Higher Resolution Codecs and Striped SSDs

Printable View

13-03-2010, 11:21 PM
SourceChild

Higher Resolution Codecs and Striped SSDs

I am recently faced with clients preferring to use more playback devices namely Doremis and Turbos and less Catalysts systems.

I agree when it comes to playing back Video rolls which contain audio. However, it's a waste for content and a severe limitation since the ability to provide effects, stack layers, and distort the images are not available.

What I am considering is configuring special HD Catalyst Servers with up to 6 striped SSDs so that I can run up to 4 layers of HD (or higher) with better quality.

I am interested in feedback and discussion regarding this.
17-03-2010, 04:40 AM
ajmaudio

what is your hesitation with audio?

I don't see why striping shouldnt provide some serious gains in capability on the HD front. The only downside off hand is a reduction in MTBF but I like my odds in that area with ssd stuff much more than normal drives. Perhaps in the not too distant future the new up and coming sata standard will help us take better advantage of our SSD drives as now it seems that the sata bus can be saturated before some of the faster SSD drives run out of steam.
17-03-2010, 12:31 PM
samsc

you need faster mac.
It's not data rate it's CPU limited beyond a threshold
and that threshold is exceeded already depending on movie data rate

2009 macs have been reported to do 5 HD layers with single ssd

striping is a rotating hard disc thing nothing to do with ssd

suggest you try putting movies on different ssd discs. As well as raid
doubt it will make much difference

Quote:

Originally Posted by SourceChild

I am recently faced with clients preferring to use more playback devices namely Doremis and Turbos and less Catalysts systems.

I agree when it comes to playing back Video rolls which contain audio. However, it's a waste for content and a severe limitation since the ability to provide effects, stack layers, and distort the images are not available.

What I am considering is configuring special HD Catalyst Servers with up to 6 striped SSDs so that I can run up to 4 layers of HD (or higher) with better quality.

I am interested in feedback and discussion regarding this.
18-03-2010, 07:45 PM
samsc

Quote:

Originally Posted by SourceChild

I am recently faced with clients preferring to use more playback devices namely Doremis and Turbos and less Catalysts systems.

I agree when it comes to playing back Video rolls which contain audio. However, it's a waste for content and a severe limitation since the ability to provide effects, stack layers, and distort the images are not available.

What I am considering is configuring special HD Catalyst Servers with up to 6 striped SSDs so that I can run up to 4 layers of HD (or higher) with better quality.

I am interested in feedback and discussion regarding this.

did you try these ssd's?

http://www.crucial.com/uk/store/part...DDAC128MAG-1G1

the C300 series is being touted as the fastest ....
07-04-2010, 06:51 AM
SourceChild

A point is that larger slower SSDs are cheaper than smaller faster ones.

I mention SSD raid For capacity reasons as opposed to speed. The speed part I can make back up by using a PCI raid controller. I am referring to Raid controllers cards with on board processing.

I'll say I'm jumping the gun a little based on rumors of the up and coming release of the new Mac Pros which are supposed to boast Xeon 5600 6 core chips, USB 3.0, and SATA r3.

As far as audio though, I have never seen a situation where an HD video file didn't drop frames at some point coming from a software based media server. Yes, it's negligible and most people will never notice but it's still not satisfactory for broadcast.
07-04-2010, 07:45 AM
samsc

I dont believe there is sufficient performance difference between 'cheaper' and 'more expensive SSD' to make any difference- 3 years ago there was - these days I havent seen anything in any 3rd party performance testing to indicate it would make a difference-
Run test G with different drives and email me-

I cant imagine any PCI raid controller being at all useful with SSD -
these things are designed to deal with things that simply do not occur with SSD- There are no access times with SSD
PCI raid controllers are designed to deal with access time latencies

----

I count drop frames in all my proper testing my performance graphs count drop frame - my testing HUD counts drop frames- i suggest you take a look at performance G - i have done very long term testing-

And run it to see where you are getting dropped frames-

There are a number of Operating system things ( the OS can go away and do something for a while - some house keeping ) - this is minimised by turning off everything you can like spotlight - on all discs

synchronisation things- VGA/DVI display rates not being the same as broadcast -

And performance things - with slower discs- performance is usually the lowest factor these days

Quote:

Originally Posted by SourceChild

A point is that larger slower SSDs are cheaper than smaller faster ones.

