Reasons for Video/PC levels. - AVS Forum
Forum Jump: 
 
Thread Tools
post #1 of 26 Old 03-19-2012, 01:36 PM - Thread Starter
Newbie
 
Shark321's Avatar
 
Join Date: Nov 2009
Posts: 5
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
As I understand PC monitors work with 24 bits, 8 bits per Channel, which results in 0-255 Levels.

On the other side TVs work with Video Levels, which are 16-235, reducing the total possible colors.

I wanted to know what is the historical reason for that. On the first look it appears to be outright crazy that they've decided to go with less than 24bit for HDTV, even if this biddepth is available since 20 years and could be easily implemented. I've googled a lot, but could find no article why they went this road.
Shark321 is offline  
Sponsored Links
Advertisement
 
post #2 of 26 Old 03-19-2012, 03:32 PM
Advanced Member
 
Rory Boyce's Avatar
 
Join Date: Feb 2000
Location: Fair Oaks, CA
Posts: 861
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Using zero to 255 makes sense in a purely digital word where all content is created as numerical values so all there is is numbers so of coarse all levels should stay as they were created. You have to remember that before digital television there was analog television where both black level and peak levels were adjustable and could easily be and often were wrong. Digital video was being created by feeding analog video through A to D converters. Also with real world video from a camera the peak video level is determined by the scene content, lighting, and lens opening. The chance that no scene content will go above the nominal peak level is for practical purposes zero. So, the people that created the digital system picked digital values that gave margin for error at both limits.

If they had used 0-255 any scene content would have been removed and never recoverable if it exceeded those limits. As it is having the black level a little low or camera level a bit high is fixable later down the chain. Also most professional equipment uses 10 bits not 8 so there are actually four times more possible levels for each component.

Rory
Rory Boyce is offline  
post #3 of 26 Old 03-19-2012, 08:21 PM - Thread Starter
Newbie
 
Shark321's Avatar
 
Join Date: Nov 2009
Posts: 5
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
"Also with real world video from a camera the peak video level is determined by the scene content, lighting, and lens opening."

You mean the peak level varies over time?
Shark321 is offline  
post #4 of 26 Old 03-20-2012, 09:31 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Another important reason for having undershoot <16 and overshoot >240 was to cope with imperfect analogue sources.

Digital equipment originally operated as islands in analogue installations, with analogue (often composite) camera and VTR feeds, as inputs, and analogue vision mixers and VTRs as destinations.

Sharp transitions in analogue signals can have overshoots and undershoots, so even if your sources had perfect black and white levels within the range allowed, it is entirely possible that by the time they hit the digital equipment they would have transitions that spiked above and below these levels. If you clip these spikes you will end up with nasty ringing artefacts, so you allow some latitude above and below.

If you were creating an entirely digital video system from scratch you might not use 16-235 level space, but in the early 80s when this standard was set, you had to handle imperfect analogue video as transparently as possible.

The bigger question is why PC manufacturers still can't properly cope with a standard that has been around for 30+ years properly...

Also - don't forget that most digital video in studios and VTRs/Edits is 4:2:2 YCrCb (8 or 10 bit), with chroma subsampling horizontally (i.e. only 16-20 bits per sample, 8/10 for the Y luminance sample and 8/10 for EITHER Cr or Cb colour difference signals alternating on a luminance sample, by luminance sample, basis)

The final leg of broadcast digital video to the home and DVD/Blu-ray is 4:2:0 YCrCb 8bit, which translates to 12 bits per sample (8 bits for luminance and then 4 bits for CrCb as 1 Cr and 1 Cb 8 bit sample accompanies 4 Y samples)
sneals2000 is offline  
post #5 of 26 Old 03-20-2012, 10:43 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Also for those who think modern TV production is an entirely digital affair - it's still very common for the signals carried over the TV camera cables from cameras to studio control rooms or production trucks to use analogue transmission techniques over triax cables.

