No more FF35 Scarlet?

[David Rasberry]

The Panasonic still records AVCHD long GOP MPEG 4, same as most current gen consumer cams. It does have HD-SDI live output, so one can use a better external recorder, but it is still just a 1080P video camera.

 

[toke lahti]

Seems to me that the line skipping is used to reduce the Jello effect while increasing iso at the same time. Half the lines means twice the scanning time per frame resulting more exposure time per frame. Yes?

Why would jello effect (cmos wobble) be more problematic when windowing?
If windowing is made with same amount of pixels, shouldn't the scan time be the same?

 

[Scott Crawley]

Why would jello effect (cmos wobble) be more problematic when windowing?
If windowing is made with same amount of pixels, shouldn't the scan time be the same?

If you are only sampling a portion of the sensor, how could you be using the same number of pixels as when you scan the whole sensor?

Seems to me that the line skipping is used to reduce the Jello effect while increasing iso at the same time. Half the lines means twice the scanning time per frame resulting more exposure time per frame. Yes?

Why would scanning fewer lines take longer? It seems to me that your logic is backward. What am I missing?

 

[toke lahti]

And like we have said several times, the DSLRs line-skip simply because they aren't powerful enough, neither were designed to, do full frame motion at 24 fps. Period. I don't understand why there's discussion about it. If they could do even 3K, the DSLR makers would obviously have implemented it.

Is there somebody here, who has suggested that these dslr's could/would/will do full frame motion at 24 fps?
All I have suggested that next canon camera/firmware could employ WINDOWING, when it uses same amount of pixels than now with line skipping.

And lets not forget that 5d2 started the whole dsmc market, so it is still 1st gen product. When they designed it, they obviously didn't have very clear knowledge what users would like from 5d2.
Otherwise there would have been 24/25 framerate from day 1.
Nobody complained about line skipping or slow read out time before it was released. Next time they could have these corrected and windowing option added.

Pana's 4/3" will be "just fullHD", but if it's something like 2 stops more sensitive than Scarlet2/3", I don't care about loosing that about .5k loss of real resolution.
But it all depends on lenses; will there be any "ENG-zoom" like models for 4/3"?

 

[toke lahti]

If you are only sampling a portion of the sensor, how could you be using the same number of pixels as when you scan the whole sensor?

Options were:
1. Line skipping
2. Windowing
Same amount of pixels read with both options.

Why windowing is so strange thing for people here?
Have you not used it with R1?
Do you guys think that cheaper/other than Red can't use it?
Have you seen/used this:
http://www.casio.com/products/Cameras/EXILIM_High-Speed/EX-FH25/
?

 

[Scott Crawley]

Options were:
1. Line skipping
2. Windowing
Same amount of pixels read with both options.

I honestly can't say if this is true, or if we are having a linguistic problem here. Let us say for the sake of argument that the camera still scans the entire sensor, and then omits data from entire lines, or the perimiter of the image. (If that is indeed the point you are making.) If that is the case then the jello factor would not be reduced as the scan time would not be different. It could however be worsened I would think. The resolution is still reduced, and you are not employing the same number of pixels.

 

[Scott Crawley]

Why would jello effect (cmos wobble) be more problematic when windowing?
If windowing is made with same amount of pixels, shouldn't the scan time be the same?

I've been trying to tell you that this basic assumption appears to be your problem. When you employ windowing you disregard pixels from the perimeter of the sensor. You do not use them. A windowed image does not contain the same number of pixel samples as a full sensor sized image unless some kind of pixel shifting or duplication is done. I don't know of another way to say it.

Maybe I misunderstand how line skipping is done. Are even lines skipped in A frames, and odd lines skipped in B frames and then the two are interlaced together? That would more accurately be called field skipping or field interpolation, wouldn't it? That might yield a perceived resolution, but not actual resolution. I digress.

My whole understanding is that Canon employs line skipping because they have to omit those pixels. I'm not clear on whether that is because the sensor is unable to generate samples that quickly, or because the camera can not handle the processing or signal throughput. Folks who know better than I appear to be saying that it is the former.