I mention SSD raid For capacity reasons as opposed to speed. The speed part I can make back up by using a PCI raid controller. I am referring to Raid controllers cards with on board processing.

I'll say I'm jumping the gun a little based on rumors of the up and coming release of the new Mac Pros which are supposed to boast Xeon 5600 6 core chips, USB 3.0, and SATA r3.

As far as audio though, I have never seen a situation where an HD video file didn't drop frames at some point coming from a software based media server. Yes, it's negligible and most people will never notice but it's still not satisfactory for broadcast.
07-04-2010, 10:57 AM
SourceChild

I'm getting some new gear when I get back in May. I plan on doing a great deal of testing.

Quote:

Originally Posted by samsc

Run test G with different drives and email me

I currently only have 5 different types of SSD drives.

Quote:

Originally Posted by samsc

I cant imagine any PCI raid controller being at all useful with SSD

My hope is that offloading the CPU etc will allow for smaller capture buffers in the future, hopefully getting to less than two frame latency.

Quote:

Originally Posted by samsc

i have done very long term testing

I'm hoping to do a big push with some of the broadcast clients. Believe me, I'll be testing thoroughly as well. I hope to collaborate with you on this.

Quote:

Originally Posted by samsc

And run it to see where you are getting dropped frames-...
synchronisation things- VGA/DVI display rates not being the same as broadcast

Converting DVI to HD-SDI is always a culprit. Maybe someday there will be reasonably priced HD-SDI video graphics cards for OS X.

As far as operating system, I have scripts to turn dozens of things off but I've found myself shot in the foot for their functionality later.
07-04-2010, 12:02 PM
samsc

what you mean 2 frame latency here?

what you referring to?

---

dvi doesnt convert nicely to hdsdi and vice versa... unless you can program dvi that work exactly at hdsdi frame rates...

Quote:

Originally Posted by SourceChild

I'm getting some new gear when I get back in May. I plan on doing a great deal of testing.

I currently only have 5 different types of SSD drives.

My hope is that offloading the CPU etc will allow for smaller capture buffers in the future, hopefully getting to less than two frame latency.

I'm hoping to do a big push with some of the broadcast clients. Believe me, I'll be testing thoroughly as well. I hope to collaborate with you on this.

Converting DVI to HD-SDI is always a culprit. Maybe someday there will be reasonably priced HD-SDI video graphics cards for OS X.

As far as operating system, I have scripts to turn dozens of things off but I've found myself shot in the foot for their functionality later.
21-04-2010, 07:11 AM
ajmaudio

What kind of issues are you guys having converting from dvi to hdsdi? I think I will more than likely be doing this in the not so distant future and since you mentioned issues I thought I would ask. On a side note I have noticed that matrox is releasing some new sdi output cards for mac but I dont know if they support openGl etc.

Here is a link: http://www.matrox.com/video/en/press..._multi_io_mac/

If you read the article it would seem that these devices are targeted toward what we do with catalyst. You guys have any input on these?

There are also input cards and cards that do i/o.... would be amazing if these were actually suitable? your thoughts Richard?

BTW... these fit in an xserve supposedly
23-05-2010, 06:23 PM
SourceChild

Quote:

Originally Posted by ajmaudio

What kind of issues are you guys having converting from dvi to hdsdi?

The biggest issue is having consistent color. Using a scope to test the HD-SDI out from a converted mac shows tons of illegal colors. Also, matching the color is a pain. Sync is an issue as well as frame delay. Since a mac is 60Hz, the multiple is 30fps which is not 59.94 or 29.97 so every 10 seconds is a subtle frame skip. It works but if your director takes Cat as a direct broadcast feed, then the producers won't be happy.

Quote:

Originally Posted by ajmaudio

....matrox is releasing some new sdi output cards for mac

I'll probably get and test the first HD-SDI solution that does 59.94 with broadcast color as an output option.

Still though, for all the companies using HD-SDI as their primary pipeline to the projection sources, a video card that works would be great.

