Questions re: recording with HDMI out and video sensors in general

[Chris7]

1. I’m trying to think of a way to get around HDV and AVCHD, which seem like are going to be here for a while. I think I heard that there is a way to use maybe the HDMI out of a camera and record to an external hard drive using a high quality codec. Is this so? What format does HDMI output (RGB, 10 bit 4:2:2 uncompressed, etc.)? Is there a way to transcode it on the fly to something like ProRes or Cineform? Are there any portable (like something you could clip onto your belt) devices like this?

2. I’m not really understanding why prosumer video camera sensors show so much noise when just adding 6-12 db of gain. As I understand it, zero gain means ISO 100, +6 db would be ISO 200, and +12 db would be ISO 400 (am I correct here?). I think both the Canon and Nikon mid level DSLRs show almost no grain at ISO 400. Intuitively, it doesn’t make sense that a $5K video camera would show much more grain at +12 db than a DSLR costing less than $1500 at ISO 400. Are the video sensors just inferior with low light performance, or am I missing something?

Thanks,
Chris

 

[LethalWolfe]

1. I’m trying to think of a way to get around HDV and AVCHD, which seem like are going to be here for a while. I think I heard that there is a way to use maybe the HDMI out of a camera and record to an external hard drive using a high quality codec. Is this so? What format does HDMI output (RGB, 10 bit 4:2:2 uncompressed, etc.)? Is there a way to transcode it on the fly to something like ProRes or Cineform? Are there any portable (like something you could clip onto your belt) devices like this?

There are a couple of portable recording devices I know of off the top of my head that will take HDMI in but you are looking at $3k-$3.5k for them. The cheaper of the two, the Convergent Design NanoFlash doesn't record into a FCP 'friendly' format so you'd still want to transcode before bring it into final cut. The other one, the AJA Ki Pro, records directly into ProRes but it is more expensive and I don't know how much of a quality improvement you are going to see because the biggest limiting factor in terms of image quality in sub-$10k cameras is typically the camera itself not the codec.

2. I’m not really understanding why prosumer video camera sensors show so much noise when just adding 6-12 db of gain. As I understand it, zero gain means ISO 100, +6 db would be ISO 200, and +12 db would be ISO 400 (am I correct here?). I think both the Canon and Nikon mid level DSLRs show almost no grain at ISO 400. Intuitively, it doesn’t make sense that a $5K video camera would show much more grain at +12 db than a DSLR costing less than $1500 at ISO 400. Are the video sensors just inferior with low light performance, or am I missing something?

ISO/ASA ratings don't really apply to video cameras as they handle light differently than film. You can do specific tests on specific cameras w/specific settings and get an idea of the ISO rating but that rating is only good under those specific conditions. With that being said, most of what I've read puts video cameras between 320 and 640. I also don't know if adding gain in a video camera is directly comparable to moving up in ISO. As far as low light sensitivity goes, much of that is dependent the size of the imager and DSLRs have massive imagers compared to prosumer video cameras. Video cameras are also much more expensive and complex machines than still cameras so that obviously factors into the price difference.


Lethal

 

[KeithPratt]

What format does HDMI output (RGB, 10 bit 4:2:2 uncompressed, etc.)?

To add to what Lethal has said, Prosumer/consumer cameras will likely output 8-bit 4:2:2 over HDMI. HD-SDI is 10-bit, but the likes of the Canon XL-H1 is 8-bit with 2 "padded" bits (i.e. not filled with picture information). The only Prosumer camera (if you can call it that) I know that does genuine 10-bit is the Sony EX1/EX3.

As I understand it, zero gain means ISO 100, +6 db would be ISO 200, and +12 db would be ISO 400 (am I correct here?). I think both the Canon and Nikon mid level DSLRs show almost no grain at ISO 400.

As said, ISO is a celluloid thing, and digital cameras only really have an equivalent. Not sure how accurately stated they are across the board. But your question of why a typical camcorder exhibits more noise than a typical DSLR is answered simply: it's all down to the size of the sensor.

 

[Chris7]

(Keith and Lethal: thanks for the info on the HDMI and HD-SDI recording options.)

...the biggest limiting factor in terms of image quality in sub-$10k cameras is typically the camera itself not the codec...

Interesting. So what I hear you saying is that, for example, the Panasonic HMC150 or Canon XH A1 would not look significantly better if they were recorded in, say AVC-Intra or ProRes 10-bit. I'm sure you're correct, but I never suspected this to be the case. Could you please expound on this point?

...most of what I've read puts video cameras between 320 and 640. I also don't know if adding gain in a video camera is directly comparable to moving up in ISO...

As said, ISO is a celluloid thing, and digital cameras only really have an equivalent. Not sure how accurately stated they are across the board.

I have no idea what the correlation is between celluloid film and DSLR sensor “ISO”, but I thought they were equivalent enough that one could use these numbers with a light meter and figure out what to set the shutter speed and aperture to. Am I correct here?

I’ve always read that 6 db of gain is one stop. What am I missing?

