Kaku's FX1 footage in Sony Vegas: personal observations at DVinfo.net

Mark Kubat · October 24th, 2004, 01:07 PM

Hi folks.

I've been working extensively with the footage in Sony Vegas 5.0b - the m2t's are imported seamlessly without any hassle. Playback on my 2.0GHz pentium 4 single processor with 1 gig of RAM is a different story, however... a bit stuttery etc.

Exporting to wmvhd 720-60p HD took almost 2 hrs. for the 1 minute dscndingbike.m2t clip!!!

Clearly, some better throughput will have to be worked out to make output to wmvhd a reality.

Much faster was "downsampling" m2t to 720x480 60i NTSC mini-dv avi. I wanted to do this to more objectively compare the potential against XL2 footage....

Sure, it's just my opinion, everyone's got one, right?

Having shot recently with XL2 quite a bit at the indie tv station where I work, I can safely say that in every case, I see more resolution and "better quality" in the FX1 downconversion compared to XL2 60i. There is more detail in wide shots - more infomration on the screen. How can I explain this? In XL2 stuff, things just all "blur together" and I'm not talking focus here. It just becomes impossible to discern things like telephone wires and individual leaves - you know... With the FX1, you feel like you're seeing EVERYTHING.

The FX1 night footage is breath-taking - I have never seen mini-dv/prosumer video shot at night that ever looked this good before - the low-light capability of the FX1 cannot be disputed.

More importantly, no motion artifacts are visible in any of the motion footage.

The daylight footage with people is most useful to me for making observations (new day0dbf6.2 and dfootbag) - in Sony Vegas, I experimented using Magic Bullet Movie Looks to punch up the colours a bit and it's great - remember, this is on the downsampled avi as Magic Bullet right now doesn't support HD resolutions unless you have the HD version... there is one, right? Check red giant's website...

*IMPORTANT OBSERVATION*
Sony Vegas has done this to me before - when converting from one resolution to another, it seems Vegas re-samples the footage and consequently, a slight pull-down type affect becomes apparent in the resulting interlaced avi - I certainly noticed this here with the 1080i footage going to 720x480 60i. There's a slight "time-slur" effect that mimicks the filmic look. I confirmed this by comparing the wmvhd 60p version I made which looked very "video-like" whereas my avi's have a slight film-feel to them.
This is NOT with the cineframe 30 or cineframe 24 footage - this is with everything shot at normal "60i"

Anyone with Vegas 5b trying the downconvert will hopefully comment on this...

Take home message is: I'd pick this cam for SD work over the XL2 hands-down - the pipeline of downsampling from HDV 1080i to SD NTSC DV 720 x 480 60i gives a sharper, better, aesthetically more pleasing image than XL2 native. There's just more info - it's the same idea as starting off with a film print and scanning to mini-dv - there's more "info" to start with that "comes through" in the final product.

The audio doesn't seem that bad...

My hunch is that HDV>SD footage from FX1 would be significantly "better" than SD>SD with FX1 - it would be interesting to see how the "downconverted" footage from the cam compares to doing it in post the way I've described above...

Time will tell.

PAL FX1E certainly would seem the be the way to go if you're considering the cam for indie filmmaking incl. film-out 25>24.
There would be less compression too: 50 interlaced frames per second at 1080 lines vs. 60 per second so theoretically a resolution advantage with the PAL - but what happens if you convert from PAL>NTSC via Vegas or someother conversion? Does conversion introduce a "generation loss" that cancels out the advantage you have gained from slight quality edge gained by shooting PAL? I guess direct comparison/tests will be the only way to know for sure... Certainly, the switchable pro version would allow such a test "in-camera" for direct comparison under controlled conditions (ie. same cam)

Joe Carney · October 24th, 2004, 06:12 PM

Your description about HD to SD is the principle behind superbit DVDs.

Kaku Ito · October 24th, 2004, 07:18 PM

Nice going Mark.

I'm going to find out with HISCO, Japanese distributor of Cineform, to lend me a copy of Connect HD for Vegas 5. Although Vegas 5 is going to support FX1 directly, Connect HD seems to give more functions, so probably it is worth the test.

I will let you know what happens.

Rob Lohman · October 25th, 2004, 04:11 AM

Joe: most DVD's are converted from an (S)HD source to an SD
source with regular resolutions of HD but most being in the 2K
range. Superbit DVD's just have a higher bitrate than most DVD's
and therefor usually go without ANY extra's on the disc including
multiple audio tracks etc.

Bill Ravens · October 25th, 2004, 09:30 AM

After downloading the files and looking at them I have several comments/observations.

