LM9801 8-Bit Greyscale/24-Bit Color Linear CCD Sensor Processor General Description Features The LM9801 is a high performance integrated signal processor/digitizer for linear CCD image scanners. The LM9801 performs all the analog processing (correlated double sampling for black level and offset compensation, pixel-by-pixel gain (shading) correction, and 8-bit analog-to-digital conversion) necessary to maximize the performance of a wide range of linear CCD sensors. The LM9801 can be digitally programmed to work with a wide variety of CCDs from different manufacturers. An internal configuration register sets CCD and sampling timing to maximize performance, simplifying the design and manufacturing processes. The LM9801 can be used with parallel output color CCDs. A signal inversion mode eases use with CIS sensors. For complementary voltage reference see the LM4041. Y Applications Y Y Y Y Y Y Y Y Y Y Y Y Key Specifications Y Y Color and Greyscale Flatbed and Sheetfed Scanners Fax and Multifunction Peripherals Digital Copiers General Purpose Linear CCD Imaging 2.5 Million pixels/s conversion rate Pixel-rate shading correction for individual pixels maximizes dynamic range and resolution, even on ‘‘weak’’ pixels Implements Correlated Double Sampling for minimum noise and offset error Reference and signal sampling points digitally controlled in 25 ns increments for maximum performance Generates all necessary CCD clock signals Compatible with a wide range of linear CCDs Supports some Contact Image Sensors (CIS) TTL/CMOS input/output compatible Y Resolution Pixel Conversion Rate Supply Voltage Supply Voltage (Digital I/O) 8 Bits 2.5 MHz a 5V g 5% a 3.3V g 10% or a 5V g 5% Y Power Dissipation 230 mW (max) Connection Diagrams TL/H/12814 – 1 TL/H/12814 – 2 Ordering Information Commercial (0§ C s TA s a 70§ C) Package LM9801CCV V52A 52-Pin Plastic Leaded Chip Carrier LM9801CCVF VEG52A 52-Pin Thin Quad Flatpack TRI-STATEÉ is a registered trademark of National Semiconductor Corporation. MICROWIRETM is a trademark of National Semiconductor Corporation. SPITM is a trademark of Motorola, Inc. C1996 National Semiconductor Corporation TL/H/12814 RRD-B30M96/Printed in U. S. A. http://www.national.com LM9801 8-Bit Greyscale/24-Bit Color Linear CCD Sensor Processor June 1996 Block Diagram TL/H/12814 – 3 http://www.national.com 2 Absolute Maximum Ratings (Notes 1 and 2) Positive Supply Voltage (V a e VA e VD e VD(I/O)) with Respect to GND e AGND e DGND e DGND(I/O) 6.5V Voltage on any Input or Output Pin 0.3V to V a a 0.3V g 25 mA Input Current at any Pin (Note 3) g 50 mA Package Input Current (Note 3) Package Dissipation at TA e 25§ C (Note 4) ESD Susceptibility (Note 5) Human Body Model 2000V Soldering Information Infrared, 10 seconds (Note 6) 300§ C b 65§ C to a 150§ C Storage Temperature Operating Ratings (Notes 1 and 2) Operating Temperature Range LM9801CCV, LM9801CCVF TMIN s TA s TMAX 0§ C s TA s a 70§ C a 4.75V to a 5.25V VA Supply Voltage a 4.75V to a 5.25V VD Supply Voltage a 2.7V to a 5.25V VD(I/O) Supply Voltage s 100 mV lVA –VDl t b 100 mV VA –VD(I/O) b 0.05V to VA a 0.05V OS, REF IN Voltage Range CD0 – CD7, MCLK, SYNC, SDI, SCLK, b 0.05V to VD(I/O) a 0.05V CS, RD Voltage Range Electrical Characteristics The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e a 5.0VDC, VD(I/O) e a 5.0 or a 3.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, RS e 25X. All LSB units are ADC LSBs unless otherwise specified. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7, 8 and 11) Symbol Parameter Typical (Note 9) Conditions Limits (Note 10) Units (Limits) 1.1 0.4 V (min) V (min) CCD SOURCE REQUIREMENTS FOR FULL SPECIFIED ACCURACY AND DYNAMIC RANGE (Note 11) VWHITE Maximum Peak CCD Differential Signal Range VRFT Maximum CCD Reset FeedThrough Amplitude VGA Gain e 0 dB VGA Gain e 9 dB 2 V (min) ADC CHARACTERISTICS Resolution with No Missing Codes 8 Bits (min) ILE Integral Linearity Error (Note 12) g 1.5 LSB (max) DNL Differential Non-Linearity g 1.0 LSB (max) 8 Bits (min) 2.8 V/V (min) 1.4 % (max) 4 Bits (min) 8.5 dB (min) g 0.15 dB (max) PGA CHARACTERISTICS Monotonicity PGA Adjustment Range GainPGA e 255 GainPGA e 0 2.95 Gain Error at any Gain (Note 14) VGA CHARACTERISTICS Monotonicity VGA Adjustment Range 20log # GainVGA e 15 GainVGA e 0 J 8.95 Gain Error at any Gain (Note 15) OFFSET TRIM CHARACTERISTICS Offset DAC LSB Size In Units of ADC LSBs Offset DAC DNL In Units of Offset DAC LSBs Offset Add Magnitude In Units of ADC LSBs 3 0.42 LSB g 0.25 g 0.9 LSB (max) 2.0 1.6 2.5 LSB (min) LSB (max) http://www.national.com Electrical Characteristics (Continued) The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e a 5.0VDC, VD(I/O) e a 5.0 or a 3.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, RS e 25X. All LSB units are ADC LSBs unless otherwise specified. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7, 8 and 11) Symbol Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) g 0.6 g 3.0 % (max) SYSTEM CHARACTERISTICS VOS1 VOS2 Full Channel Gain Error VGA Gain e 1, PGA Gain e 1 Pre-PGA Offset Error VGA Gain e 1, Offset DAC e 0 g1 LSB Post-PGA Offset Error Offset Add e 0 g1 LSB REFERENCE AND ANALOG INPUT CHARACTERSTICS OS Input Capacitance 5 OS Input Leakage Current Measured with OS e 2.45VDC RREF ADC Reference Ladder (REF OUTHI to REF IN) Impedance REF IN Reference Voltage (Note 13) pF 2 20 nA (max) 950 500 1400 X (min) X (max) 1.225 1.19 1.26 V (min) V (max) DC and Logic Electrical Characteristics The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e a 5.0VDC, VD(I/O) e a 5.0 or a 3.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, Rs e 25X. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7 and 8) Symbol Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) CD0 – CD7, MCLK, SYNC, SDI, SCLK, CS, RD DIGITAL INPUT CHARACTERISTICS VIN(1) Logical ‘‘1’’ Input Voltage VD(I/O) e 5.25V VD(I/O) e 3.6V 2.0 2.0 V (min) V (min) VIN(0) Logical ‘‘0’’ Input Voltage VD(I/O) e 4.75V VD(I/O) e 2.7V 0.8 0.7 V (max) V (max) IIN Input Leakage Current VIN e VD VIN e DGND CIN Input Capacitance 0.1 b 0.1 mA mA 5 pF DD0 – DD7, EOC, CCLK, SDO DIGITAL OUTPUT CHARACTERISTICS VOUT(1) Logical ‘‘1’’ Output Voltage VD(I/O) VD(I/O) VD(I/O) VD(I/O) VOUT(0) Logical ‘‘0’’ Output Voltage VD(I/O) e 5.25V, IOUT e 1.6 mA VD(I/O) e 3.6V, IOUT e 1.6 mA TRI-STATEÉ Output Current (DD0 –DD7 only) VOUT e DGND VOUT e VD IOUT COUT e e e e 4.75V, IOUT e b360 mA 4.75V, IOUT e b10 mA 2.7V, IOUT e b360 mA 2.7V, IOUT e b10 mA TRI-STATE Output Capacitance 2.4 4.4 2.1 2.5 V (min) V (min) V (min) V (min) 0.4 0.4 V (max) V (max) 0.1 b 0.1 mA mA 5 pF w1, w2, RS, TR DIGITAL OUTPUT CHARACTERISTICS VOUT(1) Logical ‘‘1’’ Output Voltage VD e 4.75V, IOUT e b360 mA VD e 4.75V, IOUT e b10 mA 2.4 4.4 V (min) V (min) VOUT(0) Logical ‘‘0’’ Output Voltage VD e 5.25V, IOUT e 1.6 mA 0.4 V (max) http://www.national.com 4 DC and Logic Electrical Characteristics (Continued) The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e a 5.0VDC, VD(I/O) e a 5.0 or a 3.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, Rs e 25X. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7 and 8) Symbol Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) POWER SUPPLY CHARACTERISTICS IA Analog Supply Current Operating Standby 25 50 32 mA (max) mA ID Digital Supply Current Operating MCLK e 0 6 65 8 mA (max) mA ID(I/O) Digital I/O Supply Current Operating, VD(I/O) e 5.0V Operating, VD(I/O) e 3.0V MCLK e 0, VD(I/O) e 5.0V or 3.0V 3.1 1.6 1.7 6 4 mA (max) mA (max) mA AC Electrical Characteristics, MCLK Independent The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e VD(I/O) e a 5.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, tMCLK e 1/fMCLK, tr e tf e 5 ns, Rs e 25X, CL (databus loading) e 50 pF/pin. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7 and 8) Symbol fMCLK Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) 20 1 MHz (min) MHz (max) 30 70 40 60 % (min) % (max) ns (min) Maximum MCLK Frequency Minimum MCLK Frequency MCLK Duty Cycle tA SYNC Setup of MCLK 5 10 tCDSETUP Correction Data Valid to CLK Setup 14 20 ns (min) tCDHOLD Correction Data Valid to CLK Hold b 12 0 ns (min) tD1H, tD0H RD High to DD0–DD7 TRI-STATE 5 15 ns (max) tDACC Access Time Delay from RD Low to DD0–DD7 Data Valid 15 30 ns (max) fSCLK Maximum SCLK Frequency 20 MHz (min) SCLK Duty Cycle 40 60 % (min) % (max) tSDI SDI Set-Up Time from SCLK Rising Edge 3 10 ns (min) tHDI SDI Hold Time from SCLK Rising Edge 2 15 ns (min) tDDO Delay from SCLK Falling Edge to SDO Data Valid 25 50 ns (max) tHDO SDO Hold Time from SCLK Falling Edge 30 50 5 ns (max) ns (min) tDELAY DELAY from SCLK Falling Edge to CS Rising or Falling Edge 5 10 ns (min) tSETUP Set-Up Time of CS Rising or Falling Edge to SCLK Rising Edge 0 10 ns (min) RL e 3k, CL e 50 pF 5 http://www.national.com AC Electrical Characteristics, MCLK Independent (Continued) The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e VD(I/O) e a 5.