Jump to content

Industrial CMOS camera as an entry-level HD cinema camera (1080p)


Troy Warr

Recommended Posts

Hi Cesar,

 

Nice beard you got there! :D

 

Thanks for posting the link to the Marlin F-033c images. I'm seeing some unusual results, though; here's a crop:

 

crop.jpg

 

That grid of pixels appears on the raw image, as well as the raw-bayer converted to RGB. What might be causing that?

 

Based on the measurements of the 1" sensor, it looks to be pretty comparable to a Super-16mm frame. There's more vertical resolution available on the sensor if you're shooting a more square format, but the widths in mm of the 1" sensor and Super-16mm frame seem to match relatively closely.

 

It sounds like I may even be able to go with some C-mount cine lenses, and at least get depth of field comparable to Super-16mm. I'm thinking that should even allow for digital-only SLR lenses with an appropriate adapter, since they're designed to cover the much larger APS-C format.

Link to comment
Share on other sites

  • Replies 115
  • Created
  • Last Reply

Top Posters In This Topic

  • Premium Member

Hi,

 

A 1" diagonal 16:9 CCD sensor is very considerably larger than super-16 - it's larger than super-35! S

 

Super-35 is .735" wide, which at 16:9 implies a frame height of 0.413" and change. This gives a diagonal of about 0.84".

 

I'm not quite sure what that diagram is trying to tell us, because it states that unmarked dimensions are in millimetres then shows the 1" sensor with a 16mm diagonal, whereas 1" is actually about 25.4mm.

 

Phil

Link to comment
Share on other sites

Hi Phil,

 

That just didn't seem to make sense to me based on the fact that these AVT cameras in question (all the way up to 1.2" CCD) ship with a standard C-mount. I would take that to mean that their sensors couldn't realistically be larger than Super-16mm. So, I just had to get to the bottom of this. ;)

 

According to my research and calculations, a 16:9 CCD with an actual 1" diagonal would still be slightly smaller than a Super-35mm frame with the same aspect ratio. I made a little drawing based on what I was able to find regarding the dimensions of various common formats:

 

10 pixels = 1 mm

ccd_sizes.gif

 

It looks like the AVT 1" CCD frame is just barely larger than Super-16mm, so I'm guessing that format's lenses could *probably* be used with just a slight risk of vignetting or loss of corner sharpness. On the same sensor with a 720p ROI, you'd be looking at just a tad smaller than regular 16mm.

 

I based these dimensions on what I found on Wikipedia in articles about Super 35mm (note they list 0.735" as the frame height, rather than the width) and Super 16mm. I was able to track down the actual physical dimensions of the AVT 1" sensor here, and it looks like the 1" is just a nominal term - the actual diagonal measurement is 16.39mm, or about 0.645 inches. Whether or not that renders the nominal 1" sensor an effective 2/3" inch sensor, or whether a standard 2/3" sensor would also suffer the same discrepancy, I haven't looked into yet. I arrived at the 16.39mm figure for the diagonal using the Pythagorean Theorem.

 

Using that same equation, I deduced that the diagonal for a Super 35mm frame (16:9) is about 28.56mm, while a CCD of true 1" diagonal measurement (i.e. actual physical measurement, not just nominal) would be 25.40mm, slightly smaller.

 

Let me know if I fudged anything, but assuming that the measurements that I found are accurate, I believe these other calculations should be accurate.

Link to comment
Share on other sites

Still wondering precisely what potential issue 'Note: Pixel clock frequencies higher than 48 MHz may reduce image quality.' is referring to.

 

 

Luke -- We (EPIX, Inc.) use the Micron MT9P001 sensor in the SV9T001C camera. Micron recommends a 48 MHz pixel clock rate for this sensor. The SV9T001C camera has a variable pixel clock frequency -- from 25 MHz at the low end to a maximum of 70 MHz. An increase in pixel clock frequency can result in increased noise -- which can reduce image quality. When we first published resolution / frame rate specs for the SV9T001C, we only listed the results achieved at 48 MHz. When we decided to add specs at 70 MHz, we thought it wise to indicate that the image quality may not be optimum at the faster rate -- so the user would get the hint that if they didn't like the quality achieved at the faster rate, they could reduce the pixel clock frequency to 48 MHz to reach Micron's quality standard.

