Fast IR Camera File #1 ---------------------------- Initial Rev. 12-JULY-1996 Most Recent Rev. 14-AUG-1996 Sociological: 1. Currently over committed to D-Zero Run II. We have promised to build something by a fixed date part of which we currently have no know way to meet the design specifications. 2. We now can imagine that commercial equipment is not available to operate the new Rockwell parts. The military stuff is wrong and Rockwell only has "HP based" test setups. 3. Studying: the Sun, adaptive optics, and a "new technology" 4 meter telescope all sound interesting. 4. How can there be an interesting device like this Rockwell part sitting around without some one else already having picked it up and using it ? 5. HEP paid for all of the CAD equipment that we would need to use to build anything. They purchased this equipment for their sole use. 6. We like formally defined projects where a very detailed ".txt" file describes the project before design work begins. We insist on this for both big and small projects. I know that this sounds dry and uninteresting but with everyone doing 20 different things at once this is the only way for us to deliver electronics that will actually get your job done. 7. Technology: 1. Are you picturing a "test bed" type of instrument that gets hacked up and rebuild frequently or are you picturing a stable instrument that is used for observational research ? 2. How do we save ourselves from reinventing: The control system and procedures ? The data logging, visualization, and analysis software part. 3. The system basically consists of 3 parts: A uProcessor based "Command Receiver and Executer" A Timing Generator A Pixel digitization section Is this the correct picture ? 4. The assumption is that the new Rockwell part does NOT need any precision fancy analog 3 phase clocks, i.e. the Timing Generator is purely digital. 5. How do we TEST this setup and quantify its performance: Apply stimulus: analog waveforms into the ADC, IR patterns into the Rockwell part. Analyze data: FFT, Histograms, examine the raw data, stimulate - control the source, visualize the raw data with false colors, integral and differential linearity plots,... i.e. do we need to write or are there canned programs to do all of this ? 6. What is going to absorb the data from this system. 20 Mbytes/sec will over run any IBM/PC ISA bus. Does this system need to buffer frames of data for later readout ? I can NOT imagine that just saying that this system uses RS-422 output signaling levels solves this problem. 7. What actual analog bandwidth needs to be handled and digitized to what accuracy ? This is not a question about the 10 MSPS digitization rate, rather this asks about the highest frequency Fourier components coming out of the Rockwell part that must be handled correctly. For example if in a scan line the odd number pixels are at 90% of full scale and the even numbered pixels are at 10% of full scale then the analog bandwidth that must be handled is much greater than the pixel digitization rate (e.g. 5x higher just to get the 5th harmonic that is making up the square wave signal that would be coming out of the Rockwell part). 8. What ever pixel digitization rate, analog bandwidth, accuracy this system is built for, it will NOT be easy to change it to something else. 9. Of the 14 bits, how much is just for dynamic range and how much is for the actual required accuracy ? 10. What Frame Rate vs Frame Size is actually needed ? --> MSPS required. Frames per Second at a Pixel Digitization Rate of Array Pixels Bytes per ------------------------------- Size per Frame Frame 5 MSPS 8 MSPS 10 MSPS --------- --------- --------- ------ ------ ------- 64 x 64 4,096 8,192 1,220 1,953 2,441 128 x 128 16,384 32,768 305 488 610 256 x 256 65,536 131,072 76 122 152 ------ ------ ------- 10 16 20 ----------------------------- M Bytes per Second Readout 11. What part of the system should be: At the camera, Near the camera, and At the control point. What power supplies are available ? 12. How close can we get to the dewar ? How do we pass signals through into the dewar. Can we have multiple penetrations into the dewar, i.e. digital and separate analog? The electronics in the dewar and immediately next to it will end up determining the accuracy of the ADC system. 13. Can we self ID the data as to frame size and beginning of frame and beginning of scan line by using the top two bits ? E.G. top two bits = 11 --> start of frame, top two bits = 10 --> start of line. 14. Confidence test signal source (DAC?) on board ? 15. We need to know a lot lot more about the Rockwell part. The current information tells us nothing about how to actually operate the part. I have seen a number of cases where these custom asic projects get screwed up because of a miss-understanding between the chip designers and the chip users (even over stupid things like pinout). 16. Digitizing an analog input signal that contains 50 MHz Fourier components at a 10 MSPS rate with something like 12 bits of absolute accuracy and 14 bits of dynamic range looks OK. But this is about the end of the brut force techniques. Moving up to 100 MHz Fourier components or moving to 20 MSPS (i.e. double sampling the pixels) will require a different technique (e.g. dual range conversions or something). 