N LM9812 30-Bit Color Linear CCD Sensor Processor General Description Features The LM9812 is a high performance integrated signal processor/digitizer for color linear CCD image scanners. The LM9812 performs all the signal processing (correlated double sampling, pixel rate offset and shading correction, color balance control, and 10-bit analog-to-digital conversion) necessary to maximize the performance of a wide range of linear CCD sensors. • • • • • • The LM9812 can be digitally programmed to work with a wide variety of CCDs from different manufacturers, including both 3 output RGB and 2 output GRGB CCDs. An internal Configuration Register sets CCD and sampling timing to maximize performance, simplifying the design and manufacturing processes. For complementary voltage reference see the LM4041-ADJ. • • • • • Applications • • • • • 6 million pixels/s conversion rate (2MHz/channel x 3) Pixel rate shading (gain) correction for individual pixels eliminates errors from PRNU, illumination, etc. Digitally programmed color balance controls Pixel rate offset correction for highest quality in dark regions Correlated Double Sampling for lowest noise Reference and signal sampling points digitally controlled in 20ns increments 2x and 4x analog fast preview/low resolution modes Digital control of CCD integration time Generates all necessary CCD clock signals Compatible with a wide range of color linear CCDs and Contact Image Sensors (CIS) TTL/CMOS input/output compatible Key Specifications Color Flatbed Document Scanners Color Sheetfed Scanners Multifunction Imaging Products Digital Copiers General Purpose Linear CCD Imaging • • • • • Output Data Resolution Pixel Conversion Rate (total) Supply Voltage Supply Voltage (Digital I/O) Power Dissipation (typical) 10 Bits 6MHz +5V±5% +3.3V±10% or +5V±5% 390mW CD9 GCLK VREF LO OUT VREF LO IN VREF MID OUT VREF MID IN VREF HI OUT VREF HI IN AGND VREF IN VA OSB TEST D9 D8 D7 D6 D5 VDI/O D4 DGNDI/O D3 D2 D1 D0 RD PIXEL EOC OCLK CD0 MCLK CD1 WR CD2 RD CD3 CS LM9812CCV CD4 SYNC VD RUN/STOP DGND ø1 CD5 ø2 CD6 RS CD7 TR1 7 6 5 4 3 2 1 52 51 50 49 48 47 46 8 45 9 44 10 43 11 42 12 41 13 40 14 39 15 16 38 17 37 18 36 19 35 20 34 21 22 23 24 25 26 27 28 29 30 31 32 33 CD8 TR2 OSG OSR Connection Diagrams Ordering Information Commercial (0°C ≤ TA ≤ +70°C) LM9812CCV Package V52A 52 Pin Plastic Leaded Chip Carrier TRI-STATE® is a registered trademark of National Semiconductor Corporation. ©1997 National Semiconductor Corporation 1 http://www.national.com LM9812 30-Bit Color Linear CCD Sensor Processor December 1997 Block Diagram VD VA Coarse Color Balance PGAs OSR RED OS from Sensor VDI/O Static Offset DACs +² + x1-x3 CDS OSG Offset DACR GREEN OS from Sensor Blue Gain Boost OSB BLUE OS from Sensor x1-x3 CDS 10-Bit, 6MHz ADC +² + x1-x3 CDS x1or x3 CLAMP Offset DACG +² + Offset DACB VREF IN VREF LO OUT VREF LO IN VREF MID OUT VREF MID IN VREF HI OUT VREF HI IN Pixel-Rate Offset Subtractor + 10 bits + - Pixel-Rate Shading Multiplier TEST Gain/Offset Coefficients CD0-CD9 10 GCLK OCLK Pixel Data Output/ Control Data I/O WR 10 Pixel Data Latch CS D0-D9 RD EOC RD PIXEL SYNC ø1 ø2 Control Logic MCLK RUN/STOP Configuration Register RS TR1 TR2 AGND DGND 2 DGNDI/O http://www.national.com Absolute Maximum Ratings (Notes 1 & 2) Operating Ratings (Notes 1 & 2) + Operating Temperature Range TMIN≤TA≤TMAX LM9812CCV 0°C≤TA≤+70°C VA Supply Voltage +4.75V to +5.25V VD Supply Voltage +4.75V to +5.25V VDI/O Supply Voltage +3.0V to +5.25V ≤ 100mV |VA-VD| VA-VDI/O < -100mV OSR, OSG, OSB, VREF IN Input Voltage Range -0.05V to VA + 0.05V CD0-CD9, D0-D9, MCLK, SYNC, RUN/STOP, CS, RD, WR, RD PIXEL Input Voltage Range -0.05V to VDI/O + 0.05V Positive Supply Voltage (V =VA=VD=VDI/O) With Respect to GND=AGND=DGND=DGNDI/O 6.5V Voltage On Any Input or Output Pin 0.3V to V++0.3V Input Current at any pin (Note 3) ±25mA Package Input Current (Note 3) ±50mA Package Dissipation at TA = 25°C (Note 4) ESD Susceptibility (Note 5) Human Body Model 2000V Soldering Information Infrared, 10 seconds (Note 6) 300°C Storage Temperature -65°C to +150° Electrical Characteristics The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=+5.0VDC, VDI/O=+5.0 or +3.3VDC, VREF IN = +2.0VDC, fMCLK=24MHz, Rs=25Ω. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7, 8, & 12) CCD/CIS Source Requirements for Full Specified Accuracy and Dynamic Range (Note 12) VOS PEAK Sensor’s Maximum Peak Differential Signal Range Color Balance PGA gain = 0.933 Color Balance PGA gain = 3.0 Color Balance PGA gain = 9.0 (Blue Channel, BOOST on) Sensor’s Maximum Correctable Range of Pixel-to-Pixel VWHITE Variation/Color Range 0 Range 1 Range 2 1.0 0.33 0.11 V (min) V (min) V (min) 67 (1:3) 50 (1:2) 33 (1:1.5) % % % Coarse Color Balance PGA Characteristics Monotonicity 5 bits (min) V/V (max) V/V (min) G0 (Minimum PGA Gain) PGA Setting = 0 0.93 0.90 0.97 G31 (Maximum PGA Gain) PGA Setting = 31 3.00 2.85 3.20 V/V (max) V/V (min) GBLUE BOOST Blue PGA Boost Setting = x3 3.00 2.72 3.27 V/V (max) V/V (min) ±1.0 ±4.0 % (min) % (max) 6 bits (min) Gain Error at any gain (Note 14) Static Offset DAC Characteristics Monotonicity Offset DAC LSB size In units of ADC LSBs 4.2 1.3 7.2 LSB (min) LSB (max) Offset DAC Adjustment Range In units of ADC LSBs ±130 ±118 LSB (min) 10 Bits (min) ADC Characteristics Resolution with No Missing Codes ILE Integral Linearity Error (Note 11) ±0.4 ±1.5 LSB (max) DNL Differential Non-Linearity -0.14 +0.32 -0.5 +1.0 LSB (min) LSB (max) 2.00 1.83 2.17 (min) (max) System Characteristics C Analog Channel Gain Constant 3 http://www.national.com Electrical Characteristics (Continued) The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=+5.0VDC, VDI/O=+5.0 or +3.3VDC, VREF IN = +2.0VDC, fMCLK=24MHz, Rs=25Ω. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7, 8, & 12) VOS1 Pre-PGA Analog Channel Offset Error CCD Mode In units of ADC LSBs +5 -40 +40 LSB (min) LSB (max) VOS1 Pre-PGA Analog Channel Offset Error CIS Mode In units of ADC LSBs +32 -65 +135 LSB (min) LSB (max) VOS2 Post-PGA Analog Channel Offset Error In units of ADC LSBs -5 -40 +30 LSB (min) LSB (max) Reference and Analog Input Characteristics OSR, OSG, OSB Input Capacitance 5 pF OSR, OSG, OSB Input Leakage Current Measured with OS = 2.5VDC 2 20 nA (max) VREF IN Input Leakage Current Measured with VREF IN = 2.0VDC 4 20 nA (max) DC and Logic Electrical Characteristics The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=+5.0VDC, VDI/O=+5.0 or +3.3VDC, VREF IN = +2.0VDC, fMCLK=24MHz, Rs=25Ω. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7 & 8) Symbol Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) D0-D9, CD0-CD9, MCLK, RUN/STOP, SYNC, CS, RD, WR, RD PIXEL Digital Input Characteristics VIN(1) Logical “1” Input Voltage VDI/O=5.25V VDI/O=3.6V 2.0 2.0 V (min) V (min) VIN(0) Logical “0” Input Voltage VDI/O=4.75V VDI/O=3.0V 0.8 0.7 V (max) V (max) IIN Input Leakage Current VIN=VD VIN=DGND CIN Input Capacitance 0.1 -0.1 µA µA 5 pF D0-D9, EOC, GCLK, OCLK, SYNC Digital Output Characteristics VOUT(1) Logical “1” Output Voltage VDI/O=4.75V, IOUT=-360µA VDI/O=4.75V, IOUT=-10µA VDI/O=3.0V, IOUT=-360µA VDI/O=3.0V, IOUT=-10µA 2.4 4.4 2.1 2.5 V (min) V (min) V (min) V (min) VOUT(0) Logical “0” Output Voltage VDI/O=5.25V, IOUT=1.6mA VDI/O=3.6V, IOUT=1.6mA 0.4 0.4 V (max) V (max) TRI-STATE® Output Current (D0-D9 only) VOUT=DGND VOUT=VD IOUT 0.1 -0.1 µA µA Ø1, Ø2, RS, TR1, TR2 Digital Output Characteristics VOUT(1) Logical “1” Output Voltage VD=4.75V, IOUT=-360µA VD=4.75V, IOUT=-10µA 2.4 4.4 V (min) V (min) VOUT(0) Logical “0” Output Voltage VD=5.25V, IOUT=1.6mA 0.4 V (max) Power Supply Characteristics IA Analog Supply Current Operating Standby 60 7 85 10 mA (max) mA (max) ID Digital Supply Current Operating Standby 4.7 0.1 10 1.0 mA (max) mA (max) 4 http://www.national.com DC and Logic Electrical Characteristics (Continued) The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=+5.0VDC, VDI/O=+5.0 or +3.3VDC, VREF IN = +2.0VDC, fMCLK=24MHz, Rs=25Ω. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7 & 8) Symbol IDI/O Parameter Digital I/O Supply Current Conditions Operating, VDI/O=5.0V Standby, VDI/O= 5.0V Operating, VDI/O=3.3V Standby, VDI/O= 3.3V Typical (Note 9) Limits (Note 10) Units (Limits) 12 5 2 0.3 30 20 10 3 mA (max) mA (max) mA (max) mA (max) AC Electrical Characteristics, MCLK Independent The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=VDI/O=+5.0VDC, VREF IN = +2.0VDC, fMCLK=24MHz, tMCLK=1/fMCLK, tr=tf=5ns, Rs=25¾, CL (databus loading) = 50pF/pin. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7 & 8) Symbol fMCLK Limits (Note 10) Units (Limits) Maximum MCLK Frequency Minimum MCLK Frequency 24 4 MHz (min) MHz (max) MCLK Duty Cycle 40 60 % (min) % (max) Parameter Conditions Typical (Note 9) tSETUP (OUT) Coefficient Data valid before latching edge of OCLK or GCLK GCLK and OCLK as outputs 12 20 ns (min) tHOLD (OUT) Coefficient Data held after latching edge of OCLK or GCLK GCLK and OCLK as outputs -10 0 ns (min) tSETUP (IN) Coefficient Data Valid before latching edge of OCLK or GCLK GCLK and OCLK as inputs 0 5 ns (min) tHOLD (IN) Coefficient Data held after latching edge of OCLK or GCLK GCLK and OCLK as inputs 0 5 ns (min) tGCLK-EOC Rising edge of GLCK to falling edge of EOC (GCLK as output) tGCLK-OCLK Rising edge of GLCK to falling edge of OCLK (GCLK and OCLK as outputs) 2 ns 2 bus / 2 clock mode 1 ns Rising edge of EOC to rising edge of OLCK (OCLK as output) 2 clock mode 1 ns Rising edge of OLCK to falling edge of GLCK (GCLK and OCLK as outputs) 2 clock mode 3 ns tEOC-GCLK Rising edge of EOC to falling edge of GLCK (GCLK as output) 2 bus mode 2 ns tDACC RD or RD_PIXEL low to D0-D9 data valid 15 41 ns (max) tD1H, D0H RD or RD_PIXEL high to D0-D9 data tri-state 13 20 ns (max) tCS SETUP CS setup of RD or WR 0 ns (min) tCS HOLD CS hold after RD or WR 0 ns (min) D0-D9 data valid before rising edge of WR (setup time) 5 ns (min) tEOC-OCLK tOCLK-GCLK tWR SETUP 5 http://www.national.com AC Electrical Characteristics, MCLK Independent (Continued) The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=VDI/O=+5.0VDC, VREF IN = +2.0VDC, fMCLK=24MHz, tMCLK=1/fMCLK, tr=tf=5ns, Rs=25¾, CL (databus loading) = 50pF/pin. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7 & 8) Symbol Parameter tWR HOLD D0-D9 data valid after rising edge of WR (hold time) Conditions Typical (Note 9) Limits (Note 10) Units (Limits) 0 ns (min) AC Electrical Characteristics, MCLK Dependent The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=VDI/O=+5.0VDC, VREF IN = +2.0VDC, fMCLK=24MHz, tMCLK=1/fMCLK, tr=tf=5ns, Rs=25Ω, CL (databus loading) = 50pF/pin. Refer to Table 2: Configuration Register Parameters for limits labeled C.R. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7 & 8) Symbol Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) MCLKs (max) tR/S START Rising edge of RUN/STOP to Rising edge of SYNC SYNC OUT mode 16 tSTART Rising edge of SYNC to rising edge of TR1 SYNC IN mode SYNC IN mode SYNC OUT mode MCLKs (min) 1+tGUARDBAND tGUARDBAND 13+tGUARDBAND 14+tGUARDBAND MCLKs (max) tGUARDBAND tGUARDBAND MCLKs tSYNC END Falling edge of SYNC to last rising edge of EOC SYNC IN mode SYNC OUT mode tSYNC LOW SYNC low time SYNC IN mode SYNC OUT mode 9 0 MCLKs (max) MCLKs (max) 8 # of EOL PIxels - 4 MCLKs (min) TR1, TR2 Transfer Pulse Width C.R. MCLKs TR1, TR2 Transfer Pulse Guardband C.R. MCLKs tRS WIDTH Reset Pulse Width C.R. MCLKs tCLAMP ON Start of Optical Black clamping period C.R. MCLKs tCLAMP OFF End of Optical Black clamping period C.R. MCLKs 12 24 MCLKs MCLKs tTR1 WIDTH, tTR2 WIDTH tGUARD Standard Mode Even/Odd Mode 8 0 Pixel periods tØ Ø1, Ø2 clock period tRS RS pulse position from Ø1 edge C.R. MCLKs tHOLD REF Reference hold (sample) position from RS edge C.R. MCLKs tHOLD SIG Signal hold (sample) position from RS edge C.R. MCLKs tOCLK LOW OCLK low time (OCLK as Output) 2 bus mode 1bus mode 2*tMCLK 1*tMCLK ns tOCLK HIGH OCLK high time (OCLK as Output) 2 bus mode 1 bus mode 2*tMCLK 3*tMCLK ns tGCLK LOW GCLK low time (GCLK as Output) 2 bus mode 1 bus mode 2*tMCLK 1*tMCLK ns tGCLK HIGH GCLK high time (GCLK as Output) 2 bus mode 1 bus mode 2*tMCLK 3*tMCLK ns 6 http://www.national.com AC Electrical Characteristics, MCLK Dependent (Continued) The following specifications apply for AGND=DGND=DGNDI/O=0V, VA=VD=VDI/O=+5.0VDC, VREF IN = +2.0VDC, fMCLK=24MHz, tMCLK=1/fMCLK, tr=tf=5ns, Rs=25Ω, CL (databus loading) = 50pF/pin. Refer to Table 2: Configuration Register Parameters for limits labeled C.R. Boldface limits apply for TA=TJ=TMIN to TMAX; all other limits TA=TJ=25°C. (Notes 7 & 8) Symbol Parameter Typical (Note 9) Conditions Limits (Note 10) Units (Limits) tEOC LOW EOC low time 2*tMCLK ns tEOC HIGH EOC high time 2*tMCLK ns tDATA VALID D0-D9 data valid from falling (Data Read Phase = 0°) or rising (Data Read Phase = 180°) edge of EOC 4*tMCLK-20ns ns (min) 1 bus mode w/ext OCLK tMCLK + 40ns 4*tMCLK ns (min) ns (max) 2 bus mode w/ext OCLK 40ns 3*tMCLK ns (min) ns (max) w/ext GCLK 40ns 3*tMCLK ns (min) ns (max) OCLK rising edge to EOC falling edge tOCLK-EOC 1 (OCLK IN) (Gain Coefficient Write Phase = 0°), OCLK falling edge to EOC falling edge (Gain Coefficient Write Phase = 180°) OCLK rising edge to EOC rising edge tOCLK-EOC 2 (OCLK IN) (Gain Coefficient Write Phase = 0°), OCLK falling edge to EOC rising edge (Gain Coefficient Write Phase = 180°) GCLK rising edge to EOC falling edge tGCLK-EOC (GCLK IN) (Gain Coefficient Write Phase = 0°), GCLK falling edge to EOC falling edge (Gain Coefficient Write Phase = 180°) Note 1: Absolute Maximum Ratings indicate limits beyond which damage to the device may occur. Operating Ratings indicate conditions for which the device is functional, but do not guarantee specific performance limits. For guaranteed specifications and test conditions, see the Electrical Characteristics. The guaranteed specifications apply only for the test conditions listed. Some performance characteristics may degrade when the device is not operated under the listed test conditions. Note 2: All voltages are measured with respect to GND=AGND=DGND=DGNDI/O=0V, unless otherwise specified. Note 3: When the input voltage (VIN) at any pin exceeds the power supplies (VIN<GND or VIN>VA or VD), the current at that pin should be limited to 25mA. The 50mA maximum package input current rating limits the number of pins that can simultaneously safely exceed the power supplies with an input current of 25mA to two. Note 4: The maximum power dissipation must be derated at elevated temperatures and is dictated by TJmax, ΘJA and the ambient temperature, TA. The maximum allowable power dissipation at any temperature is PD = (TJmax - TA) / ΘJA. TJmax = 150°C for this device. The typical thermal resistance (ΘJA) of this part when board mounted is 52°C/W for the V52A PLCC package. Note 5: Human body model, 100pF capacitor discharged through a 1.5kΩ resistor. Note 6: See AN450 “Surface Mounting Methods and Their Effect on Product Reliability” or the section titled “Surface Mount” found in any post 1986 National Semiconductor Linear Data Book for other methods of soldering surface mount devices. Note 7: A Zener diode clamps the OS analog inputs to AGND as shown below. This input protection, in combination with the external clamp capacitor and the output impedance of the sensor, prevents damage to the LM9812 from transients during power-up. To Internal Circuitry OS Input AGND Note 8: To guarantee accuracy, it is required that VA and VD be connected together to the same power supply with separate bypass capacitors at each supply pin. Note 9: Typicals are at TJ=TA=25°C, fMCLK = 24MHz, and represent most likely parametric norm. Note 10: Tested limits are guaranteed to National's AOQL (Average Outgoing Quality Level). Note 11: Integral linearity error is defined as the deviation of the analog value, expressed in LSBs, from the straight line that best fits the actual transfer function of the ADC. Note 12: VREF is defined as the CCD OS voltage for the reference period following the reset feedthrough pulse. VWHITE is defined as the peak CCD pixel output voltage for a white (full scale) image with respect to the reference level, VREF . VRFT is defined as the peak positive deviation above VREF of the reset feedthrough pulse. The maximum correctable range of pixel-to-pixel VWHITE variation is defined as the maximum variation in VWHITE (due to PRNU, light source intensity variation, optics, etc.) that the 7 http://www.national.com LM9812 can correct for using its internal PGA. CCD Output Signal VRFT VWHITE VREF Note 13: Reference voltages below 1.80V may decrease SNR. Reference voltages above 2.20V may cause linearity (headroom) errors inside the LM9812. The LM4041DIM3-ADJ (SOT-23 package) or the LM4041DIZ-ADJ (TO-92 package) bandgap voltage references are recommended for this application. Note 14: Gain PGA Gain Error is the maximum difference between the measured gain for any PGA code and the ideal gain calculated by using the formula PGA code 32 V ---- = G + X --------------------------- where X = ( G – G ) ------ . 0 31 0 31 PGA V 32 8 http://www.national.com Pin Descriptions General Digital I/O Sensor Driver Signals ø1 Digital Output. Clock signal, phase 1. ø2 Digital Output. Clock signal, phase 2. RS Digital Output. Reset pulse. TR1, TR2 Digital Output. Transfer pulses. MCLK Digital Input. This is the 24MHz (typical) master system clock. SYNC Digital Input (SYNC_IN mode) / Digital Output (SYNC_OUT mode). In the SYNC_IN mode, a low-to-high transition on this input begins a line scan operation. The line scan operation terminates when SYNC is taken low. In the SYNC_OUT mode, the rising edge of this output indicates the start of a line of data and the falling edge indicates the end of a line of data. RUN/STOP Digital Input. In the SYNC_OUT mode, this input should be taken high to begin converting a series of lines, and taken low to stop converting a series of lines. In the SYNC_IN mode this input is ignored. Analog I/O OSR, OSG, OSB, Analog Inputs. These inputs (for Red, Green, and Blue) should be tied to the sensor’s OS (Output Signal) through DC blocking capacitors. VREF IN Analog Input. This is the system reference voltage input and should be connected to a 2.0V voltage source and bypassed to AGND with a 0.05µF monolithic capacitor. VREF LO OUT, VREF LO IN Analog Output/Input. VREF LO OUT is a voltage equal to 0.49VA - VREF IN/2 (1.45V for VREF IN=2V) developed by the LM9812. It should be tied to VREF LO IN and bypassed to AGND with a 0.05µF monolithic capacitor. Digital Coefficient I/O CD0 (LSB) CD9 (MSB) Digital Inputs. This is the 10 bit data path for the pixel-rate gain coefficient and offset data. OCLK Digital Input/Output. This is the signal that is used to clock the Offset coefficients into the LM9812 through the CD0-CD9 databus. This can be either an output or an input depending on the state of bit 7 of Register 9. Data is latched on the rising edge of OCLK. GCLK Digital Input/Output. This is the signal that is used to clock the Shading (gain) coefficients into the LM9812 through the CD0-CD9 databus. This can be either an output or an input depending on the state of bit 7 of Register 9. Data is latched on the rising edge of GCLK. VREF MID OUT, Analog Output/Input. VREF MID OUT is a voltage equal to 0.49VA (2.45V for VREF IN=2V) develVREF MID IN oped by the LM9812. It should be tied to VREF MID IN and bypassed to AGND with a 0.05µF monolithic capacitor. VREF HI OUT, VREF HI IN Analog Output/Input. VREF HI OUT is a voltage equal to 0.49VA + VREF IN/2 (3.45V for VREF IN=2V) developed by the LM9812. It should be tied to VREF HI IN and bypassed to AGND with a 0.05µF monolithic capacitor. Configuration Register I/O CS Digital Input. This is the Chip Select signal for reading or writing to the Configuration Register through the D0-D9 databus. This input must be low in order to enable writing to or reading from the Configuration Register. RD Digital Input. A low signal on this input, when SYNC and CS are also low, places the data in the currently addressed Configuration Register on the D0-D9 databus. A RD cycle also resets the internal address latching state machine. NOTE: If this pin is taken low when CS is high, the D0-D9 databus will come out of tri-state and drive random data onto the bus. WR Digital Output I/O Digital Input. This input, when simultaneously asserted along with CS, transfers the data on the D0-D9 databus to the LM9812. If this is the first write in the cycle, this data is the address to be read or written to. If this is the second write in the cycle, this data is the data to be written to the Configuration Register at the currently latched address. Writing to the Configuration Register is independent of the state of SYNC. D0 (LSB)D9 (MSB) Digital Inputs/Outputs. When SYNC is high and RD PIXEL is low, this data bus outputs the 10 bit digital output data during line scan. This databus is also used for reading or writing to the Configuration Register using the RD, WR and CS signals. Writing to the Configuration Register can occur at any time. Reading from the Configuration Register can only occur when SYNC is low. EOC Digital Output. This is the End Of Conversion signal from the LM9812 indicating that new pixel data is available on the D0-D9 databus. RD PIXEL Digital Input. When SYNC is high, taking this input low places the digital pixel data stored in the output latch onto the D0-D9 bus. This input is ignored when SYNC is low. Test TEST 9 Analog Output. This pin can be used to view the CDS and Clamp signals. See Register 27, bits 6 and 7. http://www.national.com Analog Power VA This is the positive supply pin for the analog supply. It should be connected to a voltage source of +5V and bypassed to AGND with a 0.1µF monolithic capacitor in parallel with a 10µF tantalum capacitor. AGND This is the ground return for the analog supply. Digital Power VD This is the positive supply pin for the digital supply. It should be connected to a voltage source of +5V and bypassed to DGND with a 0.1µF monolithic capacitor. DGND This is the ground return for the digital supply. VDI/O This is the positive supply pin for the digital supply for the LM9812’s I/O. It should be connected to a voltage source of +3V to +5V and bypassed to DGNDI/O with a 0.1µF monolithic capacitor. If the supply for this pin