So analogue techniques are still part of modern TV production.
sneals2000 is offline  
post #6 of 26 Old 03-20-2012, 12:41 PM
Advanced Member
 
Rory Boyce's Avatar
 
Join Date: Feb 2000
Location: Fair Oaks, CA
Posts: 861
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:


The final leg of broadcast digital video to the home and DVD/Blu-ray is 4:2:0 YCrCb 8bit, which translates to 12 bits per sample (8 bits for luminance and then 4 bits for CrCb as 1 Cr and 1 Cb 8 bit sample accompanies 4 Y samples)

This should not be interpreted to mean that Cb or Cr are actually only carried as only 4 bits. There are simply fewer Cb/Cr samples being sent than there are Y samples. This can be done because The eye has higher resolution for luminance detail than for color detail. The eye/brain combination infers the edges of colored objects based on the luminance edges. The use of much lower resolution chroma channels than luma is a very early form of compression that has been used in color television since it began.

Rory
Rory Boyce is offline  
post #7 of 26 Old 03-20-2012, 10:14 PM
AVS Special Member
 
olyteddy's Avatar
 
Join Date: Oct 2005
Posts: 3,278
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 92 Post(s)
Liked: 209
16 to 235 comes as a result of NTSC standards. The first digital video was simply a digitized version of NTSC. Analog video required Sync pulses that are 'blacker than black' (the 0-16 range) and since the video (when broadcast) was transmitted as an Amplitude Modulated (AM) signal the video was hard limited to less than 235 out of a possible 256. If you overmodulate an AM transmitter those nicely curved sine waves that make up the carrier suddenly have flat tops (resembling square waves) that are rich in harmonics which will essentially 'spray' the airwaves stepping on any other transmission in their way. A definite No-No.
olyteddy is offline  
post #8 of 26 Old 03-21-2012, 02:58 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by olyteddy View Post

16 to 235 comes as a result of NTSC standards. The first digital video was simply a digitized version of NTSC. Analog video required Sync pulses that are 'blacker than black' (the 0-16 range) and since the video (when broadcast) was transmitted as an Amplitude Modulated (AM) signal the video was hard limited to less than 235 out of a possible 256. If you overmodulate an AM transmitter those nicely curved sine waves that make up the carrier suddenly have flat tops (resembling square waves) that are rich in harmonics which will essentially 'spray' the airwaves stepping on any other transmission in their way. A definite No-No.

That's not strictly true - the <16 level space is not used to carry sync pulses, and never has been in the CCIR 601 (which became ITU 601) standard which was the first real 16-235/16-240 level space system AIUI. There isn't' enough of a dynamic range to carry sync pulses properly is there?

CCIR 601/ITU 601 uses SAV and EAV (Start of Active Video and End of Active Video) signalling, which use 0 and 255 as reserved levels for non-video carriage. (This also allows ancillary data to be carried in horizontal and vertical blanking - and thus things like embedded audio, timecode etc. - to be carried in a manner compatible with even the earliest CCIR 656 parallel connected devices from the mid-80s - like D1 VTRs, Quantel Harry etc.)

My memory is that D2 (and the later D3) system which recorded 4fsc digital composite video also used SAV and EAV signalling. I can't remember if 4fsc was based on the same level space, or whether there was more included undershoot and overshoot to avoid subcarrier clipping on saturated low luminance signals (like the blue in 100% bars).

Similarly any suggestion that 7.5 IRE set-up (a feature of North American - but not Japanese - analogue NTSC Composite signals) is linked to the choice of 16-235/240 is also a bit flawed (not that I think you are making that point)

AIUI the main reason for sub-black and super-white was to minimise quality losses caused by analogue signals passing through digital islands (ringing caused by clipped edge transition over-shoots and under-shoots being a real issue) Certainly that was the explanation I was given at Quantel in the late 80s when I worked there on CCIR 656 parallel gear carrying CCIR 601 video.

