Behind the Scenes footage from Spoon! at DVinfo.net

Jason Rodriguez · June 30th, 2006, 02:04 PM

Go check it out:

indiefilmlive.blogspot.com

Yasser Kassana · July 1st, 2006, 09:27 AM

excellent work, glad you've made the post production a bit more efficient, I was expecting to see Rutguer Hauer! Where is he!

Also have you recieved my email?

Noah Yuan-Vogel · July 1st, 2006, 10:28 PM

Thanks for sharing all the video. Looks great and I would love to see more, longer shots of just what its like to be on set and shooting with those cameras, maybe see what setup is like.

I noticed, that you are using the xenarc ~480p 7" lcd screens (right?), are those what silicon imaging provided you with or did you get those separately? I'm just surprised because I was under the impression that they had a fancy touchscreen 7" 720p HD monitor included in their package. If you dont mind my asking, how does that work anyway? Is it always the same image running to the large director's LCD as the operators' mini LCDs (as in it has to be <1080p since the mini LCDs cant handle that resolution)? Or do they run at different resolutions? in a dual monitor setup?

Noah Yuan-Vogel · July 1st, 2006, 10:35 PM

Oh I just realized it was you, Jason, who posted this message, what is your role in the Spoon project? Are you, working for SI on their software? Sorry, I'm confused about who is doing what exactly in relation to the SI camera and Spoon, although I imagine the two teams are working pretty closely given one needs support for their cameras and the other needs a field testing and proof of the workflow for their product.

Pete Tomov · July 2nd, 2006, 06:26 AM

Looks great!I'm definitely getting one of those!
What lenses are you guys using?

Ari Presler · July 2nd, 2006, 06:54 PM

Quote:

Originally Posted by Noah Yuan-Vogel

Oh I just realized it was you, Jason, who posted this message, what is your role in the Spoon project? Are you, working for SI on their software? Sorry, I'm confused about who is doing what exactly in relation to the SI camera and Spoon, although I imagine the two teams are working pretty closely given one needs support for their cameras and the other needs a field testing and proof of the workflow for their product.

"Spoon" is the first feature being shot on Silicon (Si-1920HDVR) in South Africa. Part of the agreement with the producers to get early access was that they would post behind the scenes photos and footage during the shoot. In addition, they are beta testing new versions of software and providing feedback on operation and desirecd enhancements.

The 7" LCD monitor is a High-Brightness version of the Xenarc. We are outputing 720P to the Video Village large screen LCD and the 7" on-camera view, which scales the video to fit its native display resolution.

Jason is our lead digital cinema consultant. He has been providing our primary input on the features, operation, UI and workflow. He is intent on having us to deliver the BEST camera possible for the market and for himself...thanks Jason :-)!

As I have stated on other posts, we starting our market research on this forum. We are not only asking for input from you, the potential users/buyers, we are internalizing the input and implementing it. Please continue the input and support!

Jason Rodriguez · July 3rd, 2006, 10:26 AM

BTW, the limitation on the 720P display output is that with current CPU horsepower, we can't display two different resolutions on two different monitors. So you can do 1920x1080 *or* you can do 1280x720 but not both. If you choose the 1920x1080 approach, then the on-board Xenarc looses sync.

Since we're recording RAW, the display is really only a high-quality "preview", i.e., it's not a full-fledeged 1920x1080 signal with a high-quality demosaic . . . there's just no way to-do that in a portable platform with the processing power that's on the market right now. But we are rendering a HD signal, albeit at 1280x720 (60Hz refresh rate) so that you can use the current crop of all the HD equipment display, transmission, and conversion equipment on the market including large-screen plasmas, HD-SDI converters, downconverters, etc. to move the signal anywhere you need.

Another way of thinking of our preview screen is like watching the preview playback on a NLE computer screen display (like Final Cut's or Premiere Pro, or Avid's), not the full-resolution output from a AJA or Blackmagic card, or the EE mode on a tape deck.

Joe Carney · July 3rd, 2006, 03:26 PM

So Jason, the video being recorded from the gbEthernet connection is 1080p, just the preview is 720p? Not bad at all.

Barlow Elton · July 3rd, 2006, 03:39 PM

I'm just learning about S.I. camera so here is a basic question:

Can I get a higher than 1920x1080 @ 30 fps? Can it do 1080 60p?

Thanks for any info.