Quote:

Originally Posted by ajmaudio

BTW... these fit in an xserve supposedly

A card that fits in an xServe is great compared to trying to shoehorn a card that doesn't and having apple reject support because of it.
23-05-2010, 09:51 PM
samsc

normal hdsdi - is a lower quality output signal than rgb 4:4:4 from a dvi or vga monitor-

describing an RGB monitor as displaying illegal colour in YUV colour space is
the wrong way round-

normal HDSDI cannot display a full resolution RGB 4:4:4 signal - whatever the refresh rate-
-

HDSDI is YUV encoded and 4:2:2 - same old broadcast bandwidth limiting exercise-
the conversion from RGB to YUV is never lossless - hdsdi is downsampled- its worse than DVI/VGA not better

DVI/vga is higher quality than hdsdi...

broadcast colour is lower quality than RGB 444- i think you have been taking the wrong broadcast based medicine...

http://en.wikipedia.org/wiki/Serial_digital_interface

Quote:

For all serial digital interfaces (excluding the obsolete composite encodings), the native color encoding is 4:2:2 YCbCr format. The luminance channel (Y) is encoded at full bandwidth (13.5 MHz in 270 Mbit/s SD, ~75 MHz in HD), and the two chrominance channels (Cb and Cr) are subsampled horizontally, and encoded at half bandwidth (6.75 MHz or 37.5 MHz). The Y, Cr, and Cb samples are co-sited (acquired at the same instance in time), and the Y' sample is acquired at the time halfway between two adjacent Y samples.

the use of 29.97 and all the insane resulting confusing is just so confusing to everyone. its not an advantage of hdsdi - its an albatross.

Quote:

Originally Posted by SourceChild

The biggest issue is having consistent color. Using a scope to test the HD-SDI out from a converted mac shows tons of illegal colors. Also, matching the color is a pain. Sync is an issue as well as frame delay. Since a mac is 60Hz, the multiple is 30fps which is not 59.94 or 29.97 so every 10 seconds is a subtle frame skip. It works but if your director takes Cat as a direct broadcast feed, then the producers won't be happy.

I'll probably get and test the first HD-SDI solution that does 59.94 with broadcast color as an output option.

Still though, for all the companies using HD-SDI as their primary pipeline to the projection sources, a video card that works would be great.

A card that fits in an xServe is great compared to trying to shoehorn a card that doesn't and having apple reject support because of it.
24-05-2010, 12:26 AM
SourceChild

Quote:

Originally Posted by samsc

normal hdsdi - is a lower quality output signal than rgb 4:4:4 from a dvi or vga monitor-

Yep, it sure is. Richard is right.
Sadly, most broadcast systems use HDSDI as their primary interface.

Quote:

Originally Posted by samsc

describing an RGB monitor as displaying illegal colour in YUV colour space is the wrong way round-

Honestly, there's no problem until there is a need to have color continuity between a Playback deck and a media server.

Believe me, there's nothing worse than having to match footage and having a video engineer arguing wrong points. Then you get the content creators who complain because their creations look different from one medium to another when they themselves don't even understand Or apply Color Space.

However, theses are all issues which come into play when forced to convert to HDSDI.

Quote:

Originally Posted by samsc

normal HDSDI cannot display a full resolution RGB 4:4:4 signal - whatever the refresh rate-

But then again, neither can most codecs. It's a loosing battle.

Quote:

Originally Posted by samsc

DVI/vga is higher quality than hdsdi...

broadcast colour is lower quality than RGB 444- i think you have been taking the wrong broadcast based medicine...

It's all propaganda fed to media server people by broadcast engineers to justify the need to keep using interlaced.:mad:
THEIR ALL OUT TO GET US!!!! :eek: RUN! :eek:

Quote:

Originally Posted by samsc

the use of 29.97 and all the insane resulting confusing is just so confusing to everyone. its not an advantage of hdsdi - its an albatross.

The biggest Albatross of all is a $150k HD video projector that has a fiber input and the video company demands to use HDSDI.
24-05-2010, 02:17 AM
ajmaudio

Quote:

Originally Posted by SourceChild

The biggest issue is having consistent color. Using a scope to test the HD-SDI out from a converted mac shows tons of illegal colors. Also, matching the color is a pain. Sync is an issue as well as frame delay. Since a mac is 60Hz, the multiple is 30fps which is not 59.94 or 29.97 so every 10 seconds is a subtle frame skip. It works but if your director takes Cat as a direct broadcast feed, then the producers won't be happy.

I'll probably get and test the first HD-SDI solution that does 59.94 with broadcast color as an output option.

Still though, for all the companies using HD-SDI as their primary pipeline to the projection sources, a video card that works would be great.

