general questions about color space conversions

[jieve]

Hello,

I am relatively new to the in depth details of digital video, so a couple questions. I have captured footage from my miniDV cam over the last couple years using firewire and Adobe Premiere Pro 1.5, and at the time knew nothing about video color formats (YUV, RGB). I basically wanted an exact copy from the tape to my hard drive to edit later. I captured via firewire to the DV NTSC profile, didn’t add any effects, only made some cuts here and there so didn’t have any red bar, then exported as Microsoft DV-AVI without recompressing. I have been reading mixed information about how premiere pro 1.5 deals with color spaces. While I have read in adobe’s literature that that version supported native YUV video and only converted to RGB for playback, someone in an anime video forum mentioned that upon import premiere pro converts to 4:4:4 YUV, then I guess re-converts to 4:1:1 upon export with the Microsoft DV-avi codec. I would assume, based on the adobe literature, although it was mostly limited to advertisements, that the video would be captured at 4:1:1 YUV then exported in the same 4:1:1 color space without any conversions since all of the settings were DV NTSC. Am I correct in my original thinking that I didn’t lose anything in any color conversions?

Does it matter what DV codec I use when exporting the final video? Seems like it shouldn’t matter, bits on a tape should be converted to the same bits on the hard drive regardless of the DV codec, although people have talked about major differences between codecs.

When using RGB color correction in premiere pro, does it convert the color space then during editing?

Also, am I correct that upon playback by a media player, the codec that is responsible for the format that is playing converts from YUV to RGB if YUV is the source color scheme? For example, the DVD codec required to play DVDs converts the YUV data from the DVD to RGB for display?

Finally, I ripped a number of my DVDs using DVDShrink and DVD43 uncompressed to back them up. Does DVDShrink change the color space to RGB upon ripping or does it stay YUV? I’m assuming it stays YUV since burning the video_ts folder back to a DVD plays fine on any dvd player.

[jagabo]

I'm believe older versions of Premiere Pro convert everything to RGB for filtering. But I don't use the software so I don't know for sure.

In general terms, for YUV 4:1:1 to 4:4:4 the conversion is usually simple duplication of the U and V values over the four Y values. The reverse may take an average of the four U and V values or subsample. When starting with 4:1:1 and converting to 4:4:4 and back the conversion is lossless. When starting with 4:4:4, where the four U and V values are different, there are losses.

There are codecs where the conversion from 4:1:1 to 4:4:4 will interpolate the U and V values. In that case, back and forth conversion leads to generational losses.

What happens in a player varies from system to system. Typically, the output from a DV decoder will be either RGB (in which case the codec is doing the YUV->RGB conversion) or YUY2 (the codec converts YUV 4:1:1 to 4:2:2). Most video cards can display YUY2 video directly so it's the video card that converts the YUY2 to RGB on the monitor.

I don't remember what Microsoft's DV codec does. But avoid Panasonic DV Codec. It converts everything to RGB on decompression, and only accepts only RGB on compression. I like to use Cedocida for DV encoding and decoding. It lets you specify what comes out of the decoder: RGB (4:4:4), YUY2 (YUV 4:2:2), YV12 (YUV 4:2:0). It accepts RGB, YUY2, or YV12 for compression.

DVD Shrink works in the DCT domain. The video never leaves the YV12 colorspace.

[jieve]

Thanks, Jagabo, for you consistently knowledgable responses. Here is what I found on the adobe website regarding premiere pro 1.5:

<<< Previous Versions of adobe premiere converted video to the computer friendly RGB color scheme. Then when exported from adobe premiere, converted it back to standard tv signal called YUV. These multiple conversions caused some visual quality degredation. Now all video is stored on your hard drive in it’s native YUV scheme. Only when displaying video on your PC monitor does adobe premiere pro convert YUV to RGB. – From Adobe Digital Video curriculum guide, http://www.adobe.com/education/pdf/dvguide/DV_module_4.pdf
>>>

So I have another question about exporting my DV footage using the export with Microsoft DV-AVI codec in Premiere Pro … I was just reading through a Premiere Pro CS3 tutorial and was reading about the capturing process. What I have been doing is capturing the video, editing (cutting scenes) in the timeline, then exporting the video using the Microsoft DV-AVI codec. The method detailed in the tutorial was to actually select the In/Out points on the tape before capture, then capture the material directly to the hard drive, which does not incorporate any other codec. Despite not having actually added effects to the footage and unchecked recompress during export, have I been actually running the footage through another DV generation and 4:1:1 color compression by exporting as a Microsoft DV-AVI?