 

Charlie Dijak

EPIX Sales Dept

cad@epixinc.com

Link to comment
Share on other sites

Hi Phil,

 

That just didn't seem to make sense to me based on the fact that these AVT cameras in question (all the way up to 1.2" CCD) ship with a standard C-mount. I would take that to mean that their sensors couldn't realistically be larger than Super-16mm. So, I just had to get to the bottom of this. ;)

 

According to my research and calculations, a 16:9 CCD with an actual 1" diagonal would still be slightly smaller than a Super-35mm frame with the same aspect ratio. I made a little drawing based on what I was able to find regarding the dimensions of various common formats:

 

10 pixels = 1 mm

ccd_sizes.gif

 

It looks like the AVT 1" CCD frame is just barely larger than Super-16mm, so I'm guessing that format's lenses could *probably* be used with just a slight risk of vignetting or loss of corner sharpness. On the same sensor with a 720p ROI, you'd be looking at just a tad smaller than regular 16mm.

 

I based these dimensions on what I found on Wikipedia in articles about Super 35mm (note they list 0.735" as the frame height, rather than the width) and Super 16mm. I was able to track down the actual physical dimensions of the AVT 1" sensor here, and it looks like the 1" is just a nominal term - the actual diagonal measurement is 16.39mm, or about 0.645 inches. Whether or not that renders the nominal 1" sensor an effective 2/3" inch sensor, or whether a standard 2/3" sensor would also suffer the same discrepancy, I haven't looked into yet. I arrived at the 16.39mm figure for the diagonal using the Pythagorean Theorem.

 

Using that same equation, I deduced that the diagonal for a Super 35mm frame (16:9) is about 28.56mm, while a CCD of true 1" diagonal measurement (i.e. actual physical measurement, not just nominal) would be 25.40mm, slightly smaller.

 

Let me know if I fudged anything, but assuming that the measurements that I found are accurate, I believe these other calculations should be accurate.

 

 

Navitar is a manufacturer of C-mount lenses. Their tutorial on Low Magnification Video Lenses includes the industry definitions of the various sensor sizes. Please visit: http://machinevision.navitar.com/pages/pro...r.cfm?nav4=true. The bottom line is that a 1" sensor is MUCH smaller than 35 mm film.

a frame of 35 mm film measures 36mm H x 24 mm V. A 1" sensor with a 4:3 apsect ratio measures 12.8 mm H x 9.6 mm V with a diagonal of 16 mm.

 

Charlie Dijak

EPIX, Inc.

Link to comment
Share on other sites

Troy:

 

I cut my beard yesterday or I should say 2 days ago... :D ..Yeah I know, I looked like a cast away!

 

Thanks for noticing the grids, I did too, but the final results are not like that. I have to look more into that situation.

 

See this stereo videos (the last one). Please do not make a 400% magnification on them, they will look awful too!

 

http://www.davidrubio3d.com/view_topic.php...amp;forum_id=44

 

HD 720 having 4 times the resolution of those shots, and the smaller pixels, will look better!

 

Thanks,

Cesar Rubio.

Link to comment
Share on other sites

Troy:

 

Well done with your work, I could not make it any better, congratulations!

 

The Pike 421-c has a 1 1/2" Kodak CCD sensor, which measures 15.2 x15.2"mm. With an AOI (Area of Interest) at 16:9, is 15.2x8.5mm which is bigger than any 16mm format. Also the Pike F-210 having a native resolution of 1080x1920pxs. (HD 1080p) and a sensor of 14.3x 8mm, is bigger than the Ultra 16 format ( 1:1.89 aspect ratio and 11.66x7.49mm physical size).

