Cmosis support

HomeDragsterPrinter Friendly Version

Dragster

A record of Frequently Asked Questions on Dragster II

1. Variations

1.1. Sensor reference - what does it mean?

Awaiba's Dragster complete linescan sensor familly is the most complete range of linescan sensors on the market. 

For users to identify our products they need to get familiar with our designators.

Taking the DR2K7LCCRGB as an example: 

- DR stands for Dragster

- 2K means 2 times "1K" resolution and "1K" means 1024 pixels. If, for example, it's a "16K" it means that this sensor has 16 x 1024 = 16384 active pixels. Fhrthermore, if it's indicated 2x2K means there are 2 lines of 2K pixels one over the other.

- 7 it's the pixel size in microns. Because the pixels is square, we use only a single value to designate the pixel size. 

- LCC is the type of package ("Leadless chip carrier") and it's a special type of package. If there's no indication after the pixel size it means the sensor is offered in an INVAR package

- RGB means it is a color sensor. No RGB ending means the sensor is Black and White

1.2. How many sensors variations are available?

Awaiba has a complete set of Dragster line scan sensor family that includes not only several resolutions but also diferent pixel sizes and packages in a total of 18 variations.

All the following versions are in production and imediately available uppon confirmation.

Currently are available the following variations with 2K pixels:

DR2K7

DR2K7LCC

DR2x2K7LCC

DR2x2K7RGB

DR2x2K7LCCRGB

 

Available versions with 4K pixels:

DR4K7

DR2x4K7

DR2x4K7RGB

DR4K7LCC

DR4K3.5

DR4K3.5LCC

 

Available versions with 8K pixels:

DR8K7

DR8K3.5

DR2x8K7

DR2x8K7RGB

 

Available versions with 16K pixels

DR16K3.5

 

2. Packaging

2.1. Cooling and thermal Resistance

All characterization is made at 27ºC ambient, not junction temperature, without any cooling or forced ventilation of the sensor.

Thus the actual increase in noise will be less than the 17% from 27ºC to 80ºC, as the junction temperature is already above the 27ºC for the measurement.

The noise contribution is mainly from Thermal noise, thus it is proportional to Temperature in Kelvin. However, a part of the noise comes also from power supply noise, usually the power supply noise will also increase with temperature, again following the same physical law's. A cross section of the package is shown in the sensor specification.





However, we think we have no customer that actively cools the sensor. We would estimate that you will only achieve a marginal improvement by active cooling of the sensors compared to cooling over heat conduction by thermally coupling the Invar to a heat sink. Especially considering the high effort to provide active cooling, we would only consider this if your application requires extremely long exposure times and you would be limited by dark current integration. Furthermore, for exposure times in the range of hundreds of milli-seconds or even seconds, Dragster is maybe not the ideal sensor. It's strength is in high line rates, thus exposure times typically is less than 1 ms.

 

Thermal resistance dragster:

Material (Invar) : FeNi42 = CTE 5.3 10e-6 /K

Thermal conductivity: K =15.1 W/m/K

Thickness of Invar: 2e-3
Length (center to screw hole) 31.75e-3
Width of plate 25e-3

Cross section: 50e-6m2

Conductivity plate: 0.755e-3 Wm/K, 31mm length -> 0.023 W/K

-> center to screw hole: 42K/W
-> assumign 4 screw holes: 10.5K/W

********************************************************************


Vertical thermal resistance to cooling body with 2mm x 63mm contact area.

Cross Section of Invar (2mm x 63mm) 378e-6m

conductivity vertical plate: 0.0057 Wm/K

conductivity vertical 2mm 2.85 W/K

-> Thermal resistance vertical through 2mm Invar 0.35K/W

**********************************************************************

 

Thermal resistance 100um air gap:


Thermal conductivity Air : K
0.024 W/m/K

conductivity air gap under die: 9e-6 Wm/K
conductivity 100um air gap under die: 0.09W/K

-> thermal resistance 100um air gap 11K/W

**********************************************************************


Thermal resistance 200um thick glue (assuming low density polymere)

K = 0.16

-> thermal resistance 200um glue 3.3K/W

********************************************************************

Thermal paste (silver based) K = 8

-> 200um -> 0.06K/W

********************************************************************

Regarding the LCC version ,  we have the following :

The cross section is 3.95 e-4 m2 :

**********************************************************************

Thermal Conductivity of Tg 135 neglecting the vias contribution : 0,49 W/m/K

Thickness of Invar: 1.5e-3 m

conductivity vertical : 0.49 x 3.95e-4 = 0.193e-3 Wm/K

conductivity vertical 1.55 mm:   0.1245 W/K

-> thermal resistance :  8 K/W

2.2. How to assemble Dragster in a camera?

In this document there is a mounting concept for Awaiba's Dragster linescan sensor in a camera. Although we use a non conductive glue between INVAR and sensor due to our fabrication process, in some cases there is conductivity between ground and INVAR. You should isolate electrically the INVAR as well. 

