A naive question

[Vincent Oliver]

Does anyone have a simple explanation of what Mbps actually means in movie capture terms, My EX3 uses 35Mbps and my Canon XF305 has 50Mbps. I also notice on the Nikon D7100 it captures at 12 & 24Mbps and the Panasonic GH3 captures using a variety of Mbps from AVCHD 17, 24 & 28, MP4 4, 10 & 20 and MOV 50 to 72Mbps (ALL INTRA).

I am assuming that a higher bit rate will carry more image information, but what gets thrown out at the lower bit rates?

Any good, but not too technical, information would be much appreciated.

Thanks

[Dave Sperling]

A big topic and a big question...
With the proliferation of not only in-camera compression systems but also systems on external recorders, it becomes an equation with too many variables for a single solution.
That said -- look up inter-frame compression and intra-frame compression as a starting point. (the analaogy here is the difference between the amount of data needed to send an unsigned and a signed version of the same document -- in an intra-frame system, the entire document must be sent both times -- in an inter-frame system, the document is sent once and the signature is sent once, hence making 'better use' of the data)
More efficient codecs often require more computing power (such as h.264), while less compressed may require less work by the processor.
Within a given system, recording at a higher data rate should produce a better image. Many systems reduce color resolution to get a higher compression ratio, which becomes an important issue when filming green or blue screen mattes. Higher data/bit rates are also helpful when doing extensive color correction. I've just finished with the color correction process for a film we shot uncompressed (F3 s-log/ Gemini) and was extremely happy with the quality it produced and the amount I could adjust it in post. The producer was not quite as happy with the fact that we had 42TB of original camera data (and an equal quantity of backup) from our 20 day shoot.
Among different compression systems, just because a data rate is higher does not intrinsically mean that the image will be better than that provided by a lower data rate from a codec using a different compression system.
So now that I've totally confused you, here's a bit of info to think about:
Within a given Mbps data record rate and codec, 24fps should yield a better 'quality' image than 30p because the data per frame will be higher!
Of course I have certain clients who want high data rates, and others who want the faster transfer times and lower data storage requirements of lower data rates.
The reality is, do some critical testing of any data rate you are planning to use, to make sure it accomplishes what you need!

[Robin Probyn]

Originally Posted by Vincent Oliver

Does anyone have a simple explanation of what Mbps actually means in movie capture terms, My EX3 uses 35Mbps and my Canon XF305 has 50Mbps. I also notice on the Nikon D7100 it captures at 12 & 24Mbps and the Panasonic GH3 captures using a variety of Mbps from AVCHD 17, 24 & 28, MP4 4, 10 & 20 and MOV 50 to 72Mbps (ALL INTRA).

I am assuming that a higher bit rate will carry more image information, but what gets thrown out at the lower bit rates?

Any good, but not too technical, information would be much appreciated.

Thanks

Alot depends of what type of work is paying your bills.or that you want to get into.. broadcast HD will be minimum 50 Mbps.. (dont know who came up with that figure,but its pretty much accepted..) hence the big deal of the canon C300 when it came out being able to record 50mbps.. as opposed to F3 35Mpbs.. now rectified by the excellent F5/55 and PMW200..
Alot of non broadcast web based corp shoots are fine with 35 Mbps.. horses for courses ..

[Vincent Oliver]

Thank you Dave and Robin, you have given me a lot to think about. I will shoot some footage on all the cameras I have mentioned and see exactly what the figures mean in a real world shooting situation. (I am reviewing both the Nikon and Panasonic cameras).

My initial thought was that the various Mbps was going to be similar to JPEG compression, maybe it is as shooting at 12Mbps will give more recording time than say shooting at 50Mbps. But what I would like to know what bits of information are being sacrificed and how important are those bits of information? I also have to bear in mind that the cameras under review are going to be used by enthusiasts and semi-pro users rather than high end broadcast professionals.