 

Charlie:

 

The full frame of 35 mm film measures 36x24 mm, but it only applies for photography. For Motion pictures is 22x16mm (1:1.85 aspect ratio).

 

Thanks,

Cesar Rubio.

Link to comment
Share on other sites

  • Premium Member
Despite your experiences with a $250 camera, I think that it's dismissive to use that as a basis for criticizing the entire concept.

I think the concept is good. But to aquire a system of acceptable quality costs many times more than a basic film camera. Basically, I just don't see how it's a good deal for the kinds of projects you're talking about. Now, if your primary need was to shoot 1000fps slow motion shots all day, then there's no question that an industrial camera is the most economical way to make that happen. But, your needs are: small form factor camera for mos shorts and music videos. That sounds like a perfect match for an el cheapo film camera.

 

The type of stuff that I shoot (MOS shorts, music video, experimental) often requires the small camera form factor, so that's a big plus.

Which is sort of why I suggested a film camera to begin with. A Russian 16mm camera is well within your budget. The running costs are somewhere around $20-25 a minute, with exceptional image quality. The cameras are small and portable and are perfect for MOS shorts, music videos, and experimental films.

 

As I mentioned previously, I have been shooting video for long enough to realize that it's like formica. It's very clean, no texture, no soul. Film is lot like wood, and there are lots of things to like about that.

Link to comment
Share on other sites

Thanks for the input, Brian.

 

I'm just not sure that I accept those conclusions without following through with my own research and interpretations.

> I think the concept is good. But to aquire a system of acceptable quality costs many times more than a basic film camera.

True, you can buy a 16mm camera of acceptable quality for well under $1000 these days. But the film stock and related costs aren't cheap, and aren't getting any cheaper. I tend to shoot a lot of footage, experimental shots, etc. Neither is particularly budget-friendly with film, and you don't get instant feedback that allows you to compensate and retry certain shots and effects, either. Though the initial investment in an industrial camera may be higher, the running costs are *substantially* lower, essentially free (hard drive space) once you've made the initial purchase.

 

Personally, I think that Silicon Imaging has already illustrated the effectiveness of certain industrial cameras via their SI-2K footage. Their solution ultimately costs around $20,000, but to get that quality of results from 16mm, you'd need good film stock, processing, HD telecine, and other associated costs with film. That effectively limits your $20,000 budget to a camera and maybe 10-15 hours of quality footage. With my $5,000 budget, that's more like 3 hours total footage.

 

I'm basically interested in learning whether the same approach taken by Silicon Imaging can be scaled down efficiently to a $5,000 solution. That may or may not be possible, but I don't think we've learned enough yet to make an accurate judgment to that effect.

As I mentioned previously, I have been shooting video for long enough to realize that it's like formica. It's very clean, no texture, no soul. Film is lot like wood, and there are lots of things to like about that.

I'm totally fine with Formica. :D

 

I've been shooting video for a while, too, and I recognize the point that you're making. I've also shot film with exactly the kind of 16mm cameras that you're talking about. Ultimately, though, it boils down to a few factors - budget, practicality, and what format is right for a particular project. My projects might ideally look better on film, but they won't suffer from being shot competently on video. For my purposes, video also has the advantage in the budget and practicality departments.

Link to comment
Share on other sites

Hi,

 

This is where I misunderstand, because if it's got a 16mm diagonal why is it a 1" sensor?

 

Phil

 

 

I don't remember the theory behind it. I know it originated in ancient times -- when cameras used vacuum tubes as sensors instead of silicon --

 

You could ask Navitar -- they probably have some old guy there that would still remember.

 

Charlie

Link to comment
Share on other sites

I just think we see things differently, and that's Ok. I hope it works out well for you.

For sure - and thank you, Brian. I do appreciate your input and feedback.

 

Incidentally, if it does work out, I'd be sure to post some footage here to share the results.

