4 Flash Storage Space Requirements

The SPI flash device in the DGP system needs to be fast enough to support loading content directly from flash to the DMD. Due to the high bandwidth requirements, an octal-SPI flash device compatible with the JEDEC xSPI protocol was chosen for this application. In the DGP application all of the content for a dynamic ground projection system is pre-processed into DMD native format and compressed using RLE (run length encoding) on a PC and then loaded onto the SPI flash. Ideally an arbitrarily large flash part would be used but to optimize system bill of materials cost it's important to minimize the amount of flash storage in the system.

To estimate the amount of storage space required for each frame of video the required parameters are:

• DMD Array Width, W
• DMD Array Height, H
• Number of bit-planes, BP
• Compression Ratio, C

The storage space required (in bits) can be calculated by:

For the DLP3021-Q1 DLP Composer project and FPGA configuration in the DLP3021-Q1 product folder on TI.com, the following values can be used

• W = 608
• H = 684
• BP = 20
• C = 2.1 (estimated, depending on content see Table 4-1)

The width and height is set by the number of mirrors on the DMD. For the dynamic ground application, every mirror location must be loaded with data. The number of bit-planes is not fixed and can be adjusted depending on the system requirements. For example, in the case of a single illumination source, white light only module, you may only need to use six to eight total bit-planes. The RLE compression ratio is dependent on content. Images with fine detail such as live video may not compress as well as image with large solid fields of color such as animations. The compression ratio of 2.1:1 is used as a rough estimate of what you may expect to see across various types of content, although the worst case would always be slightly less than 1:1. Table 4-1 shows a few examples of how different image content compresses. The "1x1 checkerboard" pattern is an image where every alternating pixel is black or white. This is a type of worst case image where the content is not compressible and you still have the overhead from the compression algorithm causing the resulting compressed frame to be slightly larger than the uncompressed frame. This is for example only and is not typical of real-world images.

When the approximate storage size of each frame of video is determined, you can multiply it by the typical 25Hz frame rate to estimate the amount of storage required for your content. With the parameters above, it is possible to store approximately twenty seconds of full-color content in a 2Gb flash device. It is possible to significantly increase the amount of content stored by carefully choosing the type of content and choosing the correct illumination (RGB or single color) to match the type of content being displayed.