Sony NEX-VG10 AVCHD E-Mount Lens Camcorder

[Steve Mullen]

"AVCHD is a highly compressed acquisition codec, never intended to be an editing format.
It is very lossy and if you beat up on it in post with effects, color correction, transcodes, etc"

Don't mean to beat on you, but you can't "beat-up" on AVCHD because that's not how NLEs work.

ANY/ALL source codecs are decompressed to a frame ONCE. CF also decompresses AVCHD once.

After a frame is decompressed it is NO LONGER AVCHD/HDV, etc. It it now uncompressed RGB or YUV.

From this instant onward, ALL FX are done on this uncompressed RGB or YUV frame. You can stack as many FX as you want, they are all done on uncompressed RGB or YUV frames. (You can add as many uncompressed RGB or YUV frames from other streams of AVCHD -- it makes no difference.) With CUDA, you are doing these FX very fast.

-------

When you use CF, each CF frame during editing is uncompressed to RGB or YUV. No NLE can work on a compressed frame.

The only difference with using CF is that you originally uncompressed an AVCHD frame to YUV and then recompressed it to CF. Recompression MUST degrade the uncompressed YUV frame because the very definition of compression is DISCARDING information! It doesn't matter if you don't notice it. All compression is designed to toss out what the designer hopes you won't notice. But something MUST get removed from each uncompressed YUV frame -- or each CF frame would be the size of an uncompressed frame!

Only if you convert AVCHD to uncompressed and stored it in a file, would the uncompressed YUV frame be PRESERVED. And, that huge frame could never have a better image than that which was in the AVCHD frame.

And, when you edit, that uncompressed frame -- from a huge file -- will be identical to an AVCHD frame that is decompressed on-the-fly. So NO intermediate codec edit can ever be as good as a native edit. That's simply a math fact.

-------

When all the FX have been applied to the uncompressed RGB or YUV frame, it is sent to your monitor. Then it is discarded. It is NOT compressed as an AVCHD frame. It is not compressed as a CF frame. It is not even stored as an uncompressed frame. It is gone.

Every time you view your timeline, your NLE starts over with the untouched source frames which are once again uncompressed to RGB or YUV. That is what NATIVE real-time editing is all about..

When you export, the only difference is each uncompressed RGB or YUV frame is not discarded. It is compressed using your chosen export codec. Each exported frame will have been decompressed ONCE and recompressed ONCE. But, by using CF, each exported frame will have been decompressed TWICE and recompressed TWICE.

=========

What happens if YOU choose to render some or all of a timeline?

The one thing I can tell you is you would never ever render to AVCHD! FCP only renders to ProRes 422 (you get to chose its parameters) and MC only renders to DNxHD (you get to chose its parameters) or DVCPRO-HD. Each of these NLEs has a menu where you specify the render codec.

I have no idea where you tell Premiere what codec should be used. Since you own the CF codec, that's the one you should choose.

When you export, you decide whether or not to use rendered files. FCP will not let you. With FCP you know each exported frame will have been decompressed ONCE and recompressed ONCE.

With MC you delete the render files prior to export to force it not to use the files. I suspect the same is true of Premiere. Of course, if you are in a hurry you can use the render files even though some quality will be lost.

=======

The bottom line is that back in the days of DV someone coined the term "acquisition" codec because of the way NLEs worked in the olden days -- and it keeps being used. An acquisition codec is a SOURCE codec. It is never an EDIT codec. And, it may be an export codec -- that is incorrectly called a "distribution" codec.

[Mike Burgess]

HI Steve.

When you said in your last post that you would never, ever render to AVCHD. Can you expand on that, explain what you are talking about? Are you saying not to edit AVCHD and burn an AVCHD DVD?

Thanks,
Mike

[Robert Young]

Steve
I appreciate your exposition on what is a complex and somewhat controversial subject.
You make a good case for the durability of AVCHD.
I certainly would not rule out the possibility that I might someday switch over to editing HD in native formats.
However, presently I have a work flow with CF that is quick, and an absolute no brainer.
I can easily mix material that was originally in different formats from different cameras.
I get splendid, predictable, consistant results no matter how involved the project is, or what sort of delivery is required. It's like- if it ain't broke, why fix it.
Maybe I'm sort of like the kid who is asking his mother about the stars and planets. She says "Why don't you ask your father- after all, he's an astronomer"
Kid replys "I don't want to know THAT much about it" :)

[Steve Mullen]

Originally Posted by Mike Burgess

HI Steve.

When you said in your last post that you would never, ever render to AVCHD. Can you expand on that, explain what you are talking about? Are you saying not to edit AVCHD and burn an AVCHD DVD?

