House pisses me off

[Brian Valente]

I wanted to offer a little insight into House's choice of the 5D (we were present because they used redrock support rigs on the DSLRs). When I spoke with Gale I asked him why they chose to shoot this episode with the 5D. obviously budget isn't an issue for the world's #1 most watched television show.

He explained that as a hospital program, they often work in very confined spaces and sets. They had previously struggled to get the shallow depth of field they wanted to separate the subject/action from the background in such close quarters. Part of that also includes the size of the rig squeezing into tight spaces. With the larger chip they felt they could finally achieve the shallow DOF they wanted.

So in the end, I think they would describe it as an artistic choice. From what I understand everyone is quite pleased with the results.



Cheers

Brian

 

[Graeme Nattress]

Nitpicking, but you are comparing your latest $5000+ (today's price) sensor just begun selling a couple months ago in a $17,500 body, from a small, one camera company that presumably has a shorter time to market for new technology - versus the sensor in a $2600 camera that has been sold for a year and a half - (a full Moore's law cycle) from a large, presumably slower moving company, that has over 70 cameras in its current lineup alone. Hardly apples to apples and nothing from which to make even off-the-cuff scientific judgments.

EDIT: Graeme, not suggesting that you are making that sort of judgment - but it could be easily taken to that next step by others if my point isn't considered.

Really we're comparing the "philosophy" of doing gain as digital, over gain as analogue. With analogue gain, for each stop of gain you go up, you loose a stop of DR in the highlights. After you go up a few stops analogue gain, this becomes very noticeable.

Graeme

 

[Andrew Walker]

Here's a clip that Richard Darge and I shot when I first got my MX back from RED. Richard also had a 7D with him that we shot at the same ISO as the RED and you can see what Graeme is talking about. Its towards the middle and end of the clip.

http://rickdarge.com/mxxx.mov

 

[Brad Webb]

Here some interesting reading on how the ISO on the 7D works. Not sure if the same applies to the 5D.

http://marvelsfilm.wordpress.com/2009/10/10/canon-7d-noise-and-iso-test/

http://www.cryptobola.com/PhotoBola/Canon7D_ISO.htm

 

[chrislancaster]

Here's a clip that Richard Darge and I shot when I first got my MX back from RED. Richard also had a 7D with him that we shot at the same ISO as the RED and you can see what Graeme is talking about. Its towards the middle and end of the clip.

http://rickdarge.com/mxxx.mov

great choice of music

osti:

 

[Jaymes Poudrier]

But for HD it is not (and those of you who say 1080i is not 1080p, please look up the fact that 1080i is simply a way of delivering 1080p - it is 1080p )

I must be terribly misinformed... I think this statement should read "1080i is not 1080p, 1080i has 1080 lines of vertical resolution which is the same as 1080p, however, the I and the P are definately a distinction made to how the frames are drawn on the screen."

I don't want to nitpick but, when you make a point it is generally good to make it a complete one. The facts are that anyone reading this, trying to learn would become just as misinformed as the person who made the comment, adding to confusion and ultimately a larger mess.

In the end, 1080i & 1080p are the same vertical resolution, they are not however the SAME thing. That would be like saying time-lapse and live action are the same thing, while the end result is a video depicting motion, the methods of capturing it are entirely different. Therefore, they are not the same, but the result is SIMILIAR in many aspects.
(Bad example, but I couldn't think of anything better off hand)

 

[David Rasberry]

I must be terribly misinformed... I think this statement should read "1080i is not 1080p, 1080i has 1080 lines of vertical resolution which is the same as 1080p, however, the I and the P are definately a distinction made to how the frames are drawn on the screen."

I don't want to nitpick but, when you make a point it is generally good to make it a complete one. The facts are that anyone reading this, trying to learn would become just as misinformed as the person who made the comment, adding to confusion and ultimately a larger mess.

In the end, 1080i & 1080p are the same vertical resolution, they are not however the SAME thing. That would be like saying time-lapse and live action are the same thing, while the end result is a video depicting motion, the methods of capturing it are entirely different. Therefore, they are not the same, but the result is SIMILIAR in many aspects.
(Bad example, but I couldn't think of anything better off hand)

All current model HD flat panel displays show 1080p. 1080i is a bandwidth reduction transmission format. But when you watch it, the display shows it as a progressive scanned frame. Interlace is removed.