Your point about the 235 clipping is very well made.
sneals2000 is offline  
post #9 of 26 Old 03-21-2012, 03:02 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by Rory Boyce View Post

This should not be interpreted to mean that Cb or Cr are actually only carried as only 4 bits. There are simply fewer Cb/Cr samples being sent than there are Y samples. This can be done because The eye has higher resolution for luminance detail than for color detail. The eye/brain combination infers the edges of colored objects based on the luminance edges. The use of much lower resolution chroma channels than luma is a very early form of compression that has been used in color television since it began.

Yep - exactly - but describing this system as 24 bit is also misleading. It delivers the same bit depth as 24 bit but not at full spatial resolution.

The bitrate is based on 12 bits per pixel, but this delivers 24 bits of depth, just not at full resolution for all signals. (i.e. with 4:2:0 you get full bit depth and full resolution for the luminance, with full bit depth but at half the horizontal and half the vertical resolution for the two colour-difference signals)
sneals2000 is offline  
post #10 of 26 Old 03-21-2012, 03:14 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
http://tech.ebu.ch/docs/techreview/t...rec601_bbc.pdf
For info - there is some discussion on page 9 about meetings in the early 80s on deciding the 16-235 and 16-240 range, and it explains why the Cr/Cb and Y range are different.
sneals2000 is offline  
post #11 of 26 Old 03-21-2012, 03:41 AM
AVS Special Member
 
dr1394's Avatar
 
Join Date: May 2002
Location: Mizar 5
Posts: 3,149
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 11 Post(s)
Liked: 19
Quote:
Originally Posted by sneals2000 View Post

Your point about the 235 clipping is very well made.

Is it? If you consider that NTSC is modulated "upside down" (white is a lower modulation percentage than black and the sync tip is 100% modulation), the numbers don't add up.

Take a look at Figure 107 in this Application Note.

http://direct.www.tek.com/Measuremen...g/section6.pdf

We know that digital 16 is equal to 7.5 IRE and digital 235 is equal to 100 IRE. Therefore, each digital level is equal to 92.5 / 219 = 0.422374429 IRE.

But the zero modulation percentage is 120 IRE. If 100 IRE is digital level 235, then 120 IRE is digital level 282.351 (235 + (20 / 0.422374429)).

A properly converted digital NTSC signal cannot over modulate an analog NTSC transmitter at any digital level.

Ron

HD MPEG-2 Test Patterns http://www.w6rz.net
dr1394 is offline  
post #12 of 26 Old 03-21-2012, 04:45 AM
AVS Addicted Member
 
Ken H's Avatar
 
Join Date: Nov 1999
Location: Metro Detroit
Posts: 45,876
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 12
I love reading discourse that is so far over my head I can't even see it, let alone understand it. Impressive.

'Better Living Through Modern, Expensive, Electronic Devices'

Ken H is offline  
post #13 of 26 Old 03-21-2012, 05:54 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by dr1394 View Post

Is it? If you consider that NTSC is modulated "upside down" (white is a lower modulation percentage than black and the sync tip is 100% modulation), the numbers don't add up.

Take a look at Figure 107 in this Application Note.

http://direct.www.tek.com/Measuremen...g/section6.pdf

We know that digital 16 is equal to 7.5 IRE and digital 235 is equal to 100 IRE. Therefore, each digital level is equal to 92.5 / 219 = 0.422374429 IRE.

But the zero modulation percentage is 120 IRE. If 100 IRE is digital level 235, then 120 IRE is digital level 282.351 (235 + (20 / 0.422374429)).

A properly converted digital NTSC signal cannot over modulate an analog NTSC transmitter at any digital level.

Ron

Isn't the point not about clipped absolute levels, but clipped transitions that will create harmonics (as they will create square waves)?

The other point is that it was common, ISTR, for analogue cameras in analogue composite set-ups to have highlights above 0.7V with camera clippers set above this level, so you would want any digital islands to preserve this content.
sneals2000 is offline  
post #14 of 26 Old 03-21-2012, 12:23 PM
Advanced Member
 
Rory Boyce's Avatar
 
Join Date: Feb 2000
Location: Fair Oaks, CA
Posts: 861
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Back to the original question it seems from all of this that there were and are very good reasons why using 0 to 255 would be a bad idea for television. You can also see why from the perspective of a PC maker where the idea was to get information from the machine to a display using 0 to 255 seems an obvious choice.

