Jump to content


Photo

10 bit Uncompressed a Bit Overrated for Super-8.


  • Please log in to reply
29 replies to this topic

#1 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 29 July 2006 - 05:42 PM

10 bit does not add up for me.

DV video requires 1 gig per 4 minutes of video. 10 bit uncompressed requires 1 gig for EVERY 36 seconds, almost 8 times the memory requirement. 8 bit requires around 1 gig every 45 seconds.

My complaint is, why not something somewhere in the middle? Why does the next step up from DV have to require 6 times or 8 times the memory?

Clearly, just doubling or tripling the memory would probably provide a world of difference without forcing one to buy a lot of additional memory. Even if memory is cheap, it's the idea of being forced to use more memory than is needed that bugs me. Efficient use of memory just helps keep a computer from bogging down or the drives from crashing as often.

Is there any compromise transfer system available between DV versus 8 or 10 bit?

For instance, what about 10 bit compressed? Wouldn't that probably be exactly ideal for a film transfer?

Some claim DV is identical or slightly better than Betacam SP, others see Betacam Sp as slightly better, but what doesn't make sense is, if I want to transfer component BetaCam Sp, do I really need to use either DV codecs that probably doesn't do justice to the betacam sp signal, or do I have to use such an overkill option such as 10 bit uncompressed?
  • 0

#2 Chayse Irvin

Chayse Irvin
  • Basic Members
  • PipPipPipPip
  • 409 posts
  • Cinematographer
  • Los Angeles, CA

Posted 29 July 2006 - 06:58 PM

10 bit does not add up for me.

DV video requires 1 gig per 4 minutes of video. 10 bit uncompressed requires 1 gig for EVERY 36 seconds, almost 8 times the memory requirement. 8 bit requires around 1 gig every 45 seconds.

My complaint is, why not something somewhere in the middle? Why does the next step up from DV have to require 6 times or 8 times the memory?

Clearly, just doubling or tripling the memory would probably provide a world of difference without forcing one to buy a lot of additional memory. Even if memory is cheap, it's the idea of being forced to use more memory than is needed that bugs me. Efficient use of memory just helps keep a computer from bogging down or the drives from crashing as often.

Is there any compromise transfer system available between DV versus 8 or 10 bit?

For instance, what about 10 bit compressed? Wouldn't that probably be exactly ideal for a film transfer?

Some claim DV is identical or slightly better than Betacam SP, others see Betacam Sp as slightly better, but what doesn't make sense is, if I want to transfer component BetaCam Sp, do I really need to use either DV codecs that probably doesn't do justice to the betacam sp signal, or do I have to use such an overkill option such as 10 bit uncompressed?



I would imagine you could use a different codec but it probably wont look as good at uncompressed 10-bit. DVCPRO 50?
  • 0

#3 SteveParkr

SteveParkr

    New

  • Basic Members
  • Pip
  • 9 posts
  • Student

Posted 29 July 2006 - 07:46 PM

?

Edited by SteveParkr, 29 July 2006 - 07:47 PM.

  • 0

#4 David Mullen ASC

David Mullen ASC
  • Sustaining Members
  • 19759 posts
  • Cinematographer
  • Los Angeles

Posted 29 July 2006 - 08:39 PM

The amount of data is determined by more than just the bit depth.

Mini-DV/DVCAM/DVCPRO25 is 8-bit, but it is also 4:1:1 with 5:1 intra-frame compression. 25 Mb/sec.

Digital Betacam is usually 8-bit (but can be 10-bit from what I gather), 4:2:2, and only 2.7:1 intra-field compression. 90 Mb/sec I think?

You may want to consider DVCPRO50 (8-bit, 4:2:2, 3.3:1 compression), which is essentially double the data of DV. 50 Mb/sec instead of 25 Mb/sec. Near Digital Betacam quality and a more affordable deck.

As far as I know, the next level of recording after 8-bit is 10-bit. The jumps are exponential.

The main advantage of 10-bit is less banding/contouring artifacts around smooth gradients.

If you're asking about the places that put 10-bit uncompressed data on a hard drive, I suppose you may be able to get them to use some method of reducing the data, I don't know. Maybe compression, or going to 8-bit.
  • 0

#5 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 29 July 2006 - 11:37 PM

I could have been clearer.

If I'm mastering to betacam sp and I want a higher quality Codec than mini-dv/dv-cam when I transfer from Betacam SP into final cut pro. If I use the kona card, I am suddenly using 8 times the memory for what may be a very nominal gain in quality.

