Analysis on improving the efficiency of medium voltage boost converter in LED backlight system

Low voltage range boost converters are often used in mobile devices to boost the battery voltage (1.2V to 4.2V) to a higher voltage level (e.g., 1.5 to 20V) to power the application circuits. In this voltage range, conduction losses are the main consideration. There are many devices on the market designed specifically for these applications, and continuous conduction mode (CCM) is the main operating mode of these devices.

High voltage range boost converters are usually used as PFC converters with 90V to 270V AC input and about 400V DC output. In these applications, conduction loss is not as important as in low voltage boost converters, and more consideration needs to be given to switching loss and noise immunity. Therefore, PFC controllers usually adopt some special design elements such as critical conduction (CRM) operation mode and higher current sensing voltage. PFC controllers are widely used due to their huge market.

LED TV backlight applications require a 24V DC input, 180V DC 0.4A output boost converter. Compared to the low and high voltage range boost converters mentioned above, this type of medium voltage boost converter is rarely used in consumer electronics products. In this voltage and power rating range, conduction loss, switching loss and noise immunity all need to be considered, and it is difficult to find a suitable and cheaper device.

Topology and Component Selection Considerations

When designing consumer product solutions, it is always necessary to avoid expensive topologies and components. Moreover, since both the DC input node and the output node (LED array) are located on the secondary side, no isolation is required for the LED backlight stage. Even though we have other options such as soft-switching resonant half/full-bridge topologies, the boost topology is the best core topology for LED TV backlight power supply applications.

Considering that boost controllers for mobile devices have high PWM frequency (typically 500 KHz to 6MHz) and low noise compatibility (voltage mode or low current sensing voltage). PWM controllers for AC/DC power supplies seem to be more suitable because of their high gate drive voltage (over 10V) and high current sensing voltage (typically 0.5V-1.2V). However, most AC/DC PWM controllers operate at a frequency of 50 kHz to 100 kHz. This frequency range is suitable for power supplies with 90-270VAC input because it balances switching losses and inductor component size. However, for 24VDC input power supplies, the frequency is a bit low because the low operating frequency requires the use of a large inductor.

The CRM PFC controller is the best choice because it not only has the advantages of the AC/DC PWM controller (high gate drive voltage and high current sensing voltage), but also can set the operating frequency to the optimal value (200 kHz) by selecting the inductor. Even though the feedback loop of the CRM PFC controller works in voltage mode, its sawtooth generator and comparator are built into the chip and have a large enough amplitude. Therefore, there will be no problem in terms of noise compatibility.

Improve efficiency

Using a standard CRM PFC controller to implement a boost converter, switching losses are not an issue due to the relatively low input/output voltage and critical conduction mode operation. The issue is the conduction losses. Figure 1 shows the main sources of conduction losses in a boost converter.

We can see that the conduction loss during the conduction period comes from Rsense, Rdson and Rcoil. This article does not discuss how to reduce Rcoil. The following will discuss how to reduce Rsense and Rdson respectively.

In PFC applications, the Rsense value is determined by the maximum rated power. When an abnormal overcurrent condition occurs, the voltage on Rsense should reach the pulse-by-pulse current limit level (Vcslim), and a 10% margin range needs to be retained. Therefore, Rsense can be calculated by the following formula:

For the application discussed in this article, we should also follow this formula. The power consumption of Rsense is:

So we get:

We can see that the power dissipation of Rsense is proportional to Vcslim. The Vcslim of a standard PFC controller is about 0.5V to 1.2V to avoid false triggering caused by noise. In the FAN7930CM, Vcslim is 0.8V. This value is suitable for PFC applications because the input voltage is relatively high and IQRMS is relatively low. But for 24V input applications, this voltage is too high, making PRsense too large. For example, we use the design tool provided by Fairchild Semiconductor to calculate the power dissipation of Rsense for a 72W PFC (90VAC input, 400V/0.18A output). We get the result: Rsense = 0.289Ω, and the power dissipation of Rsense is 0.22W. The efficiency loss on Rsense is then 0.22/72×100%=0.31%. If we use the same design tool to calculate a 72W PFC controller with 24V input and 180V/0.4A output, the result is: Rsense = 0.077Ω, the power dissipation in Rsense is 0.96W, and the efficiency loss is 0.96/72×100%=1.33%, which is three times higher than the 90VAC input condition.