is different than the supply for VA and VD, it should also be bypassed with a 10µF tantalum capacitor. DGNDI/O This is the ground return for the digital supply for the LM9812’s I/O. 10 http://www.national.com Timing Diagrams RUN/STOP SYNC (OUTPUT) Ø1 Line 1 Line 2 Line n Diagram 1: SYNC OUT Mode Timing, Multiple Lines SYNC (INPUT) Ø1 Line 1 Line 2 Line n (truncated) Diagram 2: SYNC IN Mode Timing, Multiple Lines RUN/ STOP Transfer Dummy Black Valid Pixels Additional Integration Time SYNC (OUTPUT) Ø1 TR1 RS CLAMP 1 2 3 OCLK GLCK EOC Diagram 3: SYNC OUT Mode Timing, One Line 11 http://www.national.com Transfer Dummy Black Valid Pixels SYNC (INPUT) Ø1 TR1 RS CLAMP 1 2 3 OCLK GLCK EOC Diagram 4: SYNC In Mode Timing, One Line RUN/STOP tR/S START SYNC (OUTPUT) tSTART Ø1 TR1 Diagram 5: Start of Line Scan, SYNC OUT Mode, One TR Pulse SYNC (INPUT) tSTART Ø1 TR1 Diagram 6: Start of Line Scan, SYNC IN Mode, One TR Pulse tSYNC LOW 8MCLK min SYNC tSYNC END 8MCLK max EOC Diagram 7: Timing for End of Line/Start of Next Line 12 http://www.national.com tTR1 WIDTH TR1 (TR1 Polarity = 0) TR1 (TR1 Polarity = 1) tGUARD tGUARD Ø1 Ø2 Diagram 8: TR Pulse Timing, One TR Pulse tTR1 WIDTH TR1 (TR1 Polarity = 0) TR1 (TR1 Polarity = 1) tGUARD tTR2 WIDTH tGUARD TR1 (TR2 Polarity = 0) TR2 (TR2 Polarity = 1) tGUARD tGUARD Ø1 Ø2 Diagram 9: TR Pulse Timing, Two TR Pulses tRS WIDTH RS (RS Polarity = 0) RS (RS Polarity = 1) Diagram 10: RS Pulse Polarity RS n-1 n n-1 n+1 n n+1 OS tCLAMP ON Clamp Signal (Internal) n-1 n tCLAMP OFF Diagram 11: CCD Clamping Timing 13 http://www.national.com tØ Ø1 (Even/Odd Mode) Ø2 (Even/Odd Mode) tØ Ø1 (Standard Mode) Ø2 (Standard Mode) tRS tRS WIDTH RS n-1 n n-1 n+1 n n+1 OS t HOLD SIG tHOLD REF CDS (Internal) Hold Ref(n-1) Hold Signal(n-1) Hold Ref(n) Hold Signal(n) Hold Ref(n-1) Diagram 12: CDS Timing Ø1(Std Mode) Ø2 (Std Mode) Ø1(Even/Odd) Ø2 (Even/Odd) RS OS pixel 1 pixel i (dummy) (dummy) pixel i +1 pixel i +j (black) (black) (data) CDS Clamp Note: i = value programmed in Dummy Pixel Register j = value programmed in Optical Black Register Diagram 13: Dummy Pixel and Optical Black Pixel Timing 14 http://www.national.com tSETUP tGLCK-OCLK tHOLD tOCLK-GCLK OCLK (from LM9812) GCLK (from LM9812) tSETUP tGCLK-EOC tHOLD CD0-CD9 B Gainn-1 R Offn R Gainn G Offn G Gainn EOC tEOC-OCLK G Output Datan-1 D0-D9 B Output Datan-1 Diagram 14: Digital Data Timing, Output Data and Coefficient Data on separate buses, two clocks (OCLK and GCLK) for coefficient data, OCLK and GCLK generated by LM9812, RD PIXEL held low. tSETUP tHOLD GCLK (from LM9812) tSETUP tGCLK-EOC CD0-CD9 B Gainn-1 tHOLD R Offn R Gainn G Offn G Gainn EOC tEOC-GCLK G Output Datan-1 D0-D9 B Output Datan-1 Diagram 15: Digital Data Timing, output data and coefficient data on separate buses, one clock (GCLK) for coefficient data, GCLK generated by LM9812, RD PIXEL held low. tSETUP tOCLK-GCLK tHOLD OCLK (from LM9812) GCLK (from LM9812) tSETUP tEOC-OCLK tGCLK-EOC tHOLD CD0-CD9 B Gainn-1 R Offsetn R Gainn G Offsetn G Gainn EOC = RD PIXEL D0-D9 EXTERNAL DATABUS G Output Datan-1 G Output Datan-1 R Offn B Output Datan-1 R Gainn B Output Datan-1 G Offn R Outn B Gainn R Outn Diagram 16: Digital Data Timing, output data and coefficient data on same buses, two clocks (OCLK and GCLK) for coefficient data, OCLK and GCLK generated by LM9812. 15 http://www.national.com tSETUP tSETUP tHOLD tGCLK-EOC tHOLD GCLK (from LM9812) CD0-CD9 B Gainn-1 R Offsetn R Gainn G Offsetn G Gainn EOC = RD PIXEL D0-D9 EXTERNAL DATABUS G Output Datan-1 R Outn B Output Datan-1 G Output Datan-1 R Offn R Gainn B Output Datan-1 G Offn G Gainn R Outn Diagram 17: Digital Data Timing, output data and coefficient data on same buses, one clock (GCLK) for coefficient data, GCLK generated by LM9812. tSETUP tHOLD OCLK (to LM9812) GCLK (to LM9812) tSETUP tGCLK-EOC (GCLK IN) tHOLD CD0-CD9 B Gainn-1 R Offn R Gainn G Offn G Gainn EOC tOCLK-EOC 2 (OCLK IN) G Output Datan-1 D0-D9 B Output Datan-1 Diagram 18: Digital Data Timing, output data and coefficient data on separate buses, two clocks (OCLK and GCLK) for coefficient data, OCLK and GCLK input to LM9812, OCLK and GLCK phase = 0°, RD PIXEL held low. tSETUP tHOLD OCLK (to LM9812) GCLK (to LM9812) tSETUP tHOLD CD0-CD9 R Gainn-1 G Offn G Gainn B Offn B Gainn EOC tOCLK-EOC 2 (OCLK IN) D0-D9 tGCLK-EOC (GCLK IN) G Output Datan-1 B Output Datan-1 Diagram 19: Digital Data Timing, output data and coefficient data on separate buses, two clocks (OCLK and GCLK) for coefficient data, OCLK and GCLK input to LM9812, OCLK and GLCK phase = 180°, RD PIXEL held low. 16 http://www.national.com tSETUP tHOLD OCLK (to LM9812) GCLK (to LM9812) tSETUP tHOLD CD0-CD9 B Gainn-1 R Offsetn R Gainn G Offsetn G Gainn tGCLK-EOC (GCLK IN) tOCLK-EOC 1 (OCLK IN) EOC G Output Datan-1 D0-D9 R Outn B Output Datan-1 Diagram 20: Digital Data Timing, output data and coefficient data on same bus, two clocks (OCLK and GCLK) for coefficient data, OCLK and GCLK input to LM9812, OCLK and GLCK phase = 0°. tSETUP tHOLD OCLK (to LM9812) GCLK (to LM9812) tSETUP tHOLD CD0-CD9 R Gainn-1 G Offsetn G Gainn B Offsetn tGCLK-EOC (GCLK IN) B Gainn tOCLK-EOC 1 (OCLK IN) EOC D0-D9 G Output Datan-1 R Outn B Output Datan-1 Diagram 21: Digital Data Timing, output data and coefficient data on same bus, two clocks (OCLK and GCLK) for coefficient data, OCLK and GCLK input to LM9812, OCLK and GLCK phase = 180°. DATA (INTERNAL) Green Internal Datan-1 Blue Internal Datan-1 EOC RD PIXEL tDATAVALID tDACC D0-D9 Green Output Datan-1 Bn-1 tD1H, tD0H B Output Datan-1 Diagram 22: RD PIXEL Output Timing (Data Read Phase = 0°) 17 http://www.national.com Green Internal Datan-1 DATA (INTERNAL) Blue Internal Datan-1 EOC tDATAVALID RD PIXEL tDACC Green Output Datan-1 D0-D9 tD1H, tD0H B Output Datan-1 Bn-1 Diagram 23: RD PIXEL Output Timing (Data Read Phase = 180°) tCS SETUP tCS HOLD tCS SETUP tCS HOLD CS WR tWR SETUP tWR SETUP tWR HOLD D0-D7 ADDRESS tWR HOLD DATA Note: To read from the Configuration Register, SYNC must be low, either by driving SYNC low (in SYNC IN mode), or taking RUN/STOP low and waiting for SYNC to go low at the end of a line (in SYNC OUT mode). The Configuration Register can be written to independent of the state of SYNC. Diagram 24: Writing to the Configuration Register, RD held high. SYNC tCS SETUP tCS HOLD tCS SETUP tCS HOLD CS WR RD tWR SETUP tWR HOLD D0-D7 tDACC ADDRESS tD1H, tD0H DATA Note: To read from the Configuration Register, SYNC must be low, either by driving SYNC low (in SYNC IN mode), or taking RUN/STOP low and waiting for SYNC to go low at the end of a line (in SYNC OUT mode). The Configuration Register can be written to independent of the state of SYNC. Diagram 25: Reading the Configuration Register 18 http://www.national.com Table 1: Configuration Register Address Table Address (Decimal) Address (Binary) Data Bits A4 A3 A2 A1 A0 0 1 2 3 4 5 6 7 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 1 1 0 0 0 1 1 0 0 1 1 0 0 1 0 1 0 1 0 D7 D6 D5 - - Sign D3 D2 D1 D0 MSB LSB Static Offset (Green) - - Sign MSB LSB Static Offset (Blue) - - Sign MSB LSB Coarse Color Balance PGA (Red) - - - MSB LSB Coarse Color Balance PGA (Green) - - - MSB LSB Coarse Color Balance PGA (Blue) x3 - - MSB LSB Internal Offset Subtractor Coefficient MSB LSB Internal Gain Multiplier Coefficient 1 0 D4 Static Offset (Red) LSB Internal Gain Multiplier Coefficient (LM9812 only) - - - - - - MSB Misc Group 1 9 10 11 12 13 14 0 0 0 0 0 0 1 1 1 1 1 1 0 0 0 1 1 1 0 1 1 0 0 1 1 0 1 1 1 1 1 1 16 1 0 0 0 0 18 19 20 1 1 1 0 0 0 0 0 0 0 1 0 1 1 0 Number of Number of Multiplier Multiplier Multiplier Offset Coefficient Databusses Gain Range Gain Range Coefficient Coefficient Clocks in System 1 0 Source Source Number of Dummy Pixels MSB LSB Number of Optical Black Pixels MSB LSB Number of Active Pixels (LSB) LSB Number of Active Pixels (MSB) - - 1 0 1 0 MSB Number of End of Line Integration Pixels (LSB) 0 0 1 - 0 15 17 Internal External Coefficient Clock LSB Number of End of Line Integration Pixels (MSB) - - MSB - - MSB TR1 Pulse Duration LSB TR2 Pulse Duration - - MSB LSB TR1/TR2 Guardbands - - MSB LSB Optical Black Clamp Start - - - MSB LSB Optical Black Clamp End - - - 19 MSB LSB http://www.national.com Address (Decimal) Address (Binary) Data Bits A4 A3 A2 A1 A0 21 22 23 1 1 1 0 0 0 1 1 1 0 1 1 1 0 1 D7 D6 D5 D4 D3 Reset Pulse Width Bit 2 D2 D1 D0 Reset Pulse Position Bit 0 Bit 4 Bit 0 Reference Sample Position - - - Bit 4 Bit 0 Signal Sample Position - - - Bit 4 Bit 0 Misc Group 2 24 1 1 0 0 0 Signal Polarity Preview 1 Power Down 8 or 10 bit Coefficient Bus Preview 2 Internal/ External SYNC pulse TR Pulse Mode (1 or 2 Pulses) Red/Blue Order for 2 Output sensors 3 or 2 Output sensor Standard or Odd/Even sensor Misc Group 3 25 1 1 0 0 1 - TR2 Enable TR1 Enable RS Enable Ø2 Enable Ø1 Enable Misc Group 4 26 1 1 0 1 0 27 1 1 0 1 1 28 1 1 1 0 0 Gain Offset Coefficient Coefficient Data Read Ø1 Polarity TR2 Pulse Write Write Phase During TR1 Polarity Phase Phase TR1 Pulse Polarity RS Pulse Polarity Ø1/Ø2 Polarity 0 0 0 0 0 0 Test Mode 1 TB7 TB6 0 0 0 Test Mode 2 0 0 0 0 0 Table 2: Configuration Register Parameters Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) Red Offset DAC (0) Green Offset DAC (1) Blue Offset DAC (2) Control Bits D5 (SIGN) 0 0 0 ••• 0 0 1 1 1 ••• 1 1 D4 (MSB) 0 0 0 ••• 1 1 0 0 0 ••• 1 1 D3 D2 D1 0 0 0 ••• 1 1 0 0 0 ••• 1 1 0 0 0 ••• 1 1 0 0 0 ••• 1 1 0 0 1 ••• 1 1 0 0 1 ••• 1 1 20 Result D0 (LSB) 0 1 0 ••• 0 1 0 1 0 ••• 0 1 Typical Offset (in ADC LSBs) 0.00 +4.2 +8.4 ••• +126.0 +130.2 0.00 -4.2 -8.4 ••• -126.0 -130.2 http://www.national.com Table 2: Configuration Register Parameters (Continued) Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) Red PGA (3) Green PGA (4) Blue PGA (5) Control Bits D7 D4 (MSB) 0 0 0 ••• 1 1 1 0 0 0 ••• 0 0 0 D7 Coarse Color Balance PGA (Blue with x3 multiplier bit set) (5) 1 1 1 ••• 1 1 1 Result D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D4 (MSB) 0 0 0 ••• 1 1 1 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D0 (LSB) 0 1 0 ••• 1 0 1 Typical Gain (V/V) 0.93 1.00 D0 (LSB) 0 1 0 ••• 1 0 1 Typical Gain (V/V) 2.79 3.00 D0 (LSB) 0 1 0 ••• 1 0 1 Offset (in ADC LSBs) 0 1 2 ••• 253 254 255 ••• 3.00 ••• 9.00 Internal Offset Subtractor Coefficient (6) D7 (MSB) 0 0 0 ••• 1 1 1 Internal Multiplier Gain Coefficient (7, 8) D9 (MSB) 0 0 0 ••• 1 1 1 Internal/External Coefficient Clocks (9) D7 0 1 External Coefficient Clock Source Internal Coefficient Clock Source Number of Coefficient Clocks (9) D5 0 1 GCLK and OCLK GCLK only Number of Databuses in System (9) D4 0 1 2 Databuses 1 Databus D6 D5 D4 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D8 D7 ••• ••• D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 ••• 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 ••• 21 D0 (LSB) 0 1 0 ••• 1 0 1 Gain (V/V) See Equation 5 http://www.national.com Table 2: Configuration Register Parameters (Continued) Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) Control Bits Result D2 0 1 0 1 Multiplier Gain Range (9) D3 0 0 1 1 Multiplier Coefficient Source (9) D1 0 1 External CD0-CD9 Bus Internal Register Offset Coefficient Source (9) D0 0 1 External CD0-CD9 Bus Internal Register Dummy