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, tMCLK e 1/fMCLK, tr e tf e 5 ns, Rs e 25X, CL (databus loading) e 50 pF/pin. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7 and 8) Typical (Note 9) Limits (Note 10) Units (Limits) 25 50 ns (max) Symbol Parameter Conditions tS1H, tS0H Delay from CS Rising Edge to SDO TRI-STATE RL e 3k, CL e 50 pF tRDO SDO Rise Time, TRI-STATE to High SDO Rise Time, Low to High RL e 3k, CL e 50 pF 20 20 ns ns tFDO SDO Fall Time, TRI-STATE to Low SDO Fall Time, High to Low RL e 3k, CL e 50 pF 20 20 ns ns AC Electrical Characteristics, MCLK Dependent The following specifications apply for AGND e DGND e DGND(I/O) e 0V, VA e VD e VD(I/O) e a 5.0VDC, REF IN e a 1.225VDC, fMCLK e 20 MHz, tMCLK e 1/fMCLK, tr e tf e 5 ns, Rs e 25X, CL (databus loading) e 50 pF/pin. Refer to Table 2, Configuration Register Parameters, for limits labelled C.R. Boldface limits apply for TA e TJ e TMIN to TMAX; all other limits TA e TJ e 25§ C. (Notes 7 and 8) Symbol Parameter Conditions Typical (Note 9) Limits (Note 10) Units (Limits) 50 ns 1 tMCLK 400 ns 800 ns 8 16 tMCLK tMCLK tSTART MCLK to first w1 High tw w1, w2 Clock Period tTRWIDTH Transfer Pulse (TR) Width C.R. ms tGUARD w1 to TR, TR to w1 Guardband C.R. ns tRSWIDTH Reset Pulse (RS) Width C.R. ns tRS Falling Edge of w1 to RS Either Edge of w1 to RS Standard CCD Mode Even/Odd CCD Mode C.R. ns tS/HREF Falling Edge of w1 to Ref. Sample Either Edge of w1 to Ref. Sample Standard CCD Mode Even/Odd CCD Mode C.R. ns tS/HSIG Falling Edge of w1 to Sig. Sample Either Edge of w1 to Sig. Sample Standard CCD Mode Even/Odd CCD Mode C.R. ns tS/HWIDTH Sample Pulse Width (Acquisition Time) 50 ns 1 tMCLK tSYNCLOW SYNC Low Between Lines 100 ns 2 tMCLK (min) tB SYNC Setup of w1 to End Line 2 tMCLK (max) tCCLKWIDTH CCLK Pulse Width 250 ns 5 tMCLK tDATAVALID Data Valid Time from EOC Low 300 ns (min) tEOCWIDTH EOC Pulse Width 250 ns 5 tMCLK 2.5 MHz 1.25 MHz fMCLK/8 fMCLK/16 Hz Hz 50 % w1 and w2 Frequency Standard CCD Mode Even/Odd CCD Mode Standard CCD Mode Even/Odd CCD Mode w1 and w2 Duty Cycle http://www.national.com 6 Electrical Characteristics (Continued) 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 lest conditions. Note 2: All voltages are measured with respect to GND e AGND e DGND e DGND(I/O) e 0V, unless otherwise specified. Note 3: When the input voltage (VIN) at any pin exceeds the power supplies (VIN k GND or VIN l VA or VD), the current at that pin should be limited to 25 mA. The 50 mA maximum package input current rating limits the number of pins that can simultaneously safely exceed the power supplies with an input current of 25 mA to two. Note 4: The maximum power dissipation must be derated at elevated temperatures and is dictated by TJmax, iJA and the ambient temperature, TA. The maximum allowable power dissipation at any temperature is PD e (TJmax –TA)/iJA. TJmax e 150§ C for this device. The typical thermal resistance (iJA) of this part when board mounted is 52§ C/W for the V52A PLCC package, and 70§ C/W for the VEG52A TQFP package. Note 5: Human body model, 100 pF capacitor discharged through a 1.5 kX resistor. Note 6: See AN450 ‘‘Surface Mounting Methods and Their Effect on Product Reliability’’ or the section titled ‘‘Surface Mount’’ found in any National Semiconductor Linear Data Book for other methods of soldering surface mount devices. Note 7: A Zener diode clamps the OS analog input to AGND as shown below. This input protection, in combination with the external clamp capacitor and the output impedance of the CCD, prevents damage to the LM9801 from transients during power-up. TL/H/12814 – 4 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 e TA e 25§ C, fMCLK e 20 MHz, and represent most likely parametric norm. Note 10: Tested limits are guaranteed to National’s AOQL (Average Outgoing Quality Level). Note 11: For CCDs, VBLACK 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, VBLACK - VRFT is defined as the peak positive deviation above VBLACK of the reset feedthrough pulse. For CIS, VWHITE is defined as the peak CCD pixel output voltage for a white (full scale) image with respect to GND (0V). 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 LM9801 can correct for using its internal PGA. CCD Output Signal CIS Output Signal TL/H/12814 – 6 TL/H/12814 – 5 Note 12: 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 13: Reference voltages below 1.19V may decrease SNR. Reference voltages above 1.26V may cause clipping errors inside the LM9801. The LM4041EIM3-1.2 (SOT-23 package) or the LM4041EIZ-1.2 (TO-92 package) bandgap voltage references are recommended for this application. Note 14: PGA Gain Error is the maximum difference between the measured gain for any PGA code and the ideal gain calculated by using the formula V PGA code 256 e1aC GainPGA where C e (PGA RANGE b 1) and PGA RANGE e the PGA adjustment range (in V/V) of the LM9801 under test. V 256 255 # J Note 15: VGA Gain Error is the maximum difference between the measured gain for any VGA code and the ideal gain calculated by using the formula VGA code 16 GainVGA(dB) e C where C e (VGA RANGE) and VGA RANGE e the VGA adjustment range (in dB) of the LM9801 under test. 16 15 Typical Performance Characteristics w1, w2, RS, and TR Rise and Fall Times Through a Series Resistance vs Load Capacitance TL/H/12814 – 8 7 http://www.national.com Pin Descriptions CCD Driver Signals w1 Digital Output. CCD clock signal, phase 1. w2 Digital Output. CCD clock signal, phase 2. RS Digital Output. Reset pulse for the CCD. TR Digital Output. Transfer pulse for the CCD. OS Analog Input. This is the OS (Output Signal) from the CCD. The maximum peak signal that can be accurately digitized is equal to the voltage at REF IN, typically 1.225V. Digital Coefficient I/O CD0 (LSB) – Digital Inputs. Correction Coefficient CD7 (MSB) Databus. This is the 8-bit data path for the gain adjust PGA, used during line scan. CCLK Analog I/O REF IN REF OUTHI Digital Output I/O DD0 (LSB) – Digital Outputs. Pixel Output Databus. This DD7 (MSB) data bus outputs the 8-bit digital output data during line scan. Analog Inputs. These two pins are the system reference voltage inputs and should be tied together to a 1.225V voltage source and bypassed to AGND with a 0.1 mF monolithic capacitor. Analog Output. This reference voltage is developed internally by the LM9801, and is equal to 3 times REF IN. It should be bypassed to AGND with a 0.1 mF monolithic capacitor. EOC Digital Output. This is the End of Conversion signal from the ADC indicating that new pixel data is available. RD Digital Input. Taking this input low places the data stored in the output latch on the bus. When this input is high the DD0 – DD7 bus is in TRI-STATE. VA This is the positive supply pin for the analog supply. It should be connected to a voltage source of a 5V and bypassed to AGND with a 0.1 mF monolithic capacitor in parallel with a 10 mF tantalum capacitor. AGND This is the ground return for the analog supply. VD This is the positive supply pin for the digital supply. It should be connected to a voltage source of a 5V and bypassed to DGND with a 0.1 mF monolithic capacitor. DGND This is the ground return for the digital supply. VD(I/O) This is the positive supply pin for the digital supply for the LM9801’s I/O. It should be connected to a voltage source of a 3V to a 5V and bypassed to DGND(I/O) with a 0.1 mF monolithic capacitor. If the supply for this pin is different than the supply for VA and VD, it should also be bypassed with a 10 mF tantalum capacitor. DGND(I/O) This is the ground return for the digital supply for the LM9801’s I/O. Analog Power REF OUTMID Analog Output. This reference voltage is developed internally by the LM9801, and is equal to 2 times REF IN. It should be bypassed to AGND using a 0.1 mF monolithic capacitor. VTEST1, VTEST2 Analog Inputs/Outputs. These pins are used for testing the device during manufacture and should be left unconnected. Digital Power General Digital I/O MCLK Digital Input. This is the 20 MHz (typical) master system clock. SYNC Digital Input. A low-to-high transition on this input begins a line scan operation. The line scan operation terminates when this input is taken low. Configuration Register I/O SDI Digital Input. Serial Data Input pin. SDO Digital Output. Serial Data Output pin. SCLK Digital Input. This is the serial data clock, used