[edDV]

Hello,

I am relatively new to the in depth details of digital video, so a couple questions. I have captured footage from my miniDV cam over the last couple years using firewire and Adobe Premiere Pro 1.5, and at the time knew nothing about video color formats (YUV, RGB). I basically wanted an exact copy from the tape to my hard drive to edit later. I captured via firewire to the DV NTSC profile, didn’t add any effects, only made some cuts here and there so didn’t have any red bar, then exported as Microsoft DV-AVI without recompressing. I have been reading mixed information about how premiere pro 1.5 deals with color spaces. While I have read in adobe’s literature that that version supported native YUV video and only converted to RGB for playback, someone in an anime video forum mentioned that upon import premiere pro converts to 4:4:4 YUV, then I guess re-converts to 4:1:1 upon export with the Microsoft DV-avi codec. I would assume, based on the adobe literature, although it was mostly limited to advertisements, that the video would be captured at 4:1:1 YUV then exported in the same 4:1:1 color space without any conversions since all of the settings were DV NTSC. Am I correct in my original thinking that I didn’t lose anything in any color conversions?

If you are just cut editing DV format in a DV project setting (no red bar), Premiere will copy input frames to output without conversion. You have no losses.

Premiere defaults to an internal DV codec but in preferences you can change that to Microsoft or another DV codec. There is no reason to change this setting for normal editing. In cuts mode, the DV codec is only used for monitoring.

Does it matter what DV codec I use when exporting the final video? Seems like it shouldn’t matter, bits on a tape should be converted to the same bits on the hard drive regardless of the DV codec, although people have talked about major differences between codecs.

The internal Adobe codec works fine. As Jagabo said, avoid the Panasonic codec because it clips the 0-15 and 236-255 overshoot levels during conversion to RGB.

When using RGB color correction in premiere pro, does it convert the color space then during editing?

In normal mode Premiere Pro converts to "studio" RGB for filtering and then back to YUV for export. From Premiere Pro 1.5 on, a native YUV environment can be set but this is mostly used for real time SDI I/O and often with hardware assisted I/O cards. Still some filters convert to RGB. It gets complicated since various hardware does some of the filtering (e.g. color correction). Best to discuss thin in specific hardware forums.

[edDV]

Thanks, Jagabo, for you consistently knowledgable responses. Here is what I found on the adobe website regarding premiere pro 1.5:

<<< Previous Versions of adobe premiere converted video to the computer friendly RGB color scheme. Then when exported from adobe premiere, converted it back to standard tv signal called YUV. These multiple conversions caused some visual quality degredation. Now all video is stored on your hard drive in it’s native YUV scheme. Only when displaying video on your PC monitor does adobe premiere pro convert YUV to RGB. – From Adobe Digital Video curriculum guide, http://www.adobe.com/education/pdf/dvguide/DV_module_4.pdf
>>>

So I have another question about exporting my DV footage using the export with Microsoft DV-AVI codec in Premiere Pro … I was just reading through a Premiere Pro CS3 tutorial and was reading about the capturing process. What I have been doing is capturing the video, editing (cutting scenes) in the timeline, then exporting the video using the Microsoft DV-AVI codec. The method detailed in the tutorial was to actually select the In/Out points on the tape before capture, then capture the material directly to the hard drive, which does not incorporate any other codec. Despite not having actually added effects to the footage and unchecked recompress during export, have I been actually running the footage through another DV generation and 4:1:1 color compression by exporting as a Microsoft DV-AVI?

To avoid this confusion you can capture DV with a program like WinDV which simply captures the DV stream from the camcorder to a DV-AVI file. The video data in the file is identical to the data on the MiniDV tape. You can then import this file to your editor.

When set up properly, Premiere can do the same but the settings flexibility allows you to mess this up.

There should be no codec used during DV capture except for monitoring. Premiere should be in DV project mode (aka "native DV mode"). Capture simply copies the input stream to a DV-AVI file. Now for the ways to screw this up.

If your project setting is anything other than DV (e.g. HDV, IMX, uncompressed RGB, uncompressed 4:2:2 YUV, or custom settings), then it is possible to set for realtime conversion where the DV codec is used to convert incoming DV format to the project format. This usually requires hardware assist. For example, a typical SDI uncompressed YUV project setting will decompress DV to 4:2:2 YUV making a very large file that can only be realtime captured and played to a multi-disk RAID. When you stray from DV project settings, you need to know what you are doing.