Noah Yuan-Vogel · July 3rd, 2006, 04:27 PM

How flexible are the preview output options? Is the software interface designed to scale to a variety of resolutions? You mentioned there isnt enough CPU power to do a realtime high quality 1080p debayer, does that mean you resize the frames and then debayer? I guess that means a simple line-skipping resize... maybe process only 4 out of 6 pixels per line and row to get 720p? Maybe if you offered an option to do a preview with 2:6 pixel skip for 640x360 and no scaling, it would be better for those only monitoring on their 7" lcds. That appears to be all the resolution the preview on the 7" displays can do anyway (800x480 lcd with a bunch of screen space taken by controls), and it could drop cpu usage which means better battery life. and who knows how good the internal scaling of that monitor is anyway especially since 1280 doesnt go all that well into 800. This way you could run it at its native res. And then for those with high res field monitors, can it do a fast debayer for a 1600x900 preview or something? that could improve monitoring quality and still fit within the 7" lcd's max input frequency.

Oh I just realized something, is it the vga output that is 720p or the video preview resolution, since they seem to be different, with the preview window losing realestate to camera controls. Also, I recall noticing in some pictures that it looked like sometimes the controls were overlayed on top of the video preview on a big lcd, but the 7" lcd seems to always display the controls outside the preview window. are these options in the interface?

barlow - I think the SI website lists 1080p30 the maximum, and I think that would make sense given the limitation of gigE, but they list up to 72fps for 720p, although I recall a mention on indiefilmlive that the software doesnt yet allow more than 48fps. someone correct me if im wrong

Noah Yuan-Vogel · July 3rd, 2006, 04:31 PM

Whoops sorry just reread jasons post, forgot he said 1080p preview was an option, just not with a high quality debayer. Guess that means it might debayer then scale to 720p. i suppose some of my comments/questions are still relevant though

Keith Wakeham · July 3rd, 2006, 05:48 PM

Finally got around to looking at the footage and I have to say I'm impressed with the SI camera. Wow, Some more work and body design latter and it will be very very impressive.

Jason Rodriguez · July 3rd, 2006, 10:39 PM

Wow . . . great responses!

I am going to have to clear some things up here though :)

We are recording RAW image data to disk. That means you're taking RAW image data to post direct from the camera head. When you play back the AVI file . . . that's RAW image data-no messing with it other than a 12-bit linear to 10-bit LUT (which is adjustable).

Since we're recording RAW image data direct to disk, that means that you have everythign that camera head can deliver saved for you.

It seems that some of the thinking here is leaning towards the notion that we are passing data through the display, and then recording it to disk . . . so that the images you see are being reocorded to the disk . . . this is not the case. The images you see are a *preview* of what's being recorded to the disk. Think of them as real-time "thumbnails". They have a very quick demosaic applied to them called a quadlet method, which basically combinds the RED, GREEN, and BLUE channels and makes them one pixel.

Here's a simple diagram:

Code:

                                PREVIEW (Quadlet Demosaic)
                               / 
RAW DATA ---- 10-bit LUT ------
                               \
                                CINEFORM RAW DATA (direct-to-disk)

With this sort of pipeline, we can run the PREVIEW at a different resolution than what is actually being recorded to disk. The PREVIEW is giving you the controls to operate the camera and a real-time display to see what the camera is seeing. The true full-resolution RAW DATA is being saved to disk for use in post (or whenever you need to use the file). Basically our program is doing two things at once . . . it's compressing to Cineform (when you're recording), and it's giving you a real-time PREVIEW display of the RAW data coming off the camera head.

So again, the PREVIEW and the data going to the preview is on a separate path from the RAW DATA being saved to disk. That means it can be displayed independently, and at a different resolution and refresh rate. It also means we can do stuff with the PREVIEW in real-time that we save as metadata tags in the RAW DATA, so that you can later modify this metadata in post, and at high-qualities than we can do in real-time in the computer itself. It also means that we're not "baking" the RAW DATA . . . you get a LOT of flexibility this way.

Hope this helps everyone figure out what's happening.

BTW, One more note: Spoon now has 720/72P operational :)

Noah Yuan-Vogel · July 3rd, 2006, 11:18 PM

I guess maybe my questions weren't clear, I was talking only about the preview thread. I didn't think there was a question of what is written to disk, it's called Cineform RAW after all.

(still talking only about preview)
So what you are saying is that what's displayed on the monitor is a 1280x720 image containing preview video that is a 960x540 (1080p->quadlet->540p) image surrounded by control buttons for the touchscreen? Those who choose to output in 1080p then must have a slightly different interface because 1080p video doesnt leave any space for extra controls so when people preview in 1080p they get some controls overlayed on top of a full screen video preview, which instead of a quadlet debayer has maybe a nearest neighbor debayer....

Did I decipher your answer correctly, jason?

Sorry, did I turn this into a technical thread? I guess we should get back to talking about Spoon footage.

