear power supply that could supply 0VDC to 10VDC and at least 50 Amps. We were able to measure efficiency, load regulation, and current sharing using an AC/DC switching power supply. Measurements of noise at the input of the power supply were taken at no load, 30 Watts, and 70 Watts. The noise measurements are shown in figures A1, A2, & A3 of the appendix.

One of the main concerns in our implementation of six parallel modules was that some modules would supply more load current than others. In extreme cases this would cause premature failure in all modules under full load. We measured the current from each module and the total output current (I_OUT) at various load conditions and observed the results in figure 4. The module deviation was plotted against the total output current. Equation 4 was used to calculate the average module current (I_Mavg). Equation 5 was used to calculate the module deviation where I_M was the module current and I_Mavg was the average module current.

Figure 4. Current Sharing Results

We measured efficiency from no load to above 270 Watts output with an input voltage of 6.5VDC. To make this measurement we monitored the output voltage and current using the electronic load on the fuel cell test stand. Input current and voltage were taken using a clamp-on current probe and Fluke 187 Digital Multimeter. We calculated the power supply efficiency using equation 6. The results are shown in figure 5.

Figure 5. Power Supply Efficiency

Load regulation was calculated from the data recorded from the efficiency measurements. We determined load regulation to be 5.4%. Equation 7 was use to calculate the load regulation.

CONCLUSIONS
We were very pleased to report that our DC/DC converter achieved above 85% efficiency at full load and even 16% above full load. This was despite the fact that we had used a series resistor at each module that dissipated just over 2 Watts at full load. The peak efficiency was approximately 89% near the 125 Watt range.

The overall power supply suffered a relatively poor load regulation. This was due to the modules using a current sharing resistor mentioned earlier. In most 12VDC applications, this load regulation would be well within acceptable limits. Other parameters were not tested and are left for future study.

For further study, we would like to measure efficiency as a function of input voltage, input/output noise, and determine parameters outside of normal conditions.

REFERENCES
Larminie, James and Dicks, Andrew. 2002. Fuel Cell Systems Explained. Wiley and Sons Ltd., West Sussex, England.
Beta Dyne DC/DC Application Notes
Beta Dyne SR40 & SR80 Switching Regulators Data Sheet, Rev01 16-Apr-2004

APPENDIX

Figure A1. Oscilloscope Baseline Noise

Figure A2. Noise at DC/DC converter input (30 Watts)

Figure A3. Noise at DC/DC converter input (70 Watts)