AnsweredAssumed Answered

ADV212 - custom-specific HVF mode encode/decode

Question asked by gabi.furman Employee on Sep 10, 2017
Latest reply on Sep 18, 2017 by JeyasudhaMuthuPerumal

custom-specific HVF mode encode (progressive scan, CROP mode is off) :

 

1. ADV212 JPEG2000 Programming Guide Rev 2.0 on  p. 31 says: "H and V should not be asserted simultaneously". However Figure 25 on p. 33 shows  HSYNC and VSYNC asserted simultaneously. Whom should be taken  ?

2. An example on p.33 shows that XTOT register is programmed to exact number of samples between active edges of HSYNC (1600 in their example). It does not work for them.

they get good results when XTOT is programmed for at least 1 *more* sample that are actually present between active edges of HSYNC. Is it normal?

3. P. 31 says "Minimum of six lines of vertical blanking". In practice, as long as there is sufficiently long interval between frames (exact duration depends on the setting of encoder and on picture itself) it looks like 2 blank lines are sufficient: 1 before active video and 1 after active video. What do you think about it ?

4. P. 31 says " Minimum of 16 pixels of horizontal blanking". In practice it looks like 3 blank pixels are sufficient. May be, even 2, they didn't check it yet.  What do you think about it ?

 

custom HVF decode :

More specifically, they are trying to implement custom-specific HVF master decode  (progressive scan, CROP mode is off) with compressed stream fed to ADV212 via JDATA bus.

Unfortunately, ADV212 JPEG2000 Programming Guide contains no examples of such mode.

They attempted to figure it out by themselves by modifying example in Figure. 27 (p. 35), but so for results are not good.

Specifically, after they sent a single compressed frame over JDATA, ADV212 starts to generate an endless stream of output frames over VDATA. One frame looks like what they want, but the rest is garbage.

Can you, please,  give them an example of initialization of custom-specific JDATA+VDATA(HVF master) decode?

Their application is *not* a video, but high resolution still imagery, so they have to be able to establish exact one-to-one correspondence between incoming and outgoing tiles.

With exception of a single anonymous application note, all ADI documents suggest that applications like theirs served best by RAW mode.

Are you sure that doing it with custom-specific HVF master mode is at all possible?

 

Appreciate your prompt reply, as this issue is very urgent.

Outcomes