Pixels (Minimum Value is 0) (10) D7 0 0 0 ••• 1 1 1 D6 0 0 0 ••• 1 1 1 D5 0 0 0 ••• 1 1 1 D4 0 0 0 ••• 1 1 1 D3 0 0 0 ••• 1 1 1 D2 0 0 0 ••• 1 1 1 D1 0 0 1 ••• 0 1 1 D0 0 1 0 ••• 1 0 1 0 1 2 ••• 253 254 255 Optical Black Pixels (Minimum Value is 1) (11) D7 0 0 0 ••• 1 1 1 D6 0 0 0 ••• 1 1 1 D5 0 0 0 ••• 1 1 1 D4 0 0 0 ••• 1 1 1 D3 0 0 0 ••• 1 1 1 D2 0 0 0 ••• 1 1 1 D1 0 1 1 ••• 0 1 1 D0 1 0 1 ••• 1 0 1 1 2 3 ••• 253 254 255 D12 D11 ••• ••• D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 ••• ••• 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D0 (LSB) 0 1 0 ••• 1 0 1 Active Pixels Active Pixels (Minimum Value is 0) (12, 13) D13 (MSB) 0 0 0 ••• 1 1 1 D12 D11 ••• ••• D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 ••• 1 1 1 ••• 1 1 1 0 1 1 ••• 0 1 1 D0 (LSB) 1 0 1 ••• 1 0 1 End-of-Line Pixels End of Line Integration Pixels (Minimum Value is 5) (14, 15) D13 (MSB) 0 0 0 ••• 1 1 1 Range 0, 67% (1:3) Range 1, 50% (1:2) Range 2, 33% (1:1.5) Not Used ••• 22 0 1 2 ••• 16381 16382 16383 5 6 7 ••• 16381 16382 16383 http://www.national.com Table 2: Configuration Register Parameters (Continued) Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) TR1 Pulse Duration (16) TR2 Pulse Duration (17) Control Bits D5 (MSB) 0 0 0 ••• 1 1 1 TR1/TR2 - Ø1 Guardband (18) D5 (MSB) 0 0 0 ••• 1 1 1 Optical Black Clamp Start Position (Must be Š RS Pulse Width) (19) D4 (MSB) 0 0 0 ••• 1 1 1 Optical Black Clamp Stop Position (20) D4 (MSB) 0 0 0 ••• 1 1 1 Reset Pulse Width (Minimum Value Is 1) (21) D7 (MSB) 0 0 0 1 1 1 1 Result D4 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D4 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D0 (LSB) 0 1 0 ••• 1 0 1 Clamp Start (1/(2*fMCLK)) 0 1 2 ••• 29 30 31 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D0 (LSB) 0 1 0 ••• 1 0 1 Clamp Stop (1/(2*fMCLK)) 0 1 2 ••• 29 30 31 D6 D5 (LSB) 1 0 1 0 1 0 1 0 1 1 0 0 1 1 D0 (LSB) 0 1 0 ••• 1 0 1 = (x/2)tpixel, where tpixel = 12tMCLK for 3 channel mode, tpixel = 8tMCLK for 2 channel mode (except for x = 0: duration for x = 0 is 0.5tpixel, same as for x = 1) TR Pulse Width D0 (LSB) 0 1 0 ••• 1 0 1 = (x/2)tpixel, where tpixel = 12tMCLK for 3 channel mode, tpixel = 8tMCLK for 2 channel mode (except for x = 0: guardband for x = 0 is 0.5tpixel, same as for x = 1) TR Guardband 0.5 tpixel 0.5 tpixel 1.0 tpixel ••• 30.5 tpixel 31.0 tpixel 31.5 tpixel 0.5 tpixel 0.5 tpixel 1.0 tpixel ••• 30.5 tpixel 31.0 tpixel 31.5 tpixel RS Pulse Width (1/(2*fMCLK)) 1 2 3 4 5 6 7 23 http://www.national.com Table 2: Configuration Register Parameters (Continued) Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) Control Bits D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 D3 D2 D1 0 0 0 ••• 1 1 1 0 0 0 ••• 1 1 1 0 0 1 ••• 0 1 1 Result D0 (LSB) 0 1 0 ••• 1 0 1 Reset Position (1/(2*fMCLK)) 0 1 2 ••• 29 30 31 D0 (LSB) 0 1 0 ••• 1 0 1 Reference Position (1/(2*fMCLK)) 0 1 2 ••• 29 30 31 D0 (LSB) 0 1 0 ••• 1 0 1 Signal Position (1/(2*fMCLK)) 0 1 2 ••• 29 30 31 Reset Pulse Position (21) D4 (MSB) 0 0 0 ••• 1 1 1 Reference Sample Position (22) D4 (MSB) 0 0 0 ••• 1 1 1 Signal Sample Position (23) D4 (MSB) 0 0 0 ••• 1 1 1 Signal Polarity (24) D7 0 1 Preview Mode (24) D6 0 0 1 1 SYNC Pulse (24) D4 0 1 Externally Supplied (SYNC is Input) Internally Generated (SYNC is Output) TR Pulses (24) D3 0 1 1 TR Pulse (TR1) 2 TR Pulses (TR1 and TR2) Red/Blue Order for GRGB CCDs (24) D2 0 1 GR First GB First Sensor Type (24) D1 0 1 Dual Output (RG-BG) Triple Output (R-G-B) Negative, CDS On (CCD) Positive, CDS Off (CIS) D5 0 1 0 1 Normal Operation Normal Operation 2x Speed 4x Speed 24 http://www.national.com Table 2: Configuration Register Parameters (Continued) Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) Control Bits Result Sensor Type (24) D0 0 1 Standard Even/Odd Power Down (25) D7 0 1 Operating Low Power Standby Mode Coefficient Bus Width (25) D6 0 1 TR2 Enable (25) D4 0 1 TR2 Enabled TR2 Disabled TR1 Enable (25) D3 0 1 TR1 Enabled TR1 Disabled RS Enable (25) D2 0 1 RS Enabled RS Disabled Ø2 Enable (25) D1 0 1 Ø2 Enabled Ø2 Disabled Ø1 Enable (25) D0 0 1 Ø1 Enabled Ø1 Disabled Gain Coefficient Write Phase (26) D7 0 1 0° 180° Offset Coefficient Write Phase (26) D6 0 1 0° 180° Data Read Phase (26) D5 0 1 0° 180° Ø1 Polarity During TR1 (26) D4 0 1 High Low TR2 Pulse Polarity (26) D3 0 1 High Low TR1 Pulse Polarity (26) D2 0 1 High Low In 8 bit mode, coefficient data is 8 bits wide, input through CD0-CD7. 25 10 bit 8 bit http://www.national.com Table 2: Configuration Register Parameters (Continued) Note: tMCLK = 1/fMCLK = 1 MCLK period. Examples given in parenthesis are for fMCLK = 24MHz (tMCLK = 41.66ns). Parameter (Address) Control Bits Result RS Pulse Polarity (26) D1 0 1 High Low Ø1/Ø2 Polarity (26) D0 0 1 Ø1 High/Ø2 Low Ø1 Low/ Ø2 High Test Mode (27) TB7 0 0 1 TB6 0 1 1 Test Mode (28) 0 0 Test Output Low Test Output Is CDS Signal Test Output is Clamp Signal Set bits D0-D5 to 0 0 0 0 26 0 0 0 Set all bits to 0 http://www.national.com Applications Information (Ø1) makes a low to high transition (period 4), the electrons from the next pixel flow into C1. The charge across C1 now contains the voltage proportional to the number of electrons plus VRESIDUAL, an error term. If OS is sampled at the end of period 3 and that voltage is subtracted from the OS at the end of period 4, the VRESIDUAL term is canceled and the noise on the signal is reduced ([VSIGNAL+VRESIDUAL]-VRESIDUAL = VSIGNAL). This is the principal of Correlated Double Sampling. 1.0 THEORY OF OPERATION The LM9812 removes errors from and digitizes up to 3 channels of linear sensor pixel streams, while providing all the necessary clock signals to drive the sensor. Offset and gain errors are removed at the pixel rate, for individual pixels. Offset errors are removed through correlated double sampling (CDS), an analog offset DAC for large DC offsets, and finally a pixel-rate digital offset subtractor for individual pixel offsets. Gain errors (which may come from any combination of PRNU, uneven illumination, cos4 effect, RGB filter mismatch, etc.) are removed through the use of a 5 bit PGA in front of the ADC (for coarse gain correction) and a 10x10bit pixel-rate digital multiplier for individual pixel gain errors. The LM9812’s CDS circuit acquires a signal within a 1 MCLK window which can be placed anywhere in the pixel period with 0.5 MCLK precision. See Diagram 12 for more detailed timing information. 1.2 CIS Mode The LM9812 provides some support for CIS (Contact Image Sensor) devices by offering a sampling mode for capturing positive going signals, as opposed to the CCD’s negative going signal. The output signal of a CIS sensor (Figure 2) differs from a CCD signal in two primary ways: its output increases with increasing signal strength, and it does not usually have a reference level as an integral part of the output waveform of every pixel. 1.1 ANALOG SIGNAL PATH (See Functional Block Diagram) 1.1.1 Clamping and Buffering The Output Signals (OS) from the image sensor are capacitively coupled to the three (OSR, OSG, OSB) analog inputs of the LM9812. Inside the LM9812, a DC restore operation is performed by clamping the input signal to 2.5V when the input signal is known to be black (during the start of a new line of image data). To eliminate loading of the input signal, the signal is buffered through a source follower before being sent to the CDS section. OS (CIS) 1.1.2 CDS The LM9812 uses a high performance CDS (Correlated Double Sampling) circuit to remove many sources of noise and error from the CCD signal. It also supports CIS image sensors with a single sampling mode for positive-going signals. OS (CCD) 1 Figure 1 shows the output stage of a typical CCD and the resulting output waveform: VDD Q1 C1 RS (RESET) e- 2 3 4 5 Figure 2: CIS vs. CCD Output signals When the LM9812 is in CIS mode (Register 24, bit 7=1), it uses the VREF MID level as the reference (or black) voltage for each pixel. Since the LM9812 clamps the input signal to VREF MID at the black portion at the beginning of every line, the output of the sampler is an accurate and repeatable representation of the that pixel’s brightness. Q2 OS (from shift register) 1.2.1 Static PGAs The output of the CDS stage drives the PGAs (Programmable Gain Amplifiers). Each PGA provides 5 bits of fixed gain correction over a 0.93V/V to 3V/V (-0.6 to 9.5dB) range. The Blue channel has an optional x3 stage for a gain range of 2.8V/V to 9.0V/V (8.9 to 19dB). The gain of each PGA should be set during calibration to bring the maximum amplitude of the strongest pixel to a level just below the desired maximum output from the ADC. The gain is determined by the following equation: VSS Ø1 RS OS 1 2 3 4 5 PGA code V Gain ---- = 0.93 + --------------------------- 2.137 32 V Equation 1: PGA Gain Figure 1: CDS Capacitor C1 converts the electrons coming from the CCD’s shift register to an analog voltage. The source follower output stage (Q2) buffers this voltage before it leaves the CCD. Q1 resets the voltage across capacitor C1 in between every pixel at intervals 2 and 5. When Q1 is on, the output signal (OS) is at its most positive voltage. After Q1 turns off (period 3), the OS level represents the residual voltage across C1 (VRESIDUAL). VRESIDUAL includes charge injection from Q1, thermal noise from the ON resistance of Q1, and other sources of error. When the shift register clock The Blue Channel has an additional gain stage with a gain of 3V/V that may be switched on to compensate for the low amplitude blue output of some sensors. With the x3 bit set (Register 2, bit 7=1), the blue gain is: PGA code V Gain ---- = 02.79 + --------------------------- 6.411 32 V Equation 2: Blue Channel PGA Gain with x3 Bit Set 27 http://www.national.com Gain Correction Coefficient Multiplier Gain(V/V) = 1 + -----------------------------------------------------------------------N Equation 5: Shading Multiplier Gain plier gain range setting of 1:1.5 provides very accurate control of the gain of each pixel. Systems with larger variation can use gain ranges of 1:2 or 1:3. This PGA is static - it does not change at the pixel rate. The PixelRate Shading Multiplier is used to eliminate pixel-to-pixel gain errors (typically sensor PRNU and shading errors). 