Noah Yuan-Vogel · July 3rd, 2006, 11:38 PM

Oh yeah and I guess that also changes one of my previous questions to: why not also provide a preview option to run a nearest neighbor debayer that only uses a 4x4 pixel block out of each 6x6 block (what i refered to as a 4:6 pixel skip)? That would give a real 1280x720 fullscreen video preview in 720p mode. And I think my other suggestions for preview resolution options still make sense. 960x540 isnt a lot of resolution especially considering that probably ends up effectively 640x360 displayed on the 7" lcds (or less considering 1:1.77 720p isnt that compatible with 1:1.66 800x480.

*all my questions assume the camera head and hdd capture are running at 1080p*

Another thing, does compressiing bayer data end up being less efficient since all adjacent data is taken from different color channels and therefore probably looks pretty noisy to the compressor? we all know how compression handles noise... poorly

June 30th, 2006, 02:04 PM	#1
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	Behind the Scenes footage from Spoon! Go check it out: indiefilmlive.blogspot.com

July 3rd, 2006, 03:26 PM	#8
Joe Carney Inner Circle Join Date: Jul 2002 Location: Centreville Va Posts: 1,828	So Jason, the video being recorded from the gbEthernet connection is 1080p, just the preview is 720p? Not bad at all. __________________ Boycott Guinness, bring back the pint!!!

July 3rd, 2006, 05:48 PM	#12
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	Finally got around to looking at the footage and I have to say I'm impressed with the SI camera. Wow, Some more work and body design latter and it will be very very impressive. __________________ www.engr.mun.ca/~wakeham/index.htm

July 3rd, 2006, 10:39 PM	#13
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	Wow . . . great responses! I am going to have to clear some things up here though :) We are recording RAW image data to disk. That means you're taking RAW image data to post direct from the camera head. When you play back the AVI file . . . that's RAW image data-no messing with it other than a 12-bit linear to 10-bit LUT (which is adjustable). Since we're recording RAW image data direct to disk, that means that you have everythign that camera head can deliver saved for you. It seems that some of the thinking here is leaning towards the notion that we are passing data through the display, and then recording it to disk . . . so that the images you see are being reocorded to the disk . . . this is not the case. The images you see are a preview of what's being recorded to the disk. Think of them as real-time "thumbnails". They have a very quick demosaic applied to them called a quadlet method, which basically combinds the RED, GREEN, and BLUE channels and makes them one pixel. Here's a simple diagram: Code: PREVIEW (Quadlet Demosaic) / RAW DATA ---- 10-bit LUT ------ \ CINEFORM RAW DATA (direct-to-disk) With this sort of pipeline, we can run the PREVIEW at a different resolution than what is actually being recorded to disk. The PREVIEW is giving you the controls to operate the camera and a real-time display to see what the camera is seeing. The true full-resolution RAW DATA is being saved to disk for use in post (or whenever you need to use the file). Basically our program is doing two things at once . . . it's compressing to Cineform (when you're recording), and it's giving you a real-time PREVIEW display of the RAW data coming off the camera head. So again, the PREVIEW and the data going to the preview is on a separate path from the RAW DATA being saved to disk. That means it can be displayed independently, and at a different resolution and refresh rate. It also means we can do stuff with the PREVIEW in real-time that we save as metadata tags in the RAW DATA, so that you can later modify this metadata in post, and at high-qualities than we can do in real-time in the computer itself. It also means that we're not "baking" the RAW DATA . . . you get a LOT of flexibility this way. Hope this helps everyone figure out what's happening. BTW, One more note: Spoon now has 720/72P operational :)

July 1st, 2006, 09:27 AM	#2
Yasser Kassana Major Player Join Date: Oct 2005 Location: UK Posts: 223	excellent work, glad you've made the post production a bit more efficient, I was expecting to see Rutguer Hauer! Where is he! Also have you recieved my email?

July 1st, 2006, 10:28 PM	#3
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Thanks for sharing all the video. Looks great and I would love to see more, longer shots of just what its like to be on set and shooting with those cameras, maybe see what setup is like. I noticed, that you are using the xenarc ~480p 7" lcd screens (right?), are those what silicon imaging provided you with or did you get those separately? I'm just surprised because I was under the impression that they had a fancy touchscreen 7" 720p HD monitor included in their package. If you dont mind my asking, how does that work anyway? Is it always the same image running to the large director's LCD as the operators' mini LCDs (as in it has to be <1080p since the mini LCDs cant handle that resolution)? Or do they run at different resolutions? in a dual monitor setup?

July 1st, 2006, 10:35 PM	#4
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Oh I just realized it was you, Jason, who posted this message, what is your role in the Spoon project? Are you, working for SI on their software? Sorry, I'm confused about who is doing what exactly in relation to the SI camera and Spoon, although I imagine the two teams are working pretty closely given one needs support for their cameras and the other needs a field testing and proof of the workflow for their product.

July 2nd, 2006, 06:26 AM	#5
Pete Tomov Regular Crew Join Date: Sep 2005 Posts: 106	Looks great!I'm definitely getting one of those! What lenses are you guys using?