The more I look into things the more convinced I am that there is a lot of industry waffle about certain features and specifications - i.e. BBC wants a minimum of 50Mbps, yet they are more than happy to broadcast footage shot with a CCTV or mobile phone camera in their news reports.And as for sound quality, I don't think they have a standard as I have to adjust the volume level for each program. Don't get me wrong I do appreciate that a drama production which looks superb and glossy when shot in high definition, but sometimes I feel it goes over the top especially with a period production. I would like to see a slightly grity approach - which I would assume does not require high bit rates, just more creativity.

Just some extra food for thought

[Robin Probyn]

Sorry Im not that techie .. but it seems colour spacing comes into play.. in XDCAM anyway 50 mbps is 422 and XDCAM EX 35 Mbps is 420.. you need someone like Alister to give you the tech side ..
You do get longer recording times with lower bit rates .. and alot depends on if you need to do alot of post production too.. higher the better .. I have a nanoflsh that I dont use much these days but on those forums people would say at very high bit rates there was a sort of diminishing returns as the human eye couldnt actually see any difference but you had huge files.. in HD anyway..

[Vincent Oliver]

Originally Posted by Robin Probyn

people would say at very high bit rates there was a sort of diminishing returns as the human eye couldnt actually see any difference but you had huge files.. in HD anyway..

I think you are hitting the nail on the head Robin, this is what I am trying to find out.

[Steve Game]

Originally Posted by Vincent Oliver

The more I look into things the more convinced I am that there is a lot of industry waffle about certain features and specifications - i.e. BBC wants a minimum of 50Mbps, yet they are more than happy to broadcast footage shot with a CCTV or mobile phone camera in their news reports.And as for sound quality, I don't think they have a standard as I have to adjust the volume level for each program. Don't get me wrong I do appreciate that a drama production which looks superb and glossy when shot in high definition, but sometimes I feel it goes over the top especially with a period production. I would like to see a slightly grity approach - which I would assume does not require high bit rates, just more creativity.

Just some extra food for thought

Vincent,

Firstly, the technical quality of news footage is secondary to its content - every time. It doesn't matter if that on the spot record of a newsworthy event was shot on a lowly mobile phone, a consumer stills or video camera or a fixed security device. If the story is important enough it will be used. I note that the BBC news seems to have stopped putting a caption like 'Amateur Footage' up as an excuse for poor quality as it would be there so often! Only the predictable big news events get the full outside broadcast truck full of kit sent out. Much of the small items are shot with a single ENG device, (Canon XF305 etc.) with the crew shooting 'noddies' afterwards.
As far as the 50Mb/s 4:2:2 requirement is concerned, that was established as the minimum quality that would survive the broadcasting workflow including final compression down to about 9Mb/s for HD transmission. That's one of the reasons that 'Amateur Footage' looks so poor when broadcast as its initial compression induces artifacts that are multiplied as they go down the chain.
There is another issue that affects the quality of the final video, - the quality before any in-camera compression takes its toll. This is especially true with SLR footage where the crude downsampling from a stills sensor resolution of more than 12Mpixels to 1920x1080 for video frequently creates significant artifacts, notably moire. I have a Canon 550D which behaves quite badly when handling images with brickwork backgrounds. At the 40-60Mb/s that the camera compresses to, the moire can be quite intrusive. If it is then further compressed in post, it may well become unusable for anything serious, so this poses serious limitations on just what can be shot with the camera. Compare this with broadcast material where high-end studio or on-location cameras create a clean image quality which is preserved as far down the chain as possible. The final compression to H264 for transmission then has a much easier job maintaining the quality at relatively low bit rates.

[Vincent Oliver]

Steve,

You put a good case forward for the merits of using high bit rates, it all now makes sense from your explanation. Thank you for taking the time to explain things in a very clear way.

[Robin Probyn]

Originally Posted by Vincent Oliver

I think you are hitting the nail on the head Robin, this is what I am trying to find out.

This was up around180/ 200 /220 Mbps .. that the nano can record..