1.2.2 Static Offset DACs The Static Offset DACs remove DC offsets generated by the sensor and the LM9812’s analog signal chain. The DACs should be set during calibration to the lowest value that still results in an ADC output code > 0 for all the pixels. Each LSB of the offset DAC is typically 4.2 ADC LSBs, providing a total offset adjustment range of ±130 ADC LSBs. The equation for the offset DAC is: Offset ( ADC LSBs ) = 4.2 ( DAC Code ) Equation 3: Offset DAC Like the PGA, the Offset DAC is static - it does not change at the pixel rate. The Pixel Rate Offset Subtractor is used to eliminate pixel-to-pixel offset errors. Multiplier Gain Range 1:1.5 (33%) 1:2.0 (50%) 1:3.0 (67%) Minimum Gain (V/V, Multiplier Input = 0) 1 1 1 Maximum Gain (V/V, Multiplier Input = 1024) 1.5 2.0 3.0 N (used in Equation 5) 2048 1024 512 Bits D3, D2 of Register 9 1, 0 0, 1 0,0 Table 1:Shading Multiplier: 10 bit Coefficient Bus Width When the 8 bit Coefficient Bus Width is selected (Register 25, bit 6=1), the coefficient data is only 8 bits wide, supplied on CD0CD7. This reduces the accuracy of the gain correction, but allows the use of an 8 bit path to store coefficient data, instead of a 10 or 16 bit wide path. 1.2.3 ADC The 10 bit ADC quantizes the output from the analog chain and passes the result to the digital section for pixel rate error correction. The ADC’s input range is equal to VREF IN/C, where C is the gain constant compensating for all sources of gain error throughout the entire analog signal path. Multiplier Gain Range 1:1.5 (33%) 1:2.0 (50%) 1:3.0 (67%) Minimum Gain (V/V, Multiplier Input = 0) 1 1 1 Maximum Gain (V/V, Multiplier Input = 256) 1.5 2.0 3.0 1.3 DIGITAL SIGNAL PATH 1.3.1 Pixel-Rate Offset Subtractor The output of the ADC is fed to the Pixel-Rate Offset Subtractor. Each pixel of image data may have a different offset error. If bit 0 of Register 9 is set to a 0 and a byte representing the offset error code is fed to the Pixel-Rate Offset Subtractor through the CD0CD9 databus, the offset for each pixel can be subtracted from the ADC output code at the pixel conversion rate. If this feature is not desired, the Offset Subtractor can be set to a fixed value by writing the desired value to Register 6 and setting bit 0 of Register 9 to a 1. N (used in Equation 5) 512 256 128 Bits D3, D2 of Register 9 1, 0 0, 1 0,0 Table 2:Shading Multiplier: 8 bit Coefficient Bus Width 1.4 SENSOR CLOCK GENERATION Offset Subtractor Out = ADC Output – Offset Error Code Equation 4: Offset Subtractor The offset subtractor input is 8 bits wide, input on CD0-CD7, for a correction range of 0 to 255. Information on bits CD8 and CD9 is ignored. The LM9812 generates all of the clock signals required to directly drive most commercial linear CCDs and some CIS - no external clock buffers are necessary. Most linear CCDs designed for scanner applications require 0 to 5V signal swings into 20 to 500pF input loading. Series resistors are typically inserted between the driver and the CCD to control slew rate and isolate the driver from the large load capacitances. The values of these resistors are given in the CCD’s datasheet. 1.3.2 Pixel-Rate Shading Multiplier The Pixel Rate Shading Multiplier follows the Offset Subtractor. This stage compensates for nonuniformities between individual pixels (shading error). Each pixel of image data may have a different gain error. If bit 1 of Register 9 is set to a 0 and a word representing the gain correction coefficient is fed to the Pixel-Rate Shading Multiplier through the CD0-CD9 databus, the gain for each pixel is changed at the pixel rate to eliminate pixel-to-pixel gain errors. If this feature is not desired, the Shading Multiplier can be set to a fixed value by writing the desired gain value to Registers 7 and 8 and setting bit 1 of Register 9 to a 1. 1.4.1 TR1 and TR2 The LM9812 supports one or two TR (transfer, or shift) pulses as shown in Diagrams 8 and 9. This pulse is used to transfer the contents of each pixel’s photodiode to the CCD’s serial shift register for clocking out of the CCD. Configuration Registers 16, 17, and 18 control the TR1, TR2, and TR guardband pulse widths, while Register 24, bit 3 determines whether the LM9812 generates TR1 only or both TR1 and TR2 at the beginning of a line. The polarities of the TR pulses are determined by Register 26, bits 2 and 3. If not needed, one or both TR pulses can be disabled by setting Register 25, bits 3 and 4 to the appropriate value. The equation for calculating the gain of the multiplier is given in Equation 5. The Pixel Rate Shading Multiplier has three different gain ranges as shown in Table 1. For systems where the ratio of the strongest pixel to the weakest pixel is less than 1.5, the multi- 28 http://www.national.com 1.4.2 RS for tDATAVALID. The D0-D9 databus comes out of tri-state when SYNC is high and RD PIXEL is low. If SYNC is low, RD PIXEL will have no effect. The LM9812 also generates the RS (reset) pulse required to clear the CCD’s output capacitor of the previous pixel’s charge. The RS pulse’s width and position (relative to the edge of the ø1 pulse) is determined by the value in Configuration Register 21. See Diagrams 10 and 12. The polarity of the RS pulse is determined by Register 26, bit 1. If not needed, this pulse can be disabled by setting Register 25, bit 2 to a 1. When reading pixel data, RD PIXEL may be driven by EOC, putting the data on the bus only when EOC is low, and allowing other data on the bus (such as CD0-CD9 correction data) at other times. In this way the output data and correction coefficient data can share the same databus (See Diagrams 16, 17, 20, and 21). 1.4.3 ø1 and ø2 1.5.3 Correction Coefficient Data I/O ø1 and ø2 are the two phase clock for the CCD shift register. ø2 is always the inverse of ø1. For “standard” CCDs, the ø1 and ø2 frequency is equal to the pixel rate, the same frequency as the RS pulse. In Standard Mode, the RS pulse position and Sample Reference/Sample Signal edges are set with respect to the falling edge of ø1. For “even/odd” CCDs, the ø1 and ø2 frequency is equal to one half the pixel rate, a pixels is sampled after both the rising and falling edges of the ø clock. In Even/Odd Mode, the RS pulse position and Sample Reference/Sample Signal edges are set with respect to the either edge of ø1. Standard or Even/Odd mode is selected by setting bit 0 of Register 24. The absolute polarity of the ø pulses is determined by Register 26, bits 0 and 4 (both bits perform the same function). If not needed, one or both ø pulses can be disabled by setting Register 25, bits 0 and 1 to the appropriate value. Coefficient data for the pixel rate Offset Subtractor and Shading Multiplier enters the LM9812 through the CD0-CD9 databus. To maximize flexibility for the system designer, there are several clocking options available in this mode: separate or combined buses for the CD0-CD9 and the D0-D9 data, one (GLCK) or two (GLCK and OCLK) clock signals to latch the Correction Coefficient data, and the option to have the LM9812 generate the clock signals or have them supplied to the LM9812. Timing for these different options is shown in Diagrams 14 through 21, and described below. 1.5 DIGITAL INTERFACE There are three main sections to the digital interface of the LM9812: the Configuration Register interface (through which all device programming is done), the Correction Coefficient Data interface (the 10 bit-wide input databus for gain and offset correction coefficients), and the 10 bit-wide Pixel Data output databus (where the corrected digital output data appears). CD0-CD9 tied to D0-D9 (1 bus) CD0-CD9 separate from D0-D9 (2 bus) GLCK, OCLK are outputs (2 Clock) Diagram 16 Diagram 14 GLCK is output (1 Clock) Diagram 17 Diagram 15 Diagrams 20, 21 Diagrams 18, 19 GLCK, OCLK are inputs Table 3:Correction Databus Options 1.5.1 Configuration Register I/O The Configuration Register is written to and read from through the D0-D9 databus, using the CS, WR, and RD signals. To write to the Configuration Register, follow the timing shown in Diagram 24. The first byte is the address of the Configuration Register to be written to, the second byte is the data to be stored at that address. Configuration Register writes can occur at any time, independent of the state of the SYNC pin. Diagram 14 shows the case where the correction data (CD0CD9) is on a separate bus from the output data (D0-D9) (Register 9, bit 4=0). The GLCK and OCLK signals are generated by the LM9812 (Register 9, bit 5=0, bit 7=1). Gain correction data is latched on the rising edge of GCLK, and offset correction data is latched on the rising edge of OCLK. There is a one EOC clock latency between the latching of the gain coefficient for a particular pixel and the output of that pixel on the D0-D9 databus. To read from the Configuration Register, follow the timing shown in Diagram 25. The first byte is the address of the Configuration Register to be written to, the second byte is the data stored at that address. The SYNC pin must be low in order to read from the Configuration Register. To ensure that SYNC is low when in the SYNC out mode, RUN/STOP must be low until the end of the active pixels. Diagram 15 shows the case where the correction data (CD0CD9) is on a separate bus from the output data (D0-D9) (Register 9, bit 4=0). The LM9812 generates the GCLK signal only (Register 9, bit 5=1, bit 7=1). Gain correction data is latched on the rising edge of GCLK, and offset correction data is latched on the falling edge of GCLK. There is a one EOC clock latency between the latching of the gain coefficient for a particular pixel and the output of that pixel on the D0-D9 databus. If the LM9812’s RD pin is taken low, even if the CS pin is high, the D0-D9 databus will be driven. Never take the RD pin low unless you are actually doing a Configuration Register read. Diagram 16 shows the case where the correction data (CD0CD9) is on the same bus as the output data (D0-D9) (Register 9, bit 4=1). The GLCK and OCLK signals are generated by the LM9812 (Register 9, bit 5=0, bit 7=1). Gain correction data is latched on the rising edge of GCLK, and offset correction data is latched on the rising edge of OCLK. Using the EOC output to control the RD PIXEL input allows the CD0-CD9 and the D0-D9 data to exist on the same bus with no contention. There is a one 1.5.2 Pixel Data I/O The output of the multiplier is available on the D0-D9 databus. If the Data Read Phase bit (bit 5 of Register 26) is set to a 0, Data changes shortly before the falling edge of EOC and remains valid for tDATAVALID. If the Data Read Phase bit is set to a 1, Data changes shortly before the rising edge of EOC and remains valid 29 http://www.national.com To operate the LM9812 with an externally supplied OCLK and GCLK, set bit D7 of register 9 to a 0 and set the phase bit for each clock. The procedure to determine the state of the phase bit is described at the end of this section. EOC clock latency between the latching of the gain coefficient for a particular pixel and the output of that pixel on the D0-D9 databus. Diagram 17 shows the case where the correction data (CD0CD9) is on the same bus as the output data (D0-D9) (Register 9, bit 4=1). The LM9812 generates the GCLK signal only (Register 9, bit 5=1, bit 7=1). Gain correction data is latched on the rising edge of GCLK, and offset correction data is latched on the falling edge of GCLK. Using the EOC output to control the RD PIXEL input allows the CD0-CD9 and the D0-D9 data to exist on the same bus with no contention. There is a one EOC clock latency between the latching of the gain coefficient for a particular pixel and the output of that pixel on the D0-D9 databus. DIagrams 18 through 21 show the timing required when GCLK and OCLK are configured as inputs to the LM9812 (Register 9, bit 5=0, bit 7=0). This option exists to allow the LM9812 to work with ASICs designed for earlier systems, where the ASIC generates the SRAM timing (instead of the LM9812). Diagrams 18 and 19 show the 2 bus mode (Register 9, bit 4=0), diagrams 20 and 21 show the 1 bus mode (Register 