July 3rd, 2006, 10:26 AM	#7
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	BTW, the limitation on the 720P display output is that with current CPU horsepower, we can't display two different resolutions on two different monitors. So you can do 1920x1080 or you can do 1280x720 but not both. If you choose the 1920x1080 approach, then the on-board Xenarc looses sync. Since we're recording RAW, the display is really only a high-quality "preview", i.e., it's not a full-fledeged 1920x1080 signal with a high-quality demosaic . . . there's just no way to-do that in a portable platform with the processing power that's on the market right now. But we are rendering a HD signal, albeit at 1280x720 (60Hz refresh rate) so that you can use the current crop of all the HD equipment display, transmission, and conversion equipment on the market including large-screen plasmas, HD-SDI converters, downconverters, etc. to move the signal anywhere you need. Another way of thinking of our preview screen is like watching the preview playback on a NLE computer screen display (like Final Cut's or Premiere Pro, or Avid's), not the full-resolution output from a AJA or Blackmagic card, or the EE mode on a tape deck.

July 3rd, 2006, 03:39 PM	#9
Barlow Elton Major Player Join Date: Mar 2005 Location: UT Posts: 945	I'm just learning about S.I. camera so here is a basic question: Can I get a higher than 1920x1080 @ 30 fps? Can it do 1080 60p? Thanks for any info.

July 3rd, 2006, 04:27 PM	#10
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	How flexible are the preview output options? Is the software interface designed to scale to a variety of resolutions? You mentioned there isnt enough CPU power to do a realtime high quality 1080p debayer, does that mean you resize the frames and then debayer? I guess that means a simple line-skipping resize... maybe process only 4 out of 6 pixels per line and row to get 720p? Maybe if you offered an option to do a preview with 2:6 pixel skip for 640x360 and no scaling, it would be better for those only monitoring on their 7" lcds. That appears to be all the resolution the preview on the 7" displays can do anyway (800x480 lcd with a bunch of screen space taken by controls), and it could drop cpu usage which means better battery life. and who knows how good the internal scaling of that monitor is anyway especially since 1280 doesnt go all that well into 800. This way you could run it at its native res. And then for those with high res field monitors, can it do a fast debayer for a 1600x900 preview or something? that could improve monitoring quality and still fit within the 7" lcd's max input frequency. Oh I just realized something, is it the vga output that is 720p or the video preview resolution, since they seem to be different, with the preview window losing realestate to camera controls. Also, I recall noticing in some pictures that it looked like sometimes the controls were overlayed on top of the video preview on a big lcd, but the 7" lcd seems to always display the controls outside the preview window. are these options in the interface? barlow - I think the SI website lists 1080p30 the maximum, and I think that would make sense given the limitation of gigE, but they list up to 72fps for 720p, although I recall a mention on indiefilmlive that the software doesnt yet allow more than 48fps. someone correct me if im wrong

July 3rd, 2006, 04:31 PM	#11
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Whoops sorry just reread jasons post, forgot he said 1080p preview was an option, just not with a high quality debayer. Guess that means it might debayer then scale to 720p. i suppose some of my comments/questions are still relevant though

July 3rd, 2006, 11:18 PM	#14
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	I guess maybe my questions weren't clear, I was talking only about the preview thread. I didn't think there was a question of what is written to disk, it's called Cineform RAW after all. (still talking only about preview) So what you are saying is that what's displayed on the monitor is a 1280x720 image containing preview video that is a 960x540 (1080p->quadlet->540p) image surrounded by control buttons for the touchscreen? Those who choose to output in 1080p then must have a slightly different interface because 1080p video doesnt leave any space for extra controls so when people preview in 1080p they get some controls overlayed on top of a full screen video preview, which instead of a quadlet debayer has maybe a nearest neighbor debayer.... Did I decipher your answer correctly, jason? Sorry, did I turn this into a technical thread? I guess we should get back to talking about Spoon footage.

July 3rd, 2006, 11:38 PM	#15
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Oh yeah and I guess that also changes one of my previous questions to: why not also provide a preview option to run a nearest neighbor debayer that only uses a 4x4 pixel block out of each 6x6 block (what i refered to as a 4:6 pixel skip)? That would give a real 1280x720 fullscreen video preview in 720p mode. And I think my other suggestions for preview resolution options still make sense. 960x540 isnt a lot of resolution especially considering that probably ends up effectively 640x360 displayed on the 7" lcds (or less considering 1:1.77 720p isnt that compatible with 1:1.66 800x480. all my questions assume the camera head and hdd capture are running at 1080p Another thing, does compressiing bayer data end up being less efficient since all adjacent data is taken from different color channels and therefore probably looks pretty noisy to the compressor? we all know how compression handles noise... poorly