 

[Rick Burnett]

Really we're comparing the "philosophy" of doing gain as digital, over gain as analogue. With analogue gain, for each stop of gain you go up, you loose a stop of DR in the highlights. After you go up a few stops analogue gain, this becomes very noticeable.

Graeme

With current implementations of analog gain used in many cameras. If they were to use a deeper bit depth A/D converter, they could keep those highlights as well. They just don't do that though.

 

[Pierre T.]

Really we're comparing the "philosophy" of doing gain as digital, over gain as analogue. With analogue gain, for each stop of gain you go up, you loose a stop of DR in the highlights. After you go up a few stops analogue gain, this becomes very noticeable.

Graeme

Hmmm, this gets quite complicated. As is hinted in some other thread, the A/D conversion of the Mysterium X is non-linear, which allows to get 13+ stops out of it without analog gain and with 12 bit A/D conversion. If I get it right, the idea of the design is a fixed sensitivity with ample headroom in both directions.

According to tests on dpreview.com for the 5D II, is has a usable DR of 11.1 stops (in RAW, that is) which fall down to a bit less, but at least 7.8 stops at iso6400 (which is +6 stops). "At least" since the test for high iso was only in jpeg, so probably the RAW DR is somewhat higher at iso6400. The 1D IV has one stop more DR. This would mean, again under ideal conditions :

camera | overall DR | simultaneously usable DR
Mysterium X | 13.5 stops |*13.5 stops
5D II | 13.8+ stops (jpeg) | 11 stops (RAW)
1D IV | 13.9+ stops (jpeg) | 12 stops (RAW)
In spite of opposite rumours, professional iso 200 color negative film does not really have more DR than around 9 stops, it is just more graceful around the highlights (S curve).

That sounds like a case for the M X design. But a doubt remains for me: since the a/d conversion of M X is non-linear in mimicking the S curve of film, the distribution of steps would not be the same if the digital gain is changed. This would be in contrast to a uniform distribution with linear analog gain and linear A/D conversion and finally linear digital gain.

So is the golden rule for M X something like "expose to the middle" as opposed to "expose to the right" with linear A/D sensors?

 

[David Rasberry]

I wanted to offer a little insight into House's choice of the 5D (we were present because they used redrock support rigs on the DSLRs). When I spoke with Gale I asked him why they chose to shoot this episode with the 5D. obviously budget isn't an issue for the world's #1 most watched television show.

He explained that as a hospital program, they often work in very confined spaces and sets. They had previously struggled to get the shallow depth of field they wanted to separate the subject/action from the background in such close quarters. Part of that also includes the size of the rig squeezing into tight spaces. With the larger chip they felt they could finally achieve the shallow DOF they wanted.

So in the end, I think they would describe it as an artistic choice. From what I understand everyone is quite pleased with the results.



Cheers

Brian

This makes the most sense. The three major advantages of the DSLR's currently are compact form factor, low light sensitivity and large sensor DOF characteristics.

 

[Tom.Wong]

It makes sens that they are simply trying to embrace a technology that is trying to accomplish something very specific for them. At the moment R1 is probably too big in form factor to get in those tight spaces they want. What else looks good right here, right now that's as small as a 5d and can give you a pretty picture?

Nothing, when epic comes out, they won't need 5d's anymore, but for now they have to try to work around the flaws of the 5d to get the exact type of image they want. FF SHALLOW DOF in a very tight confined space.

I remember reading an article from the DP who shot the last few eps of one of the seasons of ER on RED for the first time, being one of the first shows to adapt RED in their shooting process. He said his typical process with film was shooting 500 asa film stock, and pushing it to 1000. With RED and being limited to 320 iso, he had to bring in a lot more light than he previously had to.

But the goal was to have multiple rigs pre prepped and ready to shoot at a moment's notice, and having multiples read already pre constructed for steady cam, studio style, handheld, etc made everything a lot easier in that aspect.