9, bit 4=1). In these modes, GLCK and OCLK come from an external source that may be asynchronous to the internally generated OCLK and GCLK. While these clocks may be asynchronous, they will be the same frequency, the ADC conversion rate (since offset and gain coefficients are needed for every pixel, and the pixel data rate is fixed by the ADC conversion rate). In the previous modes of operation, the offset correction clock (OCLK) and the gain (shading) correction clock (GCLK) are generated by the LM9812 and used to generate the RD pulses to coefficient RAM. These clocks can also be configured as inputs to allow compatibility with some existing ASICs or designs where it is preferred to let the ASIC generate all the timing. The offset data is always latched on the rising edge of OCLK. The gain data is always latched on the rising edge of GCLK. The circuit shown in Figure 3 is implemented inside the LM9812 to synchronize (by delaying) the external coefficient clocks and CD0-9 data with the internal LM9812 clocks. Internal OLCK phase = 0° D Q FF1O Q D Q FF2O Q 3:1 Mux Latched Offset Correction Data to Offset Subtractor D Q FF3O Q phase = 180° Ext OCLK Internal GLCK phase = 0° CD0-9 D Q FF1G Q D Q FF2G Q phase = 180° 3:1 Mux D Q FF3G Q Latched Gain Correction Data to Multiplier Ext GCLK EOC Figure 3: Using An External OCLK and GCLK clock and EOC (the LM9812’s internal coefficient clock) is known, then by selecting Q1 or Q2 as the input to FF3, the designer can guarantee that the data will be valid on the rising edge of GCLKINT, where it is latched into FF3 and synchronized to the LM9812. CD0-9 data is always latched into register FF1 on the rising edge of the external clock. The output of FF1 is then latched by FF2 on the falling edge of the external clock. So the latched data is available at the output of FF1 between the rising edges of the external clock, and is valid at the output of FF2 between the falling edges of the external clock. If the phase difference between the external GCLKINT is not externally available to the user but it is basically 30 http://www.national.com CCD sensors (this mode will not work for CIS sensors). In these modes, adjacent pixels are averaged together in the analog domain and converted at the maximum ADC conversion rate. This allows the CCD data to be clocked out and digitized 2 or 4 times faster (corresponding to a 2 to 4 times faster image scan). The image will be 1/2 or 1/4 the resolution. For example, a 600dpi sensor will look like a 300dpi sensor in the x2 preview mode, or a 150dpi sensor in the x4 preview mode. This is useful because it allows faster scans for lower resolution images and scan preview images. The quality of the lower resolution image is also significantly improved because this technique averages pixels to reduce the resolution instead of discarding pixels (which often results in a loss of image information). the inverse of EOC, plus or minus a few ns of gate delay. To use an external GLCK and OCLK, determine the phase of the external OCLK and GLCK with respect to EOC. Then look at Diagrams 18 and 19 (2 bus mode) or 20 and 21 (1 bus mode). Choose a phase setting (0° or 180°, register 26 bits 6 and 7) for OCLK and GCLK that provides the most margin with respect to the tOCLKEOC1 (OCLK IN), tOCLK-EOC2 (OCLK IN), and tGCLK-EOC (GCLK IN) specifications. It is entirely possible that either setting (0° or 180°) will work equally well. The condition that must be avoided is when FF3 is clocked (falling edge of EOC) at the same time the input data to FF3 is changing (rising edge of OCLK/GCLK for 0° phase, or falling edge of OCLK/GCLK for 180° phase). As long as this does not occur simultaneously, the data will be correctly latched. Ø1 3 (Even/Odd Mode) Note that the coefficient data must appear sooner (relative to the pixel output data) for the Clock In modes, to account for the extra latency through FF1 and FF2. This additional time is shown in Diagrams 18 through 21. 9 Ø2 (Even/Odd Mode) Ø1 (Standard Mode) This mode of operation may seem confusing at first, but it is necessary to allow the synchronous LM9812 to phase lock to external coefficient data of unknown phase. 6 1.5 1.5 3 6 Ø2 (Standard Mode) All lengths given in units of MCLK periods RS 1.5.4 MCLK This is the master clock input for the LM9812. The ADC conversion rate is fixed at 1/4 of this frequency. The pixel rate in 3 channel mode is 1/12 of the MCLK frequency. Many of the timing parameters are also relative to the frequency of this clock. OS pn pn+1 CDS (Internal) 1.5.5 SYNC Hold Ref Hold Signal Diagram 4: x2 Preview Mode Timing This input signals the beginning of a line. When SYNC goes high, the LM9812 generates a TR pulse, then begins converting pixels until the SYNC line is brought low again. If SYNC is externally applied, the LM9812 will work with sensors with any number of pixels. If SYNC is internally generated (in combination with the RUN/STOP input), sensors with up to 3 channels of 16383 pixels/channel can be used. In the x2 preview mode, the Ø clock is clocked at 2 times the normal Ø clock frequency, while the RS pulse stays at the ADC’s conversion rate. By skipping every other RS pulse, the charge for pixel n+1 will be added to the charge for pixel n on the CCD itself. Since the CCD is being clocked at 2 times the normal rate, the period between lines (and therefore the integration time) will be half as long, causing the amplitude of each pixel to be half as large. Since the output of the CCD is the sum of two pixels, the final amplitude seen by the LM9812 is very similar to what it is in normal mode. The final pixel amplitude will be equal to (pn/2 + pn+1/2) = (pn + pn+1)/2 = the average of pn and pn+1. 1.5.6 RUN/STOP The LM9812 has a “SYNC OUT” mode (Register 24, bit 4=1) that automatically generates a SYNC pulse stream based on the number of active pixels and the number of additional end-of-line integration pixels programmed into registers 12-15. When the RUN/STOP pin is brought high, the LM9812 begins generating the periodic SYNC pulse. When RUN/STOP goes low, the LM9812 will continue converting the remaining active pixels, then SYNC will go low and the part will “idle” until the next rising edge of RUN/STOP. The same principle applies in the x4 preview mode. Here the Ø clock is clocked at 4 times the normal Ø clock frequency, while the RS pulse stays at the ADC’s conversion rate. By skipping 3 out of 4 RS pulses, the charge for pixels n, n+1, n+2, and n+3 will be averaged in the CCD’s output stage. Since the CCD is being clocked at 4 times the normal rate, the period between lines (and therefore the integration time) will be one quarter as long. The final pixel amplitude will be equal to (pn + pn+1+ pn+2+ pn+3)/4 = the average of pn, pn+1, pn+2, and pn+3. Using the SYNC OUT mode provides a simple way to clock the sensor continuously with a repeatable, user-programmed integration period. By varying the number of end-of-line integration pixels (in registers 14 and 15), the integration time (and thus the amplitude of the output signal from the sensor) can be adjusted with very high accuracy (about 1 part in 5400, or 0.02%, for a 600dpi sensor). Getting the optimum performance out of the preview modes with a particular CCD may require adjusting the RS pulse width, RS pulse position, Sample Reference, and Sample Signal timing (Registers 21, 22, and 23). 2.0 PREVIEW MODE OPERATION The LM9812 supports two “preview” or low resolution modes for 31 http://www.national.com Ø1 1.5 1.5 1.5 7.5 (Even/Odd Mode) drives a transconductance amplifier with two output currents, VREF IN/2 and +VREF IN/2. These currents go into the VREF HI and VREF LO reference voltage buffers, which generate a nominal 3.45V (VREF HI = VREF MID + VREF IN/2), and 1.45V (VREF LO = VREF MID - VREF IN/2). To minimize noise, the reference amplifiers sense (IN) inputs are brought out to the chip for bypassing. Each buffer’s output and input should be tied together and decoupled to AGND through a 0.1µF monolithic capacitor. 7.5 Ø2 (Even/Odd Mode) 0.5 Ø1 6 (Standard Mode) 1 1 1 1.5 6 Ø2 (Standard Mode) 3.2 Clamp Capacitor Selection All lengths given in units of MCLK periods This section is very long because it is relatively complicated to explain, but the answer is short and simple: A clamp capacitor value of 0.01µF should work in almost all applications. The rest of this section describes exactly how this value is selected. RS pn OS pn+1 pn+2 pn+3 CCLAMP CDS (Internal) Hold Ref OS SENSOR Hold Signal Diagram 5: x4 Preview Mode Timing DOS OS NC 2.450V 3.0 ANALOG INTERFACE LM9812 3.1 Voltage Reference The VREF IN pin should be connected to a 2V±5% reference voltage. The LM4041-ADJ adjustable bandgap reference is recommended for this application, as shown in Figure 6. The inexpensive “D” grade meets all the requirements of the application and is available in a TO-92 (LM4041DIZ-ADJ) package as well as a SOT-23 package (LM4041DIM3-ADJ) to minimize board space. To reduce noise, the VREF IN pin should be bypassed to Figure 7: OS Clamp Capacitor and Internal Clamp The output signal of many sensors rides on a DC offset (greater than 5V for many CCDs) which is incompatible with the LM9812’s 5V operation. To eliminate this offset without resorting to additional higher voltage components, the output of the sensor is AC coupled to the LM9812 through a DC blocking capacitor, CCLAMP (the CCD’s DOS output, if available, is not used). The value of this capacitor is determined by the leakage current of the LM9812’s OS input and the output impedance of the sensor. The leakage through the OS input determines how quickly the capacitor value will drift from the clamp value of VREF MID, which then determines how many pixels can be processed before the droop causes errors in the conversion (±0.1V is the recommended limit). The output impedance of the sensor determines how quickly the capacitor can be charged to the clamp value during the black reference period at the beginning of every line. Reference Amplifiers VREF HI OUT VA VREF HI IN R 10.5k 0.1µF VREF MID OUT 10.1k VREF MID IN The minimum clamp capacitor value is determined by the maximum droop the LM9812 can tolerate while converting one sensor line. The following equation takes the maximum leakage current into the OS input, the maximum allowable droop (100mV), the number of pixels on the sensor, and the pixel conversion rate (fMCLK/12 for triple output sensors or fMCLK/8 for dual output sensors) and provides the minimum clamp capacitor value: 0.1µF VREF LO OUT VREF LO IN R -I +I I=VREF IN/2R 0.1µF VA VREF IN 0.1µF 3k i C CLAMP MIN = --------- dt dV leakage current (A) number of pixels = --------------------------------------------------- -----------------------------------------------------max droop(V) conversion rate (Hz) Equation 6: CCLAMP MIN Calculation 4.7k LM9812 LM4041-ADJ 3k For example, if the OS input leakage current is 20nA worst-case, the sensor has 2700 active pixels, the conversion rate is 2MHz (fMCLK = 24MHz, triple output sensor), and the max droop desired is 0.1V, the minimum clamp capacitor value is: 20nA 2700 C CLAMP MIN = -------------- --------------0.1V 2MHz = 270pF Equation 7: CCLAMP MIN Example Figure 6: Voltage Reference Generation AGND with a 0.1µF monolithic capacitor. The LM9812 generates three internal reference voltages for its analog front end: VREF MID, VREF HI, and VREF LO. VREF MID is equal to 0.49*VA, or 2.45V with a nominal 5V supply. VREF IN 32 http://www.national.com The maximum size of the clamp capacitor is determined by the amount of time available to charge it to the desired value during the optical black portion of the sensor output. The internal clamp is on for each pixel for the time specified in Registers 19 and 20 (see Diagram 11). This time can be calculated using this equation: Start-Stop t (s) = ----------------------------------DARK 2f (Hz) MCLK Equation 8: tDARK Calculation In this example a 0.01µF capacitor takes 14ms after power-up to charge to its final value, but its droop across all subsequent lines is now less than 3mV (using the previous example’s values). This wide margin is the reason a CCLAMP value of 0.01µF will work in most applications. 