BBC Analysis & Recommended Settings for EX1/EX3

[Peter Mee]

Hi All

Came across this on a recommendation from my equipment supplier. It's a full technical analysis of the EX1/EX3 from BBC engineers along with recommended settings for BBC production for both film-look and video-look. Apologies if it's been posted before.

There's A LOT of info here. In a nutshell, they recommend 1080i for HD broadcast acquisition and 720p if the target is SD.

BBC - R&D - Publications - WHP034 - Addendum 27

(click the download links on the left - the main body is the intro text only)

Peter

[Joachim Hoge]

Thanks for posting.
Interesting read.

[Brian Drysdale]

Originally Posted by Peter Mee

Hi All

There's A LOT of info here. In a nutshell, they recommend 1080i for HD broadcast acquisition and 720p if the target is SD.

Peter

Quickly looking the white paper I didn't see it recommending only 1080i, for HD broadcast, but reducing the detail setting when shooting progressive.

The "film look" settings are commonly used on the 2/3" HD cameras on BBC productions shooting progressive. This usually involves a low detail setting compared to the "video look" setting.

[Bob Grant]

Excellent read.
I think it's convinced me to shoot 720p/50 in future for anything intended for SD delivery.

[Piotr Wozniacki]

Originally Posted by Bob Grant

Excellent read.
I think it's convinced me to shoot 720p/50 in future for anything intended for SD delivery.

Even for a 50i PAL DVD, Bob?

[Peter Mee]

Originally Posted by Brian Drysdale

Quickly looking the white paper I didn't see it recommending only 1080i, for HD broadcast, but reducing the detail setting when shooting progressive.

Hi Brian,

in section 1.2.1 1080-line HQ interlace

In HQ mode, the camera records MPEG long-GoP data at 35Mb/s, with variable bit rate. In this mode, the recorded format is 1920x1080, with chroma sub-sampling at 4:2:0 thus the chroma signals have resolutions of 960x1080. This mode is what should be considered for full HDTV shooting.

and later in 1.2.2 1080-line HQ, progressive

But vertical resolution has changed significantly, there is now the same depth of modulation at 1080 vertically as there is at 1920 vertically. This level of vertical detail will cause “twittering” when viewed on a classical crt monitor, and may cause some problems in MPEG compression, because high frequency content is not expected to have high amplitudes.



It's not explicitly stated, I know but on first read through, it struck me that the flatly state 1080i 'is what should be considered for full HDTV shooting'. You are right, though, they don't state 1080i exclusively, just reads to me like a very strong recommendation?

Peter

[Steve Phillipps]

About time they got this posted, I know that Alan did it quite a while ago but wasn't able to share it!
Alan Roberts does all the set-ups for BBC NHU work and so these settings should give an image with the same aims as the Varicam set-ups etc., and so would be a good starting point for anyone shooting wildlife stuff I'd say.
Thanks for posting Peter.
Steve

[Alister Chapman]

"In HQ mode, the camera records MPEG long-GoP data at 35Mb/s, with variable bit rate. In this mode, the recorded format is 1920x1080, with chroma sub-sampling at 4:2:0 thus the chroma signals have resolutions of 960x1080."

Must be a typo as it would only be 960x540.

Overall the report is pretty complimentary. The noise measurement is interesting. I can certainly see noise but I would not have said it was -44db, that's a really terrible noise figure.

The comments on the gammas make for interesting reading. One thing I do is use Cinegamma 4, which I know produces illegal levels (109IRE at 0db), but if you use -3db gain the peak level is also reduced to around 104IRE. Using negative gain reduces the dynamic range slightly, but then by using Cinegamma 3 or 4 you can regain the loss in dynamic range by making use of the extra headroom. As stated in the paper you will need to grade or monitor your whites to keep things legal. It should also be noted that if you do use Cinnegamma 2 at -3db the camera will not go up to the full 100 IRE (only about 96 IRE) so you get a double hit on dynamic range which is not good.

The way the EX applies the gamma and gain is interesting as it implies that gain is added after gamma correction and clipping.

[Bill Ravens]

Originally Posted by Alister Chapman

"In HQ mode, the camera records MPEG long-GoP data at 35Mb/s, with variable bit rate. In this mode, the recorded format is 1920x1080, with chroma sub-sampling at 4:2:0 thus the chroma signals have resolutions of 960x1080."