 

[David Rasberry]

And lets not forget that 5d2 started the whole dsmc market, so it is still 1st gen product. When they designed it, they obviously didn't have very clear knowledge what users would like from 5d2.

Wrong. The term DSMC was invented by RED and defined as a camera that shoots both stills and motion raw with no difference in resolution or image quality between the two. It is a RED trademark. RED 1 is the first and so far only practical DSMC camera on the market, even though its form factor is not optimum as a still camera. The same idea applies to the difference between a video camera and a digital cinema camera. Digital cinema cameras, of which there are several, support raw acquisition. Video cameras do not.

The Canons are hybrid still/video cameras with consumer video features. The video is decidedly inferior to the stills capability.

 

[toke lahti]

When you employ windowing you disregard pixels from the perimeter of the sensor. You do not use them. A windowed image does not contain the same number of pixel samples as a full sensor sized image...

Same as line skipped image:
When you employ line skipping you disregard pixels from 3 out of 4 lines in the sensor. You do not use them. A line skipped image does not contain the same number of pixel samples as a full sensor sized image...

For example:
full sensor: 4k x 2k
line skipped: 4k x 1k (every other line is skipped)
windowed: 2k x 1k (500 pixels dropped from left & right, 500 lines dropped from top & bottom)

 

[Subhadip Sen]

Toke, the whole idea here is that windowing doesn't use the full sensor, whereas a line skipped image does. That means you lose out field of view, and you lose out depth of field. Depth of field being one of the biggest draws to the HDSLRs anyway.

Also, given the weakling DSLRs (compared to actual motion cameras), how much would you need to window anyway?

 

[Scott Crawley]

I think that I know where I misunderstood you now.

Why would jello effect (cmos wobble) be more problematic when windowing?
If windowing is made with same amount of pixels, shouldn't the scan time be the same?

It depends upon how the chip scan is actually done, and I honestly don't know that.

It has been my assumption that in either case the entire chip is still scanned, but the data is simply disregarded. I don't believe that windowing the image would change jello. Depending upon how the line skipped image is displayed, jello could be worse. The skipped lines are interlaced out, are they not? That would make it worse I think. Does the 5d now do true progressive frames? I didn't think so.

Now you have me wondering.

 

[toke lahti]

Wrong. The term DSMC was invented by RED and defined as a camera that shoots both stills and motion raw with no difference in resolution or image quality between the two. It is a RED trademark. RED 1 is the first and so far only practical DSMC camera on the market, even though its form factor is not optimum as a still camera. The same idea applies to the difference between a video camera and a digital cinema camera. Digital cinema cameras, of which there are several, support raw acquisition. Video cameras do not.
The Canons are hybrid still/video cameras with consumer video features. The video is decidedly inferior to the stills capability.

You can tangle up on one word or term, but that's semantics and doesn't change the point of my post.

Have you ever heard/seen any professional photography (=stills) production made with R1?
Red may own trademark for dsmc, but the term was invented long after R1 was designed and 5d2's are used right now in filming House season finale, so I guess that will be "motion" what they are recording.

Still, again and yet in the future also, this is purely speculation, but think if canon introduces 5d3 in the fall that has option in movie mode, that you can use either line skipped FF or let's say 4/3" 6 Mpx window in the center. Then they (or Pana's AG-AF100's) release 10x (or even 16x) ENG-style zoom lens that uses this 4/3".
Or you could window 2/3" and use any existing broadcast zoom.
Recording could be done in sRAW3 (meaning something compressed raw like r3d) or canon's XF codec.
(Then by your definition, it would be triple camera; sRAW for cinema, XF for video plus the normal still mode.)
Or maybe they could have made sensor read out so much faster that you could record sRAW @24fps.
And price tag on the camera might be just little higher than 5d2.

Who would want buy Scarlet 2/3" then?
5d3 might come out not until 2011, so that would give Scarlet2/3" one year ahead, but at least I'm not going to pre-order line, before it is known what canon will introduce within this year.