In order to reduce the power consumption of Rsense, we designed a “voltage block up” circuit as shown in Figure 2. The voltage divider R1 and R2 are used to introduce a voltage difference between the Vrs and Vsense pins. Through this voltage difference, Vsense can reach Vcslim with a lower Rsense voltage.

In Figure 3, we can see that by adding R1 and R2, Vsense can reach (Vcslim/1.1) even if the voltage drop on Rsense is much lower than Vcslim. This can reduce the power consumption of Rsense. For example, without using R1 and R2, if Rsense is 0.077Ω, when Ipk=10.39A, Vsense is 0.8V. If Vgate=11V, R1=10KΩ, R2=400Ω, Rsense=0.0375Ω, when Ipk=10.39A, Vsense can also reach 0.8V. However, if Rsense=0.0375Ω, the power consumption of Rsense is 0.47W, and the efficiency loss is 0.47/72×100%=0.65%, which is 0.68% higher than using 0.77Ω Rsense.

If Vdss increases, the Rdson of the MOSFET will increase when the MOSFET die size and package are the same. For example, the Rdson of the Fairchild 100V MOS device FDD86102 is 24 mΩ. But for the 250V MOS device FQD16N25C with the same package and price, the Rdson is 270 mΩ. The conduction loss of the MOSFET device is very different under the conditions of 24mΩ and 270mΩ. We calculated the conduction loss of the 24VAC input, 180V/0.4A output PFC converter Rdson using the same design tool. The values ​​are 0.9W and 10.08W respectively. Obviously, the 270mΩ Rdson is unacceptable. In the standard boost topology, in order to provide an output voltage of 180V, a 250V MOSFET is required to obtain sufficient Vdss margin. In this case, the standard way to reduce the conduction loss is to select a MOSFET device with a lower Rdson. However, at the same Vdss, MOSFET devices with lower Rdson are not only expensive, but also have larger Coss. Larger Coss means larger turn-off loss. Here, we have found another way to reduce conduction loss. That is, use 100V MOSFET devices such as FDD86102 to increase the 24V voltage to 180V. Of course, special methods must be used to solve the voltage problem, such as autotransformer.

Figure 4

Figure 4 shows a boost converter using an autotransformer instead of an inductor. During the on-time, the current flows through the red path just like a standard boost converter, and during the off-time, the current flows through the green path. The voltage on the MOSFET drain is:

If we input N1=3T, N2=7T, Vdiode=1V, Vout=180V, Vin=24V, then Vd is:

Therefore, 100V MOSFET devices can be used.

Design examples and test results

Figure 5: Shown is a schematic diagram of Fairchild Semiconductor's evaluation board for LED backlighting power supplies.

U4, Q35, T3, D36 and external components constitute this boost converter. Windings 6-10 are used to implement zero current detection (ZCD). D37, C42, R39, R40 have two functions. One function is to act as a clamping circuit to absorb the voltage pulse caused by the leakage inductance between N1 and N2. The other function is to monitor the drain voltage of Q35 and feed it back to pin 1 of U4 to achieve overvoltage protection.

Figure 6

Figure 7

Figure 6 is a photo of the top, bottom, and side of the evaluation board. We can see that the addition of R38 improves the efficiency by 1.09%. Figure 7 shows the waveform difference between using and not using the Vrsense voltage booster circuit (R38). Table 2 compares the results with and without the autotransformer. If the autotransformer is not used, D37 should be removed and the cathode of D37 should be connected to 24V Vin. We can see that the use of the autotransformer improves the efficiency by 14.06%, and Figure 8 shows the waveform comparison.

Table 1: Comparison of results with and without Vrsense voltage boost circuit (R38)

Table 2: Comparison of results with and without autotransformer

Figure 8

Conclusion

The standard CRM PFC controller is suitable for medium voltage boost converters due to its features, versatility and low price. Conduction losses are the main challenge in its application. The use of a voltage step-up circuit can reduce the peak voltage required for Rsense to improve the efficiency of the converter. The use of an autotransformer in the boost converter allows the use of low Vdss MOSFET devices to reduce Rdson, thereby significantly improving efficiency. The test results of the evaluation board confirm that this idea is feasible.