Maxim MAX17061 WLED driver recipe calls for three flavors of DAC
The IC Insider: Unity-resistor-string, binary-weighted, and R-2R ladder DAC architectures help Maxim Integrated Products? high-efficiency driver IC power large LCD panels that use an array of LEDs as a light source.
By Randy Torrance, Chipworks -- EDN, March 27, 2009
|
Light emitting diodes (LEDs) have been used commercially for 40 years. However, only in the last few years have white LEDs become commercially practical. These white LEDs are an excellent choice for backlights of small color LED displays, including those used in handheld devices such as cell phones and PDAs. White LEDs have lower power consumption and higher reliability than other backlight options. They would also be useful in larger LCD panels such as those used for notebook computers, tablet computers, and automotive displays. However, an economical method of driving these arrays of white LEDs needed to be found to enable commercial applications. Recently, single-chip drivers have come on the market that can drive multiple strings of white LEDs. Chipworks has analyzed devices from three of these manufacturers. This article discusses a device designed by Maxim.
Maxim’s MAX17061 is a high-efficiency driver for white LEDs (WLEDs). It is designed for large LCD panels that use an array of LEDs as the light source. This device has an internal switch current-mode step-up controller that drives the LED array. The LED array can be configured for eight strings in parallel with 10 LEDs per string. Each string is terminated with ballast, ensuring even LED brightness. The device has an input voltage range of 4.5V to 26V and provides a fixed or variable LED current. The MAX17061 generates an internal DPWM signal for accurate white LED dimming control. The Maxim MAX17061 white LED driver chips are manufactured using a three-metal, single-poly 0.5-µm BCD (bipolar-CMOS-DMOS) process. An annotated die photo at the metal 2 layer is shown in Figure 1.
The ability to drive up to 80 LEDs requires a high-power device. Since each LED string requires 30 mA of current, the device must be able to sink 240 mA across the eight strings. Also, since each LED can require up to 4V of ON voltage, each string needs up to 40V of headroom. The circuitry and power transistors of this device have been optimized to meet these requirements.
The MAX17061 includes many interesting analog circuitry and the layout of the DMOS transistors used for the high voltage tolerant circuits are equally interesting. The eight current source circuits include methods to ensure the LEDs in each string maintain uniform brightness. The PWM system is a surprisingly large and complex arrangement of circuits to allow for accurate dimming of the LEDs. But the blocks we will look at in this article are the DACs.
The implementation of a WLED driver requires a number of DACs. This chip has no fewer than four DACs on board. Interestingly enough, there are three different DAC architectures employed on this chip. Let’s take a look.
Resistor-string DAC
The PWM subsystem of this chip uses two DACs. One of these is shown in Figure 2. This DAC uses the simple resistor-string voltage-divider principle. Here, 32 unity resistors are connected in series. These divide the voltage across them into 32 equal voltages. The 5-bit digital input word is then applied to the pass-transistor multiplexer to choose one of the 32 taps to connect to the output.
On-chip polysilicon resistor values can easily vary by ±50% across all process, voltage, and temperature variations. However, the matching of resistors to each other in one area of a die is usually close to 1%, which is sufficient for this simple 5-bit DAC implementation. However, achieving 1% matching requires careful DAC layout. Figure 3 shows the layout of this circuit.
The array of 64 polysilicon resistors in the center of Figure 3 makes up the 32 unity resistors of the resistor string. Each unity resistor is made up of two resistor segments, each of which is 2.5-µm wide, five times the minimum polysilicon width. This close grouping and large width should allow good matching between the resistors. Interestingly, the metal 1 layer layout shows that the resistors are connected in a linear pattern from the bottom to the top of the left column, then wrapped immediately back to the bottom of the second column and continuing up that column (Figure 4). We would have expected the left column to wrap to the top of the second column in order to improve integral nonlinearity (INL). INL must not have been a problem in this application for Maxim.
Binary-weighted DAC
The MAX17061 includes a reference-adjustment circuit that uses an 8-bit DAC, as shown in Figure 5. This DAC is implemented as a binary-weighted resistor string. The 8-bit digital input word is used to selectively bypass individual resistor elements, allowing any set of the eight binary-weighted elements to be included in the string, and hence allowing any of 256 values of resistance to be selected. Note the scaling of the bypass transistors, which scales the impedance of these transistors with the same weighting as that of the resistors. This scaling is required to allow for accurate overall resistance values, which include the impedances of the switches. The resistors in the top right of the schematic are not used and are likely included because of last-minute design changes or metal mask changes.
Once again, the layout of this DAC shows the use of an array of unity resistors (Figure 6). Each of the binary-weighted resistors is actually made up of an appropriate number of unity resistors. This configuration is required to get the matching necessary for 256 levels of resistance. Once again we can see that the resistors are very wide—in this case, 5-µm wide. As such they are 10 times the minimum dimension. These resistors need to be wider than those in the previous DAC due to the greater accuracy required from this 8-bit DAC versus the 5-bit design. Also visible on the left side of this image is an array of unity NMOS transistors. These are used to implement the DAC’s binary-weighted NMOS bypass transistors. An array of unity MOS transistors is required for matching purposes (analogous to the resistor segments).
As is evident from the schematic, a binary-weighted DAC reduces the number of DAC components required. On the other hand, the layout shows that since the components still use groups of unity resistors and transistors, the silicon area used is not actually reduced.
R-2R ladder DAC
One of the reference-voltage generators on the MAX17061 requires a 7-bit DAC, and for this implementation Maxim chose to use an R-2R ladder architecture, as shown in Figure 7. As can be seen, all the resistors along the top of this schematic are 24 squares of polysilicon, and all those in the vertical branches are 48 squares (hence, R-2R). The input 7-bit digital word controls the seven switches along the bottom.
The R-2R DAC implementation can be explained as follows. At each node along the top resistor string the current is split in half. Exactly half of the current comes from the parallel-series collection of resistors to the left of each node, and half comes from the single 2R resistor below. As such, each vertical branch carries a binary-weighted current, and the digital input can connect any of these branches to either the reference current applied to the top of this cell (ITOP) or to the reference current connected to the bottom (VBOTTOM). As such, VOUT can be set to any of 128 voltage levels by the resistive divider. An advantage of this type of DAC is that it requires only two different resistor values, so the DAC can theoretically be smaller than the architectures described above given the same matching requirements.
The layout of this DAC at the polysilicon layer is shown in Figure 8. The array of 22 unit resistors is clearly visible in the center of the screen. The seven switches are visible just to the right of the resistors. These resistors are only 1-µm wide, which is only two times the minimum feature size. It is clear from this layout that this R-2R DAC architecture is much more space-efficient than the previous architectures shown.
The Maxim MAX17061 white LED driver is a very interesting chip. This single chip allows 80 white LEDs to be driven and controlled. As discussed in this article, the chip includes a collection of DACs using a variety of architectures. Chipworks has also analyzed a Texas Instruments and a Freescale white LED driver, and these three chips together show a good cross-section of the state of the art in drivers for backlighting in today’s medium sized LCD screens, such as laptop PCs. Given some recent announcements from large-screen TV suppliers, it looks like white LED backlighting, and possibly these chips, will soon show up in your living room.
| Author Information |
 Randy Torrance leads the Circuit Analysis team for the Technical Intelligence group at Chipworks. During 22 years in the technology industry he has held senior technical and management positions in the IC design and electronic systems areas. He holds bachelor’s and master's degrees in electrical engineering from the University of Waterloo.
|