For me Epics will forever be too expensive to own and I really want to own the camera to experiment and use it also non-commercially. Scarlet2/3" will never be a real DSMC because of its resolution and sensor size. Maybe there will be ScarletS35 at the end of this year or next year. It will maybe be too expensive for me at least if new canon comes out with specs above and half the price. I hope there still would be ScarletFF in the future (EpicFF 2011, so maybe 2012?) to really top the canons and hope it will not be triple the canon's price. And I still hope there would be Scarlet4/3", since that would be sweet spot for video camera for me by lens sizes and sensor sensivity. S35 by sensor size, could be the first real DSMC, but I wouldn't pay many times more for it than for canon, since I already have FF canon.

In the long run it will be the lenses that decide which camera will be success.
Already with Arri, full prime set costs more than the camera.
For DSMC, there should be lenses witch are better for stills and lenses that suit better for run&gun video like newsgathering. For stills you could use something like Canon's EF lenses but for r&g that can be a little problematic. If you'd like to use 2/3" zooms, that area with 6 micron pixel size does not give even full HD resolution. And there are still no compact over 10x zooms with f around 2 for bigger sensor than 2/3". If Red would make such a lens, it would be a real market changer with Scarlet4/3".

 

[Terry VerHaar]

And what if...

awww... never mind.

 

[David Rasberry]

Have you ever heard/seen any professional photography (=stills) production made with R1?

http://www.esquire.com/cover-detail?year=2009&month=6
http://www.esquire.com/the-side/video/megan-fox-images-0609

Just one outstanding example.

The term DSMC was first used in RED's original announcement of the Epic and Scarlet cameras.

till, again and yet in the future also, this is purely speculation, but think if canon introduces 5d3 in the fall that has option in movie mode, that you can use either line skipped FF or let's say 4/3" 6 Mpx window in the center. Then they (or Pana's AG-AF100's) release 10x (or even 16x) ENG-style zoom lens that uses this 4/3".
Or you could window 2/3" and use any existing broadcast zoom.
Recording could be done in sRAW3 (meaning something compressed raw like r3d) or canon's XF codec.
(Then by your definition, it would be triple camera; sRAW for cinema, XF for video plus the normal still mode.)
Or maybe they could have made sensor read out so much faster that you could record sRAW @24fps.
And price tag on the camera might be just little higher than 5d2.

Who would want buy Scarlet 2/3" then?

I would. I am looking specifically for a 16mm film camera replacement. The bottom of the barrel AVCHD 1080p video quality out of anything in production now or announced for release or even rumored for 2011 at under $10k, much less under $5k is no longer of any interest. The question is not can Canon or Sony or Panasonic make a high speed higher than HD motion camera. The question is will they? So far the answer is no they won't anytime in the near future. Panasonic didn't even use their higher end P2 AVCintra codec in the newly announced 4/3 camcorder because their spokesperson said it would raise the price by $2000 or more and they wanted to stay under $6k.
Shooting well beyond HD resolutions at high speed raw frame rates is not even on their radar in the near future. Maybe that capability is not of interest to you, but it is specifically what is of interest to me. So far when it comes to high speed raw motion at affordable to own prices RED is the only game in town and likely to remain that way for some time to come. What they are doing with Scarlet and Epic is not easy to accomplish and not cheap to develop. If it were the other guys would have already done it.

 

[Subhadip Sen]

You can tangle up on one word or term, but that's semantics and doesn't change the point of my post.

Have you ever heard/seen any professional photography (=stills) production made with R1?
Red may own trademark for dsmc, but the term was invented long after R1 was designed and 5d2's are used right now in filming House season finale, so I guess that will be "motion" what they are recording.