Thanks,
Mike

Render has at least two meanings:

1) Render FX, means perform the FX math on the RGB/YUV frames. No compression is used -- the frames are only displayed.

2) Render FX, means perform the FX math on the RGB/YUV frames AND compress the resulting frames to a file. One would never compress using any long-GOP codec.

Thus, during editing, one would not "render" to AVCHD.

During EXPORT, however, one can certainly compress to AVCHD. Many call exporting "rendering" which it is. But, rending during export can be to ANY codec.

PS: "It's like- if it ain't broke, why fix it." That's true, but most modern NLEs already let you mix anything on a Timeline. That's what's called an "open timeline." All the sources are native. They can be SD and HD. They can be progressive and interlaced. They can have different frames sizes. They can even have different frame rates.

Now I'm not claiming you can do these things with Premiere. The code base for CS5 -- with the exception of CUDA -- is still the old old Premiere Pro with bug fixes. So YOU may need to use CF. I'm only saying the reason you gave for using CF is not valid. And, the CineForm marketing materials have not been valid for years -- and are really not valid with CUDA.

And, while nothing may be broken, transcoding to CF is a huge waste of time and space because you have CUDA.

=======

I am curious about the report of gamma issues. Is this all AVCHD? Just Sony? Just Pana? Just Canon? Just 24Mbs? It if it's so major, why hasn't Adobe fixed it by now?

[Robert Young]

Originally Posted by Steve Mullen

And, while nothing may be broken, transcoding to CF is a huge waste of time and space because you have CUDA.

Transcoding to CF on today's systems is way faster than RT and can be done directly off of the camera card if you want to cut it to the minimum. Time is not a real issue.
As for space- space is the cheapest part of this whole deal anymore. That's hardly a problem either.

[Mike Burgess]

Originally Posted by Steve Mullen

Render has at least two meanings:

1) Render FX, means perform the FX math on the RGB/YUV frames. No compression is used -- the frames are only displayed.

2) Render FX, means perform the FX math on the RGB/YUV frames AND compress the resulting frames to a file. One would never compress using any long-GOP codec.

Thus, during editing, one would not "render" to AVCHD.

During EXPORT, however, one can certainly compress to AVCHD. Many call exporting "rendering" which it is. But, rending during export can be to ANY codec.

PS: "It's like- if it ain't broke, why fix it." That's true, but most modern NLEs already let you mix anything on a Timeline. That's what's called an "open timeline." All the sources are native. They can be SD and HD. They can be progressive and interlaced. They can have different frames sizes. They can even have different frame rates.

Now I'm not claiming you can do these things with Premiere. The code base for CS5 -- with the exception of CUDA -- is still the old old Premiere Pro with bug fixes. So YOU may need to use CF. I'm only saying the reason you gave for using CF is not valid. And, the CineForm marketing materials have not been valid for years -- and are really not valid with CUDA.

And, while nothing may be broken, transcoding to CF is a huge waste of time and space because you have CUDA.

=======

I am curious about the report of gamma issues. Is this all AVCHD? Just Sony? Just Pana? Just Canon? Just 24Mbs? It if it's so major, why hasn't Adobe fixed it by now?

Thanks Steve.

Little by little I am learning, although I have a long ways to go. It will take me at least 5 more times reading your post before it begins to become somewhat clearer. I am, after all, a very slow learner. I do appreciate your patience and explanations. Thanks again.
Mike

[Robert Young]

Originally Posted by Steve Mullen

When all the FX have been applied to the uncompressed RGB or YUV frame, it is sent to your monitor. Then it is discarded. It is NOT compressed as an AVCHD frame. It is not compressed as a CF frame. It is not even stored as an uncompressed frame. It is gone.

Every time you view your timeline, your NLE starts over with the untouched source frames which are once again uncompressed to RGB or YUV. That is what NATIVE real-time editing is all about..

When you export, the only difference is each uncompressed RGB or YUV frame is not discarded. It is compressed using your chosen export codec. Each exported frame will have been decompressed ONCE and recompressed ONCE.

Good call on the technology of native format editing.
I have been quite underestimating the process and it's potential.
This little discussion has been an eye opener for me- I appreciate your patience.

[Steve Mullen]

Originally Posted by Robert Young

Transcoding to CF on today's systems is way faster than RT and can be done directly off of the camera card if you want to cut it to the minimum. Time is not a real issue.
As for space- space is the cheapest part of this whole deal anymore. That's hardly a problem either.

By the way, I am surprised at how long it takes to read AVCHD files WITHOUT conversion. And that's because as each file is read, it is processed to yield two additional files. One has non-video data and the other I suspect is a "hint" or "index" file that NLEs can use for editing.

Even Sony Vegas processes the data during import and creates a waveform file.