[Garrett Low]

Hmm. I've never thought of trying to explain this in a non highly technical way that went beyond higher bitrates with the same codec will give you more detail. First, you can't make a comparison between the various cameras looking at Mbps only. The Nikon uses variable bitrates so the 24 is a maximum when the camera determines that the frame needs it. AVC is a much more compressed codec than even Sony's HDCAM EX or HDCAM HD (which is what the nanoFlash uses). The Canon XF305 and Sony EX3 both use comparable codecs and algorithms so they are directly comparable. One the bigger differences between those two though are that the EX3 internally records 4:2:0 color space while the XF305 records 4:2:2. So the extra 15Mbps generally will go to the expanded color space and you won't see any increase in detail.

Even within AVC encoding things vary quite a bit. The more compressed a codec is, the lower the bitrate needed to capture the same quality as a less compressed codec.

Vincent, you are correct in the basic assumption that higher bitrates carry more image information. An easy way to think of this is to look at how colors are produced in the digital world. the visible light spectrum actually contains an infinite number of colors if you think of a color as specific wavelength. If we only had two ways of representing color, either a zero for black or a one for white, we would be able to represent two colors. so every color would need to be approximated to either black or white. So the precision of colors are being approximated. now we have to also introduce time into the capture.

So if we say that the dots in a picture can either be 0 or 1 and when we put a bunch of dots together we can get either a big black spot or a complete white spot, or if we vary some white and some black our eyes will see a grey spot. Each spot it one bit. So if I want to have a 3 x 3 spot I need 9 bits. If I want to show this spot once every second I need 9 bits per second. So now I want to make my picture bigger, say 9 x 9, I would need 81 bits per second if I wanted to have every dot accounted for. So if I only have 9 bits available per second, I need to to somehow use 1 bit to cover more than one dot (if I have no compression). One of the most noticeable results usually is usually macro blocking. A way to see some extreme examples of this is to take a short clip into your NLE. Render using a high bitrate, then render the same holding the resolution (say 1920 x1080) but use a very low bitrate. That will give a good example of what happens when you lower the bitrate.

There is a point at which you will not be able to perceive any improvement with a higher bitrate but it all depends on a number of factors.

For the higher bitrates in devices like the nanoFlash, they are using I-frame only or intra-frame which in itself requires a higher bitrate because the compression ratios are less (no temporal compression).

Don't know if that's what you were asking or if that helps but it's an attempt at an explanation.

[Steve Game]

Garrett's assertion that the increase in bandwidth from 35Mb/s XDCAM EX to the various flavours of 50Mb/s 4:2:2 compression (e.g. Canon's MPEG-2 MXF CODEC and now Sony's XDCAM HD 422) are true for the limit of definition in the colour components of the image. However, the luminance information is determined by the resolution of the camera front end, i.e. the lens and the sensors. Assuming that those parts are designed to maximise the performance of a 1920x1080 system, the resulting images will have a resolution approaching 1000 LPPH. With ANY compression, there will be some degradation and artifacts. How relevant this is depends on the manner of processing undertaken in post production, and the limitations of the final playout system.
This is where the broadcasters' rules come into play. Acceptable footage may undergo recoding in ingestion, recoding at grading and further compromise during editing/assembly into the final work. Then comes the big squeeze at distribution/transmission to a fraction of the original camera bandwidth.
As I said in my last post, 'amateur footage' (or more importantly footage from non-professional cameras) is likely to contain artifacts that arise from consumer level front end processing in-camera particularly where the sensor does not have a 1:1 ratio with the number of horizontal and vertical image samples. Hence compliant cameras tend either be of the 3 chip design with each colour captured at the full 1920x1080 samples per frame, or with single sensors, greater than 1 inch in size, where the sensor resolution and image collection method is does not degrade the picture with artifacts.
Also, along with the minimum bitrate and colourspace requirement is the CODEC used. Here, for prime HD programming, MPEG-2 for inter-frame compressed material and sAVC for intra-frame compressed material is specified. There are concessions in allowed CODEC types for the lower 'Tier' cameras, (i.e. those used for news and special, maybe non-critical footage) where even AVCHD can be used, but such use comes with restrictions on the entire post processing regime that are unlikely to be acceptable for general programme generation.
The camera CODEC issues can of course be avoided by the use of external video recorders, like Nanoflash etc., but if the camera has a front end that emblazons the raw video with significant aliasing artifacts, it doesn't matter how pure the external recording CODEC is, those defects are burnt into the image and will come out and bite the production possibly during post, and certainly during transmission.

