CCD Camera using a KAF-1600 sensor

Coming Soon Someday Someday Soon

Some time back I ran across a KAF-1600 sensor for $50. They used to sell for $1300 or so. I sure hope it's good, because I am putting a lot of effort into getting it going.


I have a CCD camera ca. 2000 with a Kodak KAF-0400 sensor in it - 1/4 the sizeof this 1600. The camera takes about 19 seconds to digitize the 768 x 512 pixels of the KAF-0400. It needs a PC parallel port to control everything. (Try to find a laptop with a parallel port.) That would be nearly 80 seconds to digitize an image with the 1536 x 1024 KAF-1600 sensor. I've been playing around with the AD7667 1MSPS analog to digital converter from Analog Devices and thinking how much of an improvement it would be over the old 100kSPS AD976 converter. It can digitize faster than the clocks can get data out of the CCD. What is needed is a way to transfer the data off of the camera at the rate it can be digitized.

The FTDI UM232H is a dev board for the FT232H USB to serial/parallel/I2C/SPI converter that can theoretically transfer 8MB/S async parallel, which could help out. With a 3.042µS pixel time we are digitizing 1536 pixels (3072 bytes) at 656kB/S for one line. The average is 649kB/S for a download time of 4.84 seconds per full frame. No problem for the UM232H.

Camera Timing

The timing graph above shows the beginning of a line. This was captured from the breadboard below.


I built the digital and analog circuits on a breadboard, and wrote the firmware. I did not include the MAX333 CCD clock circuits, since they require all of the power supplies. I also wrote a program for OS X that can read the camera and display the image.

I'm going to go through this ambitious project and try to break it into manageable pieces. I've decided on four 100mm x 100mm boards for the camera, since the 100mm x 100mm format is the cheapest by magnitudes, and a pair of 150mm x 100mm boards for the switching power supplies.

  • Bottom - (cpu-r0) MCU, USB I/O, temperature probes
  • Middle - (ps-r0) Power supply - CCD voltages, video ±12V, ±19V, +5V
  • Top - (anl-r0) ADC, video Amp, diagnostic connector
  • CCD - (ccd-r0) Clocks, video buffer, CCD

  • External power supplies

    I decided this was going to run from a 12V battery in the field. That means a 12V supply needs to provide the input power to the external power supply at home. It also means the power supply itself must be a switcher, or in this case, two switchers - one for the pre-regulated camera supply and one for the variable TEC supply.

    • Switcher 1: ±20.5V and +9V @ 100mA or so
    • Switcher 2: +15V @4A for the TEC
    • (The water pump will run from the 12V input power)

    The purpose of pre-regulators is to keep as much of the ripple and noise as possible out of the camera, limiting it to relatively clean DC inputs. The pre-regulated output voltage needs to be comfortably higher than the dropout voltage of the regulators in the camera. The LM3x7 regulators have a 3V dropout, so I allowed 4V of headroom. The ±15V supply only needed ±19V to work, but the LM3524D feedback resistor values worked out such that I ended up with 20.5V.

    Parts List
    • 2 ea. PCB $60.00/5
    • 2 ea. LM3524D $5.92
    • 4 ea. IRF520 $3.56
    • 8 ea. 2N3904 NPN transistor $3.20
    • 4 ea. 1N4148 Signal diode $0.40
    • 4 ea. Vishay 1.5KE51CA-E3/54 $2.32
    • 2 ea. LM317T Pos. Adj. Regulator $1.04
    • 1 ea. LM337T Neg. Adj. Regulator $0.66
    • 28 ga. Magnet wire (secondaries)
    • 24 ga. Magnet wire (primaries and secondaries)
    • 2 ea. Amidon PC3622-77 pot core w/bobbin $12.50
    • 12 ea. 80SQ045NRLG 8A/45V Schottky diode $7.70
    • 4 ea. Nichicon UPM1V102MHD6 1000/35 Electrolytic $5.48
    • 8 ea. Nichicon UPW1V471MPD 470/35 Electrolytic $4.64
    • 3 ea. 10kΩ 1% load resistor for rectifier output
    • 1 ea. 1kΩ 1% Freq LM3524D
    • 2 ea. 0.01 Xicon 23PS310 Polystyrene Capacitor $0.84 Freq LM3524D
    • 3 ea. 240Ω 1%
    • 2 ea. 3.6kΩ 1%
    • 1 ea. 1.5kΩ 1%
    • 3 ea. 10µF Ceramic Capacitor $0.24
    • 1ea. 11.9" x 7.5" x 2.9" Low-Profile Aluminum Instrument Enclosure $17.95
    • and so on...
    • Total about $123.38
    • Weight: ~3 pounds
    Reference Schematic (revised) PCB Layout

    The boards were made by at $60 for 5 boards.