It's all give and take, and they have a clear goal, and right now 5d is the only thing that offers that right now. Another year from now hopefully, and you won't have to see this kind of stuff anymore, and we'll be moving back to high standard high quality equipment.

 

[Jim Retar]

Rob Ruffo.... do you not see any irony in constantly mentioning misinformation.... and in the same breath, accusing Canon of paying the House DP to use their camera?

 

[Pierre T.]

Here's a clip that Richard Darge and I shot when I first got my MX back from RED. Richard also had a 7D with him that we shot at the same ISO as the RED and you can see what Graeme is talking about. Its towards the middle and end of the clip.

http://rickdarge.com/mxxx.mov

That's odd. What I basically see is roughly a 2 stop mismatch between the two "iso6400" modes. Numbers may not mean that much, you should have exposed the 7D correctly (no RAW headroom) and then adapt the Red One footage to it.

Besides, I'm astonished by the still present skew, I thought this was more or less history with M X?

 

[Rob Ruffo]

I must be terribly misinformed... I think this statement should read "1080i is not 1080p, 1080i has 1080 lines of vertical resolution which is the same as 1080p, however, the I and the P are definately a distinction made to how the frames are drawn on the screen."

I don't want to nitpick but, when you make a point it is generally good to make it a complete one. The facts are that anyone reading this, trying to learn would become just as misinformed as the person who made the comment, adding to confusion and ultimately a larger mess.

In the end, 1080i & 1080p are the same vertical resolution, they are not however the SAME thing. That would be like saying time-lapse and live action are the same thing, while the end result is a video depicting motion, the methods of capturing it are entirely different. Therefore, they are not the same, but the result is SIMILIAR in many aspects.
(Bad example, but I couldn't think of anything better off hand)

the end result is the same because virtually no 1080i television is filmed at 1080i (where each of the 60 fields are snapshots of 60 different moments in time). Source material is shot at either 24p (most frequent) or 30p. Looking that the example of 30p, each frame of the source material is split into 2 fields, then broadcast. Your TV waits for 2 fields, stitches them together then display a progressive ("p") frame, 30 times a second, exactly as it would a native 30p signal - so the end user normally would see no difference.

Only CRTs can actually display a 1080i signal as 1080i (so actually displaying 60 fields), as only CRTs can become native interlaced displays when fed that kind of signal. "hard pixel" displays like plasmas, LCDs and DLP fornt-projectors can only display a progressive image - they are physically incapable of anything else.. CRT HDTVs are quite rare.

1080i only looks different to the viewer from 1080p in two cases:
1- The TV does a poor job of de-interlacing the 1080i signal, and thus does not properly stitch together the two fiields into 1 progressive frame. This used to be a common problem, but TVs sold within the last 2 years, particularly 1080p models, most often do this perfectly.
2 - On a TV that can properly natively display 1080p24, 1080p24 will appear smoother, as if 24fps media is delivered via 1080i the TV must do 2:3 pulldown, so the viewer will see one 24p frame displayed for 3/60 second, the next frame for 2/60 second, and so on. The effect (called "pulldown judder" is subtle, and we are used to it, but it's still there.
3 - 1080p60 does exist, however few people shoot or master this, blu-ray cannot do it, and very displays can accurately display it, but it is possible 1080i has a "speed limit" of 60i (so 30p seen by end-viewers) in North America and 50i in Europe (so 25p seen by home-viewers).

Of course in Europe, 1080i is 50 fields, not sixty, and local TV would be shot at 25p.

 

[David Rasberry]

Hmmm, this gets quite complicated. As is hinted in some other thread, the A/D conversion of the Mysterium X is non-linear, which allows to get 13+ stops out of it without analog gain and with 12 bit A/D conversion. If I get it right, the idea of the design is a fixed sensitivity with ample headroom in both directions.