to clock data in through SDI and out through SDO. SCLK is asynchronous to MCLK. Input data is latched and output data is changed on the rising edge of SCLK. CS NC NC Digital Input. This is the Chip Select signal for writing to the Configuration Register through the serial interface. This input must be low in order to communicate with the Configuration Register. This pin is used for serial I/O only–it has no effect on any other section of the chip. http://www.national.com Digital Output. This is the signal that is used to clock the Gain coefficients into the LM9801. Data is latched on the rising edge of CCLK. 8 All pins marked NC (no connect) should be left floating. Do not tie NC pins to ground., power supplies, or any other potential or signal. Timing Diagrams TL/H/12814 – 9 DIAGRAM 1. Line Scan Timing Overview TL/H/12814 – 10 DIAGRAM 2. Pixel Pipeline Timing Overview 9 http://www.national.com Timing Diagrams (Continued) TL/H/12814 – 11 DIAGRAM 3. Timing for Start of Line Scan TL/H/12814 – 12 DIAGRAM 4. Timing for End of Line/Start of Next Line TL/H/12814 – 13 DIAGRAM 5. TR Pulse Timing TL/H/12814 – 14 DIAGRAM 6. RS Pulse Polarity http://www.national.com 10 Timing Diagrams (Continued) TL/H/12814 – 15 Note: Clamp signal only active during optical black pixels at beginning of line. DIAGRAM 7. CCD Timing TL/H/12814 – 16 Note: Clamp signal only active during optical black pixels at beginning of line. DIAGRAM 8. CCD Timing (Even/Odd CCDs) 11 http://www.national.com Timing Diagrams (Continued) TL/H/12814 – 17 Note: i e value programmed in Dummy Pixel Register - 1 (for example: Dummy Pixel Register e 17 j e value programmed in Optical Black Register. x i e 16 x 16 dummy pixels). DIAGRAM 9. Dummy Pixel and Optical Black Pixel Timing TL/H/12814 – 18 DIAGRAM 10. Coefficient Data Timing TL/H/12814 – 19 DIAGRAM 11. Output Data Timing http://www.national.com 12 Timing Diagrams (Continued) TL/H/12814 – 20 DIAGRAM 12. Data Timing (Output and Coefficient Data Sharing Same Bus) Serial Configuration Register Timing Diagrams TL/H/12814 – 21 DIAGRAM 13a. Configuration Register Write Timing using CS, Continuous SCLK (16-Bit Word) TL/H/12814 – 22 DIAGRAM 13b. Configuration Register Read Timing using CS, Continuous SCLK (16-Bit Word) 13 http://www.national.com Serial Configuration Register Timing Diagrams (Continued) TL/H/12814 – 23 DIAGRAM 14. SDO Timing TL/H/12814 – 24 DIAGRAM 15a. Configuration Register Write Timing with CS Continuously Low (16-Bit Word) TL/H/12814 – 25 DIAGRAM 15b. Configuration Register Write Timing with CS Continuously Low (Two 8-Bit Bytes) TL/H/12814 – 26 DIAGRAM 15c. Configuration Register Read Timing with CS Continuously Low (16-Bit Word) TL/H/12814 – 27 DIAGRAM 15d. Configuration Register Read Timing with CS Continuously Low (Two 8-Bit Bytes) http://www.national.com 14 TABLE I. Configuration Register Address Table A2 0 A1 0 A0 0 D7 D6 Standard Mode or Even/Odd Mode 0 0 0 1 1 RSW1 0 1 1 1 0 1 1 RSW0 RSPOL SR2 SR1 SR0 0 w2 Enable RS Enable TR Enable w1EN w2EN RSEN TREN 1 Signal Polarity D0 RS Pulse Position RSPOS3 RSPOS2 RSPOS1 RSPOS0 SS3 SS2 SS1 SS0 TR – w1 Guardband TR Polarity TRGRD TRPOL BLS1 BLS0 BLL2 BLL1 BLL0 TR Pulse Width TRW1 TRW0 Dummy Pixels (Minimum Register Value is 2) BLS6 BLS5 BLS4 BLS3 BLS2 Optical Black Pixels (Minimum Register Value is 1) 0 BLL6 BLL5 BLL4 BLL3 PGA Gain Coefficient 1 1 D1 Sample Signal Position w1 Enable 0 D2 RS Pulse Polarity SR3 BLL7 1 D3 Sample Reference Position (Maximum Register Value is 14) SIGPOL 1 D4 RS Pulse Width MODE 0 D5 GAIN7 GAIN6 GAIN5 GAIN4 GAIN3 GAIN2 GAIN1 GAIN0 PGA Gain Source Offset DAC Sign Powerdown Offset Add VGA Gain MSB VGA Gain VGA Gain VGA Gain LSB PGASRC ODSIGN PD OFFADD VGA3 VGA2 VGA1 VGA0 Offset DAC MSB Offset DAC Offset DAC Offset DAC LSB VOS3 VOS2 VOS1 VOS0 Test Modes 0 0 0 0 TABLE II. Configuration Register Parameters Parameter Control Bits Result MODE MODE RS Pulse Width (tRSWIDTH) 0 1 Standard CCD (w frequency e fMCLK/8) Even/Odd CCD (w frequency e fMCLK/16) RS1 RS0 0 0 1 1 0 1 0 1 1 tMCLK (50 ns) 2 tMCLK (100 ns) 3 tMCLK (150 ns) 4 tMCLK (200 ns) RSPOL RS Pulse Polarity 0 1 RS RS Note: tMCLK e 1/fMCLK e 1 MCLK period. Examples given in parenthesis are for fMCLK e 20 MHz (tMCLK e 50 ns). 15 http://www.national.com TABLE II. Configuration Register Parameters (Continued) Parameter RS Pulse Position (tRS) Sample Reference Position (tS/HREF) Sample Signal Position (tS/HSIG) Control Bits Result RSPOS3 RSPOS2 RSPOS1 RSPOS0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 SR3 SR2 SR1 SR0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 SS3 SS2 SS1 SS0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 Note: tMCLK e 1/fMCLK e 1 MCLK period. Examples given in parenthesis are for fMCLK e 20 MHz (tMCLK e 50 ns). http://www.national.com 16 0.0tMCLK (0 ns) 0.5tMCLK (25 ns) 1.0tMCLK (50 ns) 1.5tMCLK (75 ns) 2.0tMCLK (100 ns) 2.5tMCLK (125 ns) 3.0tMCLK (150 ns) 3.5tMCLK (175 ns) 4.0tMCLK (200 ns) 4.5tMCLK (225 ns) 5.0tMCLK (250 ns) 5.5tMCLK (275 ns) 6.0tMCLK (300 ns) 6.5tMCLK (325 ns) 7.0tMCLK (350 ns) 7.5tMCLK (375 ns) 0.0tMCLK (0 ns) 0.5tMCLK (25 ns) 1.0tMCLK (50 ns) 1.5tMCLK (75 ns) 2.0tMCLK (100 ns) 2.5tMCLK (125 ns) 3.0tMCLK (150 ns) 3.5tMCLK (175 ns) 4.0tMCLK (200 ns) 4.5tMCLK (225 ns) 5.0tMCLK (250 ns) 5.5tMCLK (275 ns) 6.0tMCLK (300 ns) 6.5tMCLK (325 ns) 7.0tMCLK (350 ns) Not Valid 0.0tMCLK (0 ns) 0.5tMCLK (25 ns) 1.0tMCLK (50 ns) 1.5tMCLK (75 ns) 2.0tMCLK (100 ns) 2.5tMCLK (125 ns) 3.0tMCLK (150 ns) 3.5tMCLK (175 ns) 4.0tMCLK (200 ns) 4.5tMCLK (225 ns) 5.0tMCLK (250 ns) 5.5tMCLK (275 ns) 6.0tMCLK (300 ns) 6.5tMCLK (325 ns) 7.0tMCLK (350 ns) 7.5tMCLK (375 ns) TABLE II. Configuration Register Parameters (Continued) Parameter Control Bits Result w1EN w1 Enable 0 1 w1 Output Off w1 Output On w2EN w2 Enable 0 1 w2 Output Off w2 Output On RSEN RS Enable 0 1 RS Output Off RS Output On TREN TR Enable 0 1 TRW1 TRW0 0 0 1 1 0 1 0 1 TR Pulse Width (tTRWIDTH) TR-w1 Guardband (tGUARD) TR Output Off TR Output On 20 tMCLK (1.0 ms) 30 tMCLK (1.5 ms) 40 tMCLK (2.0 ms) 50 tMCLK (2.5 ms) TRGRD 0 1 1 tMCLK (50 ns) 2 tMCLK (100 ns) TRPOL TR Polarity 0 1 TR TR SIGPOL Signal Polarity Dummy Pixels Note: Minimum Register Value is 2. Actual number of dummy pixels in CCD should be one less than number in this register. Optical Black Pixels Note: Minimum Register Value is 1. 0 1 Positive (CIS) Negative (CCD) BLS6 BLS5 BLS4 BLS3 BLS2 BLS1 BLS0 Dummy Pixels 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 1 Not Valid Not Valid 1 2 # # # # # # # # 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 0 1 124 125 126 BLL7 BLL6 BLL5 BLL4 BLL3 BLL2 BLL1 BLL0 Optical Black Pixels 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 1 Not Valid 1 2 3 # # # # # # # # # 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 0 1 253 254 255 Note: tMCLK e 1/fMCLK e 1 MCLK period. Examples given in parenthesis are for fMCLK e 20 MHz (tMCLK e 50 ns). 17 http://www.national.com TABLE II. Configuration Register Parameters (Continued) Parameter Internal PGA Gain Coefficient Control Bits Result GAIN7 GAIN6 GAIN5 GAIN4 GAIN3 GAIN2 GAIN1 GAIN0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 # # # # # # # # 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 0 1 PGA Gain Coefficient Source PGASRC Offset DAC Sign ODSIGN 0 1 dB [V/V] (typical) 0.00 0.07 0.13 1.000 1.008 1.015 # 9.35 9.37 9.40 2.935 2.942 2.950 Internal External 0 1 Negative Positive PD Power Down 0 1 Operating Powered Down OFF ADD Offset Add VGA Gain 0 1 Offset E 0 LSB Offset E a 2 LSB VGA3 VGA2 VGA1 VGA0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 Note: tMCLK e 1/fMCLK e 1 MCLK period. Examples given in parenthesis are for fMCLK e 20 MHz (tMCLK e 50 ns). http://www.national.com 18 dB [V/V] (typical) 0.00 0.60 1.20 1.79 2.39 2.99 3.59 4.19 4.79 5.38 5.98 6.58 7.18 7.78 8.38 8.97 1.00 1.07 1.15 1.23 1.32 1.41 1.51 1.62 1.74 1.86 1.99 2.13 2.29 2.45 2.62 2.81 TABLE II. Configuration Register Parameters (Continued) Parameter Offset DAC Control Bits ODSIGN VOS3 VOS2 VOS1 VOS0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 Result Offset (LSB) (typical) 0.00 b 0.42 b 0.84 b 1.26 b 1.68 b 2.10 b 2.52 b 2.94 b 3.36 b 3.78 b 4.20 b 4.62 b 5.04 b 5.46 b 5.88 b 6.30 0.00 a 0.42 a 0.84 a 1.26 a 1.68 a 2.10 a 2.52 a 2.94 a 3.36 a 3.78 a 4.20 a 4.62 a 5.04 a 5.46 a 5.88 a 6.30 Note: tMCLK e 1/fMCLK e 1 MCLK period. Examples given in parenthesis are for fMCLK e 20 MHz (tMCLK e 50 ns). Block Diagram of LM9801-Based System TL/H/12814 – 28 Note: Power supplies and bypass capacitors not shown for clarity. FIGURE 1. LM9801 System Block Diagram 19 http://www.national.com Applications Information An approximately 2 LSB (29 mV) offset can be added at the output of the PGA stage if necessary to ensure that the offset is greater than zero. This eliminates the possibility of a negative offset clipping the darkest output pixels. For more information on the Offset Add Bit, see Section 4.8. Finally, the output of the PGA is digitized by the ADC and made available on the DD0 – DD7 bus (Section 4.9). Three reference voltages are used throughout the signal path: the externally supplied REF IN (1.225V), and the internally generated REF OUTMID (2.45V) and REF OUTHI (3.675V). 