[David Heath]

Originally Posted by Vincent Oliver

I will shoot some footage on all the cameras I have mentioned and see exactly what the figures mean in a real world shooting situation. (I am reviewing both the Nikon and Panasonic cameras).

I also have to bear in mind that the cameras under review are going to be used by enthusiasts and semi-pro users rather than high end broadcast professionals.

The more I look into things the more convinced I am that there is a lot of industry waffle about certain features and specifications - i.e. BBC wants a minimum of 50Mbps, yet they are more than happy to broadcast footage shot with a CCTV or mobile phone camera in their news reports.

Vincent, you seem to be considering that bitrate in itself is one of the most important factors in determining how good or bad the images from any camera will be. That is not the case.

Firstly, all codecs in common usage tend to be "good enough" that for first generation pictures you're unlikely to see much difference due to codec differences. Any differences you see in basic pictures between the Panasonic and Nikon cameras you talk of are far more likely to be down to front end differences than codec and/or bitrate.

Secondly, bitrate is only one factor in codec quality. It may be true to say that *IF ALL ELSE IS EQUAL* the higher the bitrate, the higher the quality - but if all else is not equal, sensible conclusions from bitrate alone are impossible. And factors which may be involved are not simply the codec itself (MPEG2 v AVC-HD, say), but interframe v intraframe, and a host of other factors. Most significantly, the coder hardware itself. It's the case that a high quality coder (ie expensive!) will do a much better job than a cheap one, even if both have the same bitrate, same compression scheme etc.

As far as your last paragraph above, then the "rules" that broadcasters are applying are based on far more than whims, and it is totally wrong to regard them as just meaningless waffle. As Steve has said, the point is not as regards first generation quality - but to ensure the images pass well through the post and broadcast chain. It's not enough to just look at first generation pictures and think "well, that looks OK!" You could have two cameras, and both may look fine originally - but put both sets of pictures through a broadcast chain and one may stand up well, the other may turn to mush.

If that sounds improbable, then the best analogy I've ever heard compares it to food poisoning. (!!) You may have two meals side by side, and both may look, smell and taste delicious - but it's conceivable one may make you very ill indeed if certain bacteria are present! You won't be able to detect them in advance without lab tests - but they'll still affect your body. And in the case of video it's similar. Without going into full technicalities, aliasing is one of the worst culprits, and one of the most misunderstood. Suffice to say it can be hardly visible in the original image, persist through the broadcast chain, then cause problems in the final compression stages.

OK, yes, mobile phone footage will get used if it's all that's available, but that's no reason to deliberately use equipment which is known may give problems on a pre-planned shoot.

Incidentally, the BBC will want a minimum of 50Mbs with an interframe codec, but a minimum of 100Mbs with an intraframe codec, for very good reasons. See how little bitrate alone tells you?

If that wasn't all bad enough, then how good/bad any compression scheme will perform will depend on the type of pictures themselves. Lots of movement? Lot's of fine detail? Saturated colours? High levels of noise? See what a can of worms the whole subject becomes, and why simplistic comparisons are meaningless?

[Vincent Oliver]

Thank you everyone for taking the time to answer my seemingly naive question, you have all given me plenty of information to digest. I hadn't realised that there would be so many variables to achieve a quality end result. I have so much more to learn.