3. Prices and volume

3.1. Price and volume request

Please always send quotation requests to info@awaiba.com or sales@awaiba.com

Awaiba has minimum quantities of all variations Dragster of all variations available for imediate delivery, please get in contact with our team.

3.2. What is the minimum order quantity?

Awaiba is commited to customer satisfaction so we support any volume customer needs, from 1 single sample to 1 million!

Of course availability must be checked in advance.

4. EvalBoard

4.1. What is it? Do I really need one?

To evaluate all Dragster features the customer can confortably use our Dragster Evaluation Board. Is a simple system that provides means to configurate the sensor registers and outputs data via Camera Link. Is also a quite flexible system which firmware can be adapted to customer needs.

As a new customer do you really nead one? It depends on your flexibility to create and adapt electronics. If you are used to image sensor testing you maybe have your own "Evaluation board" or "Image processing Unit" and you need only to design an easy connector adapter to our sensor and your electronic evaluation board.

If don't include yourself in this category, Awaiba recommends to acquire one because of the following advantages:

 - pre tested and fully functional state machine with all control signals created

 - possibility to test several sensors (all resolutions, all pixel sizes) with only one piece of hardware

 - full remote customer support ( also on site, if required)

 - stable system that Awaiba develloped and knows completelly and that as constantly beeing improved

 - easy to debug and easy to re-configure according to customer needs

 - compatible will all commercial grabbers on the market

 - communication cable and power supply included. Camera link cables not included

 

If you don't have any electronics and you don't know what do you need to test our sensor, here is a short list:

 - Complete PC (or laptop) with digital frame grabber with Camera Link input - we recommend the customer to use/buy a grabber with at least one Camera Link Full interface - no special request on the computer but the faster the better!

 - Evaluation Board - communication cable and power supplu included

 - At least one Camera Link cable. The Evaluation board supports up to 4 Camera Link outputs and the more outputs that can be connected the faster the image sensor can run and output the full image

 - Lens Holder can also be provided if requested

 - No lens or image target is included

 

 

4.2. Evaluation Board firmware - what do I get?

One of the big advantages of Awaiba's Evaluation Board is that is highly configurable. This means Awaiba can adjust it's firmware so that it achieves the performance expected.

Dragster linescan sensors are known to have a huge data amount that has to be handled. This amount of data is proportional to the resolution of the sensor and operating speed and has to be matched to the Camera Link interface used and to the grabber limitations. 

Customers are free to get in contact with Awaiba devellopment team (via support@awaiba.com) and request for a suitable firmware. Most cases is something Awaiba already develloped and tested and it's easy to include in the Evaluation Board prior to shipping. 

By default we have an FPGA code that reads out all sensors over 1 Camera Link, base connector, capable of operating the sensor with maximal line rate however discarding lines if the data rate coming from the sensor exceeds the data rate that can be transmitted over one base Camera Link connection. Alternatively we have specific FPGA codes that can support the sensors with full data transmission using up to 4 Camera Link base or 2 Camera Link medium configurations.

4.3. Is the image processed in the Evaluation Board?

Awaiba's Evaluation board does not make any kind of digital correction to the image. The Evaluation Board is only a highly flexible piece of hardware that transfers data to the computer via Camera Link and that configures all sensor registers. There are no images processed although that could be done. 

4.4. What Frame Grabber do I need?

Although not tested with all grabbers available, Awaiba guarantees that the Eval Board works with commercial available grabbers that support Camera Link standard. 

Awaiba has experience with the following grabbers and can distribute configuration files:

 - Silicon Software Me4 Full x1

 - BitFlow Karbon

 - Matrox Radient

4.5. I have an Evaluation Board. How to get started?

Once all hardware is installed, cables are connected and the sensor is plugged please follow these guides:

- be familliar with your grabber software and setting options and configure it according the firmware you got in the Evaluation Board

- be familliar with Dragster architecture, readout modes and register parameters

4.6. I can't grab images or they are corrupted, what to do?

All Evaluation Boards are configured and tested prior to shipping, so it's guaranteed that the system works with the sensor required.

Before getting in contact with Awaiba support team, please check the following points:

 - use the stock power supply or, if possible, a laboratory power supply with 12V output and at least 2A current limitation

 - check all cable connectivity and if the Green LED in the Power board is "ON".