First, there most definitely ARE motion artifacts. They appear as the horrid jaggies on all moving objects, to the point of being distracting from the overall excellent rez. Playing back on different players also seems to minimize or amplify the jaggies, depending on the player. FFDSHOW decoder incorporates an anti-aliaser, and this seems to make the jaggies look more like motion blur.

When I bring this footage into Vegas 5, I notice two things. Vegas 5 detects it as "upper field first" interlaced video at 29.97. If I manually change the footage to "lower field first", the jaggies don't go away, but, look less pronounced. V5 is also detecting the PAR as 1.33.

Downconverting and rendering to DV format-letterboxed, the images look no better nor worse than my XL2 images. In fact, with progressive scan on my XL2, it looks better since I don't get those jaggies. I'm still not convinced this is a step for the better.

Jim Arthurs · October 25th, 2004, 11:06 PM

The pixel aspect ratio of 1.33 is correct, as the Sony footage is 1440 by 1080i. By using 1.33 instead of square pixel, the 1440 wide image fills the 1920 wide frame without rendering.

I use this "trick" all the time when rendering HD animation to save a bit of render time. This isn't too bad a cheat, as even the mighty HDCam is recorded to tape as 1440 by 1080i.

All HD is "upper field first" to my knowledge, at least all the pro formats. I wouldn't expect HDV to be different... I'd leave this setting alone and experiment with your output field order...

I hope this helps,

Jim Arthurs

<When I bring this footage into Vegas 5, I notice two things. <Vegas 5 detects it as "upper field first" interlaced video at <29.97. If I manually change the footage to "lower field first", the <jaggies don't go away, but, look less pronounced. V5 is also <detecting the PAR as 1.33.

Kevin Dooley · October 26th, 2004, 06:37 AM

I don't want to oversimplify anything...but it sounds like what you're talking about is the "jaggies" associated with watching interlaced footage on a progressive computer screen...not MPEG2 motion artifacts. If that's the case then I'm a lot less worried about your comments... I noticed the interlaced jaggies, but I had yet to see any serious MPEG2 artifacts and I thought maybe I had missed something...

Bill Ravens · October 26th, 2004, 07:00 AM

Kevin...

You may be right. My first inclination to think they were motion artifacts is based on the observation that interlaced footage tends to show field lines thruout the image. These images show field lines only in areas where there is motion. The "still" part of the image looks quite smooth and regular. Quite nice, actually. Either way, motion artifacts or interlaced field lines, I didn't like what I saw. A shame Sony didn't provide progressive scan on this machine. It'll never stand up to transfer to film.

Kevin Dooley · October 26th, 2004, 07:04 AM

<--A shame Sony didn't provide progressive scan on this machine. It'll never stand up to transfer to film.-->

Couldn't agree with you more...I know there are ways to make it work...but the loss of resolution will still occur during deinterlacing... But then again--if they gave us everything we wanted in one camera, we wouldn't need to buy anything from them again, would we? Stupid capitalism...=)

Jacques Ligou · October 27th, 2004, 07:24 PM

I have found that TMPGEnc 3.0 XPress can make any conversion of m2t files. Of course this is not a real editing program, only cuts are possible but many filters can be chosen before encoding.

I have also played with the dscndingbike.m2t file. I could directly display this file on the LCD monitor of my PC (3.2 GHz and 1GB RAM). I was impressed by the video quality although I could see a few artifacts because the deinterlacing of the VLC player is not the best.

I have produced several wmvhd files (1280*720, 1440*1080, etc..) and the encoding time was around 15 min, not two hours as mentioned by Mark Kubat.

As recommended by Microsoft, 5Mbps is a good bitrate. The size of wmv files is less than 45 MB while the initial m2t file was 207 MB. More, there are no longer artifacts since I have used in TMPGEnc the filter "deinterlace only when necessary".

Cieonatechori Michelieve · October 27th, 2004, 08:40 PM

<<<-- I'm still not convinced this is a step for the better. -->>>

You might not be, but the first time buyer is sure going to be. Less price, ,more(or even equally compared power/functionality[compared to dvx or xl2]-[a step for better!)) . Things can never go wrong when you buy it compared to DVX or XL2.

