How much can I compress the image before it's too bad.
That is sort of a trick question. What is "acceptable" is completely subjective and the quality versus compression ratio is also highly dependent on the content. Images with lots of high frequency content will get "bad" at smaller compression ratios than images with lots of low frequency content.
The ADV2x2 has two different rate control methods that the customer can set:
1) Target size
In this mode, the user sets the maximum number of bytes that will be output per compressed image. This is only a cap-- the part will produce a codestream no more than 5% over the target size but it could be much much less if the image is very simple.
2) Target quality
The user sets a lambda value which defines the quality. The lower the number the higher the quality but it's just a number in an equation rather than some objective visual quality scale. The part will then produce a codestream where the distortion metrics will have a lambda no worse than the target set. How this is calculated is defined in the JPEG2000 standard. This produces codesteams that can vary wildly in size depending on the content.
Can you comment on the compression ratio of the ADV212 in regards to standard PC video SVGA 800x600 60f/s quality? And again, commenting on fast changing video vs. say a menu screen where the video output is not changing very much.
Thank you in advance for your help.
A couple comments:
1) fast moving versus static makes no difference at all. There is no temporal compression-- it's frame by frame so there is no motion.
2) There really isn't anything to say related to a particular format for FPS-- FPS means nothing to quality as in 1). Quality is in the eye of the beholder-- it's frequency content that effects the visual quality at a given compression ratio. The resolution doesn't really mean anything other than a larger tile is obviously more bits to compress.
Thank you. I agree that the “quality” of the compressed video would be subject to the “you have to see it to believe it” experience of the end user.
My next question is not about quality, but about quantity. How much compression can it get from the ADV212? Example:
SVGA =~ 691Mb/s
I would like to compress that SVGA video stream down to say
Compressed video1 =~ 400Mb/s
Compressed video2 =~ 40Mb/s
For a compression ratio of
Compression ratio1 =~ 40% (400/691)
Compression ratio2 =~ 94% (40/691)
Can the ADV212 compress the images to these types of levels? What levels can the ADV212 compress to?
Thank you in advance for your help,
As I mentioned in the initial answer, the target size mode lets you arbitrarily set whatever bit rate you want. It's unrelated to the input size-- you are setting the amount of bytes per frame in the output. There is a practical limit to how low you can set it because of the header size... and a practical limit on the high end based on bus bandwidth (we spec it at 200Mbps) but that's it.
40Mb/s is around 87k bytes per frame which would mean an RCVAL of around 0x15000 as an example.
You could set RCVAL to 1k bytes which would be about 480kbps for example.
We are almost there. We are happy with the compression information of the ADV212. A few more questions yet to go:
One is whether there is a software version of the ADV212 (and it appears to be so, from the files I downloaded from the AD ftp site associated with the ADV212/202), that can be used to run on a PC, that would allow being able to observe what images look like with given levels of compression. If I’m misunderstanding what I’m seeing, that would be useful to know.
The second question is whether the ADV212 can accept full-rate SVGA (which is 691.2Mb/s). You mention 200Mb/s as an absolute limit; so is this an output limit, or is it also an input limit?
The third question, and this is also new, is whether I should be using two ADV212 chips instead of one. If I’m reading the datasheets correctly, it appears that for doing 720i (HD video) images that two chips are required, but 720i is half of 720p, and 720p is 720 lines at a 9:16 ratio, while SVGA is 600 lines at a 3:4 ratio; doing the math, that means that 720p is 921600 pixels, and SVGA is 480000 pixels, only a shade over half as many – so if 720i requires two chips, does SVGA require two chips?
Thank you in advance for your help in understanding the ADV212 and it’s development environment.
There is no software version of ADV212. There are several software JPEG2000 compression engines-- we typically use Kakadu from www.kakadusoftware.com. We also have a tool on our FTP site (linked from the ADV212 Design Support Files page) where you can indeed give it BMPs and try out different compression ratios.
On the input side... we spec in samples per second (65 MSPS is max per chip), not Mbps. SVGA is 28.8 MSamples * samples per pixel. Normally computer resolutions are RGB 4:4:4 which you would either need to run each component through sequentially in single component mode which at 60Hz would be a bit more than 1 chip could do... or use 3 chips with each one handling a component.
If you converted to YCbCr 4:2:2, then 1 ADV212 could handle SVGA 60Hz with no problem.
The 200mbps spec is compressed data on the host bus.
Good morning Dave,
What Eval Boards are currently available for the ADV212?
AD has a schematic for their “ADV212HD PCI”. However, going to the ADV212 page
selecting the “Evaluation Boards & Kits <http://www.analog.com/en/audiovideo-products/video-compression/adv212/products/EVAL-ADV212/eb.html> ” link gets me to
Are there other available boards or kits that have the ADV212, or is this the only one?
Thank you in advance,
Only the PCI board is still available, though it's currently out of stock temporarily. That board is really only useful if your application matches how it's designed though. It can't be used to test any modes or configurations other than standard 720p/1080i in on VDATA and out of HDATA in a 2-chip configuration.
Retrieving data ...