1.0 THEORY OF OPERATION The LM9801 removes errors from and digitizes a linear CCD pixel stream, while providing all the necessary clock signals to drive the CCD. Offset and gain errors for individual pixels are removed at the pixel rate. Offset errors are removed through correlated double sampling (CDS). 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 8-bit programmable gain amplifier (PGA) in front of the ADC. 1.1 The Analog Signal Path (See Block Diagram) The analog output signal from the CCD is connected to the OS input of the LM9801 through a 0.01 mF (typical, see Section 4.2, Clamp Capacitor Selection ) DC blocking capacitor. During the CCD’s optical black pixel segment at the beginning of every line, this input is clamped to the REF OUTMID voltage (approximately 2.45V). This DC restore operation fixes the reference level of the CCD pixel stream at REF OUTMID. The signal is then buffered and fed to a digitally-programmed 4-bit VGA (variable gain amplifier). The gain of the VGA is digitally programmable in 16 steps from 1V/V to 3V/V. The VGA is used to compensate for peak white CCD outputs less than the 1.225V full-scale required by the LM9801 for maximum dynamic range. When used with parallel output CCDs, the VGA can fine-tune the amplitude of the red, green, and blue signals. For a detailed explanation of the VGA, see Section 4.3. The output of the VGA goes into the CDS (Correlated Double Sampling) stage, consisting of two sample/hold amplifiers: S/H Ref (Reference) and S/H Signal. The reference level of the signal is sampled and held by the S/H Ref circuit and the active pixel data is sampled and held by the S/ H Signal circuit. The output of S/H Ref is subtracted from the S/H Signal output and amplified by 2. The full-scale signal range at this point is approximately 2.45Vp-p. CDS reduces or eliminates many sources of noise, including reset noise, flicker noise, and both high and low frequency pixel-to-pixel offset variation. For more information on the CDS stage, see Section 4.4. At this point an offset voltage can be injected by the 5-bit (4 bits a sign) Offset DAC. This voltage is designed to compensate for any small fixed DC offset introduced by the CDS S/Hs and the x2 amplifier. The LSB size of the DAC is approximately 0.42 ADC LSBs (4 mV). The adjustment range is g 6.3 ADC LSBs. For a detailed explanation of the Offset DAC, see Section 4.6. The next stage is the PGA. This is a programmable gain amplifier that changes the gain at the pixel rate to correct for gain errors due to PRNU, uneven illumination (such as cos4 effect), RGB filter mismatch, etc. The gain adjustment range is 0 dB to 9 dB (1V/V to 3V/V) with 8 bits of resolution. The gain data (correction coefficients) is provided on the CD0 – CD7 bus. The gain may also be fixed at any value between 0 dB and 9 dB with the PGA Gain Coefficient configuration register. For additional information on the PGA, see Section 4.7. http://www.national.com 1.2 The CCD Clocking Signals To maximize the flexibility of the LM9801, the CCD’s w1, w2, RS, and TR pulses are internally generated, with a wide range of options, making these signals compatible with most commercial linear CCDs. In many cases, these output signals can drive the CCD clock inputs directly, with only series resistors (for slew rate control) between the LM9801’s outputs and the CCD clock inputs. 1.3 The Digital Interface There are three main sections to the digital interface of the LM9801: a serial interface to the Configuration Register, where all device programming is done, an 8 bit-wide input databus for gain correction coefficients with a synchronous clock output (CCLK), and an 8-bit output databus for the final pixel output data with a synchronous end of conversion output signal (EOC) and an output enable input (RD). Please note that the CS input affects only the serial I/O – it has no effect on the output databus, input coefficient bus, or any other section of the LM9801. 2.0 DIGITAL INTERFACE 2.1 Reading and Writing to the Configuration Register Communication with the Configuration Register is done through a standard MICROWIRETM serial interface. This interface is also compatible with the Motorola SPITM standard and is simple enough to easily be implemented in custom hardware if desired. The serial interface timing is shown in Diagrams 13a – 13b and Diagrams 15a – 15d. Data is sent serially, LSB first. (Please note that some microcontrollers output data MSB first. When using these microcontrollers the bits in the configuration register are effectively reversed.) Input data is latched on the rising edge of SCLK, and output data changes on the falling edge of SCLK. CS must be low to enable serial I/O. If SCLK is only clocked when sending or receiving data from the LM9801, and held low at all other times, then CS can be tied low permanently as shown in Diagrams 15a – 15d. If SCLK is continuous, then CS must be used to determine the beginning and the end of a serial byte or word (see Diagrams 13a – 13b). Note that CS must make its high-to-low and low-to-high transitions when SCLK is low, otherwise the internal bit counter may receive an erroneous pulse, causing an error in the write or read operation. Data may be transmitted and received in two 8-bit bytes (typical with microcontroller interfaces) or one 16-bit word (for custom serial controllers). 20 Applications Information (Continued) control slew rate and isolate the driver from the large load capacitances. The values of these resistors are usually given in the CCD’s datasheet. The Configuration Register is programmed by sending a control byte to the serial port. This byte indicates whether this is a read or a write operation, and gives the 3-bit address of the register bank to be read from or written to. If this is a read operation, the next 8 SCLKs will output the data at the requested location on the SDO pin. If this is a write operation, the data to be sent to the specified location should be clocked in on the SDI input during the next 8 SCLKs. Data is sent and received using the LSB (Least Significant Bit) first format. For maximum system reliability, each configuration register location can be read back and verified after a write. If the serial I/O to the configuration register falls out of sync for any reason, it can be reset by sending 8 or more SCLKs with CS held high. 4.0 ANALOG INTERFACE 4.1 Voltage Reference The two REF IN pins should be connected to a 1.225V g 2% reference voltage capable of sinking between 2 mA and 5 mA of current coming from the 500X – 1400X resistor string between REF OUTHI and REF IN. The LM4041-1.2 1.225V bandgap reference is recommended for this application as shown in Figure 2 . The inexpensive ‘‘E’’ grade meets all the requirements of the application and is available in a TO-92 (LM4041EIZ-1.2) package as well as a SOT-23 package (LM4041EIM3-1.2) to minimize board space. Due to the transient currents generated by the LM9801’s ADC, PGA, and CDS circuitry, the REF IN pins, the REF OUTMID pin and the REF OUTHI pin should all be bypassed to AGND with 0.1 mF monolithic capacitors. 2.2 Writing Correction Coefficient Data on the CD0 – CD7 Bus Correction coefficient data for each pixel is latched on the rising edge of the CCLK output signal (see Diagram 10). Note that there is a 3 pixel latency between when the coefficient data is latched and when the output data is available. As Diagram 2, Pixel Pipeline Timing Overview shows, coefficient data for pixel n is latched shortly before the output data for pixel n-2 becomes available on the output databus (DD0 – DD7). Note that there is no way to provide a correction coefficient for pixel 1, the first pixel in the CCD array. This is not a problem since the first several pixels of the CCD are used for clamping. 