Must be a typo as it would only be 960x540.

uhhh...no. It's not a typo.

[Alister Chapman]

But later in the paper he states that when used in progressive HQ the chroma sampling is 960x540 and that in SP Mode (interlace 1440 x 1080) the sampling is 960x540 (which I think should be 720x540). So are we to believe that there is some magic sampling when your in HQ interlace or a mistake has been made.

Do we indeed have 4:2:2 sampling (960x1080) or 4:2:0 sampling (960x540).

The implication is that the sampling is increased when in interlace HQ compared to progressive HQ, which we know not to be the case. My guess is it's a typo, either that a clear mistake as he gives two different amounts of samples for the same 4:2:0 scheme.

I'm also a bit surprised that the noise figure was obtained at 6db gain. I realise that a "calculation" was added to compensate for the gain, but that then assumes that the gain is entirely linear and exactly 6db.

[Steve Phillipps]

In "Another Place" Alan has said it was a typo, it should be 960x540.
Steve

[Jay Gladwell]

Originally Posted by Alister Chapman

The comments on the gammas make for interesting reading. One thing I do is use Cinegamma 4, which I know produces illegal levels (109IRE at 0db), but if you use -3db gain the peak level is also reduced to around 104IRE. Using negative gain reduces the dynamic range slightly, but then by using Cinegamma 3 or 4 you can regain the loss in dynamic range by making use of the extra headroom. As stated in the paper you will need to grade or monitor your whites to keep things legal. It should also be noted that if you do use Cinnegamma 2 at -3db the camera will not go up to the full 100 IRE (only about 96 IRE) so you get a double hit on dynamic range which is not good.

Alister, that's exactly what Doug Jensen suggests in his training video on the EX3. Oddly enough, it does provide a better contrast range while preventing the highlights from being blown out. His adjustments are very subtle, but they provide (for my taste) an outstanding image.

[Piotr Wozniacki]

Originally Posted by Alister Chapman

The comments on the gammas make for interesting reading. One thing I do is use Cinegamma 4, which I know produces illegal levels (109IRE at 0db), but if you use -3db gain the peak level is also reduced to around 104IRE. Using negative gain reduces the dynamic range slightly, but then by using Cinegamma 3 or 4 you can regain the loss in dynamic range by making use of the extra headroom. As stated in the paper you will need to grade or monitor your whites to keep things legal. It should also be noted that if you do use Cinnegamma 2 at -3db the camera will not go up to the full 100 IRE (only about 96 IRE) so you get a double hit on dynamic range which is not good.

What do you think about Cinegamma 1 - it's the one used widely for its deeper mids and blacks (as compared to the 3 and particularly 4, which stretches blacks); will it benefit from using the negative gain as well (other than reduced noise, of course)? I should think so (seeing how easily it clips the highs), but just wanted others' opinion as it is not mentioned in the article...

[Brian Drysdale]

Originally Posted by Peter Mee

Hi Brian,

in section 1.2.1 1080-line HQ interlace

In HQ mode, the camera records MPEG long-GoP data at 35Mb/s, with variable bit rate. In this mode, the recorded format is 1920x1080, with chroma sub-sampling at 4:2:0 thus the chroma signals have resolutions of 960x1080. This mode is what should be considered for full HDTV shooting.

and later in 1.2.2 1080-line HQ, progressive

But vertical resolution has changed significantly, there is now the same depth of modulation at 1080 vertically as there is at 1920 vertically. This level of vertical detail will cause “twittering” when viewed on a classical crt monitor, and may cause some problems in MPEG compression, because high frequency content is not expected to have high amplitudes.



It's not explicitly stated, I know but on first read through, it struck me that the flatly state 1080i 'is what should be considered for full HDTV shooting'. You are right, though, they don't state 1080i exclusively, just reads to me like a very strong recommendation?

Peter

That possibly could be the case, although I suspect it will depend on if you wish to shoot progressive in camera or the de-interlace the 1080i in post to create a progressive effect and I can't see the latter having any advantages. He does refer to the "mode" as in I assume "HQ mode" rather than interlace as such. Certainly, HD mode would make sense for a broadcaster shooting HDTV and that's how I would read his intentions.

Reducing the detail on the progressive setting should cut back on the "twittering".

[Brian Luce]

I saw an analysis of 1080p versus 720p downrezzed for SD. I don't recall the specifics and the hows, but the 1080p looked better.