Related Articles

Sony 3ClearVid CMOS Technology Primer

Oct 23, 2006 12:59 PM, Steve Mullen


         Subscribe in NewsGator Online   Subscribe in Bloglines  

In this installment of HDV@Work, we look at the technology behind Sony's ClearVid CMOS chips, found in the new HVR-V1U camcorder. In the last installment, we looked at five aspects of CMOS technology. First, we looked at how CMOS chips function. Second, we saw how exposure control is implemented — including how the ability to simultaneously read multiple elements from a CMOS chip reduces the possibility of the “rolling shutter” artifact. Third, Correlated Double Sampling (CDS) was shown to be a common method of reducing Fixed Pattern Noise (FPN) that results from each CMOS element having its own on-chip amplifier. Fourth, we explored how extra circuits on a chip can increase latitude. (ClearVid does not use this technology.) Last, we learned that the chip element-count determines the cut-off frequency for the anti-alias low-pass filter that is what really limits image resolution.

In the illustration below, we can see that only a portion of a CMOS element actually captures light. The area of the photodiode divided by the element’s total area is called its “fill factor.” (It’s possible that when Sony claims V1’s ClearVid elements have the “same area” as the Z1’s CCD elements, they are talking about fill factor percentages.)

Most CMOS imagers today use Active Pixel Sensor technology (APS), which requires at least three transistors to implement. (These are called 3T devices.) There is no reason to believe ClearVid is any different. Sony illustrates its ClearVid technology using the diagram below.

Each CMOS element is rotated 45 degrees. Despite the rotation, the elements are “square.” Because the ClearVid chips used in the V1/FX7 have a resolution of about 1000x1000 and have a 16:9 aspect ratio, we know the idealized elements above cannot be an accurate representation.

ClearVid elements most likely look like the squat diamond shown below. The photodiode area is shown outlined in yellow. (This example has about a 33-percent fill factor. Sony claims a very high fill factor.)

In my crude diagram below, you can see how such an element can support an equal number of elements on each axis and fill a 16:9 space.

Despite the rotation of elements, elements are read out row-by-row as shown below. Elements are read out into Sony’s Enhanced Image Processing chip (EIP).

According to Sony Japan, four elements — from four columns — are read simultaneously from each ClearVid chip into each of the EIP’s three 2-million cell buffers. By outputting four elements at a time, read-out speed is increased by a factor of four, thereby eliminating, or at least significantly reducing, the rolling-shutter artifact. (Read-out speed is “X” in the last installment’s section on exposure control.)

When a row is output, each element sends its signal down its column bus where it is input to a Sample & Hold circuit that briefly stores the value. (There is a Sample & Hold for each column.) Next, a second cycle is performed on the same row. Immediately, all row elements are sent down to a second set of Sample & Hold circuits. The second set of values is measures of each element’s inherent noise. The first set of values is measures of element signal+noise values. Now all stored noise values are subtracted from all signal+noise values to yield a row of signal values. This system is called Correlated Double Sampling (CDS). Sony uses this type of CDS. The output from a ClearVid chip is an analog signal that is converted to a digital signal by a 14-bit A-to-D converter in the Digital eXtended Processor (DXP). The DXP chip uses 14-bit words for baseband signal processing — for example, for noise reduction.

The diagram below shows eight elements that lie in two rows. The elements in odd rows — labeled A, B, C, and D — are read out simultaneously. This occurs 135 times for the 540 elements in each row. All elements in Row 1 are read out before all elements in Row 2 are read out.

The elements in even rows — labeled W, X, Y, and Z — are also read out simultaneously. Again, this occurs 135 times for the 540 elements in each even row.

 

 

 

1

2

3

4

5

6

7

8

ROW 1

LINE 1

 

1A

 

1B

 

1C

 

1D

 

ROW 2

LINE 2

 

2W

 

2X

 

2Y

 

2Z

ROW 3

LINE 3

3A

 

3B

 

3C

 

3D

 

ROW 4

LINE 4

 

4W

 

4X

 

4Y

 

4Z

ROW 5

LINE 5

5A

 

5B

 

5C

 

5D

 

ROW 6

LINE 6

 

6W

 

6X

 

6Y

 

7Z

ROW 7

LINE 7

7A

 

7B

 

7C

 

7D

 

ROW 8

LINE 8

 

8W

 

8X

 

8Y

 

8Z

ROW 9

 

9A

 

9B

 

9C

 

9D

 

 

The first two rows to be read out are Rows 1 and 2. The next rows to be read out are Rows 3 and 4. The read-out process works down the chip as shown below.

 

 

 

1

2

3

4

5

6

7

8

ROW 1073

LINE 1073

 

1073A

 

1073B

 

1073C

 

1073D

 

ROW 1074

LINE 1074

 

1074W

 

1074X

 

1074Y

 

1074Z

ROW 1075

LINE 1075

1075A

 

1075B

 

1075C

 

1075D

 

ROW 1076

LINE 1076

 

1076W

 

1076X

 

1076Y

 

1076Z

ROW 1077

LINE 1077

1077A

 

1077B

 

1077C

 

1077D

 

ROW 1078

LINE 1078

 

1078W

 

1078X

 

1078Y

 

1078Z

ROW 1079

LINE 1079

1079A

 

1079B

 

1079C

 

1079D

 

ROW 1080

LINE 1080

 

1080W

 

1080X

 

1080Y

 

1080Z

ROW 1081

 

1081A

 

1081B

 

1081C

 

1081D

 

 

The last two rows that are read-out are Rows 1080 and 1081. By the end of this process, all 960 elements in each of the 1081 rows have been read out and