October 24th, 2004, 01:07 PM	#1
Mark Kubat Major Player Join Date: Jan 2003 Location: Toronto, Canada Posts: 375	Kaku's FX1 footage in Sony Vegas: personal observations Hi folks. I've been working extensively with the footage in Sony Vegas 5.0b - the m2t's are imported seamlessly without any hassle. Playback on my 2.0GHz pentium 4 single processor with 1 gig of RAM is a different story, however... a bit stuttery etc. Exporting to wmvhd 720-60p HD took almost 2 hrs. for the 1 minute dscndingbike.m2t clip!!! Clearly, some better throughput will have to be worked out to make output to wmvhd a reality. Much faster was "downsampling" m2t to 720x480 60i NTSC mini-dv avi. I wanted to do this to more objectively compare the potential against XL2 footage.... Sure, it's just my opinion, everyone's got one, right? Having shot recently with XL2 quite a bit at the indie tv station where I work, I can safely say that in every case, I see more resolution and "better quality" in the FX1 downconversion compared to XL2 60i. There is more detail in wide shots - more infomration on the screen. How can I explain this? In XL2 stuff, things just all "blur together" and I'm not talking focus here. It just becomes impossible to discern things like telephone wires and individual leaves - you know... With the FX1, you feel like you're seeing EVERYTHING. The FX1 night footage is breath-taking - I have never seen mini-dv/prosumer video shot at night that ever looked this good before - the low-light capability of the FX1 cannot be disputed. More importantly, no motion artifacts are visible in any of the motion footage. The daylight footage with people is most useful to me for making observations (new day0dbf6.2 and dfootbag) - in Sony Vegas, I experimented using Magic Bullet Movie Looks to punch up the colours a bit and it's great - remember, this is on the downsampled avi as Magic Bullet right now doesn't support HD resolutions unless you have the HD version... there is one, right? Check red giant's website... IMPORTANT OBSERVATION Sony Vegas has done this to me before - when converting from one resolution to another, it seems Vegas re-samples the footage and consequently, a slight pull-down type affect becomes apparent in the resulting interlaced avi - I certainly noticed this here with the 1080i footage going to 720x480 60i. There's a slight "time-slur" effect that mimicks the filmic look. I confirmed this by comparing the wmvhd 60p version I made which looked very "video-like" whereas my avi's have a slight film-feel to them. This is NOT with the cineframe 30 or cineframe 24 footage - this is with everything shot at normal "60i" Anyone with Vegas 5b trying the downconvert will hopefully comment on this... Take home message is: I'd pick this cam for SD work over the XL2 hands-down - the pipeline of downsampling from HDV 1080i to SD NTSC DV 720 x 480 60i gives a sharper, better, aesthetically more pleasing image than XL2 native. There's just more info - it's the same idea as starting off with a film print and scanning to mini-dv - there's more "info" to start with that "comes through" in the final product. The audio doesn't seem that bad... My hunch is that HDV>SD footage from FX1 would be significantly "better" than SD>SD with FX1 - it would be interesting to see how the "downconverted" footage from the cam compares to doing it in post the way I've described above... Time will tell. PAL FX1E certainly would seem the be the way to go if you're considering the cam for indie filmmaking incl. film-out 25>24. There would be less compression too: 50 interlaced frames per second at 1080 lines vs. 60 per second so theoretically a resolution advantage with the PAL - but what happens if you convert from PAL>NTSC via Vegas or someother conversion? Does conversion introduce a "generation loss" that cancels out the advantage you have gained from slight quality edge gained by shooting PAL? I guess direct comparison/tests will be the only way to know for sure... Certainly, the switchable pro version would allow such a test "in-camera" for direct comparison under controlled conditions (ie. same cam)

October 25th, 2004, 04:11 AM	#4
Rob Lohman RED Code Chef Join Date: Oct 2001 Location: Holland Posts: 12,514	Joe: most DVD's are converted from an (S)HD source to an SD source with regular resolutions of HD but most being in the 2K range. Superbit DVD's just have a higher bitrate than most DVD's and therefor usually go without ANY extra's on the disc including multiple audio tracks etc. __________________ Rob Lohman, visuar@iname.com DV Info Wrangler & RED Code Chef Join the DV Challenge \| Lady X Search DVinfo.net for quick answers \| Buy from the best: DVinfo.net sponsors

October 26th, 2004, 06:37 AM	#7
Kevin Dooley Major Player Join Date: Dec 2003 Location: Fort Worth, TX Posts: 414	Question I don't want to oversimplify anything...but it sounds like what you're talking about is the "jaggies" associated with watching interlaced footage on a progressive computer screen...not MPEG2 motion artifacts. If that's the case then I'm a lot less worried about your comments... I noticed the interlaced jaggies, but I had yet to see any serious MPEG2 artifacts and I thought maybe I had missed something...