As for the references to composite digital the quantizing levels chosen were entirely different because they were digitizing the entire analog waveform incuding sync and subcarrier. I went looking and found my copy of SMPTE 244M "System M/NTSC Composite Video Signals-Bit-Paraellel Interface" with a note written by me on it saying current 8-14-92. Per 244M for 8 bit systems sync tip is at 04h (4), blanking at 3Ch (60), and white is at C8h (200). These allowed for the passage of 100% saturated chroma.

Rory
Rory Boyce is offline  
post #15 of 26 Old 03-21-2012, 12:38 PM
AVS Addicted Member
 
scowl's Avatar
 
Join Date: Nov 2003
Location: Portland, OR
Posts: 10,440
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 18
Quote:
Originally Posted by sneals2000 View Post

Sharp transitions in analogue signals can have overshoots and undershoots, so even if your sources had perfect black and white levels within the range allowed, it is entirely possible that by the time they hit the digital equipment they would have transitions that spiked above and below these levels. If you clip these spikes you will end up with nasty ringing artefacts, so you allow some latitude above and below.

By limiting the values to the 16-235 range, doesn't that make clipping and undershooting more likely? If an analog source goes outside of this range, it has to be clipped to 235 or raised to 16 when converted to this acceptable range, right?

It seems like it would better serve to prevent a digital source from clipping an analog input.

I must be looking at this problem backwards!

NOW: my post on AVS Forum.
NEXT: someone else's post on AVS Forum.
scowl is offline  
post #16 of 26 Old 03-21-2012, 01:42 PM
Advanced Member
 
Rory Boyce's Avatar
 
Join Date: Feb 2000
Location: Fair Oaks, CA
Posts: 861
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
The idea is to prevent clipping and to leave head and foot room. Nothing limits the digital values to 16 and 235 those are just the nominal values. Per the applicable digital video standards the digital value is allowed to as low as 1 and as high as 254. The values 0 and 255 are reserved for timing information and the digital value from an A to D output must be hard limited to prevent it from ever being 0 or 255 no matter what the incoming analog signal may do.

Rory
Rory Boyce is offline  
post #17 of 26 Old 03-21-2012, 02:27 PM
Senior Member
 
gbynum's Avatar
 
Join Date: Dec 2006
Location: Greenville SC
Posts: 458
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Ken H View Post

I love reading discourse that is so far over my head I can't even see it, let alone understand it. Impressive.

keeps us humble ...

BUT ... I still learn something every day, even if I won't match the REAL experts.
gbynum is offline  
post #18 of 26 Old 03-21-2012, 03:48 PM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by scowl View Post

By limiting the values to the 16-235 range, doesn't that make clipping and undershooting more likely? If an analog source goes outside of this range, it has to be clipped to 235 or raised to 16 when converted to this acceptable range, right?

No - the whole point of 16-235 nominal range is NOT to clip <16 and >235 to avoid this. That's precisely why black is at 16 not 0/1, and white at 235 not 254/255.

The values between 1 and 16 and 235 and 254 SHOULD be preserved to avoid clipping at 16 and 235.

Quote:


It seems like it would better serve to prevent a digital source from clipping an analog input.

I must be looking at this problem backwards!

The issue is (was) analogue signals with dirty edges that had excursions below black and above white not being clipped as they were digitised.
sneals2000 is offline  
post #19 of 26 Old 03-21-2012, 03:52 PM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by Rory Boyce View Post

Back to the original question it seems from all of this that there were and are very good reasons why using 0 to 255 would be a bad idea for television. You can also see why from the perspective of a PC maker where the idea was to get information from the machine to a display using 0 to 255 seems an obvious choice.