17. If the three component description of this project is correct (i.e. the uProcessor based "Command Receiver and Executer", the Timing Generator, and the Pixel Digitization Section) then the first step is to write a functional description of each of these sections, a description of the connections between the sections, and a description of the software and hardware connections to the outside world. (see point #6 under the sociological heading) 18. Expressions of converter performance in dB Ratio of LSB Ratio Bits to the Full Span Expressed in dB ------ ------------------ ----------------- 8 1 in 256 48.16 9 1 in 512 54.19 10 1 in 1,024 60.21 11 1 in 2.048 66.23 12 1 in 4,096 72.25 13 1 in 8,192 78.27 14 1 in 16,384 84.29 15 1 in 32,768 90.31 16 1 in 65,536 96.33 17 1 in 131,072 102.35 18 1 in 262,144 108.37 SNR_dB - 1.761 Effective_Bits = -------------- 6.021 SNR_dB = 10 Log ( (12/8) x (2**(2 x Effictive_Bits)) ) Effictive_Bits SNR_dB ---------------- -------- 8 49.93 9 55.95 10 61.97 11 67.99 12 74.01 13 80.03 14 86.05 15 92.07 16 98.09 17 104.11 18 110.13 19. The 6 Component Description of the Hardware for the Fast_IR Project: The uProcessor Command Receiver and Executor The Timing Generator The Pixel Digitization Section The Timing Signal Receivers and Array Drivers The IR Array Cold Support Circuit Board (aka the Rock) The various Power Supplies +------------------+ | | | uProcessor | +-----------------+ | Command Receiver | | | | and Executor | | Timing Signal | | | | Receivers and | +------------------+ | Array Drivers | +-----------------+ +-----------------+ | | +------------------+ | IR | | | | Array | | Timing | +-----------------+ | | | Generator | | | +-----------------+ | | | Pixel | +------------------+ | Digitization | | | +------------------+ +-----------------+ | | | Power | | Supplies | | | +------------------+ The Command Receiver and Executor, the Timing Generator, and the mains feed power supplies are located together in a VME environment that is about 20 feet from the cryostat. The Timing Signal Receivers and Array Drivers are located in an RF box on the cryostat. The Pixel Digitization Section is located in its own RF box on the Cryostat. The IR Array is located on the Rock circuit board in the cryostat. 20. Connections among the 6 boxes in the Fast_IR_1 system: The uProcessor Command Receiver and Executor: Receives ASCII Commands from the Host System via RS-232 Examples of Commands: Setup Rockwell array size Setup light integration time Obtain "N" Frames Select the Confidence Test Source or the Rockwell Array as input to the Pixel Digitizer. Return Status to the Host System. Examples of this are: Command (Sequence Number XYZ) Understood and it's being Executed Command (Sequence Number XYZ) contains an Error: Parameter Error, Command Verb Error Ready for next command. Note that the Return Status from the Command Receiver and Executor reflects only that boxes view of the command (as a whole) at the time that the the command is parsed. Besides the Host System, the Command Receiver and Executor communicates directly ONLY with the Timing Generator. It can program the Timing Generator and it uses hardware on the Timing Generator circuit card to implement PIO lines. All signals that are needed to setup the Rockwell Array (e.g. serial strings to setup the array size) are generated directly by CODE in the Command Receiver and Executor and rendered into electrical signals via PIO hardware on the Timing Generator. Timing Generator: The Timing Generator is controlled by the Command Receiver and Executor. The Command Receiver and Executor sets up the Timing Generator via loading registers that are in the Timing Generator. The Timing Generator creates all of the waveform that are required to obtain one Frame of array data. Once the Timing Generator has been setup, the Command Receiver only needs to access the "GO" bit in one of the Timing Generators Control-Status Registers to cause a Frame of Array Data to be acquired. The Timing Generator maintains a "BUSY" status bit via which the Command Receiver and Executor can tell when the Timing Generator has finished acquiring a Frame of Array Data. --> Executive Summary: All timing and control at the Frame rate and slower are implemented by the Command Receiver and Executor itself (i.e. by CODE running on this uProcessor). The frame rate looks like it will be under 5kHz so this is practical. All timing and control that is at the Frame rate and faster is implemented by the Timing Generator (i.e. by hardware in the Timing Generator). Power Supplies: All of the mains feed power supplies are located at the Command Receiver and Executor, Timing Generator, VME location. VME power is feed directly to the Command Receiver and Executor and to the Timing Generator. The power supplies also feed power to the Timing Signal Receivers and Array Driver which are located in their own RF box on the cryostat. Secondary