A card that fits in an xServe is great compared to trying to shoehorn a card that doesn't and having apple reject support because of it.

I havent heard back from the sales engineer I tried to contact regarding those cards... you know anything about them? Or any other solution for that matter. That sync issue you speak of is killing me as well. A broadcast friendly output from catalyst needs to happen.
24-05-2010, 05:39 AM
SourceChild

Quote:

Originally Posted by ajmaudio

That sync issue you speak of is killing me as well. A broadcast friendly output from catalyst needs to happen.

I don't know yet. Too many things on my plate to jump at right away.

Honestly, use an ImagePro or a TVOne Scaler. Richard said it best. The whole thing is an albatross. If you can stay DVI or VGA then do so. Fibre extension is a wonderful thing.

If you Must integrate into broadcast, use an ImagePro.
24-05-2010, 08:19 AM
samsc

Quote:

Originally Posted by ajmaudio

I havent heard back from the sales engineer I tried to contact regarding those cards... you know anything about them? Or any other solution for that matter. That sync issue you speak of is killing me as well. A broadcast friendly output from catalyst needs to happen.

frame rates depend entirely on whether you can set the edid to work at hdsdi frame rates.

that depends on your scan convertor-

edid tells computer what it wants to output

---

All colour space conversions discussions are moot/pointless/endless and affect all systems across the board - in every industry with a colour output device-

printer- video deck - monitor - lcd -projector -art -its been this way since the first time anyone tried to make a colour image... even painters have colour space problems

how that conversion is done is dependant on the box you are using to do the conversion from dvi/vga to hdsdi-

There is no universal/lossless colour space it affects everyone in broadcast print or post -production.

doesnt depend on quality or cost.

http://en.wikipedia.org/wiki/Color_space

http://dx.sheridan.com/advisor/cmyk_color.html

and in video world there is the YUV RGB conversion problem as well and colour sub-sampling-

http://en.wikipedia.org/wiki/YUV

----

ultimately colour is perceptual - its not technical problem.

it kindof really doesnt matter because the eye will do its own thing anyway-

vision is all about context -
there is no absolute colour in the eye.
24-05-2010, 08:26 AM
samsc

look at this optical illusion

http://www.michaelbach.de/ot/lum_ade...dow/index.html

the eye cant even determine the tonal value of 2 squares accurately in the same image....

the 2 squares look completely different tonally - but are the same...
24-05-2010, 10:54 AM
samsc

Quote:

Originally Posted by ajmaudio

I havent heard back from the sales engineer I tried to contact regarding those cards... you know anything about them? Or any other solution for that matter. That sync issue you speak of is killing me as well. A broadcast friendly output from catalyst needs to happen.

you need to check out whether the dvi parrot can do hdsdi refresh rates from edid - if thats what you need to do - sync is not a graphics card thing - its a monitor -> graphics card - they talk to each other

http://www.tmb.com/products/parrot/
http://www.tmb.com/products/parrot/r...rot-A4-web.pdf
05-06-2010, 02:57 AM
SourceChild

Wow That's cool HEHE:D

Quote:

Originally Posted by samsc

look at this optical illusion

http://www.michaelbach.de/ot/lum_ade...dow/index.html

the eye cant even determine the tonal value of 2 squares accurately in the same image....

the 2 squares look completely different tonally - but are the same...
06-06-2010, 02:22 AM
ajmaudio

Quote:

Originally Posted by samsc

you need to check out whether the dvi parrot can do hdsdi refresh rates from edid - if thats what you need to do - sync is not a graphics card thing - its a monitor -> graphics card - they talk to each other

http://www.tmb.com/products/parrot/
http://www.tmb.com/products/parrot/r...rot-A4-web.pdf

I am suprised you havent checked that out allready! I looked into it briefly some months ago. I dont recall the reason, but I recall coming to the conclusion that it was not a probable fix. I dont think it lets you get super specific on refresh rate maybe.. ie 59.94 instead of 60. Since the problem seems very sensitive to even the smallest change in this area I figured it was not a solid solution I think? Something like that.. like I said its been a good bit. I do understand the graphics card>monitor thing. I have not found any good solution to clearing up the hiccups when taking video in from AS cards and spitting it back out.... similar to the problem posted here when converting to sdi... Always get some level of glitching... some more offensive than others but none really acceptable.