Back to DV project mode:

When you capture DV to DV project mode, you are capturing native compressed DV so it can be captured to a non-RAID environment. You can drag files to the timeline and freely cut and trim without the red line indicating need for "render" (i.e. conversion to project format) since the source and project formats are the same. Since you aren't disk capacity bound, you can capture the whole tape, drag it to the timeline, and cut edit freely without need to render.

The uncompressed realtime SDI "big boys" have problems you don't have. They capture to DV-AVI, but every file that gets dragged to the timeline gets the red bar and must be "rendered" before it can previewed. In other words, every clip needs conversion to uncompressed YUV. For this reason, they "log" the tape before capture and only capture the good stuff, then they trim clips in the trim window before dragging to the timeline. All this is to minimize the frames that need to be rendered. If they have enough disk space and processing power, they can set to realtime convert DV to project format on capture thus avoiding the need to software render.

Despite not having actually added effects to the footage and unchecked recompress during export, have I been actually running the footage through another DV generation and 4:1:1 color compression by exporting as a Microsoft DV-AVI?

If you kept to DV project mode and exported DV (with "recompress unchecked) you should still have first generation DV (identical to the data on tape).

If you changed the project to YUV, or filtered in any way, then the video would have been processed. A quick reveal that this is happening is the appearance or the red bar. If you can preview the timeline without render, then you are still native DV.

[jieve]

Thanks edDV for your detailed response. So I just decided to test these questions for the hell of it, I took 19 generations of a short clip, then saved 3 frames from both the 1st generation and 19th generation as TIFF, and compared them in photoshop. No difference, and the video file sizes were all the same, just as you had said and I had suspected. Then I captured DV footage from the camera directly to the hard drive, then took that same footage, and exported it using the Microsoft DV-AVI export. The exported footage was 107kB larger, actually. Any idea why this might be? Unfortunately the video was shot at night so tough to tell the difference if any between two clips. I guess I was a little confused, 'cause everytime I export, the dialog box displays "rendering frames".

Also, what exactly do you mean the codec is only used for monitoring? You mean in case any recompressions/renderings due to effects are necessary? That would make sense.

<<If you changed the project to YUV, or filtered in any way, then the video would have been processed. A quick reveal that this is happening is the appearance or the red bar. If you can preview the timeline without render, then you are still native DV.>>
I thought the DV was captured to the hard drive as YUV? I'm using Premiere Pro CS3 now.
<<<Premiere defaults to an internal DV codec but in preferences you can change that to Microsoft or another DV codec. There is no reason to change this setting for normal editing. In cuts mode, the DV codec is only used for monitoring.>>>

Hmmm, I looked for this setting in Preferences and couldn’t find it. Maybe it shows up under export when you have the other codecs installed? The only useful option I always have is export to Microsoft DV-AVI so I assumed it was using the Microsoft codec by default.

<<<In normal mode Premiere Pro converts to "studio" RGB for filtering and then back to YUV for export. From Premiere Pro 1.5 on, a native YUV environment can be set but this is mostly used for real time SDI I/O and often with hardware assisted I/O cards. Still some filters convert to RGB. It gets complicated since various hardware does some of the filtering (e.g. color correction). Best to discuss thin in specific hardware forums.>>>

What exactly do you mean a native YUV environment can be set? Isn’t the video simply captured in YUV 4:1:1 and kept that way? Sure not totally sure I understand how this works while one is editing. For example, if I were to apply a color correction RGB filter to say a short clip from a sequence to correct a blue color cast, would the program then convert that section to RGB, apply the filter, then convert back to YUV? Or is that reconversion another filter I need to add? Seems like the multiple conversions could create inconsistencies in color between the rendered and unrendered parts of the sequence. What is the best way to go about doing editing (applying filters) under these conditions without multiple color space conversions?

Thanks!

[jagabo]

Then I captured DV footage from the camera directly to the hard drive, then took that same footage, and exported it using the Microsoft DV-AVI export. The exported footage was 107kB larger, actually. Any idea why this might be?

Within an AVI file there are several optional data chunks that can effect the size of the file. Some AVI writers will include an ODML header, some won't. Some will align chunks on 1K boundaries, some on 4k boundaries, etc. (they add JUNK chunks to achieve this). The interleaving of audio chunks can vary. All these things can cause the size of the file to vary even if the audio and video are identical.

Also, what exactly do you mean the codec is only used for monitoring?