According to tests on dpreview.com for the 5D II, is has a usable DR of 11.1 stops (in RAW, that is) which fall down to a bit less, but at least 7.8 stops at iso6400 (which is +6 stops). "At least" since the test for high iso was only in jpeg, so probably the RAW DR is somewhat higher at iso6400. The 1D IV has one stop more DR. This would mean, again under ideal conditions :

camera | overall DR | simultaneously usable DR
Mysterium X | 13.5 stops |*13.5 stops
5D II | 13.8+ stops (jpeg) | 11 stops (RAW)
1D IV | 13.9+ stops (jpeg) | 12 stops (RAW)
In spite of opposite rumours, professional iso 200 color negative film does not really have more DR than around 9 stops, it is just more graceful around the highlights (S curve).

That sounds like a case for the M X design. But a doubt remains for me: since the a/d conversion of M X is non-linear in mimicking the S curve of film, the distribution of steps would not be the same if the digital gain is changed. This would be in contrast to a uniform distribution with linear analog gain and linear A/D conversion and finally linear digital gain.

So is the golden rule for M X something like "expose to the middle" as opposed to "expose to the right" with linear A/D sensors?

Bit depth does not define dynamic range limits. The minimum noise level voltage to maximum clip point voltage output of the analog sensor does. A/D bit depth defines how many discrete steps that analog range is divided into.
Red records linear A/D at a fixed optimum gain level. Non-linear modifications are added later as meta data and don't change the raw sensor data as recorded. Only scene illumination and exposure settings affect the raw data.

 

[Rob Ruffo]

Rob Ruffo.... do you not see any irony in constantly mentioning misinformation.... and in the same breath, accusing Canon of paying the House DP to use their camera?

I did not accuse them of this, specifically - I accused them of misinformation. They achieve this in various ways,

 

[Brad Webb]

Rob Ruffo.... do you not see any irony in constantly mentioning misinformation.... and in the same breath, accusing Canon of paying the House DP to use their camera?

Rob never said Canon was paying them.

I said it, however, I have made it very clear that it is 100% speculation on my part.

 

[Rob Ruffo]

Didn't you just get finished saying:



It seems to me you've already made up your mind and don't want to get confused by actual results. Much like a lot of people who dismissed Red prior to actually looking at the images.

What results? Tell me a broadcast time in 1080i over cable - that's what I am talking about. I have seen results - the SNL opening credits look really awful. I don't care how things look on Vimeo - this is not the market of my work not most of the work here.

 

[Rob Ruffo]

Bit depth does not define dynamic range limits. The minimum noise level voltage to maximum clip point voltage output of the analog sensor does. A/D bit depth defines how many discrete steps that analog range is divided into.
Red records linear A/D at a fixed optimum gain level. Non-linear modifications are added later as meta data and don't change the raw sensor data as recorded. Only scene illumination and exposure settings affect the raw data.

But... Bit depth does affect how far you can push footage and thus, in an end sense, does affect how much dynamic range you can "cheat" from footage using a lot of noise reduction.

 

[Rob Ruffo]

It makes sens that they are simply trying to embrace a technology that is trying to accomplish something very specific for them. At the moment R1 is probably too big in form factor to get in those tight spaces they want. What else looks good right here, right now that's as small as a 5d and can give you a pretty picture?

Nothing, when epic comes out, they won't need 5d's anymore, but for now they have to try to work around the flaws of the 5d to get the exact type of image they want. FF SHALLOW DOF in a very tight confined space.

I remember reading an article from the DP who shot the last few eps of one of the seasons of ER on RED for the first time, being one of the first shows to adapt RED in their shooting process. He said his typical process with film was shooting 500 asa film stock, and pushing it to 1000. With RED and being limited to 320 iso, he had to bring in a lot more light than he previously had to.

But the goal was to have multiple rigs pre prepped and ready to shoot at a moment's notice, and having multiples read already pre constructed for steady cam, studio style, handheld, etc made everything a lot easier in that aspect.

It's all give and take, and they have a clear goal, and right now 5d is the only thing that offers that right now. Another year from now hopefully, and you won't have to see this kind of stuff anymore, and we'll be moving back to high standard high quality equipment.

Is the 1.5 or so extra feet of the R1 really going to make that much difference? I dunno, I'd rather lose that foot than lose so much image quality.