regulators on the Timing Signal Receiver and Array Driver circuit board provided filtered Digital power to the Rockwell Array. The power supplies also feed power to the Pixel Digitization Section which is located in its own RF box on the Cryostat. Secondary regulators on the Pixel Digitization Section circuit board provided filtered Analog power to the Rockwell Array. All expensive components (e.g. the Rockwell Array and the ADC) have the necessary protection: power sequence, polarity check, and over voltage clamp. Timing Signal Receivers and Array Drivers: The Timing Signal Receiver and Array Driver circuit board is located in an RF box that is mounted on the cryostat. This circuit board receives differential ECL signals from the Timing Generator, converts these signals to the appropriate digital signaling voltage levels and delivers these signals to the Rockwell Array with the appropriate filtering and series termination. This circuit board also contains the secondary regulators for supplying digital power to the Rockwell Array. Pixel Digitization Section: The Pixel Digitization Section is located in its own RF box that is mounted on the cryostat. The analog signal(s) from the Rockwell Array come into the Pixel Digitization Section RF box via their own penetration through the cryostat. The dsafasfd The Pixel Digitization Section also contains the secondary regulators that supply the Analog power to the Rockwell array. The Pixel digitization section contains the "confidence check" signal source. The Pixel Digitization Section may require forced air cooling. IR Array Cold Support Circuit Board (aka the Rock): The Rock circuit board provides all of the support functions and required connections for the IR Array. These may/will include: The "cold connection". Digital power supplies and their bypassing. Analog power supplies and their bypassing. The required analog control voltages and their bypassing. The required ground and power planes. Required stainless steel heat stops. Required guard traces and shields to isolate the analog vs digital signals. Required connectors so the the IR Array can be removed from the Rock and so that the Rock can be removed from the cryostat. 21. Information that we need to learn about the Rockwell IR Array part: Physical Specifications and Characteristics. How cold to get it. How to control getting it cold. Power Supplies: What Power Supplies are required. Is there a required power supply sequencing. Special controllable power supplies. Power supplies that must be trimmed and the procedure for adjusting them. Fatal power supply conditions. Recommended bypassing of the various power supplies. Recommended pcb layout e.g. what ground power planes. Digital control - programming of the array size. Digital control of: photon integration period clocking out the pixels other Analog output: Source Follower vs AB Amplifier Electrical characteristics of each Pixel to Pixel gain variation, how much is this? Dynamic response of the analog section ? Can 4 of these analog sections actually operate from 5.8ma (ColHi = 1.1 ma + ABAmpHi = 2.0 ma = AmpHi = 2.7 ma) and provide 12-14 bit accuracy at the 5 mega pixel per second readout speed? 22. Analog Part Companies: Burr-Brown (but no fast ADC's anymore) Analog Devices Analogic Datel Maxim Linear Technology National (e.g. LM6171) Comlinear (opamp and ADC) Elantec 800-882-2109 Harris 23. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Some ADC numbers: 12 Bit Fast ADC's ComLinear CLC935, CLC936, CLC937, CLC938, CLC949 Analog Devices AD9221, AD9223, AD9220 Analog Devices AD9042 Analog Devices AD872A 14 Bit Fast ADC's Analogic ADC3120 Not Released, Not know if it will ever be released Analogic ADC3121 $2500 6-8 week delivery Burr-Brown ADC614 Burr-Brown got out of this business Datel ADS-946 $650 sampling July 1996 Datel ADS-945 $1,154 3 week delivery Analog Devices AD9014 $3,200 1 week delivery Some ADC numbers: Analogic Analog Devices Comlinear ------------------ --------------- --------------- ADC3120 ADC3121 AD9042 AD9220 CLC93% CLC949 ------- ------- ------ ------ ------ ------ Bits Resolution: 14 14 12 12 12 12 MSPS Range: 10:20 10:20 5:41 .01:10 0:31 .01:30 Convrtng @ MSPS: 10?20 10 20 41 10 15 30 20 SNR DC:10 MHz 75 70 72 - - - - SNR 0.4 MHz - - - - - 65 65 SNR 1.2 MHz - - - 68 70 - - SNR 5.0 MHz - - - - 68 65 64 68 SNR 7.2 MHz - - - - - 65 64 68 SNR 9.6 MHz - - - 67 - - - SNR 19.5 MHz - - - 67 - - - SNR 50.0 MHz - - - - - - 60 SFDR DC:10 MHz 90 - - - - SFDR 0.4 MHz - - - - - 82 - SFDR 1.2 MHZ - - - 81 88 - - SFDR 2.0 MHz - 81 80 - - - - SFDR 4.8 MHz - 81 73 - 75 80 75 72 SFDR 9.6 MHZ - - - 81 - 75 70 72 SFDR 19.5 MHZ - - - 81 - - - THD 1.0 MHz 85 - - - 84 - - THD 2.0 MHz - 78 74 - - - - THD 4.8 MHz - 78 70 - 72 - - THD 10.0 MHz 82 - - - - - - %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Stability Harris rules of thumb: Small signal (e.g. 200 mV) pulse response, if the overshoot is less than 40% then the circuit will be stable. Small signal frequency response, if the high frequency peaking is less than +6 dB above the low frequency gain then the circuit will be stable.