2.3 Reading Output Data on the DD0–DD7 Bus The corrected digital output data representing each pixel is available on the DD0–DD7 databus. The data is valid after the falling edge of the EOC output. The RD input takes the databus in and out of TRI-STATE. RD can be held low at all times if there are no other devices needing the bus, or it can be used to TRI-STATE the bus between pixels, allowing other devices access to the bus. Diagram 12, Data Timing (Output and Coefficient Data Sharing Same Bus) , shows how EOC can be tied to RD to automatically multiplex between coefficient data and conversion data. TL/H/12814-29 FIGURE 2. Voltage Reference Generation 4.2 Clamp Capacitor Selection This section is very long because it is relatively complicated to explain, but the answer is short and simple: A clamp capacitor value of 0.01 mF should work in almost all applications. The rest of this section describes exactly how this value is selected. 2.4 MCLK This is the master clock input that controls the LM9801. The pixel conversion rate is fixed at 1/8 of this frequency. Many of the timing parameters are also relative to the frequency of this clock. 2.5 SYNC This input signals the beginning of a line. When SYNC goes high, the LM9801 generates a TR pulse, then begins converting pixels until the SYNC line is brought low again. Since there is no pixel counter in the LM9801, it will work with CCDs of any length. 3.0 DIGITAL CCD INTERFACE TL/H/12814 – 30 FIGURE 3. OS Clamp Capacitor and Internal Clamp The output signal of many CCDs rides on a large DC offset (typically 8V to 10V) which is incompatible with the LM9801’s 5V operation. To eliminate this offset without resorting to additional higher voltage components, the output 3.1 Buffering w1, w2, RS, and TR The LM9801 can drive the w1, w2, RS, and TR inputs of many CCDs directly, without the need for external buffers between the LM9801 and the CCD. Most linear CCDs designed for scanner applications require 0V to 5V signal swings into 20 pF to 500 pF input loading. Series resistors are typically inserted between the driver and the CCD to 21 http://www.national.com Applications Information (Continued) Where n e 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 CCD has 18 black reference pixels, the output impedance of the CCD is 1500X, the LM9801 is configured to clamp for 300 ns, the worst case initial voltage across the capacitor is 10V, and the desired voltage after clamping is 0.1V (accuracy e 10/0.1 e 100), then: of the CCD is AC coupled to the LM9801 through a DC blocking capacitor, CCLAMP (the CCD’s DOS output is not used). The value of this capacitor is determined by the leakage current of the LM9801’s OS input and the output impedance of the CCD. The leakage through the OS input determines how quickly the capacitor value will drift from the clamp value of REF OUTMID, which then determines how many pixels can be processed before the droop causes errors in the conversion ( g 0.1V is the recommended limit). The output impedance of the CCD determines how quickly the capacitor can be charged to the clamp value during the black reference period at the beginning of every line. The minimum clamp capacitor value is determined by the maximum droop the LM9801 can tolerate while converting one CCD line. The following equation takes the maximum leakage current into the OS input, the maximum allowable droop (100 mV), the number of pixels on the CCD, and the pixel conversion rate (fMCLK/8) and provides the minimum clamp capacitor value: 18 300 ns 1500X In(100) e 514 pF The final value for CCLAMP should be less than or equal to CCLAMP MAX, but no less than CCLAMP MIN. A value of 470 pF will work in this example. In some cases, depending primarily on the choice of CCD, CCLAMP MAX may actually be less than the CCLAMP MIN, meaning that the capacitor cannot be charged to its final voltage during the black pixels at the beginning of a line and hold its voltage without drooping for the duration of that line. This is usually not a problem because in most applications the CCD 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: CCLAMP MAX e i dt dV leakage current (A) number of pixels e max droop (V) conversion rate (Hz) For example, if the OS input leakage current is 20 nA worstcase, the CCD has 2700 active pixels, the conversion rate is 2.5 MHz (fMCLK e 20 MHz), and the max droop desired is 0.1V, the minimum clamp capacitor value is: 20 nA 2700 CCLAMP MIN e 0.1V 2.5 MHz e 216 pF 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 CCD output. The internal clamp is on for each pixel from the rising edge of the S/H ref pulse to the falling edge of the S/H signal pulse (see Diagrams 7 and 8). This time can be calculated using the values stored in the Sample Signal and Sample Reference configuration registers and the MCLK frequency. For normal CCDs: CCLAMP MIN e tDARK(s) e J # J # J # J At a 2.5 MHz conversion rate, this is about 14 ms. In this example a 0.01 mF capacitor takes 14 ms after power-up to charge to its final value, but its droop across all subsequent lines is now less than 2 mV (using the previous example’s values). This wide margin is the reason a CCLAMP value of 0.01 mF will work in most applications. 2 a SS – SR 2 fMCLK(Hz) And for even/odd CCDs: 4.3 VGA The LM9801 has a VGA (Variable Gain Amplifier) that can be used to increase the amplitude of the CCD signal prior to sampling, correction, and digitization. The gain of the VGA is 0 dB to 9 dB and is determined by the codes in the 4-bit VGA Gain register, as given by the equation: 18 a SS – SR 2 fMCLK(Hz) Where SS is the value in the Sample Signal Position register (0 – 15), SR is the value in the Sample Reference Position register (0 – 14), 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 CCD’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: tDARK(s) e VGA code 9.55 16 This gain may be changed at the line rate (not the pixel rate) by writing to the configuration register. You can write to the configuration register to change the gain at any time, but if you write during a line, the remaining pixels of that line may be corrupted. It is best to change the gain after all active pixels have been read out or while SYNC is low. GainVGA (dB) e 1 t R In(accuracy) tDARK(s) n e ROUT(X) In(accuracy) CCLAMP MAX e http://www.national.com # ROUT CCLAMP Initial Voltage In n tDARK Final Voltage Using the values shown before and a clamp capacitor value of 0.01 mF, this works out to be: 1500X 0.01 mF 10V e 12.8 lines lines e In 18 300 ns 0.1V lines e 22 Applications Information (Continued) 4.4 Correlated Double Sampler (CDS) Figure 4 shows the output stage of a typical CCD and the resulting output waveform: TL/H/12814 – 32 FIGURE 5. CIS vs CCD Output Signals While CIS devices do not usually have a reference level with which to perform correlated double sampling, many have a very repeatable reset level which can be used as a black reference allowing the LM9801 to perform pseudo CDS on the signal. When the Signal Polarity bit is set to a zero, the LM9801 expects a positive going signal, typically from a CIS device. When the Signal Polarity bit is set to a one, the LM9801 expects a negative going signal, typically from a CCD sensor. 4.6 Offset DAC The 4 bit plus sign Offset DAC is used to compensate for DC offsets due to the correlated double sampling stage. The offset can be corrected in 31 steps of 0.42 ADC LSB size between b6.3 and a 6.3 LSBs. Note that the DAC comes betore the PGA, so any offset errors at this stage are multiplied by the gain of the PGA. The calibration procedure described in Section 5.0 demonstrates how to use the DAC to eliminate offset errors before scanning begins. Note that this DAC is programmed during LM9801 calibration/configuration and is not meant to compensate for pixelto-pixel CCD offset errors. CDS cancels the pixel-rate offset errors. TL/H/12814 – 31 FIGURE 4. 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 maximum. 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 (w1) 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 a VRESIDUAL] b VRESIDUAL e VSIGNAL). This is the principal of Correlated Double Sampling. The LM9801 implements CDS with two switched-capacitor S/H amplifiers. The S/Hs acquire a signal within a 50 ns window which can be placed anywhere in the pixel period with 25 ns precision. See Diagrams 7 and 8 for more detailed timing information. 4.7 Programmable Gain Amplifier (PGA) The PGA provides 8 bits of pixel-to-pixel gain correction over a 0 dB to 9 dB (x1 to x3) range. After the input signal is sampled and held by the CDS stage, it is amplified by the gain indicated by the data (‘‘PGA Code’’) on the CD0 – CD7 databus using the formula: Gain #VJ V e1a PGA code 1.95 256 4.8 Offset Add Bit In addition to the Offset DAC, there is a bit in the configuration register which, when set, adds a positive 2 LSB offset at the output of the PGA. This offset ensures that any offset between the output of the PGA and the ADC is positive, so that no dark level information is lost due to negative offsets. The calibration procedure described in Section 5.0 demonstrates how to set this bit. 4.5 CIS Mode The LM9801 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. 