What he meant was that, while capturing, the software only has to get data from the firewire port and write it to a file. But most capture programs will also display the video so you can see what's going on. That is where the codec comes into play. The video has to be decompressed in order for you to see it. This has no effect on what is going into the file.

[edDV]

Thanks edDV for your detailed response. So I just decided to test these questions for the hell of it, I took 19 generations of a short clip, then saved 3 frames from both the 1st generation and 19th generation as TIFF, and compared them in photoshop. No difference, and the video file sizes were all the same, just as you had said and I had suspected. Then I captured DV footage from the camera directly to the hard drive, then took that same footage, and exported it using the Microsoft DV-AVI export. The exported footage was 107kB larger, actually. Any idea why this might be? Unfortunately the video was shot at night so tough to tell the difference if any between two clips. I guess I was a little confused, 'cause everytime I export, the dialog box displays "rendering frames".

Also, what exactly do you mean the codec is only used for monitoring? You mean in case any recompressions/renderings due to effects are necessary? That would make sense.

Jagabo answered most of this. Besides video, you have PCM audio and metadata variations. Also, DV transfer and files can be Type 1 or Type 2. Type 1 has audio interleaved and requires a demux (deinterleave) to play. Type 2 has Type 1 interleave plus an additional separate copy of audio. This speeds access for edit programs and reduces CPU load. Thus Type 2 files are larger than Type 1. In the past when CPUs were weak, Type 2 was required. Today most edit programs can read either.

In the DV cut case of "rendering frames", it is just a copy and wrapper processing. If you process the source frames (e.g. filter, composite or transition), they normally would be converted to RGB, processed and rendered for preview display (these are stored in the Adobe temp files directory). On export, these would be converted to your export format.

<<If you changed the project to YUV, or filtered in any way, then the video would have been processed. A quick reveal that this is happening is the appearance or the red bar. If you can preview the timeline without render, then you are still native DV.>>

I thought the DV was captured to the hard drive as YUV? I'm using Premiere Pro CS3 now.

DV is captured to the hard drive as DV which is a form of 4:1:1 YUV (actually YCbCr). HDV is 4:2:0 YCbCr MPeg2. In most editors "YUV" means uncompressed 4:2:2 YCbCr. If that was your project format setting, both DV and HDV would need format conversion (red line) whereas SDI (SMPTE 259M/292M) import would not.

<<<Premiere defaults to an internal DV codec but in preferences you can change that to Microsoft or another DV codec. There is no reason to change this setting for normal editing. In cuts mode, the DV codec is only used for monitoring.>>>

Hmmm, I looked for this setting in Preferences and couldn’t find it. Maybe it shows up under export when you have the other codecs installed? The only useful option I always have is export to Microsoft DV-AVI so I assumed it was using the Microsoft codec by default.

It really doesn't matter which DV codec is used in Premiere, Vegas, AVID, etc. They all come with licensed optimized 8bit/10bit codecs and performance is similar. The Microsoft DVSD codec is more generic and not as optimized. In contrast, if you were using Virtualdub for example, you would need to pay for these premium DirectShow compatible codecs. The best free one is Cedocida for the older "Video for Windows". Avoid the free Panasonic DV codec.

<<<In normal mode Premiere Pro converts to "studio" RGB for filtering and then back to YUV for export. From Premiere Pro 1.5 on, a native YUV environment can be set but this is mostly used for real time SDI I/O and often with hardware assisted I/O cards. Still some filters convert to RGB. It gets complicated since various hardware does some of the filtering (e.g. color correction). Best to discuss thin in specific hardware forums.>>>

What exactly do you mean a native YUV environment can be set? Isn’t the video simply captured in YUV 4:1:1 and kept that way? Sure not totally sure I understand how this works while one is editing. For example, if I were to apply a color correction RGB filter to say a short clip from a sequence to correct a blue color cast, would the program then convert that section to RGB, apply the filter, then convert back to YUV? Or is that reconversion another filter I need to add? Seems like the multiple conversions could create inconsistencies in color between the rendered and unrendered parts of the sequence. What is the best way to go about doing editing (applying filters) under these conditions without multiple color space conversions?

DV is captured as 4:1:1 YCbCr and kept that way as source files. Processed frames are stored as RGB temp files separate from the original DV source. They aren't converted from RGB (except for timeline preview) until you specify an export format. This prevents repeated conversion should you next take your color corrected clip, transform it and then composite over a background, etc. All serial (concatenated) processing is done in RGB and then export converted once.