As for the references to composite digital the quantizing levels chosen were entirely different because they were digitizing the entire analog waveform incuding sync and subcarrier. I went looking and found my copy of SMPTE 244M "System M/NTSC Composite Video Signals-Bit-Paraellel Interface" with a note written by me on it saying current 8-14-92. Per 244M for 8 bit systems sync tip is at 04h (4), blanking at 3Ch (60), and white is at C8h (200). These allowed for the passage of 100% saturated chroma.

Yep - apart from the parallel digital output of D3 VTRs (useful for cloning D3 recordings) I never had much to do with digital composite stuff. D2 VTRs were popular as replacements for ACR25 2" cart machines for advert play out ISTR - and some broadcasters used them instead of 1" VTRs to allow for early automated programme playout (using robot cart machines) - but in many set-ups they were used purely as analogue sources/destinations rather than using digital interconnects.

ISTR that GVG had digital composite NTSC vision mixers, but they never really launched any PAL digital composite models as digital component was already on the rise.
sneals2000 is offline  
post #20 of 26 Old 03-21-2012, 04:36 PM
Advanced Member
 
Rory Boyce's Avatar
 
Join Date: Feb 2000
Location: Fair Oaks, CA
Posts: 861
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
I designed television equipment for GVG in a prior life. Back in the early 90's I designed the modular composite NTSC A to D and D to A converters that GVG sold. I also designed the SMS series component converters shortly after that.

The only thing I ever designed that had issues in the field was that NTSC A to D board. A couple of years after they began shipping the things began failing in the field. I think NBC was one of the unhappy customers. The problem was that the SONY serilizers were dying. I went through considerable grief trying to figure out what the problem was. We sent schematics and eventually a board to SONY but could never get any answers. I later redesigned the board to get rid of the SONY serializer and all was good. Years later I found out that we were not the only company having issues with the SONY part. I few years back I met a guy that had been at SONY in that era and found out that whoever the head guy was in the video division had designed the parts and in the Japanese culture it was simply not possible there could be anything wrong with them.

Rory
Rory Boyce is offline  
post #21 of 26 Old 03-22-2012, 06:07 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by Rory Boyce View Post

I designed television equipment for GVG in a prior life. Back in the early 90's I designed the modular composite NTSC A to D and D to A converters that GVG sold. I also designed the SMS series component converters shortly after that.

In another life I was in R&D at Quantel... Seems like a lifetime ago now. I evaluated some of the early SDI chipsets ISTR, as well as spending quite a time designing wipe patterns...
sneals2000 is offline  
post #22 of 26 Old 03-22-2012, 12:02 PM
Advanced Member
 
Rory Boyce's Avatar
 
Join Date: Feb 2000
Location: Fair Oaks, CA
Posts: 861
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by sneals2000 View Post

In another life I was in R&D at Quantel... Seems like a lifetime ago now. I evaluated some of the early SDI chipsets ISTR, as well as spending quite a time designing wipe patterns...

We had a number of Quantel still stores and a Hal at the TV station I worked at after leaving GVG.

Rory
Rory Boyce is offline  
post #23 of 26 Old 03-23-2012, 01:56 PM
AVS Addicted Member
 
scowl's Avatar
 
Join Date: Nov 2003
Location: Portland, OR
Posts: 10,440
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 18
Quote:
Originally Posted by sneals2000 View Post

The values between 1 and 16 and 235 and 254 SHOULD be preserved to avoid clipping at 16 and 235.

If I'm allowed to preserve values outside of 16-235 then I should freely be able to use those values. There is no way a digital value will change therefore there is no more chance of clipping.

I know I must be looking at this backwards somehow.

NOW: my post on AVS Forum.
NEXT: someone else's post on AVS Forum.
scowl is offline  
post #24 of 26 Old 03-24-2012, 04:27 AM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by scowl View Post

If I'm allowed to preserve values outside of 16-235 then I should freely be able to use those values. There is no way a digital value will change therefore there is no more chance of clipping.

I know I must be looking at this backwards somehow.

You are looking at it from a purely digital point of view - and aren't considering analogue signals and how they aren't as 'clean' as purely digitally originated content. Even today video that has passed through an analogue link in the chain is not uncommon (analogue HD triax is used by both Sony and Philips/Thomson/GrassValley/Whateverytheyarecalledthis week system cameras in studios and on location, and many shows include archive from 2"/1"/BetaSP/UMatic etc.)