 - check the grabber configuration options according to sensor readout methods

 

In case the problem is found in the sensor or with the power board or in case the Eval Board is malfunction, please get in contact with Awaiba's support team.

4.7. Evaluation Board CameraLink Configurations

The Awiaba Evaluation Board has 4 Camera Link Ports that can be configured as shown below :

5. Sensor

5.1. LVAL stuck at High or Low

Let me mention that we have observed an issue, where LVAL remained "stuck high" in some conditions, when the sensor is running at 85MHz or 100MHz readout clock and high die temperature.

Those issues had been solved with increasing the duty cycle of the main clock to be 55% or 60% high.

5.2. Characterization Settings

  • <Address>1</Address>
  • <DefaultValue>169</DefaultValue>
  • <Address>2</Address>
  • <DefaultValue>18</DefaultValue>
  • <Address>3</Address>
  • <DefaultValue>26</DefaultValue>
  • <Address>4</Address>
  • <DefaultValue>128</DefaultValue>
  • <Address>5</Address>
  • <DefaultValue>19</DefaultValue>
  • <Address>6</Address>
  • <DefaultValue>01</DefaultValue>
  • <Address>7</Address>
  • <DefaultValue>06</DefaultValue>
  • <Address>8</Address>
  • <DefaultValue>79</DefaultValue>
  • <Address>9</Address>
  • <DefaultValue>128</DefaultValue>
  • <Address>10</Address>
  • <DefaultValue>15</DefaultValue>

5.3. Pixel 500/600 Artifact

The effect appears at around pixel 570 N_RST_CDS and pixel 603 N_INT, mainly due to signal coupling. Our recommend is to apply a PRNU and DSNU correction.  Please have a look in the attached document.

Please check this DR2K7 line profile.

 

5.4. Readout - what is the sensor output?

All Dragster sensors have the following details regarding black pixels:

- sensors with 3.5um pixel always have 64 black pixels in the first 4 taps

- sensors with 7um pixel always have 32 black pixels in the first 2 taps

These pixels are read the same way as normal active pixels and so that means the first taps have extra pixels flowing on them compared to all other taps.

 

In the attached document there is an example of pixel output once the readout phase is started. This example is for a DR4K7 but can be extended to other resolutions with other amount of taps.

5.5. How to calibrate it?

The final goal of calibration is to get an uniform image with unnoticeable artifacts related to gain or offset mismatches. To achieve a good calibration is mandatory to tune the ADC offset and gain between segments and that can be made in at least i2 necessary steps. One calibration step without light and another with the light, preferentially green (~ 520nm) at about 80% of saturation.

On XK3.5 sensor versions  calibration should be between Even and Odd Pixels 

 

Odd Pixels : SPI Block CD1 CD2  GH1 GH2

Even Pixels :  SPI Block AB1 AB2  EF1 EF2

 

Colored versions calibration should be done between the Green Colored channels :

Green Channel :

  • Tap D1 , D2 , H1 , H2 ;
  • Tap A1 , A2 , E1 , E2  ;

 

Black level calibration:


The goal is to get both segment histograms right  one over the other.





Prevent any light of getting to the sensor by switching off the light and placing a black cap over it. Make a "Send All" of the default settings with DragsterComm. First we'll make a rough tuning so that all segments show a histogram over 0. Once this is achieved a fine tuning of the individual black levels has to be made so all histograms overlap and look like a single one. Activate the image histogram. If it is clamped to zero, decrease the black level in "All segments" and send this register. If both segments are far away from zero, increase the black level until one of the histograms start getting close to 0.

After this the histogram should look like this :

 

 

 

ADC Gain calibration :


To calibrate the gain the best is to use a very stable light source with the possibility to regulate the light energy. LEDs are recommended. Use the maximum light energy available or tune it so that the histogram is between 3000DN and 3500DN Set the sensor to the operating point, especially the gain programmed on the sensor to the desired setting in your application. Tune the analogue gain on/off ( bit 5 Control 2) and theinverse ADC gain register.


The histogram should look like this:

 

 

Tune the Inverse ADC gain of the individual segments. Set again a value ±10 of what was set in the Inverse ADC gain of All segments so that we can identify
which segment to tune. Once the histograms are one over the other the calibration is done and the sensor shows a good response over all range, just with the remark that the saturation level is not 4096 any more. If the on chip offset subtraction is activated, the ADC range is smaller . The "End of range" register should be increased so that the ADC covers again a 12
bit range. In the Illustration 1 it's shown that the black level was set around 100DN so this is the value that is recommended to be increased in the "End of range" register.