The so called "studio RGB" format allows conversion from YCbCr to RGB and back without levels shift and with minimal color space conversion artifacts. A more detailed explanation gets into 16-235 vs. 0-255 digital levels and 4:1:1 or 4:2:0 to 4:4:4 color processing artifacts.

You don't need to worry about the internal RGB. Premiere does all the conversions internally. That is why you are paying the big bucks. A typical user doesn't interact with the temp files.

Setting up an SDI SMPTE 259M "YUV" project environment gets more complicated. It is usually done with special hardware acceleration cards that take over portions of the filter/translation/compositing process to allow real time editing for supported modes. This also gets into 10bit vs 8bit 4:2:2 YCbCr processing. This is rarely done on a home machine although some use the affordable Matrox cards. In that case, you would follow Matrox instructions.

[jieve]

edDV and Jagabo, thanks guys for taking the time to answer my questions.

So I’m curious, then, about two things. When compressing to MPEG2 or using some other spatial / temporal compression scheme from 4:4:4 YUV, when does the color conversion take place? For example, when MPEG2 is looking at frames to see where there are changes in motion, does the color conversion to YUV 4:2:0 take place first or afterwards?

Also, I’m curious about the general way that DV compression works. From what I understand, the sensor records the pixels RGB, then the information is sent through an algorithm to compress to 4:1:1 YUV in the case of NTSC, then stored on tape as such. However, is this color conversion enough to compress the video 1:5 (or was it 1:7) as it is from uncompressed or is there something else going on here as well?

[jagabo]

So I’m curious, then, about two things. When compressing to MPEG2 or using some other spatial / temporal compression scheme from 4:4:4 YUV, when does the color conversion take place? For example, when MPEG2 is looking at frames to see where there are changes in motion, does the color conversion to YUV 4:2:0 take place first or afterwards?

Maybe edDV can answer about Premiere in particular. In general, it depends on the software you're using. An editor may pass RGB 4:4:4 to the MEPG encoder and then the MPEG encoder converts to YV12 (4:2:0) and compresses. Or the editor might pass YUY2 (4:2:0) and the encoder converts to YV12. Or the editor may pass YV12 directly.

Also, I’m curious about the general way that DV compression works. From what I understand, the sensor records the pixels RGB, then the information is sent through an algorithm to compress to 4:1:1 YUV in the case of NTSC, then stored on tape as such. However, is this color conversion enough to compress the video 1:5 (or was it 1:7) as it is from uncompressed or is there something else going on here as well?

It's much more complicated than that. The sensors in the camera may be RGB, YUV, or some other format. There isn't usually a 1:1 mapping of RGB cells to pixels. There are typically many more green cells than red or blue. There are all different kinds of arrangements of the the subpixels.

In DV 4:1:1 you have a 720x480 Y plane, a 180x480 U plane, and a 180x480 V plane. That's total of 720x480 + 180x480 + 180x480, or 518,400 bytes per frame. Multiply by 30 fps and you get 15,552,000 bytes per second.

We know that a DV AVI runs about 3,600,000 bytes per second. About 192,000 bytes of that is audio. So 3,408,000 bytes per second for the video.

15,552,000/3,408,000 is roughly 5. So the 5:1 compression that's usually quoted for DV is relative to the YUV 4:1:1 source. This comes from DCT compression.

[jieve]

Understood.

<<<DV is captured as 4:1:1 YCbCr and kept that way as source files. Processed frames are stored as RGB temp files separate from the original DV source. They aren't converted from RGB (except for timeline preview) until you specify an export format. This prevents repeated conversion should you next take your color corrected clip, transform it and then composite over a background, etc. All serial (concatenated) processing is done in RGB and then export converted once.

The so called "studio RGB" format allows conversion from YCbCr to RGB and back without levels shift and with minimal color space conversion artifacts. A more detailed explanation gets into 16-235 vs. 0-255 digital levels and 4:1:1 or 4:2:0 to 4:4:4 color processing artifacts.>>>

So can RGB24 actually be mapped directly to YUV 4:4:4 (24bit) and back without color conversion loss, is that what studio RGB is? Or is it like mapping back and forth between RGB and CYMK, where values get changed? I'm getting from this that mapping from YCrCb 4:1:1 to 4:4:4 entails no loss, but from 4:4:4 to 4:1:1 the chrominance values get re-averaged, but what about the 4:4:4 RGB conversion and back? Or am I getting into too complicated of an explanation here?