23 http://www.national.com Applications Information (Continued) 4.9 ADC The ADC converts the normalized analog output signal to an 8-bit digital code. The EOC output goes from high to low to indicate that a new conversion is ready. ADC data can be latched by external memory on the rising edge of EOC. The RD input takes the ADC’s output buffer in and out of TRI-STATE. RD may be tied to EOC in many applications, putting the data on the bus only when EOC is low, and allowing other data on the bus (such as CD0–CD7 correction data) at other times. In this way the output data and correction coefficient data can share the same databus (see Diagram 12). TL/H/12814 – 33 FIGURE 6. Offset Calibration ply by turning off the scanner’s illumination). If this voltage is known with a PGA gain of 1.00V/V (0 dB) and 2.95V/V (9 dB), then the offset errors (VOS1 and VOS2) can be determined from the following two equations: VADC1 e 1(VOS1 a VDAC1) a VOS2 a VDAC2 5.0 CALIBRATION Calibration of a CCD scanner is done to normalize the pixels of a linear CCD so that each pixel produces the same digital output code at the output of the scanner when presented with the same image light intensity. This intensity ranges from black (no light) to white (maximum light intensity). The CCD’s analog output may have large pixel-to-pixel DC offsets (corresponding to errors on black signals) and pixel-topixel variations in their output voltage given the same white image (corresponding to errors on brighter signals). If these offsets are subtracted from each pixel, and each pixel is given its own gain setting to correct for different efficiencies, then these errors can be eliminated. Ideally the digital output code for any pixel would be zero for a black image, and some code near fullscale for an image with maximum brightness. For an 8-bit system like the LM9801, that code might be 250. This code will be called the Target Code. The LM9801 eliminates these global and pixel-to-pixel offset and gain errors with its Correlated Double Sampling (CDS), Offset DACs, Variable Gain Amplifier, and pixel-rate Programmable Gain Amplifier. This section describes how to program the LM9801 and the coefficient RAM being used with it to eliminate these errors. Calibration of a LM9801-based system requires 3 steps. The first, described in Section 5.1, Offset Calibration , takes a black image and normalizes the digital output code for each pixel to a code at or near 0. The second step, Section 5.2, Coarse Gain (VGA) Calibration , finds the optimum gain setting that places the output voltage of all the pixels within the 9 dB adjustment range of the PGA. The final step, described in Section 5.3, PGA Correction Coefficients (Shading Calibration) , describes how to calculate the gain required to normalize the output of each pixel to the desired output code (the Target code). (PGA gain e 1) VADC2 e 2.95(VOS1 a VDAC1) a VOS2 a VDAC2 (PGA gain e 2.95) Solving for VOS1 and VOS2: VOS1 e (VADC2 –VADC1)/1.95 – VDAC1 VOS2 e (2.95VADC1 –VADC2)/1.95 – VDAC2 These equations were used to produce this procedure for cancelling the LM9801’s offset errors. Please note that all voltages and measurements are in units of ADC LSBs to simplify calibration. 1. Set the VGA Gain to 1V/V (VGA code e 0 LSBs). 2. Set the Offset DAC (VDAC1) to its maximum value ( a 6.3 LSBs) to ensure the total offset is positive and therefore measurable by the ADC. 3. Set the Offset Add bit (VDAC2) to 0. 4. Set the PGA Gain to 1V/V (PGA code e 0). 5. Digitize a black line. 6. Calculate the average (in ADC LSBs) of all the valid pixels in the black line and store that number as VADC1. 7. Set the PGA Gain to 2.95V/V (PGA code e 255). 8. Digitize a black line. 9. Calculate the average (in ADC LSBs) of all the valid pixels in the black line and store that number as VADC2. 10. Calculate VOS1: VOS1 e (VADC2 –VADC1)/1.95 – 6.3 11. Program the Offset DAC register using the formula: Offset DAC code e -(VOS1)(15/6.3) e (6.3 a (VADC1 –VADC2)/1.95)(15/6.3) e 15 a 1.22(VADC1 –VADC2) (Note: This calculation can be done as 15 a 39(VADC1 –VADC2)/32 for ease of programming in 8-bit microcontrollers) 12. If 3VADC1 l VADC2, then set the Offset Add bit to 0. If 3VADC1 k VADC2, set the Offset Add bit to 1. 13. The final value of the offset present at the ADC input can be used for the shading calibration calculations. Calculate the final value of the ADC input offset (VOFFSET) using: VOFFSET e (3VADC1 –VADC2)/2 (if the Offset Add bit is 0), or VOFFSET e (3VADC1 –VADC2)/2 a 2 (if the Offset Add bit is 1) 5.1 Offset Calibration This procedure corrects for static offsets generated by the CCD and the LM9801. Because the LM9801 uses CDS to eliminate the pixel-to-pixel offset errors of the CCD, no pixel-rate offset correction is required. To use the Offset DAC and Offset Add bit for offset correction, the offset errors (VOS1 and VOS2) must first be determined, as shown in Figure 6 . This is done be measuring the voltage at the PGA output, using the ADC with a black image on the CCD (a black image can usually be created sim- http://www.national.com 24 Applications Information (Continued) 5.2 Coarse Gain Calibration The LM9801’s PGA corrects for up to 9 dB of varIation in the CCD output signal’s white level intensity. That 9 dB range has to be centered inside the 9 dB window of correction as shown in Figure 7 . The window’s upper limit is determined by the Target code, and the lower limit by the Target code divided by 2.8 (this corresponds to the minimum gain range of the PGA). To allow proper calibration, the amplitude of all the pixels in the CCD should be inside this range when those pixels are scanning an image corresponding to the Target code. The placement of the pixels inside the 9 dB window can be controlled by any of three ways: changing the gain of the VGA, changing the integration time of the CCD, or changing the intensity of the light source. In most designs, the output waveform of the CCD can be brought into the 9 dB correction range of the PGA by adjusting the gain of the VGA. This is the next step in system calibration. TL/H/12814 – 34 FIGURE 7. CCD Input Signal In Range Figure 8 is a flowchart of one technique to find the optimum VGA gain setting during calibration. Calibration begins with a VGA gain setting of 1V/V and increments the VGA gain until one of the four possible results occur. Result 1 is the desired outcome, where the signal falls into the range shown in Figure 7 and the VGA calibration has been successful. TL/H/12814 – 35 FIGURE 8. VGA Calibration Flowchart 25 http://www.national.com Applications Information (Continued) The final problem that can occur during VGA calibration (Result 4) is the ‘‘Signal too weak: increase light intensity or integration time’’ condition, shown in Figure 11 . In this case, even with the VGA gain set to a maximum of 2.8, the amplitude of one or more pixels is less than the minimum required for shading correction. The solution is to increase the intensity of the light source or lengthen the integration time of the CCD to increase the CCD’s output amplitude. To ensure that a scanner system is manufacturable, the result of the VGA calibration must always be State 1. States 2, 3, and 4 must be eliminated either by ensuring that the total variation in light intensity (from all sources) from system to system to a maximum of 9 dB, or by being able to adjust the light source’s intensity and/or the CCD’s light integration time. There are several conditions that can cause the VGA gain calibration routine to fail. Result 2, ‘‘Signal is too strong: Decrease light intensity or integration time’’ is shown in Figure 9 . This condition indicates that the amplitude of one or more of the white pixels coming from the CCD is greater than the maximum input voltage that the LM9801 is capable of accepting (about 1.2Vp-p). In this case the amplitude of the analog CCD output must be reduced before it enters the LM9801’s OS input to prevent clipping. This can be done by reducing the intensity of the light source or shortening the integration time of the CCD. 5.3 PGA Correction Coefficients (Shading Calibration) Once the input signal has been centered inside the range the LM9801 can correct for, correction coefficients must be generated for each pixel to compensate for the gain error of that pixel. 1. Set Offset DAC and Add Bit as determined in Section 5.1. 2. Set the VGA gain to the value determined in Section 5.2. 