I can pretty much guarantee that I'm not gaining 8 times the quality by using the Kona card, I doubt if I'm gaining more than 10-15% the quality over using the DV codec, 8 times the memory for a 15% gain in quality (just a guess on my part) does not add up, it seems very inefficient to me.

Is it possible that if one used a DVC Pro-card that had component inputs that the gigs required per minute would be significantly less than the Kona Card requires?

Does final cut pro recognize DVC-Pro as a superior input to DV, and if so, are there any DVC-Pro cards that take betacam sp component inputs?
  • 0

#6 David Mullen ASC

David Mullen ASC
  • Sustaining Members
  • 19759 posts
  • Cinematographer
  • Los Angeles

Posted 29 July 2006 - 11:58 PM

I could have been clearer.

If I'm mastering to betacam sp and I want a higher quality Codec than mini-dv/dv-cam when I transfer from Betacam SP into final cut pro. If I use the kona card, I am suddenly using 8 times the memory for what may be a very nominal gain in quality.


You're still not being very clear. Could you lay out the actual post chain because I don't understand what you mean by "mastering". If you are "mastering" to betacam-SP, wouldn't you be going OUT of FCP to it, not INTO it? Or by "mastering" do you mean the original transfer from film, not the finished edited piece?

Do you mean that you are taking film material transferred to betacam-SP, inputting into FCP, and putting the final edit out onto a digital tape format like DVCPRO50? Or are you talking film material transferred to DVCPRO50, let's say, inputting that into FCP, and outputting the final edit to betacam-SP?

And the other question is, why use analog tape anywhere in the post chain at all, expect maybe as one of your final output formats, since it would not be dubbed anymore from there.

I'd say that DVCPRO50 is the next higher digital codec over Mini-DV/DVCAM/DVCPRO(25).
  • 0

#7 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 30 July 2006 - 04:11 AM

Workflow is.....Super-8 transferred to Betacam SP.

I would then like to have the Betacam SP input to Final Cut Pro via a higher codec than mini-dv but a lower memory requirement than the Kona Card requires via it's 8 or 10 bit uncompressed option. The DVC Pro codec might be what I'm looking for but if a DVC Pro input card to final cut pro has no component input itself, than I could not feed it a betacam sp signal.

Why Betacam Sp versus a digital format? Film origination material transfered to betacam sp holds up better than camera acquisition betacam sp because film delivers 24 different images per second along with 30 A & B fields that are not motion challenged, whereas betacam sp video acquistion is delivering 60 different fields with dissimilar interlacing fields. Since I like Betacam SP camera acquisition anyway, then film transferred to betacam sp will hold up even better over a few generations.

I like that I can directly edit tape to tape when necessary. I can also make real time image adjustments directly from the betacam sp tape via the video deck without need of a computer of any kind. The betacam SP machines are cheaper than DVC pro. Betacam SP offers Time-Code and user bit adjustability so I can number reels from hour 1 to hour 24. I can also adjust audio output if necessary directly from the deck. Betacam SP also has RS-422.

My perception is that grainier film transfered to betacam sp looks more realistic and never pixely versus grainy film transferred to Mini-dv or DV-Cam.

As for the Kona Card, I'm flummoxed that it requires such a high amount of memory when compared to mini-dv for purposes of importing into final cut pro.
  • 0

#8 Stephen Williams

Stephen Williams
  • Sustaining Members
  • 4708 posts
  • Cinematographer
  • Europe

Posted 30 July 2006 - 09:45 AM

Digital Betacam is usually 8-bit (but can be 10-bit from what I gather), 4:2:2, and only 2.7:1 intra-field compression. 90 Mb/sec I think?


David,

Digital Betacam is always 10 bit, however some facilities may run their Flames etc @ 8 Bit!

Stephen
  • 0

#9 Robert Hughes

Robert Hughes
  • Basic Members
  • PipPipPipPip
  • 873 posts
  • Sound Department
  • Minneapolis

Posted 30 July 2006 - 02:09 PM

Workflow is.....Super-8 transferred to Betacam SP.


My impression is that your workflow image structure is not properly balanced. Generally, in order to retain the most image quality in a workflow you want your transfers to go from the highest quality to lesser quality image formats. But you're intentionally placing BetaSP (with 340 lines horizontal resolution) earlier in the workflow than the higher resolution digital formats. Everything downstream from the BetaSP transfer will bear the BSP stamp; it will be bandwidth limited to whatever the BSP deck provides. Essentially, you're using the BSP deck as a low-pass resolution filter. Yes, the graininess will be less, because it will be smudged out in the BSP stage.