Link to comment
Share on other sites

Troy:

 

Before you buy anything, please wait for the images and videos of the Pike F145c. At $3490 and with those specs. (720p)...it might be worth buying!

Will do, Cesar. I still have a lot of research to do before buying all of the components that I'd need to make this happen.

 

I'm eager to see footage from all of your HD cams (720p & 1080p) if you're planning to shoot the latter.

 

By the way, have you played around with ROI on any of your cameras? I'm still trying to learn whether I can shoot "overcranked" footage for slow motion on any of these cameras - ideally 48fps or higher. It looks that that can be very expensive to achieve in 1080p, but if I can do it at 720p without breaking the bank, that would be a huge plus.

 

Thanks, Cesar!

Link to comment
Share on other sites

As a person very much out of touch with current technology (typing this on an '03 Powerbook), I am asking an honest question here. Is it even possible to record uncompressed 1080x1920 images to a laptop computer 24 times a second? Or is this some kind of studio rig you guys are talking about here? How many minutes of uncompressed data can you fit on a 100GB internal laptop hard drive?

I don't believe that it's possible with a laptop, or certainly at least not practical or affordable. From what I understand, the data rate for uncompressed footage will be *very* high, probably 50MB/second or higher. It's not so much an issue of capacity with laptop drives (some are up to 160GB these days, and getting bigger with PMR technology), but rather with bandwidth. Even fast single desktop hard drives (even SCSI or 10K rpm) likely couldn't keep up with uncompressed 1080p24, at least not past the innermost portions of the platter(s), where data transfer rate is highest.

 

I would plan to do this with a desktop PC, either in a studio/home setting or in the field with a generator or other power source. It would likely need a RAID setup to reach the needed minimum sustained transfer rate.

 

As I understand it, though, with appropriate software and a powerful enough processor, you can compress the footage real-time as it's fed into the computer. I'm basing that partly on feedback from Blackmagic Design (who makes the Infinity card for HDMI capture). The idea with their card, at least, is that you can bypass the harsh compression of an HDV camera by capturing the uncompressed HDMI stream live, and either keep it uncompressed or choose your own, more appropriate (and lower) compression codec to apply real-time.

Link to comment
Share on other sites

By the way, have you played around with ROI on any of your cameras? I'm still trying to learn whether I can shoot "overcranked" footage for slow motion on any of these cameras - ideally 48fps or higher. It looks that that can be very expensive to achieve in 1080p, but if I can do it at 720p without breaking the bank, that would be a huge plus.

 

Thanks, Cesar!

 

 

Troy:

 

With the Pike F-145c you can achieve:

 

At 1388x1038= 25 fps.

 

And to calculate other resolutions this is the formula:

 

1388x1038= 1,440,744 pxs.

 

1280x720 (720p)= 921,600 pxs.

 

Now divide 1,440,744/ 921,600= 1.56

 

Multiply 1.56 for the full frame resolution at 25 fps....1.56x25= 39 fps.

 

 

For the Pike F-210c is this.

 

1928x1084= 2,089,952 pxs. at 30 fps.

 

1280x720= 921,600

 

2,089,952/ 921,600= 2.26x30=67.8 fps!

 

That formula is the same for any MVC and any resolutions.

 

Thanks,

 

Cesar Rubio.

Cambridge Wisconsin, USA.

http://www.davidrubio3d.com/

Link to comment
Share on other sites

As a person very much out of touch with current technology (typing this on an '03 Powerbook), I am asking an honest question here. Is it even possible to record uncompressed 1080x1920 images to a laptop computer 24 times a second? Or is this some kind of studio rig you guys are talking about here? How many minutes of uncompressed data can you fit on a 100GB internal laptop hard drive?

 

Brian:

 

You can read with more detail the data transfers calculations in this Topic (seventh post from top to bottom).

 

http://www.davidrubio3d.com/view_topic.php...amp;forum_id=44

 

But for a short answer, I would say that with currently Laptops HD's speeds is not possible.