October 27th, 2004, 07:24 PM	#10
Jacques Ligou New Boot Join Date: Oct 2004 Location: Lausanne, Switzerland Posts: 5	m2t file conversion I have found that TMPGEnc 3.0 XPress can make any conversion of m2t files. Of course this is not a real editing program, only cuts are possible but many filters can be chosen before encoding. I have also played with the dscndingbike.m2t file. I could directly display this file on the LCD monitor of my PC (3.2 GHz and 1GB RAM). I was impressed by the video quality although I could see a few artifacts because the deinterlacing of the VLC player is not the best. I have produced several wmvhd files (1280720, 14401080, etc..) and the encoding time was around 15 min, not two hours as mentioned by Mark Kubat. As recommended by Microsoft, 5Mbps is a good bitrate. The size of wmv files is less than 45 MB while the initial m2t file was 207 MB. More, there are no longer artifacts since I have used in TMPGEnc the filter "deinterlace only when necessary". __________________ Jacques Ligou Retired Physicist

October 24th, 2004, 06:12 PM	#2
Joe Carney Inner Circle Join Date: Jul 2002 Location: Centreville Va Posts: 1,828	Your description about HD to SD is the principle behind superbit DVDs.

October 24th, 2004, 07:18 PM	#3
Kaku Ito Trustee Join Date: Jun 2004 Location: Chiyoda-ku, Tokyo Posts: 1,382	Nice going Mark. I'm going to find out with HISCO, Japanese distributor of Cineform, to lend me a copy of Connect HD for Vegas 5. Although Vegas 5 is going to support FX1 directly, Connect HD seems to give more functions, so probably it is worth the test. I will let you know what happens.

October 25th, 2004, 09:30 AM	#5
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	After downloading the files and looking at them I have several comments/observations. First, there most definitely ARE motion artifacts. They appear as the horrid jaggies on all moving objects, to the point of being distracting from the overall excellent rez. Playing back on different players also seems to minimize or amplify the jaggies, depending on the player. FFDSHOW decoder incorporates an anti-aliaser, and this seems to make the jaggies look more like motion blur. When I bring this footage into Vegas 5, I notice two things. Vegas 5 detects it as "upper field first" interlaced video at 29.97. If I manually change the footage to "lower field first", the jaggies don't go away, but, look less pronounced. V5 is also detecting the PAR as 1.33. Downconverting and rendering to DV format-letterboxed, the images look no better nor worse than my XL2 images. In fact, with progressive scan on my XL2, it looks better since I don't get those jaggies. I'm still not convinced this is a step for the better.

October 25th, 2004, 11:06 PM	#6
Jim Arthurs Regular Crew Join Date: Sep 2004 Location: Colorado Springs CO Posts: 120	The pixel aspect ratio of 1.33 is correct, as the Sony footage is 1440 by 1080i. By using 1.33 instead of square pixel, the 1440 wide image fills the 1920 wide frame without rendering. I use this "trick" all the time when rendering HD animation to save a bit of render time. This isn't too bad a cheat, as even the mighty HDCam is recorded to tape as 1440 by 1080i. All HD is "upper field first" to my knowledge, at least all the pro formats. I wouldn't expect HDV to be different... I'd leave this setting alone and experiment with your output field order... I hope this helps, Jim Arthurs <When I bring this footage into Vegas 5, I notice two things. <Vegas 5 detects it as "upper field first" interlaced video at <29.97. If I manually change the footage to "lower field first", the <jaggies don't go away, but, look less pronounced. V5 is also <detecting the PAR as 1.33.

October 26th, 2004, 07:00 AM	#8
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Kevin... You may be right. My first inclination to think they were motion artifacts is based on the observation that interlaced footage tends to show field lines thruout the image. These images show field lines only in areas where there is motion. The "still" part of the image looks quite smooth and regular. Quite nice, actually. Either way, motion artifacts or interlaced field lines, I didn't like what I saw. A shame Sony didn't provide progressive scan on this machine. It'll never stand up to transfer to film.

October 26th, 2004, 07:04 AM	#9
Kevin Dooley Major Player Join Date: Dec 2003 Location: Fort Worth, TX Posts: 414	<--A shame Sony didn't provide progressive scan on this machine. It'll never stand up to transfer to film.--> Couldn't agree with you more...I know there are ways to make it work...but the loss of resolution will still occur during deinterlacing... But then again--if they gave us everything we wanted in one camera, we wouldn't need to buy anything from them again, would we? Stupid capitalism...=)

October 27th, 2004, 08:40 PM	#11
Cieonatechori Michelieve Regular Crew Join Date: Aug 2004 Location: Washington DC Posts: 25	<<<-- I'm still not convinced this is a step for the better. -->>> You might not be, but the first time buyer is sure going to be. Less price, ,more(or even equally compared power/functionality[compared to dvx or xl2]-[a step for better!)) . Things can never go wrong when you buy it compared to DVX or XL2.