But why concern yourself with a higher resolution computer capture downstream if you bandwidth limited the image in the previous step? What you're proposing is similar to using a softening filter on the lens, then running the image through an electronic sharpening filter to compensate for loss at the first stage. That may be an interesting effect, but otherwise it's not an appropriate workflow for keeping the highest fidelity video image & signal.

Edited by Robert Hughes, 30 July 2006 - 02:12 PM.

  • 0

#10 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 30 July 2006 - 04:39 PM

My impression is that your workflow image structure is not properly balanced. Generally, in order to retain the most image quality in a workflow you want your transfers to go from the highest quality to lesser quality image formats. But you're intentionally placing BetaSP (with 340 lines horizontal resolution) earlier in the workflow than the higher resolution digital formats. Everything downstream from the BetaSP transfer will bear the BSP stamp; it will be bandwidth limited to whatever the BSP deck provides. Essentially, you're using the BSP deck as a low-pass resolution filter. Yes, the graininess will be less, because it will be smudged out in the BSP stage.

But why concern yourself with a higher resolution computer capture downstream if you bandwidth limited the image in the previous step? What you're proposing is similar to using a softening filter on the lens, then running the image through an electronic sharpening filter to compensate for loss at the first stage. That may be an interesting effect, but otherwise it's not an appropriate workflow for keeping the highest fidelity video image & signal.


Because BetaCam Sp is not less quality than DV, thats why. This whole idea that any noise is bad news is a load of crap. Nominal amounts of Noise adds dimension and depth to the picture IF it's a very minor portion of the image. DV is 4:1:1 sampling with 5-1 compressiom can be called noiseless if you want, but that doesn't make it better.
  • 0

#11 David W Scott

David W Scott
  • Basic Members
  • PipPipPip
  • 132 posts
  • Director
  • Toronto

Posted 01 August 2006 - 10:49 AM

Because BetaCam Sp is not less quality than DV, thats why. This whole idea that any noise is bad news is a load of crap. Nominal amounts of Noise adds dimension and depth to the picture IF it's a very minor portion of the image. DV is 4:1:1 sampling with 5-1 compressiom can be called noiseless if you want, but that doesn't make it better.



You can argue about the quality of BCSP vs DV25 all day if you like... they both produce "pretty good" images. They are both lower quality than the formats or codecs you plan to use further down the chain. Why put your footage on BCSP or DV25 to start with, if you plan on bumping up the resolution later?

There seem to be two conflicting thoughts here:

1) Using BCSP or DV is convenient, tape based and relatively cheap. If these concerns outweigh final output quality, then you are simply seeking a way to capture and edit BCSP while preserving as much quality as possible. But preferably not at the huge requirements of Uncompressed 10-bit.

OR

2) You are seeking the best quality possible, regardless of tape format. You are favouring BCSP because you perceive it as higher quality than DV25. But if quality alone is what you are seeking, then using BCSP or DV25 as your telecine master is self-defeating. You are willing to use better quality formats later in post, so why not use them for telecine recording?


If thought #1 is closer to your intention, here is my suggestion:
- Get as good an analog-to-digital video capture card as you can. This might be inside a system (like a used AVID or Media 100) or it might be an add-on card (like a DPS/Leitch card.) You want something with component inputs, has a codec that is supported by your editing software, and will have adjustable compression settings. This way you can capture 4:2:2 or 4:4:4 SD video anywhere between 5:1 compression and 1:1 (uncompressed), adjusting it to get a data rate, file size, and quality that you like.

If thought #2 is more accurate, then here are two other suggestions:
- Have the telecine recorded to DVCPRO50 (which is a 4:2:2 DV codec, compressing at 3.3:1). DVCPRO50 is supported in Final Cut Pro, and I believe can be input directly by firewire. No special card required. You will need to own or rent a DVCPRO50 deck.
- Have the telecine recorded directly to harddrive, using a good-quality codec that is supported by BOTH the transfer house and your non-linear software. That really means something built in to Final Cut Pro. If file size is your concern, have it recorded in 8-bit Uncompressed instead of 10-Bit. The real advantage (beyond quality) is that you don't have to own or rent any deck or special hardware. This will work with a fast PowerBook and off-the-shelf Lacie FW800 drives.
  • 0

#12 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 01 August 2006 - 02:10 PM

1) Using BCSP or DV is convenient, tape based and relatively cheap. If these concerns outweigh final output quality, then you are simply seeking a way to capture and edit BCSP while preserving as much quality as possible. But preferably not at the huge requirements of Uncompressed 10-bit.