 

The fastest Laptop HD speed is 7200 rpm, but even with this speed the Minimum Transfer Rate is 26.2 MB/s. Please see this chart:

 

http://www23.tomshardware.com/storage25.ht...4&chart=146

 

That speed works fine for 720p. This is the formula:

 

1280x720= 921,600 pxs. x 24 (fps)= 22,118,400 or 22.1 MB/s.

 

But for that speed to be achievable and "clean" so to speak, the video stream recording has to be made in a separate Hard Drive from the OS (Operating System).

 

A Firewire external HD might work, but I have not made tests to be 100% sure about this.

 

This is a "chart" of HD space that you need to record 1 hour of Raw uncompressed video:

 

The HD's capacity is calculated using the number 1,048,576 instead of just 1,000,000.*

 

720p

1280 x 720 x 24 fps x 1 byte (8 bit) = (2,2118,400 / 1,048,576 )=21.09 MB/s* x 60 sec x 60 mins. = 75,924 MB's or 75.9 GB's

1280 x 720 x 24 fps x 2 bytes (10-16bit) = 42.19 MB/s*x60x60= 151.8 GB's.

 

1080p

1920 x 1080 x 24 fps x 1 byte (8 bit) = 47.46 MB/s*x60x60= 170.8 GB's

1920 x 1080 x 24 fps x 2 bytes (10-16 bit) = 94.92 MB/s*x60x60 = 341,712 GB's

 

Thanks,

 

Cesar Rubio.

Cambridge Wisconsin, USA.

http://www.davidrubio3d.com/

Link to comment
Share on other sites

I don't believe that it's possible with a laptop, or certainly at least not practical or affordable. From what I understand, the data rate for uncompressed footage will be *very* high, probably 50MB/second or higher. It's not so much an issue of capacity with laptop drives (some are up to 160GB these days, and getting bigger with PMR technology), but rather with bandwidth. Even fast single desktop hard drives (even SCSI or 10K rpm) likely couldn't keep up with uncompressed 1080p24, at least not past the innermost portions of the platter(s), where data transfer rate is highest.

 

Troy:

 

With a single Western Digital Raptor HD, you can record 1080p Raw uncompressed video at 24 fps!

 

http://www.newegg.com/Product/Product.asp?...N82E16822136012

 

Or with almost any HD's out there in a 2 Raid set-up!

 

http://www23.tomshardware.com/storage.html...42&chart=38

 

Thanks,

 

Cesar Rubio.

Cambridge Wisconsin, USA.

http://www.davidrubio3d.com/

Link to comment
Share on other sites

Thanks for the reply. What kind of limits would that impose on a filmmaker?

It mostly causes issues with portability. The camera has to be tethered to a PC, since that's where all of the data is going. There are only two ways that I can imagine doing it - either setting up the PC in a nearby location (like on a table in the corner of a room that you're shooting in), or building a very small form-factor PC and using it as an actual part of the camera. The latter approach is what Silicon Imaging does - their SI-2K is basically an industrial camera head with a small PC attached; together they form the whole camera body, and it's surprisingly not any bigger than many professional 16mm or 35mm motion picture cameras out there.

 

Also, though I haven't used frame grabber or other live recording software before, I don't think that it's designed with a filmmaker's workflow in mind. Many of the better industrial cameras do include ample controls for image adjustment, and some even work with LUTs. But, they're mostly targeted for industrial use, so an engineer or technician would probably have an easier time than a DP.

 

So, most kinds of documentary work and run-and-gun style shooting are out, but I don't think that it would pose any particularly serious challenges for studio or location shooting, provided that you don't need to move the camera through crowds, attach it to moving vehicles, etc. The IEEE-1394 or Ethernet "tether" can be quite long - up to a couple hundred feet with the proper equipment, as I understand - so you can still do all of the pans, tracking shots, motion control, even handheld to a large degree.

