Codec Shootout – BlackMagic uncompressed Hyperdeck Shuttle vs. FS100 AVCHD

Okay, here comes the promised comparison between uncompressed recording via BlackMagic Hypershuttle and Sony´s flavor of AVCHD recorded to SD card in camera.

http://vimeo.com/27529215

As you can see, you can´t see barley any difference at all. Only in fast motion some minor differences are visible. Even with excessive grading to the material (much more that you would normally do) both material holds up about the same, before it falls apart.

That is good and bad. Good, because you can use cheap SD cards and AVCHD almost every time. Bad because I was spending money on the Shuttle and SSDs that I will use very seldom.

Actually I´m shocked how good Sony´s implementation of AVCHD is.

UPDATE:
I figured out a new workflow that involves grading in linear color space.
That´s where the uncompressed BM material blows the AVCHD out of the water.
It´s much richer, smoother and… well hard to describe, but I would call it much fatter.
Also if you got much motion or detail or motion and much detail in the shot the differences are there and visible.
I also did some green screen tests, where the BM really shines.

12 thoughts on “Codec Shootout – BlackMagic uncompressed Hyperdeck Shuttle vs. FS100 AVCHD

  1. It’s hard to tell from the file I downloaded. The compression is very high, 2 generations of GOP already, some shadows are so artifacted they have turned in to patchwork, but the material is obviously good for compression. Flat surfaces of uniform color, smooth changes in luma, no real chroma detail (no real chroma at all really.) If you record waves, action, explosions, fireworks or even closeups with skin and hair texture or fabrics in motion, uncompressed will be superior. A simple closeup of an actress with some motion in her hair shows how plastic it looks compared to 1:1 pixel intraframe and uncompressed.

    Judging by the texture and the chroma, I thought it was a $1k camcorder but I see that it’s a $7K one with a 24K GOP codec. The handles are deep red/brown, and the wall is actually greenish, right? I would love to see a digital still camera still of the same shot.

  2. The motion blur on the right is much better. On the left we just gets blocks and artifacted “detail” that shouldn’t be there.

    The right side is interlaced but even after a quick interpolation to fix that (on the right side only), the chroma has a much higher resolution, no blocking, realistic motion blur etc. Does it matter? For this type of shot, probably not.

    http://imageshack.us/photo/my-images/694/chroma1.png/

  3. Hey Frank-
    I think a LOT of folks would be VERY interested in the PHYSICAL aspects of the Shuttle operation!
    -size/fitment/mounting
    -how easily it integrates with the camera output; record ON/OFF controls
    -Time code?
    -battery life/or taping of the camera
    -SSD performance, costs
    Stu Aull
    Alaska

  4. Hi Frank…….
    I am interested in the process used for pulldown removal of fs100 to black magic I have been trying to removed it with cineform hdlink but the result is a split screen type jagged video with a green bar at the bottom …is there a better way?

  5. Hey Frank, thanks for the test. What got me really interested was the update, though. Could you be more specific, please? Grading in linear color space?

    -Jori

  6. Hi Frank,

    I hope this discussion isn’t dead. Your article is the first I came across on this subject. I’ve seen several reviews and tests of external recorders used with the FS100 and most come to the same conclusion. In most the reviewer can’t see the difference between the un-compressed (or less-compressed) footage and the AVCHD from the camera’s card. Or they actually like the AVCHD better because it munges high frequency detail and hides the sensors noise a bit, or because they haven’t properly removed the pulldown, or haven’t interpreted the 24p inside of 60i footage properly in their NLE. The camera’s AVCHD is very good, but I broke it’s compression the first day I shot with it. All I had to do was a fast aperture pull from f16 to f1.4, and viola! COMPRESSION artifacts everywhere!
    I decided to do some tests for myself using my MXO2 to record un-compressed, ProRes 422 HQ, DNxHD 220 10bit, 8bit and DNxHD 145 at 1080i (camera recording simultaneously to 1080p 24 FX). I then brought everything into AE and synced the less-compressed files above the native AVCHD in a comp. I could then switch on and off the top layer to see the difference for any given frame. I also switched the composite mode to difference and screened over a grey solid so I could see only the difference between the two. My conclusions are much different from most tests I’ve seen. Just looking from a distance, the AVCHD looks great. But compare it frame for frame with any of the other formats I recorded and it falls apart when stressed. I did however do my best to torture the codecs. AVCHD dose great with static (motion and luminance) images. Film a static frame of a talking guy without lighting changes, or a landscape and AVCHD looks amazing. In my stress test, some of the AVCHD frames (probably the I-frames) are almost identical, but most suffer from smudged detail and other artifacts. I do see huge value in a less compressed recording device that would actually simplify the workflow. I’d actually prefer Sony to give us an in camera all I-frame version of AVCHD as I think it could hold up as good as ProRes and DNxHD. I don’t know why camera manufacturers try to stick with using a delivery format for a capture format.
    So on to the Hyperdeck. Now I’m not afraid of doing a little work in post, but I am not willing to slave away for days trying to remove non-flagged pulldown. After all, these recorders are supposed to simplify your workflow, not severely compound it’s complexity. With 24p AVCHD you can set up a batch transcode, walk away from your computer, and come back ready to edit proxies. If you record 24p in a 60i file that has pulldown flags, most NLEs can detect the pulldown cadence and remove it for editing or transcoding. Cinema tools can quickly strip the interlaced frames without transcoding the file, leaving you with pristine 24p in it’s untouched original codec. From what I can tell, the Hyperdeck and the products from ATOMOS cannot see or record the FS100’s timecode or pulldown flags from the HDMI. This means I would have to manually identify the interlaced frames and remove them from every shot to get my 24p out. The cadence will start in a different place in every shot because the recorder and the deck aren’t synced in any way. No batch processing! I don’t care how wonderful a recorded file’s compression (or lack of compression) is, if I have to manually asses and assign the pulldown pattern in order to remove it from every shot, it’s not worth my time. It’s a shame. I would love the Hyperdeck, but until Sony update the firmware to send 24p over HDMI (and Hyperdeck records it), or the Hyperdeck records the timecode with pulldown flags, or better yet, removes pulldown before recording, the Hyperdeck is useless with the FS100 at 24p! Unless you just plan on doing just an occasional chroma key shot with it. Trying to get 24p out of all the shots, even for just a short film, would be a huge nightmare. I guess it’s a law of the universe. If it’s affordable, it has to be crippled.

    Here is a link to some tiffs

    http://www.filedropper.com/fs100avchdvslesscompressed

    • Shiloh, you are very right about there being a difference in IQ, but I would say it is a rather huge difference. Like you said, the AVCHD can be very muddy in the shadows when a lot of motion is present. Also anywhere there is a lot of detail like when shooting in the woods, etc.

      But, I disagree that shooting 24p is worthless when using the FS100. All you have to do is go into After Effects and interpret footage to remove the pull down and then encode to a high quality DNxHD or Prores. Yes it is an extra step, rather than just dropping it into your timeline, but it is well worth the extra step to get that higher quality.

      What I do is leave the uncompressed footage on the drive in the dock and take it straight into Ae and remove the pull down. Then encode from the disc to my desktop or where ever the project files are.

  7. set of Zeiss Super Speed lenses is setting you back well over $20,000, while a set of Zeiss ZF SLR lenses are under $6,000. For the extra money you get a focus that goes clockwise, bigger, unified housing, longer throw and smooth (declicked) aperture. All this is nice to have, but how about sheer image quality. Do the “cheap” ZF SLR lenses hold up against professional cineglass?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s