Still, again and yet in the future also, this is purely speculation, but think if canon introduces 5d3 in the fall that has option in movie mode, that you can use either line skipped FF or let's say 4/3" 6 Mpx window in the center. Then they (or Pana's AG-AF100's) release 10x (or even 16x) ENG-style zoom lens that uses this 4/3".
Or you could window 2/3" and use any existing broadcast zoom.
Recording could be done in sRAW3 (meaning something compressed raw like r3d) or canon's XF codec.
(Then by your definition, it would be triple camera; sRAW for cinema, XF for video plus the normal still mode.)
Or maybe they could have made sensor read out so much faster that you could record sRAW @24fps.
And price tag on the camera might be just little higher than 5d2.

Who would want buy Scarlet 2/3" then?
5d3 might come out not until 2011, so that would give Scarlet2/3" one year ahead, but at least I'm not going to pre-order line, before it is known what canon will introduce within this year.

For me Epics will forever be too expensive to own and I really want to own the camera to experiment and use it also non-commercially. Scarlet2/3" will never be a real DSMC because of its resolution and sensor size. Maybe there will be ScarletS35 at the end of this year or next year. It will maybe be too expensive for me at least if new canon comes out with specs above and half the price. I hope there still would be ScarletFF in the future (EpicFF 2011, so maybe 2012?) to really top the canons and hope it will not be triple the canon's price. And I still hope there would be Scarlet4/3", since that would be sweet spot for video camera for me by lens sizes and sensor sensivity. S35 by sensor size, could be the first real DSMC, but I wouldn't pay many times more for it than for canon, since I already have FF canon.

In the long run it will be the lenses that decide which camera will be success.
Already with Arri, full prime set costs more than the camera.
For DSMC, there should be lenses witch are better for stills and lenses that suit better for run&gun video like newsgathering. For stills you could use something like Canon's EF lenses but for r&g that can be a little problematic. If you'd like to use 2/3" zooms, that area with 6 micron pixel size does not give even full HD resolution. And there are still no compact over 10x zooms with f around 2 for bigger sensor than 2/3". If Red would make such a lens, it would be a real market changer with Scarlet4/3".

I don't understand what you are trying to say at all. What if, what if... I don't think any of what you are "predicting" is realistic. You are making far too much assumptions about windowing. For starters, how do you know the sensor can just skip pixels in the middle? Can't they skip entire lines only? You are saying skip 500 pixels on all four sides. Reality suggests that only the top and bottom can be skipped. Besides, you are introducing over a 2x crop factor at 4/3". And, let's not forget - these HDSLRs are nowhere near capable of doing 6 MP per 1/24th of a second continuously, other than a short burst. And don't even get me started on how you would expect them to record a usable footage. The AVCHD standards end at 1080p. No offense, but your suggestions and predictions all appear too naive and unrealistic, and I don't understand what your point is at all.

You seem to be too hung up about sensor size, pixel size, etc. to realize size isn't everything for image quality. Besides the image quality, 2/3" is practically the perfect size - overlapping with 16mm. You have a massive variety of lenses with zoom ranges that are simply impractical with 35mm. You can keep a deep focus, and you can also have a shallow depth of field.

Having said all of this, I hope I am terribly wrong and your predictions are right and indeed we can shoot full 5+K (Or even 6MP - 3+K) on a DSLR in the near future. It's a faint hope though.

 

[toke lahti]

You are making far too much assumptions about windowing. For starters, how do you know the sensor can just skip pixels in the middle? Can't they skip entire lines only? You are saying skip 500 pixels on all four sides. Reality suggests that only the top and bottom can be skipped.