3. Set the PGA gain to 0 dB. 4. Scan a reference line corresponding to all white or light grey and store it in memory. 5. Calculate the required gain correction coefficients for each pixel using the formula: TL/H/12814–36 FIGURE 9. CCD Input Signal Too Strong The second possible failure mode of the VGA calibration (Result 3) occurs if there is ‘‘Too much variation’’ in the amplitude of the pixels coming from the CCD (Figure 10) . The LM9801 can correct for up to a 2.8 to 1 variation in pixel amplitude. If the variation is greater than this than it must be reduced before it can perform shading correction on all the pixels. Typically this is done by using a better light source that has more uniform illumination, higher quality lenses, or other opto-mechanical techniques to reduce variation across all the active pixels. TL/H/12814–37 FIGURE 10. CCD Input Signal Range Too Wide TL/H/12814–38 FIGURE 11. CCD Input Signal Too Weak http://www.national.com # J 256 Target Code b1 1.95 Uncorrected Coden Where Uncorrected Coden is the ADC output code for pixel n with the PGA gain e 0 dB, Target Code is the number that corresponds to the desired output from the ADC with the given reference line input, and Correction Coefficientn is the gain correction number that is sent to the CD0 – CD7 correction databus to provide gain correction for pixel n when digitizing a line with the LM9801’s PGA gain correction operating. If it is difficult or undesirable to do the division, subtraction, and multiplication operations shown above for every pixel, then a lookup table can be generated in advance that will return the Correction Coefficient for any Uncorrected Code. This table can be stored in ROM or RAM and can speed up the calibration process. The disadvantage of this technique is that the Target Code must be fixed when the table is generated, so only one Target Code can be used (unless multiple tables are generated). All the Correction Coefficients must be stored and sent to the LM9801 through the CD0 – CD7 databus for every line scanned. Correction Coefficientn e 26 Applications Information (Continued) 6.0 POWER SUPPLY CONSIDERATIONS 6.3 Power Down Mode 6.1 General Setting the Power Down bit to a ‘‘1’’ puts the device in a low power standby mode. The CCD outputs (w1, w2, RS, and TR) 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 LM9801 enters the Power Down mode. Recovery from Power Down typically takes 50 ms (the time required for the reference voltages to settle to 0.5 LSB accuracy). The LM9801 should be powered by a single a 5V source (unless 3V-compatible digital I/O is requiredÐsee Section 6.2). The analog supplies (VA) and the digital supplies (VD and VD(I/O)) are brought out individually to allow separate bypassing for each supply input. They should not be powered by two or more different supplies. In systems with separate analog and digital a 5V supplies, all the supply pins of the LM9801 should be powered by the analog a 5V supply. Each supply input should be bypassed to its respective ground with a 0.1 mF capacitor located as close as possible to the supply input pin. A single 10 mF tantalum capacitor should be placed near the VA supply pin to provide low frequency bypassing. To minimize noise, keep the LM9801 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. 7.0 COLOR There are two primary ways to use the LM9801 in a color system with a triple output (RGB) CCD. The first is to use one LM9801 with an external multiplexer. This is the simplest solution. The second technique is to use one LM9801 per RGB color. 7.1 Parallel Output CCD, One LM9801 Figure 12 is an example of how to use a single LM9801 with a triple-output RGB CCD. In this case an entire line of red is digitized, followed by an entire line of green, then blue. This solution provides a 2.5 Mpixels/sec (for an effective 830k RGB pixels/sec after de-interleaving) pixel rate using a high performance triple output color CCD. The Mux 1 multiplexer, located between the CCD’s OS outputs and the LM9801’s OS input, selects the color to be digitized according to the states of the A and B inputs (described below). The multiplexer’s speed requirements are minimal because the mux switches at the line rate, not the pixel rate. Also, since the output of the mux goes into a high impedance, low-capacitance input, the ON resistance of the mux is not critical. The 74HC4052 is a good choice for this application. 6.2 3V Compatible Digital I/O If 3V digital I/O operation is desired, the VD(I/O) pin may be powered by a separate 3V g 10% or 3.3V g 10% supply. In this case, all the digital I/O pins (CD0–CD7, CCLK, MCLK, DD0–DD7, EOC, RD, SYNC, CS, SCLK, SDO, and SDI) will be 3V compatible. The CCD clock signals (w1, w2, RS, and TR) remain 5V outputs, powered by VD. In this case the VD(I/O) input should be bypassed to DGND(I/O) with a parallel combination of a 0.1 mF capacitor and a 10 mF tantalum capacitor. TL/H/12814 – 39 FIGURE 12. Parallel Output CCD Application Circuit 27 http://www.national.com Applications Information (Continued) TL/H/12814 – 40 FIGURE 13. Parallel Output CCD Timing To maximize the integration time for the Red, Green, and Blue photodiodes, the transfer (TR) pulses should be staggered as shown in Figure 13 . This is done by a demultiplexer (Mux 2) between the TR output of the LM9801 and the transfer gate inputs of the CCD. If the CCD’s transfer gate input capacitance is relatively low (see the CCD datasheet for this specification and the requirements for TR pulse rise and fall time), then the other half of the 74HC4052 may be used to switch the TR pulses as shown. If the TR gate input capacitance is so large that the minimum TR rise and fall times can not be met because of the 200X max on resistance of the 74HC4052’s switches, then the 74HC4052 can not be used to multiplex the TR output and should be replaced with an active device such as the 74HC155 dual 2to-4 demultiplexer. Two signals (A and B) must be generated to choose which color is going to be digitized and receives the TR pulse. These signals can be as simple as the output of a two bit counter that counts from 0 to 2 (0, 1, 2, 0, 1, 2, etc.). This counter should be incremented after the end of the previous line and before the first transfer pulse of the next line. Also, since each color will need a different VGA gain, the appropriate VGA gain value for each color should be sent to the LM9801 during this time. 7.2 Parallel Output CCD, Three LM9801s Figure 14 uses three LM9801s to achieve a 7.5 Mpixel/sec (2.5M RGB pixels/sec) pixel rate. The three LM9801s are synchronized by applying the same MCLK and SYNC signals to all three devices. One LM9801 provides the clock signals required for the CCD. Since the coefficient data for all three LM9801s will be latched simultaneously on the rising edge of CCLK, the correction coefficient bus must either be at least 24 bits wide (8 correction coefficient bits by 3 LM9801s) or run at a 7.5 MHz rate and be latched into a buffer between the correction coefficient databus and each LM9801. Similarly, the output data for all three LM9801s will be available simultaneously at the 3 output databusses. Since each LM9801 is dedicated to one color, the VGA gain does not change during line scan. TL/H/12814 – 41 FIGURE 14. Parallel Output CCD, Three LM9801 http://www.national.com 28 Applications Information (Continued) The Mode is set to Even/Odd, RS Pulse Width is set to its minimum value, and RS polarity is positive. The timing, shown in Figure 16 , is determined by the RS, SR, and SS registers. The RS pulse position (RS) is set to 10, dividing the pixel period so that the signal portion is available for the first 5 MCLKs following a w1 clock edge and the black reference portion appears during the last 2 MCLKs (following the 1 MCLK wide reset pulse). Sample Reference (SR) is set to 14, so it samples the black reference just before the next w1 clock edge. Sample Signal (SS) is set to 8, so it samples the black reference just before the next reset pulse. These values can be adjusted to account for differences in CCDs, CCD data delays, settling time, etc., but this is often not necessary. 