Bottom-line, native editing may well not be faster than is conversion to an Intermediate file.

And, in favor of intermediate editing, those without CUDA, will find it difficult to edit more than one stream with no more than 1 second transitions. EDIUS does very well, but you need four real cores -- maybe 8 cores.

So now that I've probably confused folks, let me sort it out. Native IMHO is better, but only if you have a monster computer -- which means not a laptop. Intermediate editing can use a far less powerful computer, but you need tons of storage -- which once again is not a laptop.

For those of us with laptops, 720p30 is a better match because each frame is half as big. Laptops simply aren't ready for 720p60 or 1080p30. It will take several more years to get 6 to 8 real cores in a laptop. And, it will likely burn your lap. :)

So far, with my MBP, I've found the free copy of Vegas running under XP under Bootcamp on a 2.53GHz I5 to be "reasonable" -- assuming you like editing with Vegas.My other favorite is far from free. Avid Media Composer v5 is REALLY sweet. It is now far more like FCP -- without FCP's negatives. For the classically trained editor, MC is very intuitive. For someone without decades of editing, Vegas is likely to be EZ to learn.

[Noa Put]

By the way, I am surprised at how long it takes to read AVCHD files WITHOUT conversion. And that's because as each file is read, it is processed to yield two additional files. One has non-video data and the other I suspect is a "hint" or "index" file that NLEs can use for editing.

I am a premiere pro cs3 user and bought canopus edius pro specifically to handle my dslr footage on my 3 year old pc, importing those native 1080p files and dragging and dropping it onto the timeline is just a matter of seconds. Right after that I can view my footage, not in realtime though but sufficient to do some very rough cutting.

[Ron Evans]

I mainly edit with Edius 5.5 but also have Vegas 8 and Vegas 9 as well as CS3 on the PC at the moment. The philosophy of Edius is to always edit at the choosen project properties. If that is 1920x1080 then the output is always 1920x1080. This made earlier versions of Edius not work with AVCHD as there was too much processor power needed with the then software algorithms. Vegas from Vegas 8 would edit AVCHD native by reducing the preview resolution to maintain frame rate of project or one can select the preview resolution and frame rate will drop based on PC power/resources.

My sources are from SR11, XR500 and NX5U both Edius 5.5 and Vegas will edit these native if there is only one track on my Q9450 quad core, 8G RAM running WIN Vista 64. For multicam I still need to encode to the Canopus HQ intermediate but I am aware that those with faster processors can do multicam native now in Edius 5.5.

I agree with Steve that native is preferable and went this way with HDV as soon as the editors were able which for Vegas and Edius was some time ago. I will also go full native for AVCHD as soon as I get a PC upgrade.

What one must look at though is the quality of the conversions through to the output. For editing the conversion is to RGB/YUV for preview but for export the conversion is from source through to selected output format. It may well be that choosing intermediates at some point leads to a better outcome depending on the NLE. As an example I prefer TMPGenc encode so I export a HQ file from Edius and do my encoding for discs in TMPGenc.

Ron Evans

[Steve Mullen]

I was reading about the firmware update to the nex 5 today. It will finally get aperture priority mode when shooting video. The comment was that in bright light an nd filter will be needed to keep shutter speed low.

This is true, of course. But, with the vg10 even when you use shutter priority, you still need an nd filter, for two reasons.

At the ideal 1/60, the aperture may be f/22 which is too small. Your video will be soft.

To get a shallow dof, you need to keep the aperture near, but not at, the largest f-stop at the current focal length.

Because you need to do this under varying light levels, it seems a variable density nd filter is a must. That being said, I've found an nd8 (3 stops) to work most of the time outdoors.

It is amazing that with the exception of the new nikon 7000. None of these cameras have a built in nd filter.


Bottom line, getting manual control isn't enough.

[Robert Young]

Originally Posted by Steve Mullen

Because you need to do this under varying light levels, it seems a variable density nd filter is a must. That being said, I've found an nd8 (3 stops) to work most of the time outdoors.

I shot a big, long event gig today completely with the VG10.
Almost all of the shooting was ENG style, a lot of AF and AE. All of the daylight shots were @ 1/60, and using the VariND. I could easily & quickly stay b/t f 3.5 and f 8 depending on what I needed.
It all seemed to work very well. The cam was easy to shoot with under pressure, and the VariND provided the missing link for outdoor shooting.
I'm just now downloading the footage- we'll see what I've got.

[David Bankston]

Originally Posted by Bill Koehler

Yes. But I would like to know first:
1. If the camera outputs uncompressed HDMI when shooting.
2. The camera shoots interlaced or progressive, even if the latter is most likely in psf.