Lethal, where I’m really confused is when you say video camera sensors are between 320 and 640. For example, I’ve got a HF100, similar in functioning to a HV30. (It’s a mess to try to set exposure manually because I don’t always know if it’s adding gain. But lets say that, like a prosumer camera, I know how much gain is added and can even dial in a specific amount of gain.) If there is no gain added, is it already shooting at least ISO 320? It seems would need to factor this in if I were to make any sense of using a light meter (like dial in ISO 320 even if there were no gain). But I think I might be missing you here.

...as far as low light sensitivity goes, much of that is dependent the size of the imager and DSLRs have massive imagers compared to prosumer video cameras...

But your question of why a typical camcorder exhibits more noise than a typical DSLR is answered simply: it's all down to the size of the sensor.

Is it the size of the sensor or the size of the photosites? It would seem, for example if Panasonic would have just cut off about a third of the 5D Mk. II’s sensor and used it on its GH1, it would have been able to shoot ISO 1600 with minimal grain (albeit with plenty of jello). Would this be reasonable to expect?

It seems like there’s a bit of argument over the exact relationship between photosite size and sensitivity, with most saying that bigger is better and few saying not necessarily. Before I posted again I just read that the 5D Mk. II’s photosites are 6.25 microns in diameter (.00625mm), and it looks like the photosites on a three 1/3” CCD would be no more than 2.5 microns in diameter (4.8mm/1920 = .0025mm). So the photosites on the 5D Mk. II would have about 6.5 times the surface area of those on a standard 3CCD chip. If the relationship between photosite size (area) and sensitivity were linear, that would be about two and a half stops more ISO. Come to think of it, if a 1/3" sensor had as much grain at ISO 270 as the 5D Mk. II sensor does at ISO 1600, the relationship between photosite size and sensitivity would look relatively linear. But anyway, there's no comparison. Darn.

Many thanks for your time,
Chris

 

[LethalWolfe]

Interesting. So what I hear you saying is that, for example, the Panasonic HMC150 or Canon XH A1 would not look significantly better if they were recorded in, say AVC-Intra or ProRes 10-bit. I'm sure you're correct, but I never suspected this to be the case. Could you please expound on this point?

I haven't done any tests so I can't say for certain but I wouldn't expect a drastic difference. If you shot a lot of green screen it would probably make life easier because the compression won't be as harsh, but if you are shooting green screen you are probably on a stage of some sort and you could just run the HDMI into your Mac for a lot less money.

All in all if you are looking at around $3000 for a camera plus another $3000-3500 for an external storage device you'd probably be better off spending $6k for a better camera.

I have no idea what the correlation is between celluloid film and DSLR sensor “ISO”, but I thought they were equivalent enough that one could use these numbers with a light meter and figure out what to set the shutter speed and aperture to. Am I correct here?

Long story short, you don't need to use a light meter to set exposure w/a video camera because the video camera is a light meter. Pro cameras, and a growing number of consumer cameras, have built in exposure tools such as zebra stripes to guide you when setting up the camera (some newer cameras even have the option to show a histogram). Also, on a shoot you should be feeding the camera to a properly setup monitor so you see exactly what the video you are recording looks like. Ideally you'd be using video scopes as well but that gets expensive fast.

And I believe you are correct that 6db=1 stop.

Lethal, where I’m really confused is when you say video camera sensors are between 320 and 640. For example, I’ve got a HF100, similar in functioning to a HV30. (It’s a mess to try to set exposure manually because I don’t always know if it’s adding gain. But lets say that, like a prosumer camera, I know how much gain is added and can even dial in a specific amount of gain.) If there is no gain added, is it already shooting at least ISO 320? It seems would need to factor this in if I were to make any sense of using a light meter (like dial in ISO 320 even if there were no gain). But I think I might be missing you here.

I haven't used a light meter since college (i.e. many years) but if you google 'video camera asa rating' you'll get a lot of good hits on the first page on the topic as well as instructions on how to use a light meter to find the approx. ASA rating of your video camera.

Is it the size of the sensor or the size of the photosites?

It's the size of the photosite but it's typically true that bigger imagers will have bigger photosites. Most prosumer and consumer HD cameras still use the same size imagers as their SD ancestors which is why these sub-$10k HD cameras have inferior low-light performance compared to similar SD cameras. The balancing act is overall image resolution vs light sensitivity and many times overall image resolution 'wins' because the marketing department likes wowing people w/big numbers.


Lethal

 

[knello]

There are a couple of portable recording devices I know of off the top of my head that will take HDMI in but you are looking at $3k-$3.5k for them.

The Blackmagic Design Intensity captures HDMI to ProRes for $199.

Unfortunately, it's a PCIe card, so it won't work with a portable. For that, you can get something like the Aja IO express, or a Matrox MXO2 Mini. Both of these units require an express card slot, so make Apple hasn't taken away yours yet.

 

[KeithPratt]

I have no idea what the correlation is between celluloid film and DSLR sensor “ISO”, but I thought they were equivalent enough that one could use these numbers with a light meter and figure out what to set the shutter speed and aperture to.