8.0 A TYPICAL GREYSCALE APPLICATION Figure 15 shows the interface between the LM9801 and a typical greyscale even/odd output CCD, the TCD1250. The interface for most other CCDs will be similar, the only difference being the values for the series resistors (if required). The clamp capacitor value is determined as shown in Section 4.2. The resistor values are usually given in the CCD’s datasheet. If the datasheet’s requirement is given as a particular rise/fall time, the resistor can be chosen using the graph of w1, w2, RS and TR Rise Times Through a Series Resistance vs Load Capacitance graph in the Typical Performance Characteristics section. Given the required rise time and the input capacitance of the input being driven, the resistor value can be estimated from the graph. TL/H/12814 – 43 FIGURE 16. Typical Even/Odd Timing All 4 digital outputs (w1, w2, RS, and TR) are enabled. The TR pulse width is set to the minimum, 20 MCLKs, as is the guardband between w1 and TR. Either of these settings can be increased if necessary. The TR polarity is positive, as is the RS polarity. Some CCDs may require one or both of these signals to be inverted, in which case the corresponding bit can be set to a ‘‘1’’. If there is an inverting buffer between the LM9801 and the CCD, these bits may be used to correct the output polarity at the CCD. Note that if w1 and w2 are inverted, then w2 should be used as w1 at the CCD, and w1 should be used as w2 at the CCD (Figure 17 ). TL/H/12814 – 42 FIGURE 15. Greyscale CCD Interface Example These are the Configuration Register parameters recommended for use as a starting point for most even/odd CCDs: Mode e 1 (Even/Odd mode)* RS Pulse Width e 0 (1 MCLK) RS Pulse Polarity e 0* RS Pulse Position e 10 Sample Reference Position e 14 Sample Signal Position e 8 w1/w2/RS/TR Enable e 1/1/1/1 TR Pulse Width e 0 TR-w1 Guardband e 0 TR Polarity e 0* Signal Polarity e 1 Dummy Pixels e 2* Optical Black Pixels e 5* (*Value given in CCD datasheet) TL/H/12814 – 44 FIGURE 17. w1 and w2 After Inversion Since this is a CCD sensor, the Signal Polarity is set to a 1 (inverted) to match the CCD’s output signal. The number of dummy pixels and optical black reference pixels are given in the CCD’s datasheet. The dummy pixel register should be programmed with the number of dummy pixels in the CCD a 1 (for example, if the CCD has 16 dummy pixels then the register should contain 17). The optical black reference register should be programmed with the number of optical black pixels in the CCD. 29 http://www.national.com Applications Information (Continued) The final ‘‘trick’’ required to interface a CIS to the LM9801 is the generation of optical black pixels for the LM9801 to clamp to at the beginning of a line. Unlike CCDs, CIS devices do not have a sequence of optical black pixels at the beginning of a lineÐthe first pixel out of a CIS is valid image data. There are several ways to create black pixels for the LM9801 to clamp to. The PGA gain coefficient register and PGA Gain Source bit are used during calibration (see Section 5.0). The Power Down bit should be set to 0 for normal operation. The Offset Add bit is also programmed during calibration. The VGA and Offset DAC bits are programmed during calibration (Section 5.0). The Test Mode bits should always be set to ‘‘0’’. 9.0 TYPICAL CIS APPLICATION Many CIS sensors (such as those made by Dyna Image Corporation) have only one clock input, a transfer signal, and an output signal that is referred to ground (Figure 18) . Figure 19 shows the analog and digital circuitry required to connect a typical Dyna CIS to the LM9801. TL/H/12814 – 47 FIGURE 20. CIS Interface Digital Timing The simplest solution is to physically place a light shield (black plastic, tape or metal) over the first 10 or so pixels. This reduces the voltage output of the CIS to nearly 0V, which is adequate for the LM9801 to clamp to. This has the side effect of slightly reducing the number of active pixels available for image capture. A second option is to artifically generate ‘‘black’’ pixels by holding the CLOCK input high for 10 or so RS pulses (Figure 21) . This forces the output voltage to zero for the time that the CLOCK input is high, and only one active image pixel is lost. The BLACK signal could be generated by the ASIC/external logic that generates a pulse on the first rising edge of RS after the TR pulse. TL/H/12814–45 FIGURE 18. CIS Waveforms TL/H/12814–46 FIGURE 19. Minimum CIS Interface Because the CIS requires only one clock with a duty cycle of less than 50%, the LM9801’s RS output is used as the CIS’s CLK source. w1 and w2 are not used. The 74HC74 D flip-flop is used to lengthen the transfer pulse (SI, or ‘‘Shift In’’ on the CIS) so that it overlaps the first RS pulse and meets the timing requirement of the CIS (see Figure 20 ). TL/H/12814 – 48 FIGURE 21. Generating Artificial Black Pixels http://www.national.com 30 Applications Information (Continued) Suggested timing for CIS devices is: Mode e 0 (Standard Mode)* 10.0 HINTS AND COMMON SYSTEM DESIGN PROBLEMS RS Pulse Width e 0 (2 MCLKs) RS Pulse Polarity e 0 (or 1 if circuit of Figure 21 is used)* RS Pulse Position e 0 Sample Reference Position e 2 Sample Signal Position e 14 w1/w2/RS/TR Enable e 0/0/1/1 TR Pulse Width e 0 TR-w1 Guardband e 0 TR Polarity e 1* Signal Polarity e 0 Dummy Pixels e 2 Optical Black Pixels e 10 (*Value given in CCD datasheet) As CIS sensors approach pixel rates of 1 MHz and above (corresponding to MCLK frequencies of 8 MHz and above), the voltage during the reset level becomes less stable, making it difficult to perform CDS on the output (Figure 22) . The solution is to create the ground reference externally, shorting the LM9801’s input to ground for half of the time using the w1 clock, as shown in Figure 23 . 10.1 Reading and Writing to the Configuration Register The Configuration Register sends and receives data LSB (Least Significant Byte) first. Some microcontrollers send out data MSB (Most Significant Byte) first. The order of the bits must be reversed to when using these microcontrollers. Note: Unlike the LM9800, the SYNC pin does not have to be held high to send or receive data to or from the Configuration Register. 10.2 Setting the Dummy and Optical Black Pixel Registers The minimum value in the Dummy Pixels register is 2 (a value of 0 or 1 will cause errors in the EOC and CCLK timing). Note that the value in this register should be equal to 1 plus the actual number of dummy pixels in the CCD. For example, if the CCD being used with the LM9801 has 12 dummy pixels, this register should be set to 13. The minimum number in the Optical Black Pixels register is 1. 10.3 Stretching the TR-w1 Guardband Some CCDs (Sony’s ILX514, ILX518, and ILX524, for example) require a TR to w1 guardband greater than the 100 ns (2 MCLKs) provided by the LM9801. The circuit shown in Figure 24 produces a 1 ms wROG (transfer) pulse with a guardband between the end of the wROG pulse and the next edge of w1. This is done by setting the LM9801’s TR pulse width register to 2 ms and using the 74HC4538 to generate a 1 ms pulse inside that TR period to send to the CCD. TL/H/12814 – 49 FIGURE 22. High Speed CIS Waveforms TL/H/12814 – 51 FIGURE 24. Stretching the TR-w1 Guardband TL/H/12814 – 50 FIGURE 23. High Speed CIS Interface 31 http://www.national.com Applications Information (Continued) Figure 25 shows a different technique for increasing the TRw1 guardband and/or increasing the length of the TR pulse by stopping the s MCLK during the TR period. When TR initially goes high, the first one-shot (U1A) triggers, effectively disabling the LM9801’s MCLK for & 2 ms, thereby lengthening the TR pulse width by & 2 ms over the value programmed in the configuration register. On the falling edge of TR, the second one-shot (U1B) fires, disabling the LM9801’s MCLK for & 1 ms and increasing the TR-w1 guardband by that amount. TL/H/12814 – 52 FIGURE 25. Stretching TR and the TR-w1 Guardband http://www.national.com 32 Physical Dimensions inches (millimeters) unless otherwise noted 52-Pin Plastic Leaded Chip Carrier (PLCC) Order Number LM9801CCV NS Package Number V52A 33 http://www.national.com LM9801 8-Bit Greyscale/24-Bit Color Linear CCD Sensor Processor Physical Dimensions inches (millimeters) unless otherwise noted (Continued) 52-Pin Thin Quad Flatpak Order Number LM9801CCVF NS Package Number VEG52A 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. National Semiconductor Corporation 1111 West Bardin Road Arlington, TX 76017 Tel: 1(800) 272-9959 Fax: 1(800) 737-7018 http://www.national.com 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 Europe Fax: a49 (0) 180-530 85 86 Email: europe.support @ nsc.com Deutsch Tel: a49 (0) 180-530 85 85 English Tel: a49 (0) 180-532 78 32 Fran3ais Tel: a49 (0) 180-532 93 58 Italiano Tel: a49 (0) 180-534 16 80 National Semiconductor Hong Kong Ltd. 13th Floor, Straight Block, Ocean Centre, 5 Canton Rd. Tsimshatsui, Kowloon Hong Kong Tel: (852) 2737-1600 Fax: (852) 2736-9960 National Semiconductor Japan Ltd. Tel: 81-043-299-2308 Fax: 81-043-299-2408 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.