OR

2) You are seeking the best quality possible, regardless of tape format. You are favouring BCSP because you perceive it as higher quality than DV25. But if quality alone is what you are seeking, then using BCSP or DV25 as your telecine master is self-defeating. You are willing to use better quality formats later in post, so why not use them for telecine recording?


Betacam SP probably captures all the quality that Super-8 has to offer. Reason one is exactly the issue I am concerned about. The Kona card offers 8 bit in addition to 10 but but it's still only approximately 20% less than 10 bit, yet quite a bit more more memory than the firewire input. It's unnecessary overkill, otherwise fire wire is a really awful codec, which I don't think it is.

In other words, either the dv codec is awful, or the Kona card 8 bit uses way too much memory for betacam sp by comparison, it's one or the other, unless there are specs that are lower in terms of memory that I am not aware of as it relates to the Kona card and 8 bit and 10 bit compression.
  • 0

#13 Robert Hughes

Robert Hughes
  • Basic Members
  • PipPipPipPip
  • 873 posts
  • Sound Department
  • Minneapolis

Posted 01 August 2006 - 07:56 PM

Well, it sounds like you decided before you even signed on. So why did you ask for responses, then call them a load of crap? Seems pretentious and self-defeating to me.
  • 0

#14 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 01 August 2006 - 08:57 PM

Well, it sounds like you decided before you even signed on. So why did you ask for responses, then call them a load of crap? Seems pretentious and self-defeating to me.


Maybe you should reacquaint yourself with the first post in this topic before making such an inaccurate accusation towards me. Just reread the first post in this topic and you'll see that your statement is not substantiated in the least.

I clearly stated in that first post that I was unhappy with what my research was showing in regards to a compromise bit rate between a DV codec and 8 or 10 bit umcompressed. I then asked what alternatives exist.

DVC-Pro may be a logical compromise, but the lack of decks as compared to Betacam SP makes doesn't make that viable a choice.
  • 0

#15 Chris Graham

Chris Graham
  • Basic Members
  • PipPipPip
  • 127 posts
  • Director

Posted 01 August 2006 - 09:33 PM

10-bit direct to hard drive = convenience and those willing to pay. I don't own any sort of decks to layoff to anyways. If it's as good as DV that's enough more.
  • 0

#16 Chris Keth

Chris Keth
  • Sustaining Members
  • 4427 posts
  • 1st Assistant Camera
  • Los Angeles

Posted 02 August 2006 - 01:34 AM

Nominal amounts of Noise adds dimension and depth to the picture IF it's a very minor portion of the image.



No, it doesn't. It adds noise. My eyes don't see a noiseless world and let me tell you: there's quite a bit of depth and dimension out there. ;)
  • 0

#17 Alessandro Machi

Alessandro Machi
  • Sustaining Members
  • 3318 posts
  • Other
  • California

Posted 02 August 2006 - 04:07 AM

No, it doesn't. It adds noise. My eyes don't see a noiseless world and let me tell you: there's quite a bit of depth and dimension out there. ;)


Your phrase "My eyes don't see a noiseless world and let me tell you: there's quite a bit of depth and dimension out there" seems to imply you agree with my statement.
  • 0

#18 Chris Keth

Chris Keth
  • Sustaining Members
  • 4427 posts
  • 1st Assistant Camera
  • Los Angeles

Posted 02 August 2006 - 09:57 AM

Your phrase "My eyes don't see a noiseless world and let me tell you: there's quite a bit of depth and dimension out there" seems to imply you agree with my statement.



You know what I mean...that'll teach me for posting at 2:30 am :P
  • 0

#19 David W Scott

David W Scott
  • Basic Members
  • PipPipPip
  • 132 posts
  • Director
  • Toronto

Posted 02 August 2006 - 10:19 AM

Reason one is exactly the issue I am concerned about. The Kona card offers 8 bit in addition to 10 but but it's still only approximately 20% less than 10 bit, yet quite a bit more more memory than the firewire input. It's unnecessary overkill, otherwise fire wire is a really awful codec, which I don't think it is.

In other words, either the dv codec is awful, or the Kona card 8 bit uses way too much memory for betacam sp by comparison, it's one or the other, unless there are specs that are lower in terms of memory that I am not aware of as it relates to the Kona card and 8 bit and 10 bit compression.


The DV codec isn't awful -- it's actually delivers a pretty good image at a very low data-rate. Remember, the DV codec was defined 10 years ago (!!!) to allow full-resolution SD recording to a 6.5mm tape (!!!) and then for it to be ingested and processed by the average computer of 1998 (!!!) DV really was a miracle, because it offloaded the analog-to-digital conversion duties to the camcorder, and negated the need for good quality A-to-D capture cards.