What I meant by that was, I'm not sure how close ISO 200 is on a Canon 5D to ISO 200 on a Nikon D700.

It seems like there’s a bit of argument over the exact relationship between photosite size and sensitivity, with most saying that bigger is better and few saying not necessarily.

It can get vicious. One analogy I'd heard (intending to side with the photositists) goes like this:

Imagine a small box and how many marbles you can fit in it. Now imagine a bigger box and how many more marbles you could fit in that.

The marbles are photons of light and the box is the photosite. Initially you may think that means larger photosite=greater sensitivity; but then consider how many big boxes versus how many small you could space out on the floor in any one room...

You may end up with the same number of marbles overall.

 

[bigbossbmb]

I have no idea what the correlation is between celluloid film and DSLR sensor “ISO”, but I thought they were equivalent enough that one could use these numbers with a light meter and figure out what to set the shutter speed and aperture to. Am I correct here?

Yes.

Lethal, where I’m really confused is when you say video camera sensors are between 320 and 640. For example, I’ve got a HF100, similar in functioning to a HV30. (It’s a mess to try to set exposure manually because I don’t always know if it’s adding gain. But lets say that, like a prosumer camera, I know how much gain is added and can even dial in a specific amount of gain.) If there is no gain added, is it already shooting at least ISO 320? It seems would need to factor this in if I were to make any sense of using a light meter (like dial in ISO 320 even if there were no gain). But I think I might be missing you here.

You would really need to run tests with a gray card and a light meter to figure out the ISO of your cam's sensor. On a video camera like yours, it will likely be between 320-500. My dvx was between 500-640 IIRC (which is very very good for a prosumer cam). If you can't tell if/how much gain is being added, you won't be able to accurately gauge your cam's ISO.

 

[Chris7]

Most prosumer and consumer HD cameras still use the same size imagers as their SD ancestors which is why these sub-$10k HD cameras have inferior low-light performance compared to similar SD cameras. The balancing act is overall image resolution vs light sensitivity and many times overall image resolution 'wins' because the marketing department likes wowing people w/big numbers...

Thanks. I wonder what kind of ISO ratings the prosumer SD cams had. I think what I want most in a camera is one that can shoot in low light with minimal grain. But based on what’s out there, I’m starting to wonder how much most videographers care about low light performance. I understand that if you’re taping a set, you control the lighting. But it would seem that the prosumer crowd would be taping weddings and other indoor events where light would be a real issue. And what about those making documentaries?

The ...Unfortunately, it's a PCIe card, so it won't work with a portable. For that, you can get something like the Aja IO express, or a Matrox MXO2 Mini. Both of these units require an express card slot, so make Apple hasn't taken away yours yet.

Thanks for the rec. I called AJA, and it looks like this might actually work, and the price is right, when the time comes to invest. And yes, my ExpressCard34 slot is still here, not replaced by a SD card yet (Sometimes I can’t tell if Apple is marketing savvy or just crazy. I’m glad Apple’s marketing experiment of removing the FW port from their 13” laptop didn’t work for them, but removing the ExpressCard34 and replacing it with a SD card slot is just crazy. They could have just added FW800 or ExpressCard34 adaptor for SD cards with the package, and still have allowed standard import from the FW800 directly to a hard drive attached to the ExpressCard34 or FW800).

...It can get vicious. One analogy I'd heard (intending to side with the photositists) goes like this...

Yeah, I just found another thread on this on the reduser site yesterday – it gets complicated fast.

...You would really need to run tests with a gray card and a light meter to figure out the ISO of your cam's sensor...

Interesting. The manufactures don't publish this info?

 

[LethalWolfe]

Thanks. I wonder what kind of ISO ratings the prosumer SD cams had. I think what I want most in a camera is one that can shoot in low light with minimal grain. But based on what’s out there, I’m starting to wonder how much most videographers care about low light performance. I understand that if you’re taping a set, you control the lighting. But it would seem that the prosumer crowd would be taping weddings and other indoor events where light would be a real issue. And what about those making documentaries?

It's a trade off. If you want all the things HD does better than SD and you can't afford an expensive camera then you have to compromise somewhere. The EX1 though, from what I've read, offers low light sensitivity that rivals even the best SD prosumer cameras.

Yeah, I just found another thread on this on the reduser site yesterday – it gets complicated fast.
Interesting. The manufactures don't publish this info?

Video cameras measure light sensitivity w/a lux rating and that info is published, but it's from the manufacture so it's typically a BS number anyway that doesn't reflect a real world use of the camera.


Lethal

 

[Chris7]

...The EX1 though, from what I've read, offers low light sensitivity that rivals even the best SD prosumer cameras...

Thanks. I read that the sensor is about ISO 400 and glass is f/1.9-2.8 (although it falsely displays 1.9 at full zoom in). I guess that's not bad compared to the best. A Canon 5D Mk. II shooting ISO 1600 w/ a f/2.8 zoom is only two stops faster. (Just using the Canon as a theoretical "best" for low light only -- I realize it will not do 24P and is very difficult to control for video).