Analogue stuff can have spikes on sharp transitions (which will over / under shoot and if clipped can ring).

Also black is still set at 16 - anything below that level shouldn't be visible on a properly set-up display. However clipping stuff at 16 and not allowing undershoots on transitions can generate ringing that would be visible.
sneals2000 is offline  
post #25 of 26 Old 03-25-2012, 12:26 PM
AVS Addicted Member
 
scowl's Avatar
 
Join Date: Nov 2003
Location: Portland, OR
Posts: 10,440
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 3 Post(s)
Liked: 18
Quote:
Originally Posted by sneals2000 View Post

You are looking at it from a purely digital point of view - and aren't considering analogue signals and how they aren't as 'clean' as purely digitally originated content.

All the more reason to have the widest range of values to digitally sample these analog signals with.

Quote:


Also black is still set at 16 - anything below that level shouldn't be visible on a properly set-up display.

I always thought black shouldn't be visible on a properly set-up display. :What should black look like? confused:

Quote:


However clipping stuff at 16 and not allowing undershoots on transitions can generate ringing that would be visible.

Are you saying that analog equipment will have problems if a signal from a digital device is at zero volts instead of something slightly higher?

NOW: my post on AVS Forum.
NEXT: someone else's post on AVS Forum.
scowl is offline  
post #26 of 26 Old 03-27-2012, 03:32 PM
AVS Special Member
 
sneals2000's Avatar
 
Join Date: May 2003
Location: UK
Posts: 7,029
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 46 Post(s)
Liked: 46
Quote:
Originally Posted by scowl View Post

All the more reason to have the widest range of values to digitally sample these analog signals with.

Yes - assuming that clipping stuff outside the display range won't adversely impact stuff within the display range - but that isn't always the case.

Quote:


I always thought black shouldn't be visible on a properly set-up display. :What should black look like? confused:

I agree - black should be that. Nothing should appear blacker (hence PLUGE using sub-black to assist you in aligning black correctly) However clipping stuff you can't see can create artefacts that you can see.

Quote:


Are you saying that analog equipment will have problems if a signal from a digital device is at zero volts instead of something slightly higher?

What I'm saying is that if you agree on 0V as black level, and have an analogue signal generated in the real world, that has passed through real cables, DAs, EQs etc., then you would not be surprised to see some transient excursions below 0V in a 'real world' signal, particularly on sharp edge transitions.

Clipping these excursions rather than preserving them will introduce further artefacts (thinking about the signal in the frequency domain you are adding square wave frequency components when you clip something - adding frequency components to a signal that wouldn't otherwise be there), particularly when you converted the clipped signal back to analogue (or filtered the signal in the digital domain in a DVE or similar)

Just as you allow headroom in an analogue audio signal when you digitise it (to avoid clipping and creating square wave distortion on peaks) you sort of need to do the same with digital video.

Even if your camera black and white levels are set perfectly in the CCD or tube (this stuff dates back to broadcast tubed cameras), an analogue signal will have been filtered as it passes through a cable/RF channel, and this will cause sharp edges to be 'softened' and they will have transients that over/undershoot.

If you have the output from an analogue remote OB coming on an analogue microwave circuit, and used a digital frame store to synchronise it to lock it (synchronisers were one of the early applications of digital video in the mid-70s) to local syncs, and then this was used to output a composite analogue signal for use elsewhere you wouldn't be surprised to see that the source video signal wasn't a nice, clean perfect 'nothing below black, nothing above white' signal.

If you clipped this analogue signal when digitising it, so no information was preserved below standard black and white levels, you'd clip transients (spikes if you will, or imperfect edges), and this would introduce ringing in many cases when you converted back to analogue.

I have dim recollections of seeing this demo-ed in the 80s when I was learning about digital video.
sneals2000 is offline  
Reply HDTV Technical

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off