4.0 CALIBRATION System calibration is required to correct the many sources of error in a scanner and optimize image quality. There are many different ways to calibrate a system, some take a long time but produce better results, others are faster with potentially lower quality images. The method described below should produce very good results, but it is by no means the only way to approach scanner calibration. Some calibration steps could be eliminated to speed up the calibration procedure. Other steps could be improved by iteratively repeating them, verifying that the previous calculation achieved the desired result. Every scanner system is different, so every system may benefit from optimization of the calibration routine. Where Start is the Optical Black Clamp Start Position (Register 19), Stop is the Optical Black Clamp Stop Position (Register 20), fMCLK is the MCLK frequency, and tDARK is the amount of time (per pixel) that the clamp is on. The following equation takes the number of optical black pixels, the amount of time (per pixel) that the clamp is closed, the sensor’s output impedance, and the desired accuracy of the final clamp voltage and provides the maximum clamp capacitor value that allows the clamp capacitor to settle to the desired accuracy within a single line: t 1 C CLAMP MAX = ------ -------------------------------R ln(accuracy) t (s) n DARK = ------------------ -------------------------------R OUT ln(accuracy) Equation 9: CCLAMP MAX Calculation 4.0.1 LM9812 Configuration Sequence After Power-On The power-on reset circuit of the LM9812 may take several hundred microseconds. Wait 1ms after power-on before writing to the Configuration Registers. Where n = the number of optical black pixels, tDARK is the amount of time (per pixel) that the clamp is on, ROUT is the output impedance of the CCD, and accuracy is the ratio of the worst-case initial capacitor voltage to the desired final capacitor voltage. For example, if a sensor has 18 black reference pixels, the output impedance of the sensor is 1500Ω, the LM9812 is configured to clamp for 375ns, the worst case initial voltage across the capacitor is 10V, and the desired voltage after clamping is 0.1V (accuracy = 10/0.1 = 100), then: 18 375ns C CLAMP MAX = ------------- -----------------1500 ln(100) = 977pF Equation 10: CCLAMP MAX Example The final value for CCLAMP should be less than or equal to CCLAMP MAX, but no less than CCLAMP MIN. A value of 470pF will work in this example. Make sure the SYNC and RUN/STOP inputs are low before writing to the Configuration Registers. When SYNC is high, you can not read the configuration register (but you can still write to it). The LM9812 configuration sequence: • Set bit 2 of Test Register 28 to 0 (to allow writing of configuration data). This register is reset to 0 by the power-on reset, but this step is still recommended to ensure that writes will work even if this register was corrupted. • Program bit 4 of register 24 to determine if the LM9812 will be used in the SYNC OUT or SYNC IN mode. • Cycle the Powerdown bit to reset the LM9812’s state machines. (Take bit 7 of register 25 to a 1, then back to a 0). This procedure will completely reset the part. At this time the LM9812 is ready for data to be written to all 28 configuration registers. Data can also be read back from all 28 registers if write confirmation is desired. In some cases, depending primarily on the choice of sensor, CCLAMP MAX may actually be less than CCLAMP MIN, meaning that the capacitor can not be charged to its final voltage during the black pixels at the beginning of a line and hold it’s voltage without drooping for the duration of that line. This is usually not a problem because in most applications the sensor is clocked continuously as soon as power is applied. In this case, a larger capacitor can be used (guaranteeing that the CCLAMP MIN requirement is met), and the final clamp voltage is forced across the capacitor over multiple lines. This equation calculates how many lines are required before the capacitor settles to the desired accuracy: R OUT C CLAMP Initial Voltage lines = ----------------- ------------------------- ln ------------------------------------- t DARK Final Voltage n Equation 11: Line Settling Formula After all 28 configuration registers have been programmed, scanning can begin by taking SYNC high (in the SYNC IN mode) or RUN/STOP high (in the SYNC OUT mode). 4.1 Calibration Initialization • Set the Offset DACs to their maximum positive value (Offset DAC registers 0, 1, and 2 =31) • Set the PGA gains to 1V/V (PGA gain registers 3, 4, and 5 =1) • Set the Pixel Rate Offset Adder Source to internal (register 9, bit 0=1) • Set the Pixel Rate Multiplier Source to internal (register 9, bit 1=1) • Set the Internal Pixel Rate Offset Adder value to 0 (register 6=0) • Set the Internal Pixel Rate Multiplier value for a gain of 1 (registers 7 and 8 = 0) Using the values shown before and a clamp capacitor value of 0.01µF, this works out to be: 1500 0.01µF 10V lines = ------------- ------------------ ln ------------ = 10.2 lines 18 375ns 0.1V Equation 12: Line Settling Example At a 2MHz conversion rate, this is about 14ms. 33 http://www.national.com • Set the Multiplier Gain Range to 1:3 (register 9, bits 2 and 3 = 00) Target Code WhiteMAX ADC Output Code 4.2 Setting the Static Offset DACs 018 014 ADC 010 Output 00C Code 008 004 000 BlackMIN Data Pixels Figure 10: Finding WhiteMAX gain too high Target Code Data Pixels gain correct Figure 8: Finding BlackMIN gain too low • Scan one line of a pure black test pattern, or scan one line with the light source turned off. (Caution: turning the light source on and off may cause thermal drift in light intensity as the light source warms up and cools down.) Scanning multiple black lines and averaging each pixel from each line will reduce noise and increase the accuracy of this step. • Find the minimum pixel value of the scanned black image (BlackMIN) for each color (Figure 8). • Reduce the Offset DAC value for each color so that the minimum black pixel is close to 0, but still greater than 0 (Figure 9). One offset DAC code = ~4.2 ADC LSBs, so change the DAC code to (31 - BlackMIN/4.2). 018 014 ADC 010 Output 00C Code 008 004 000 ADC Output Code Data Pixels Figure 11: CCD Input Signal In Range gain equation (Equation 1), the PGA code required is: Target Code 32 PGA code = Integer --------------------------------- – 0.94 --------------- White 2.137 MAX Equation 13: PGA Code Calculation If the Blue channel requires a gain greater than 3 (Target Code/WhiteMAX is >3), then set register 2, bit 7=1 and use Equation 14: BlackMIN = 13: too high 32 Target Code PGA code = Integer --------------------------------- – 2.82 --------------- 6.411 White MAX Equation 14: Blue PGA Code Calculation, x3 Bit Set • Program the calculated PGA values for each color into configuration registers 3, 4, and 5. • Because the above equations are based on typical LM9812 performance, the new PGA settings should be verified. Scan another white line(s) with the new PGA values and recalculate WhiteMAX and WhiteMIN (the minimum ADC output code for the white image) for each color. Verify that WhiteMAX is near but still below the Target Code. If WhiteMAX is greater than or equal to the target code, decrement that PGA value by 1. • BlackMIN = 6: correct Data Pixels BlackMIN = 0: too low Figure 9: Setting the Static Offset DACs • Because the above equation is based on typical LM9812 performance, the new DAC settings should be verified. Scan another black line(s) with the new DAC values and verify that all the pixels are still above 0. If any pixels are equal to 0, increment the DAC value by 1. If the minimum pixel is above 8, consider decrementing the DAC value by 1. Repeat this step until the minimum pixel value of the scanned black image (BlackMIN) for each color is greater than 0 but less than 9. BlackMIN values greater than 9 are acceptable (they will be removed by the Pixel Rate Offset Subtractor), but will reduce the dynamic range of the system. 4.4 Determining the Multiplier Correction Range Setting • For best performance, repeat Steps 4.2 and 4.3 before proceeding. This will maximize the available dynamic range of the ADC and ensure that all the pixels are in range. • Digitize another white line (or multiple lines) and calculate the minimum and maximum pixels from that line (WhiteMIN and WhiteMAX). Verify that WhiteMAX is less than the Target Code. 4.3 Setting the Static PGAs • Decide on the Target Code for your scanner. This is the maximum output code from the ADC for a pure white image. This code may be as large as 1023, but this value could cause clipping if the light source drifts and becomes brighter after calibration. A value of 1000 allows for moderate light source drift between calibration and image scan. • Scan one line of a pure white test pattern. Scanning multiple white lines and averaging each pixel from each line will reduce noise and increase the accuracy of this step. • Find the maximum white pixel value of the image data (WhiteMAX) for each color (Figure 10). • Program the PGA value for each color so that the maximum white pixel is close to, but still less than the Target Code (Figure 11). The PGA gain required to meet this condition is equal to the Target Code divided by WhiteMAX. Using the PGA Target Code WhiteMAX WhiteMIN ADC Output Code Data Pixels Figure 12: Determining Correction Range Setting • The correction range required (Range) = WhiteMAX/WhiteMIN • If Range < 1.5, then set the Multiplier Range bits (register 9, bits 3 and 2) to 10 (33%). 34 http://www.national.com • If 1.5 < Range < 2.0, then set the Multiplier Range bits (register 9, bits 3 and 2) to 01 (50%). • If 2.0 < Range < 3.0, then set the Multiplier Range bits (register 9, bits 3 and 2) to 00 (66%). • If Range > 3.0, then set the range is too large and the system can not be calibrated. channel of the LM9812. The constant C (typically 2, given in the Electrical Characteristics table) represents the total gain error from Vin through DOUT with G = 1V/V (PGA register setting = 1). To further simplify the model, the signal is always assumed to be 0V for black and increasing in the positive direction as pixel brightness increases. In reality, black may be offset from ground, and white may be negative (CCD) or positive (CIS) with respect to black, but this is all taken care of by the CDS function and the correct setting of the Signal Polarity bit (Register 24, bit 7). 