    One copy of this board will have the ±20V and +9V supplies on it, and another will produce the +15VDC for the TEC. The LM3524D has an override that allows one to force a duty cycle with an input voltage, making it a voltage controlled power supply so the temperature error can drive the TEC. No linear voltage regulator is needed because the LM3524 will regulate the voltage adequately.

    Problems found with prototype switcher board:

    • Q3,4,5,6 were P2n2222 not 2n2222. Wrong pinout. Had to make a new 2N3904 in CAD.
    • Ground was not connected to the LM3524D.
    • Shutdown was pulled high, shutting down the converter.
    • The MOSFET drivers Q3,4,5,6 had no Vcc on their Vcc net.
    • There are no ceramics bypassing the output filters.
    • There was no load on the supply.
    • I should not have put a gate resistor on the MOSFETS.

    One interesting thing. The MOSFETS are hot as the blazes when the supply is run with no load. Adding a tiny 2mA load cools them right off. I've run it at 12 watts output, which due to the linear regulators' overhead voltage is more like 16W at the MOSFETS, and all is fine. That is about all I can do with my 12V 1A (12W) power supply feeding it. The input droops to just over 7.5V, while the outputs stay at ±20.5V and +9V. The duty cycle is at 50%, which is not at all surprising considering the low input voltage.

    I changed the operating frequency from ~25kHz to ~36kHz to help with efficiency.

  • MCU and USB-FIFO
    • MCU
    • USB FIFO Controller
    • Ports for 3 DS18B20 1-wire temperature sensors

    Parts List:
    • 1 ea. ATmega1284 $5.15
    • 1 ea. UM232H USB FIFO module $20.00
    • 2 ea. 2N3904 NPN $0.80
    • 1 ea. 20MHz crystal $0.18
    • 3 ea. 0.1µF/50V ceramic $0.20
    • 2 ea. 22pF/50V ceramic $0.20
    • 4 ea. 10kΩ 5%
    • 1 ea. PCB $17.00/10
    • Total $

    I've had problems with serial programming adapters powering up the MCU through their transmit pin, and that causes problems with the AD7667 ADC latching up when you really power it up. I added two transistors to isolate the receive input from the USB-Serial converter transmit pin. Other than that, the CPU board is just the MCU and the USB to Parallel converter.

    There are ports for three DS18B20 temperature sensors, but since they are 1-wire devices, any number of them may be paralleled on a single connector. I envision one for the cold finger, one for the water plate and one for ambient. The cold finger temperature will be used to set the TEC voltage.

    Reference Schematic PCB Built

  • Power Supply and Bias

    The camera takes 5 fixed voltages and 9 variable voltages to run, counting power, bias and clocks. Some are positive and some are negative. The fixed voltages use LM317 and LM337 regulators. The variable voltages come from pots hung off of the ±15V supplies. Each voltage is adjustable within the range specified in the data sheet, and buffered by an LM324 opamp. In theory, swapping this board out with one having different ranges will allow the board set to be used with different KAF sensors. Perhaps only adjusting the value within the existing ranges will suffice.

    The ADC takes 3 different 5V supplies - DVDD, AVDD, and OVDD. OVDD has to be sequenced to come on last or latchup might will occur in the ADC. There is a MOSFET to switch the OVDD supply on and off under control of the MCU. Once the chip has latched up, only a power cycle will unlatch it. And with power applied it gets very hot.

    Schematic PCB
    • CCD voltage supplies (with resistor values):

      I wrote a little script that takes the high and low voltage limits and turns it into the resistor values required for either end of the 2k pot.

      • +11V (+10.5 to +11.5 Reset Drain) (6.65k, 21k)
      • +9V (+8.0 to +12.0 Guard Ring) (1.43k, 4.02k)
      • +6V (+5.0 to +6.5 Horizontal Clock High) (11k, 6.65k)
      • +4V (+3.75 to +5.0 Output Gate) Combined with below - keep 3.75V minimum (15.8k, 6.04k)
      • +4V (+3.5 to +5.0 Reset Clock High) Combined with above for 3.75V minimum
      • +2V (+1.5 to +2.5 Output Amp Return, Reset clock high) (24.9k, 3k)
      • +0.5 (+0.0 to + 1.0 Vertical Clock High) (27.7k, 0)
      • -4V (-5.0 to -3.5 Horizontal Clock Low, Reset Clock low) (10k, 3k)
      • -8V (-8.5 to -7.5 Vertical Clock Low) (12.7k, 15k)
    • The CCD VDD is not adjustable but it needs to be switched. I changed the voltage on the MAX333A switches to ±19V to accomodate switching 15V. This is accomplished by series diodes in the ±20V lines. Switching VDD off prevents "electroluminescence" which lights up one corner of the image during exposure. It is switched back on to do any readout functions (clear, read).
      • +15V fixed (+14.5 to +15.5) CCD VDD
    • The AD8021 opamps require a power supply no greater than ±12V. The output of the CCD is +10.5V, so that pretty much defines the range of voltages that can work as the video amp supply as being between 10.5V and 12V.
      • ±12V Video Amp VCC

    The five regulators and one MOSFET are closely spaced and so require a heat spreader to cool them off. I found some aluminum washers to space them off the board a little bit, and a Home Depot 0.125" x 0.5" aluminum bar will be the spreader.