If you want to continue using BCSP, you are moving the responsibility for A-to-D back to the computer. There are still lots of choices to do that:

For about $5000:
VelocityQ boardset and NLE software

For about $1000:
AJA Kona LS

For about $600:
BlackMagic Decklink SP

Simply use adjustable compression with one of these cards -- usually M-JPEG or JPEG. The quality of 4:2:2 JPEG at 3MB/s is really good. I used to edit with a DPS Perception card, offlining at 1MB/s, onlining at 7MB/s. At 7MB/s the 4:2:2 M-JPEG footage looked as good as Digital Betacam.

Just because a card comes with uncompressed (the holy grail of non-linear editing in the 90's) doesn't mean that you have to use it uncompressed.

Here's an excerpt from the Kona LS manual about the choices available for it:

PhotoJPEG
Data rate: approximately 1-3 MB/second?supported by internal system drive
Quality: Very Good
The Apple PhotoJPEG codec offers an excellent compressed media choice for
on-line quality at low data rates. PhotoJPEG can use the full-raster at 4:2:2
sampling. Final Cut Pro allows you to adjust quality using a PhotoJPEG control
panel. KONA LS allows for PhotoJPEG monitoring and/or output in SD.
KONA LS can capture from almost any input, directly to PhotoJPEG media.

DV (DV25)
Data rate: 3.13 MB/second (megabyte/second)?supported by internal system
drive
Quality: Good
In this workflow, DV is usually input to a Power Mac running Final Cut Pro
through its FireWire port. DV offers good quality, but it has lower Chroma
resolution when compared to DV50, JPEG, or uncompressed. You can use
KONA LS to playback DV projects to uncompressed?in real time?for
monitoring and/or output. Alternatively, KONA LS can capture uncompressed
from any input, directly to DV clips.

DV50
Data rate: 6.26 MB/second?supported by internal system drive
Quality: Very Good
Like DV25, Final Cut Pro also supports the Panasonic DV50 standard
definition codec. DV50 is a 4:2:2 compressed format and therefore has higher
chroma resolution when compared to DV25. Also like DV25, you can use
KONA LS to playback DV50 projects to uncompressed?in real time?for
monitoring and/or output. KONA LS can capture uncompressed from any
input, directly to DV50 clips.

Uncompressed 8-bit
Data rate: 21 MB/second ? requires SCSI, Fibre Channel, or ATA drive array
Quality: Excellent
Uncompressed media is KONA LS?s native storage format, offering the highest
quality available. Capturing in uncompressed results in no compression
artifacts, and video is sampled over the full raster at a 4:2:2 rate. Using
uncompressed maintains a higher quality in your project from capture all the
way through effects rendering. Final Cut Pro supports realtime effects with
uncompressed media using RT Extreme. KONA LS supports capture of
uncompressed through any of its inputs, and projects are output to all of its
outputs simultaneously.

Uncompressed 10-bit
Data rate: 28 MB/second ? requires SCSI, Fibre Channel or ATA drive array
Quality: Excellent, very high quality
Offering all the benefits noted previously for 8-bit uncompressed, 10-bit
additionally offers the very highest quality available. With 10-bit media and Final
Cut Pro?s 32 bit Floating Point YUV effects rendering, video quality is second to
none?at any price.

Source
  • 0

#20 Chris Graham

Chris Graham
  • Basic Members
  • PipPipPip
  • 127 posts
  • Director

Posted 02 August 2006 - 10:20 AM

Imagine if there was 128-bit transfer for S8 with a hard drive being the size of a building. That would be awesome! :o
  • 0


The Slider

FJS International, LLC

Ritter Battery

rebotnix Technologies

Opal

Gamma Ray Digital Inc

CineTape

Broadcast Solutions Inc

Aerial Filmworks

Willys Widgets

Tai Audio

Metropolis Post

Abel Cine

Glidecam

Visual Products

Technodolly

Wooden Camera

CineLab

Rig Wheels Passport

Media Blackout - Custom Cables and AKS

Paralinx LLC

Rig Wheels Passport

CineTape

Glidecam

Tai Audio

Willys Widgets

Aerial Filmworks

Visual Products

Paralinx LLC

Broadcast Solutions Inc

The Slider

Gamma Ray Digital Inc

Opal

Metropolis Post

Abel Cine

CineLab

Wooden Camera

Technodolly

Media Blackout - Custom Cables and AKS

FJS International, LLC

rebotnix Technologies

Ritter Battery