4.5 Calculating the Pixel Rate Offset Coefficients • Scan one line of a pure black test pattern, or scan one line with the light source turned off. (Caution: turning the light source on and off may cause thermal drift in light intensity as the light source warms up and cools down.) Scanning multiple black lines and averaging each pixel from each line will reduce noise and increase the accuracy of this step. • The value of each pixel is the offset error for that pixel. Store each pixel value into that pixel's offset correction RAM location. • Set the Pixel Rate Offset Adder Source to external (register 9, bit 0 = 0) VREF IN C VIN + Σ + VOS1 10 Bit ADC + Σ +Σ + + VOS2 VOD G DOUT Offset DAC Figure 13: Simplified Model of One Channel of the Analog Front End The analog front end’s transfer function is shown in Equation 18. 4.6 Calculating the Pixel Rate Multiplier Coefficients • Scan one line of a pure white test pattern. Scanning multiple white lines and averaging each pixel from each line will reduce noise and increase the accuracy of this step. • The equation for the multiplier gain is: D (V + V )G + V +V OS 2 OD IN OS 1 = 1024 C ---------------------------------------------------------------------------------------V REFIN Equation 18: AFE Transfer Equation OUT The typical values of C (2) and VREFIN (2V) cancel each other out. It is also useful to convert all the voltages to ADC codes, since that is the domain in which they will be seen by the calibration software. VIN (in ADC codes) = VIN (in V)*1024. Calibration can be more easily accomplished using the simplified model of Equation 19. x Multiplier Gain = 1 + ---N Equation 15: Multiplier Gain Equation where: x is the 10 bit multiplier gain coefficient (0 to 1023), N = 2048 (Multiplier range = 1:1.5), N = 1024 (Multiplier range = 1:2), N = 512 (Multiplier range = 1:3). • Determine the multiplier gain for each pixel n so the result is the Target Code: D OUT = (V IN +V OS 1 )G + V OS 2 +V OD Equation 19: AFE Transfer Equation (units of ADC LSBs) Measurement of the two offset voltages (VOS1 and VOS2) is done by measuring the ADC output, with a black image on the sensor (if a black reference image is not available, it may be created simply by turning off the scanner’s illumination). In this example, VIN is considered to be 0V, and any offsets from the sensor are lumped in with the LM9812’s VOS1. If the ADC output is measured with a PGA gain of 1V/V and 3V/V, then the offset errors (VOS1 and VOS2) can be determined from the following two equations: Target Code Gain(n) = --------------------------------pixel datan Equation 16: Gain Calculation for Each Pixel • The multiplier coefficient required is therefore: Multiplier Coefficent (n) = Integer(N(Gain(n) - 1)) Target Code = Integer N --------------------------------- – 1 pixel datan Equation 17: Multiplier Coefficient Calculation for Each Pixel VADC1 = 1VOS1 + VOS2 + VDAC2 (PGA gain = 1V/V) VADC2 = 3VOS1 + VOS2 + VDAC2 (PGA gain = 3V/V) • Store the Multiplier Gain Coefficients for each pixel into that pixel's offset correction RAM location. • Set the Pixel Rate Multiplier Source to external (register 9, bit 1=0) Solving for VOS1 and VOS2: VOS1 = (VADC2 - VADC1)/2 VOS2 = (3VADC1 - VADC2)/2 CALIBRATION COMPLETE 5.0 POWER SUPPLY CONSIDERATIONS 4.7 Additional Calibration Information 5.1 General The procedure detailed in sections 4.1 through 4.6 is only one of many approaches to scanner calibration. In some approaches it may be useful to use a model for the analog front end and actually measure VOS1 and VOS2. The LM9812 should be powered by a single +5V source (unless 3V-compatible digital I/O is required-see Section 5.2). The analog supplies (VA) and the digital supplies (VD and VDI/O) are brought out individually to allow separate bypassing for each supply input. They should not be powered by two or more different supplies. Figure 13 is a mathematical model of the analog front end of one 35 http://www.national.com STATED. If the voltage on the input pin is at the threshold, large amounts of current can be drawn from the digital I/O supply because the N and the P channel of the input buffer are simultaneously on. These latches are very weak (sourcing and sinking about 50µA typically) and can be easily overdriven. In systems with separate analog and digital +5V supplies, all the supply pins of the LM9812 should be powered by the analog +5V supply. Each supply input should be bypassed to its respective ground with a 0.1µF capacitor located as close as possible to the supply input pin. A single 10µF tantalum capacitor should be placed near the VA supply pin to provide low frequency bypassing. From Multiplier To minimize noise, keep the LM9812 and all analog components as far as possible from noise generators, such as switching power supplies and high frequency digital busses. If possible, isolate all the analog components and signals (OS, reference inputs and outputs, VA, AGND) on an analog ground plane, separate from the digital ground plane. The two ground planes should be tied together at a single point, preferably the point where the power supply enters the PCB. 10 D0-D9 From Config Register To Config Register 5.2 3V Compatible Digital I/O If 3V digital I/O operation is desired, the VDI/O pin may be powered by a separate 3V±10% or 3.3V±10% supply. In this case all the digital I/O pins (CD0-CD9, D0-D9, MCLK, SYNC, RUN/STOP, CS, RD, WR, EOC, GLCK, OCLK, and RD PIXELI) will be 3V compatible. The CCD clock signals (Ø1, Ø2, RS, TR1 and TR2) remain 5V outputs, powered by VD. In this case, the VDI/O input should be bypassed to DGNDI/O with a parallel combination of a 0.1µF capacitor and a 10µF tantalum capacitor. Figure 14: Weak Latches On Databus 6.3 Don’t Decrement the Reference Sample Position Register by 1 A write to the Reference Sample Position Register (Configuration Register 22) containing a value that is 1 less than the current value stored in the Reference Sample Position Register will sometimes make one of the LM9812’s internal state machines to fail, causing erratic and improper operation of the LM9812. For example, if the value currently stored in the Reference Sample Position Register is 12, then writing an 11 can cause this problem. 5.3 Power Down Mode Setting the Power Down bit to a “1” puts the device in a low power standby mode. The CCD outputs (Ø1, Ø2, RS, TR1 and TR2) are pulled low and the analog sections are turned off to conserve power. The digital logic will continue to operate if MCLK continues and SYNC is held high, so for minimum power dissipation MCLK should be stopped when the LM9812 enters the Power Down mode. Recovery from Power Down typically takes 50µs (the time required for the reference voltages to settle to 0.5 LSB accuracy). To avoid this condition, never decrement the existing value stored in the Reference Sample Position Register by 1. Incrementing by any amount or decrementing by 2 or greater or writing the same number over again will not cause any problems. 6.0 RULES, HINTS, AND COMMON DESIGN PROBLEMS To get out of this mode, write a new number to the Reference Sample Position Register that is greater than the original number. 6.1 Ignore MCLK When Designing a System While the MCLK input is the master clock for all the LM9812 timing, it should not be used to predict when other clocks or events will occur. To get the highest possible timing resolution, the LM9812 uses double-edged flip flops for many functions. CCD clock output signals, coefficient clocks, and internal clocks may change state on the rising or falling edge of MCLK, depending on the state of MCLK and some internal free running state machines when the RUN/STOP or SYNC input goes high. Normal process variations from part to part result in different delays between an MCLK edge and other events in the LM9812. Since incrementing and decrementing this register should only be done during the development phase of the system until an ideal value for this register is chosen. This issue should cause absolutely no problems in production, since the value of this register will be fixed, so the firmware should never write more than the final ideal value to this register. 6.4 Taking RD Low With CS High Will Drive the Databus In many digital systems, the RD signal is ignored when CS is held high. This is not the case with the The LM9812. Taking RD low with CS high will take the D0-D9 databus out of tristate and put random data on the databus. Do not assume that RD is ignored when CS is high. For a reliable design, synchronize your system to EOC, GCLK, and OCLK. These signals are repeatable from device to device. Do not synchronize your system to MCLK: treat it only as a frequency that sets the pixel rate of the system. If you need to generate additional CCD timing signals, synchronize them to the LM9812’s TR and RS outputs. 6.2 Weak Latches On Databus The D0-D9 databus has weak latches on its output pins to keep the databus from drifting through the input’s trip point when TRI- 36 http://www.national.com Physical Dimensions inches (millimeters) 52-Pin Plastic Leaded Chip Carrier (PLCC) Order Number LM9812CCV NS Package Number V52A LIFE SUPPORT POLICY NATIONAL'S PRODUCTS ARE NOT AUTHORIZED FOR USE AS CRITICAL COMPONENTS IN LIFE SUPPORT DEVICES OR SYSTEMS WITHOUT THE EXPRESS WRITTEN APPROVAL OF THE PRESIDENT OF NATIONAL SEMICONDUCTOR CORPORATION. As used herein: 1. Life support devices or systems are devices or systems which, (a) are intended for surgical implant into the body, or (b) support or sustain life, and whose failure to perform, when properly used in accordance with instructions for use provided in the labeling, can be reasonably expected to result in a significant injury to the user. N 2. A critical component is any component of a life support device or system whose failure to perform can be reasonably expected to cause the failure of the life support device or system, or to affect its safety or effectiveness. National Semiconductor National Semiconductor Europe Corporation Fax: a49 (0) 180-530 85 86 National Semiconductor Hong Kong National Semiconductor Ltd. Japan Ltd. 1111 West Bardin Road Email: europe.support @ nsc.com 13th Floor, Straight Block, Tel: 81-043-299-2308 Arlington, TX 76017 Deutsch Tel: + 49 (0) 180-530 85 85 Ocean Centre, 5 Canton Rd. Fax: 81-043-299-2408 Tel: 1(800) 272-9959 English Tel: + 49 (0) 180-532 78 32 Tsimshatsui, Kowloon Fax: 1(800) 737-7018 Francais Tel: + 49 (0) 180-532 93 58 Hong Kong Italiano Tel: + 49 (0) 180-534 16 80 Tel: (852) 2737-1600 http://www.national.com Fax: (852) 2736-9960 37 http://www.national.com National does not assume any responsibility for use of any circuitry described, no circuit patent licenses are implied and National reserves the right at any time without notice to change said circuitry and specifications.