    Parts List:
    • 9 ea. Bourns 3006-1-202LF side adjust 2kΩ $18.72 (6.79 on ebay)
    • Resistors for adjustable voltages:
    • 2 ea. 6.65kΩ 1%
    • 1 ea. 21kΩ 1%
    • 1 ea. 1.43kΩ 1%
    • 1 ea. 4.02kΩ 1%
    • 1 ea. 11kΩ 1%
    • 1 ea. 15.8kΩ 1%
    • 1 ea. 6.04kΩ 1%
    • 1 ea. 24.9kΩ 1%
    • 2 ea. 3.01kΩ 1%
    • 1 ea. 27.7kΩ 1%
    • 1 ea. 10kΩ 1%
    • 1 ea. 12.7kΩ 1%
    • 1 ea. 15kΩ 1%
    • Resistors for voltage regulators:
    • 5 ea. 240Ω 1%
    • 2 ea. 2.61kΩ 1%
    • 1 ea. 715Ω 1%
    • 2 ea. 2.05kΩ 1%
    • 10 ea. 0.1µF/50V Ceramic
    • 5 ea. 10uF Ceramic
    • 5 ea. 1uF Ceramic
    • 3 ea. LM324AN $1.14
    • 1 ea. PCB 100x100mm $17.00/10
    • Total $

  • Analog Board Schematic PCB Layout Parts List:
    • 2 ea. AD8021ARZ $6.74
    • 1 ea. AD7667 $23.88
    • 1 ea. Schmartboard $6.00
    • 1 ea. DG413
    • 2 ea. 4.7uF Tantalum
    • 5 ea. 1uF Ceramic
    • 3 ea. 0.1uF Ceramic
    • 1 ea. 0.01uF Ceramic
    • 1 ea. 0.001uF Ceramic
    • 1 ea. 10Ω 1/4W 1%
    • 4 ea. 75Ω 1/4W 1%
    • 1 ea. 10kΩ 1/4W 1%
    • 1 ea. PCB 100x100mm $17.00/10
    • Total $

    The analog board has a connector for the logic analyzer so the timing can be accurately set. A 24-pin cable runs to the CCD board. There are three problems with this board:

    1. The AMP_EN line does not go to the dewar board connector.
    2. The AD8021 LOGIC_REFERENCE pins go to a no-connect pin on the analog switch, and not to ground.
    3. PDBUF and PDREF on the AD7667 are floating and should be grounded.

    All of these problems are fixable without a new board.

    Schmartboard has a fire sale on their 48-pin carrier - $1.00 each without headers. I was alerted by a reader that he had found them in the clearance bin on their website. I don't use the 0.025" sq. post headers anyway. I use the machine pin headers because the mating part has a much lower profile.


  • CCD Board PCB Layout
    • CCD dewar

      This is the hardest part for me. Too mechanical.

      • 1 ea. 1/4" x 12" x 12" PVC for walls
      • 2 ea. 0.125" x 2.5" x 2.5" Aluminum
      • 1 ea. 0.5" x 0.5" x 2.25" Aluminum "cold finger" part 1
      • 1 ea. 0.125" x 1.9" x 1.9" Aluminum "cold finger" part 2
      • 1 ea. 1" x 2" x 2" Aluminum water block
      • 1 ea. 1/32" x 6" x 36" Neoprene 30A for gaskets $4.14
      • 2 ea. 1/4" NPT to 1/4" hose barb (aluminum) $5.60
      • 1 ea. 39mm coated UV filter $6.00
      • 2 ea. MAX333ACPP+ $18.48
      • 1 ea. 2N3904 $ 0.40
      • 1 ea. 24-pin male header (no shroud)
      • 24 ea. PCB socket pins for the CCD (Salvage from machine-pin sockets)
      • 1 ea. KAF-1600A CCD
      • 1 ea. PCB 100mm x 100mm x 0.8mm black mask $27.00/10

  • Case to hold it all
    • 2 ea. Aluminum sheet 7" x 9" x 0.125" from local metal supply house
    • Some kind of a rotary shutter
    • A 4-pos 50mm filter wheel