Windowing basics 101:

http://www.arridigital.com/creative/camerabasics/3
"Windowing – Readout of a Reduced Image Area
Reading out an area smaller than the full sensor size is called windowing. This reduces the amount of information coming from the sensor and, in turn, allows capturing higher frame rates. If a camera is designed to make use of windowing, it is able to run e.g. 75 fps when using the full sensor area and 150 fps when using half of it. CCD design only allows limited use of Windowing. CMOS sensors allow readout of individual photocells and thus, unrestrained use of windowing."

http://broadcastengineering.com/hdtv/ccd-cmos/index2.html
"A CMOS device has the unique capability to employ both row and column addressing to create a window within the photodiode matrix. A small window can be used to capture slow-motion video. For example, a 960 × 360 pixel window within a 1920 × 1080 sensor requires less than 17 percent of the entire sensor be read out, allowing the readout to occur up to 6X faster."

http://www.extremetech.com/article2/0,2845,1156453,00.asp
"The charge readout from the AP CMOS sensor is done using parallel circuitry, which allows the signal from single pixels or columns of pixels to be directly addressed. This direct random access ability allows a CMOS sensor to intelligently choose to readout select groups of pixel charges (rather than the entire sensor array). This is called window-of-interest or windowing readout. A CMOS sensor can intelligently subsample (reduce the size of) an image when it is captured. It also offers the potential of increased readout speed, as compared to CCD's, which must offload all its charge through a single horizontal shift register."

Are we clear at least on this subject?

 

[Subhadip Sen]

Are we clear at least on this subject?

No. How would you explain the Red Epic doing 96 fps for 16:9 and 100 fps for 2:1 then? Note: 16:9 is 4800x2700 12.9 MP; 2:1 5120x2560 is 13.1 MP.

It was clarified that it is because the max frame rate depends on the number of rows rather than actual pixels - 16:9 uses 2700 rows as opposed to 2560 for 2:1. If your theory were true, the 16:9 would be 104 fps, not 96 fps.

Now, maybe not every sensor is like this, but how do you know for certain that the Canon 5D can switch off pixels independently. If it could, don't you think a strategic line/column skip would produce far better results than just line skipping?

Anyway, perhaps someone can step in and clarify.

 

[toke lahti]

I would. I am looking specifically for a 16mm film camera replacement.

So, would you buy Scarlet4/3" if it were only $1000-$1500 more expensive than Scarlet2/3"?
How nice would it be to use 16x ENG-zoom with 2/3" window when that suits the needs and have nice set of 4/3" primes with f1.8 and 5x zoom with f2.8 for more cinematic work?
4/3" would be even usable for stills, so that would be entry level DSMC.

 

[David Rasberry]

I would buy a 4/3 Scarlet if it had 4k resolution like RED 1 and same level of performance as the 2/3". I am very much looking forward to shooting some 120-150fps high frame rate experimental footage. Olympus makes some excellent 4/3 F2 and F2.8 constant speed zooms in their premium line, though they are not cheap. S35 cine lenses would work well, and many s16mm lenses 25mm and longer will cover 4/3 format.
Not so concerned about stills, I have still cameras.

 

[toke lahti]

No. How would you explain the Red Epic doing 96 fps for 16:9 and 100 fps for 2:1 then? Note: 16:9 is 4800x2700 12.9 MP; 2:1 5120x2560 is 13.1 MP.

Oh, I didn't notice that.
Really strange.
Most cmos cameras do have several outputs from column buses, which enables ROI.
Eg. cheap Casio EX-FH25 has one row cut to pieces of 224 pixels, so different frame rates have horizontal resolution multiplied by this (224px/1000fps, 448px/240fps, 640px/120fps).
Maybe this explanes why Red's sensors are so affordable...
Btw, is there somewhere in these forums Epic's formats posted (resolutions, fps)?

Now, maybe not every sensor is like this, but how do you know for certain that the Canon 5D can switch off pixels independently. If it could, don't you think a strategic line/column skip would produce far better results than just line skipping?

When canon put 5d2 out there wasn't even term DSMC yet.
Maybe they didn't know what customers would want.
Or maybe they just thought that FF size is the thing people wants and don't mind line skipping artifacts.
And maybe they didn't think that movie mode was used professionally at all. Same thing when Sony put first dv(cam) in 1995-96. It wasn't meant for broadcasting, but it is still used for that.
I don't know if 5d2's sensor could handle windowing and how well, but seems to be quite natural that when next model is out, that would be the most important new feature of the camera. Or maybe second to raw recording.