 

The guide attached contains the basics for the calibration process.

5.6. Can a lower pixel depth be used?

Dragster sensor family has a true 12 bit per pixel ADC that can be used in lower bit configuration. The several possibilites are explained in this document.

5.7. Which Frame Rates can I get ?

We recommend that Line Period must in any case be longer than the maximum of:

 

  • ADC time (controlled by sensor (register 0x09 * 32)/M_Clock ) +1us ;

  • Integration Time +2us ;

  • LVAL period (Readout Time = 1040 clks) + 8 + 4 ;

 

The following table states all relevant timing conditions , depending on the operating frequency and ADC resolution :

Resolution Frequency [MHz] ADC Time* [us] +1us Readout Time [us] Min Line Period [us] MAX Integration Time* [us] FPS
12 Bit 100 41,96 10,52 41,96 39,96 23832
80 52,2 13,15 52,2 50,2 19157
40 103,4 26,3 103,4 101,4 9671
20 205,8 52,6 205,8 203,8 4859
10 Bit 100 11,24 10,52 11,24 9,24 88968
80 13,8 13,15 13,8 11,8 72464
40 26,6 26,3 26,6 24,6 37594
20 52,2 52,6 52,6 50,6 19011
8 Bit 100 3,56 10,52 10,52 8,52 95057
80 4,2 13,15 13,15 11,15 76046
40 7,4 26,3 26,3 24,3 38023
20 13,8 52,6 52,6 50,6 19011

By activating the Companding Mode the ADC conversion time is significantly reduced while no information is lost .  ADC thresholds  recommended in this  configuration are the following :

  Threshold 1 Threshold 2 Threshold 3 ADC END RANGE ADC Main Clk Times
Reg Set HEX 1 4 10 7F 620

Resulting then in the following timings:

Resolution Frequency [MHz] ADC Time* [us] +1us Readout Time [us] Min Line Period [us] MAX Integration Time* [us] FPS
12 Bit 100 7,2 10,52 10,52 8,52 95057
80 8,75 13,15 13,15 11,15 76046
40 16,5 26,3 26,3 24,3 38023
20 32 52,6 52,6 50,6 19011

5.8. LCC package alignment

The LCC package holes were not designed for alignment they don't have any specific tolerance . The most suitable way to align it , is through the DIE reference marks or the sensor pixel line.

 

5.9. Signals Latching

Pixel Clock is inverted relatively to Main Clock . So if using Main Clock to latch data please have this in consideration .  

Control Signals :

RST_CVC - FALLING EDGE PIXEL CLOCK

RST_CDS - FALLING EDGE PIXEL CLOCK

SAMPLE - FALLING EDGE PIXEL CLOCK

LOAD_PULSE - FALLING EDGE PIXEL CLOCK

N_RESET- Goes OFF on FALLING EDGE MCLK

Output Signals :

L_VAL - RISING EDGE PIXEL CLOCK

END_ADC - FALLING EDGE PIXEL CLOCK

BIT_XX - RISING EDGE PIXEL CLOCK

 

 

5.10. Temperature Specs

 

  min   max
  Tj Junction Temperature 0 27 80 ºC
Ta Ambient Temperature    

 

depends 

on the housing

Operating Temperature Range 0 27 80 ºC  
Operating Humidity 0   99 %  
Storage Temperature -40 20 80 ºC  

5.11. Regx01 bit 1

this is a bug in the sensor although internally the bit is self cleared , otherwise whatever you wrote to register 1 you would get an update of all registers . But in fact the shadow regster that contains the actual value of the register is not cleared so you always get this bit to one when in fact it is not .

5.12. 1V Swing explanation

The reason to tune the ADC Gain, in respect of the ADC end of range or companding mode configuration. 

5.13. How to distinguish between a color RGB and BW sensor ?

5.14. Blooming


Blooming resistance of Dragster sensors is extremely high.

Indeed, each pixel is isolated form the other with a specific charge drain, thus blooming from line one to line 2 can not be observed.

If a large part of the sensor is overexposed, one can see a small effect to the un or low exposed pixels of this very same line, however this effect is not due to pixel Blooming , but due to an electrical crosstalk on the ADC reference. (this is a known effect we doubted "horizontal crosstalk".

Provided that the reset time (time reset_CVC =1) and the minimum sampling time (time SAMPLE = 1) are respected as specified, there is no image lag, or blooming from line N to line N-1 or N+1.

5.15. Power Consumption

5.16. Glass cleaning instructions

Please find attached the document related to glass cleaning instructions.