30P or 60i – Which Is It? - Quoted from Luminous-Landscape Site

Most people interested in video know that interlaced is old tech, that all modern TVs are progressive devices, and that therefore Progressive video recording is to be preferred, because it potentially offers twice the vertical resolution of interlaced. But Sony AVCHD video cameras (at least until one gets to their higher-end prosumer and pro gear) is speced as recording 60i. But, when you read the fine print you discover that qt least some of these cameras (such as the VG10) really aren't recording 60i, they're capturing 30P and placing it in a 60i "wrapper".

Confused? Well, you're not the only one. I'm unable to write a knowledgeable treatise on the subject, but in brief, what's going on is that the AVCHD standard doesn't include 30P, but it does include 60i. This is relevant for those that want to burn a Blu-Ray disk because Blu-Ray uses the AVCHD standard. (Ok – everyone that burns Blu-Ray disks please hold up your hands. Humm. I thought so.)

So what we have is confusion. Most Sony's AVCHD cameras (including the VG10) capture 30P, but make it appear to other devices as 60i. Most non-linear editors, such as Final Cut, figure this out by themselves, and when you check the Info screen will confirm that the footage is 30P.

Other manufacturers who adhere to the AVCHD standard, for reasons best known only to themselves, try and avoid this confusing state of affairs, capture 30P, and also record as 30P. No interlace confusion.

So if you are looking at a Sony AVCHD camcorder that is speced as 60i, be assured that where the rubber meets the road (in your NLE) you will likely be working with true 30P footage. I can't imagine how many camcorder sales Sony loses each year because of this confusion.

[Ron Evans]

David,
Most Sony AVCHD camcorders do not capture at 30P they are in fact true interlaced 60i with the temporal motion of 60 exposure a second( the field rate) . Other than the VG10 and a few others like the Bloggies( which record at 30P and 60P but not 60i) the Sony's are all 60i.

I for one like this as I really do not like the temporal motion of slow frame rates. I too would love progressive but at 60p or faster. Some of the latest Sony AVCHD camcorders like the CX550 will output 60P from the HDMI essentially deinterlacing in camera to match most displays which are 60P or faster in NTSC lands. A VG10 with selectable progressive frame rates would be nice for those who want a film look and those who want a real look. But neither 30P or 60P full resolution( 1920x1080) are part of the spec!!!!

Ron Evans

[Steve Mullen]

"30P or 60i – Which Is It? - Quoted from Luminous-Landscape Site"

When "reviewers" ask a question like this it indicates that still photographers don't understand video. And, in most cases when they review the video capabilities of these new cameras they start to babble nonsense. But, it is understandable.

When Sony USA PR doesn't "understand," it shows how Sony/Pana/JVC Japan works. The engineers in Japan understand that their chip can't run faster than 30Hz. They know BD and AVCHD don't record at 30p. And, they know if you put the 30p in a file tagged as 60I -- that the file will be compatible with EVERYTHING. So they send Sony USA the specs that say "60i." Since all the consumer marketing groups are staffed with low-paid boys and girls (mostly the latter) they simply print what Japan sends them. (They don't know to ask questions or are too afraid to ask questions.) And, reviewers print what the PR sheet says.

The same confusion came with AVCHD Lite which claimed to be 60p. But, the chip runs at 30Hz.

In both cases, the goal was to make files that are compatible with equipment and NLE. There is no 720p24, 720p25, or 720p30 in the BD/AVCHD world. There are only 720p50 and 720p60. Which by the way are what are broadcast. Likewise 50i and 60i.

All the NLEs I've tested see the clips as 50i or 60i. Make a 50i or 60i Sequence and all will be well. You can go right to AVCHD DVDs or BDs.

Unfortunately, for the non-consumer buyer, all will not be well. We'll notice the odd motion from 25fps and 30fps. We'll wonder how much quality will be lost when our NLEs needlessly perform interlace scaling for FX. Alternately, if we drop the clips in a 25p or 30p Sequence, we'll be pained that each clip will be erroneously deinterlaced.

We'll worry that if we upload an interlace file to Vimeo, will they automatically deinterlace it causing a loss in vertical resolution. And, we'll ask "should we deinterlace when making 720p25/720p30 for the Net."

Lastly, we'll wonder if the deinterlacer in our HDTVs are adaptive or not. Will they Weave deinterlace or only bob deinterlace?

So for the intended consumer market, calling it 50i/60i makes a good deal of sense. This group will buy because it is very sexy looking and has the magic buzzword "BIG CHIP." And, it does shoot beautiful video!

When we try to buy a $2000 camcorder rather than a $6000 product, we doing exactly what Sony does not want us to do. This likely explains why no 24p. We are supposed to buy products from the PRO division!