 

There is a blog (now defunct, I think - it hasn't been updated since I started visiting) that chronicled the use of an early version of the Silicon Imaging SI-2K, and it talks a lot about these kinds of workflow issues - that's one of the things that got me thinking about this concept more seriously.

Link to comment
Share on other sites

With the Pike F-145c you can achieve:

 

At 1388x1038= 25 fps.

 

And to calculate other resolutions this is the formula:

 

1388x1038= 1,440,744 pxs.

 

1280x720 (720p)= 921,600 pxs.

 

Now divide 1,440,744/ 921,600= 1.56

 

Multiply 1.56 for the full frame resolution at 25 fps....1.56x25= 39 fps.

For the Pike F-210c is this.

 

1928x1084= 2,089,952 pxs. at 30 fps.

 

1280x720= 921,600

 

2,089,952/ 921,600= 2.26x30=67.8 fps!

 

That formula is the same for any MVC and any resolutions.

Very nice, and thank you, Cesar. I had actually been using that same math to guesstimate the maximum frame rate, but I hadn't realized that it really was that straightforward!

 

I'm curious, why did you decide to go with the Pike F145C over the Marlin F131C for your 720p package? Since the F145C is significantly more expensive, I would assume that the image quality has to be higher, and the higher pixel depth is definitely a plus. But, I was looking at the specifications for the F131C, and it seems to have some things in common - obviously it's CMOS vs. the F145C's CCD, but the pixel size, sensor size, and resolution are roughly comparable, plus is has a global shutter so there shouldn't be any image skew.

 

Would something like the F131C have potential as a lower-cost digital cinema camera? I came across a PDF that lists its specifications, and I think (if I'm interpreting it correctly) it should be capable of 47fps if I'm shooting at 1280x535, which would be in the 2.39:1 ratio.

Link to comment
Share on other sites

Troy:

 

I tested the Marlin F-131c and I did not like it, even thought they share "similar" specs (as the Pike F-145c) the CMOS sensors work differently collecting light. CMOS sensors aren't very good performers in low light.

 

I love CCD's technology and I don't really like CMOS.

 

The price indicates quality in things (most of the time anyways) The Marlin F-033c has the same price tag as the F-131c, and I prefer the F-033c in a heart beat!

 

Thanks,

Cesar Rubio.

Link to comment
Share on other sites

Thanks, Cesar! That's pretty much what I'd guessed. I'm glad to hear that you were actually able to test the F-131c in comparison to the F-145c to make the comparison first-hand. Since I tend to shoot a lot in low light, it sounds like I may need to avoid CMOS if possible.

 

I really wish that I could afford the F-210c (assuming that the video looks as I'm hoping it does), but that would be stretching my budget, so the F-145c might be a good way to compromise if I can't quite finance 1080p. If I shoot 720p in 2.39:1, it looks like I'd be able to achieve 60fps, which would be fantastic.

Link to comment
Share on other sites

Troy:

 

I apologize because I don't have clips taken with the Pike F-145 yet, there's some problems with my distribution channel that need to be resolved. As soon as I get my camera, I will post some footage on my forum.

 

I think that 720 is enough resolution for most circumstances, that is 4 times as much information of SD with letterbox (16:9 aspect ratio 0r 360 x640 pxs.)!

 

If we consider that we are recording uncompressed RAW video, that we could convert to RGB 4:4:4 color space later in post...that is a LOT of info, and then if we shoot at 14 bit, oh well needless to say!

 

I think that the folks that make the Viper, CineAlta, etc won't be very happy with us!

 

Even Sony is the one who produces the CCD for the Pike F-145c...

 

Thanks,

Cesar Rubio.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Forum Sponsors

Visual Products

Film Gears

BOKEH RENTALS

CineLab

CINELEASE

Gamma Ray Digital Inc

Broadcast Solutions Inc

Metropolis Post

New Pro Video - New and Used Equipment

Cinematography Books and Gear



×
×
  • Create New...