[edDV]

And for DV format that DCT compression is all intraframe (inside each individual frame similar to JPEG) so there is no motion compression over GOPs like in MPeg. Each DV frame is intact and can be cut edited without processing.

[edDV]

edDV and Jagabo, thanks guys for taking the time to answer my questions.

So I’m curious, then, about two things. When compressing to MPEG2 or using some other spatial / temporal compression scheme from 4:4:4 YUV, when does the color conversion take place? For example, when MPEG2 is looking at frames to see where there are changes in motion, does the color conversion to YUV 4:2:0 take place first or afterwards?

It depends on the project. Let's say you are editing a DV project in Premiere and have a mix of cuts and some processed frames on the timeline. The processed frames were converted to RGB before processing and are stored in RGB Temp files on your "scratch disk". The unprocessed frames are still referenced from your captured DV source files which are still 4:1:1 YCbCr DV.

If you export to 4:2:0 YCbCr MPeg2 as 720x480i, the encoder needs to convert DV YCbCr 4:1:1 frames to MPeg2 YCbCr 4:2:0. It will also convert the 4:4:4 RGB temp frames to MPeg2 YCbCr 4:2:0 as it steps through the timeline.

Note that for a DV project, the RGB temp frames preserve interlace, so if the MPeg2 file is specified as 720x480i (lower field first) then interlace is preserved through Premiere processing. Uncompressed "YUV" projects would be handled similarly. MPeg or progressive project settings would be handled differently.


Note that everything I've said about Premiere Pro applies equally to Vegas and other similar DV native editors.

[edDV]

...For example, when MPEG2 is looking at frames to see where there are changes in motion, does the color conversion to YUV 4:2:0 take place first or afterwards?

This gets into the inner workings of the MPeg2 encoder. The Mainconcept MPeg encoder used in Premiere and Vegas has the ability to track motion over a mix of project specified frames (in this case 4:1:1 480i DV) and 480i RGB temp frames. I'm not sure how Mainconcept does this but my guess is the motion detector looks at luminance from RGB->YCbCr converted frames. This would eliminate influence from chroma on motion detection.

[edDV]

Understood.

<<<DV is captured as 4:1:1 YCbCr and kept that way as source files. Processed frames are stored as RGB temp files separate from the original DV source. They aren't converted from RGB (except for timeline preview) until you specify an export format. This prevents repeated conversion should you next take your color corrected clip, transform it and then composite over a background, etc. All serial (concatenated) processing is done in RGB and then export converted once.

The so called "studio RGB" format allows conversion from YCbCr to RGB and back without levels shift and with minimal color space conversion artifacts. A more detailed explanation gets into 16-235 vs. 0-255 digital levels and 4:1:1 or 4:2:0 to 4:4:4 color processing artifacts.>>>

So can RGB24 actually be mapped directly to YUV 4:4:4 (24bit) and back without color conversion loss, is that what studio RGB is? Or is it like mapping back and forth between RGB and CYMK, where values get changed? I'm getting from this that mapping from YCrCb 4:1:1 to 4:4:4 entails no loss, but from 4:4:4 to 4:1:1 the chrominance values get re-averaged, but what about the 4:4:4 RGB conversion and back? Or am I getting into too complicated of an explanation here?

No. CYMK is never used for video (long color theory discussion).

"Studio RGB" is a set of conversion equations that allows Rec601 (8bit or 10bit/component) YCbCr (4:1:1, 4:2:0, 4:2:2 or 4:4:4) to be converted to 4:4:4 (24 or 30bit) RGB and then back without levels shift or corruption from clipping. I'm trying to avoid breaking into math equations.

Math here--> http://books.google.com/books?id=ra1lcAwgvq4C&pg=RA1-PA309&dq=%22studio+RGB%22+Poynton...age&q=&f=false

All conversion has some effect on color balance or color resolution. If 4:4:4 YCbCr source was used, these color effects would be small to none. When 4:2:2, 4:2:0 or 4:1:1 YCbCr->RGB->YCbCr sources are used, luminance resolution gets "corrupted" to some degree by lower resolution chroma channels. Chroma balance may see shifts. This is one reason why film source is often used for color sensitive commercial shoots. 4:4:4 sourced video (from film or YCbCr video cameras) can be color matched to "Pantone" colors. If the processing path (editor project timeline) is also 4:4:4, a reasonable Pantone color match can be held through video effects processing. This was first done in the 80's with a 4:4:4:4 YCbCrA four level realtime digital compositing processor called the D/FX Composium which received a 1990 Technical Emmy for Best Technology.