First Helium 8k footage

[Goran Ljubuncic]

@Mike Nagel,

The piece was graded with our LUTs. We'll have an interview with Jonny about the workflow and the grading process soon over on our blog.

Thanks.

Goran from Lutify.me

 

[Mike Nagel]

Hello Mike Nagel,

The piece was graded with our LUTs. We'll have an interview with Jonny about the workflow and the grading process soon over on our blog.

Thanks.

Goran from Lutify.me

Hi Goran,

sounds great.

post a link to the blog here once it's up !

 

[Goran Ljubuncic]

Hi Goran,

sounds great.

post a link to the blog here once it's up !

Sure thing!

 

[Filipe Abreu]

I think we still have problems with highlight fringing

 

[Andy White]

I think we still have problems with highlight fringing

Well, a similar often pointed out issue for the Dragon was the horizontal banding/Cmos smear when talent stood in front of blown out or high key light sources (neon strips etc) - everyone jumped up and down and moaned that it was 'only on Dragon etc'. Well, watching episode 1 of Season 4 of Ray Donovan, there's shot in there with the same type of horizontal banding - yet the show is Alexa.

Basically, the purple fringing on highlights is the same type of thing... really, a lot is down to exposure strategy, available shoot time, budget and (of course) skill and experience levels.

This short, BTW, is great - and for only a day/night's work, even better ... but you only have to look at the comments section to see all the 'opinions' on noise etc. Haters will hate - LOL

 

[Filipe Abreu]

Well, a similar often pointed out issue for the Dragon was the horizontal banding/Cmos smear when talent stood in front of blown out or high key light sources (neon strips etc) - everyone jumped up and down and moaned that it was 'only on Dragon etc'. Well, watching episode 1 of Season 4 of Ray Donovan, there's shot in there with the same type of horizontal banding - yet the show is Alexa.

Basically, the purple fringing on highlights is the same type of thing... really, a lot is down to exposure strategy, available shoot time, budget and (of course) skill and experience levels.

This short, BTW, is great - and for only a day/night's work, even better ... but you only have to look at the comments section to see all the 'opinions' on noise etc. Haters will hate - LOL

I also think this short is great, i have been following Abandon Visuals for some time on instagram, loving their stuff, great vision. The thing is i am still surprised about fringing all this years and it is still presente on many sensors, my first real problema was with FS700. Of course it is not very presente on this shot but still i would think by now sensors would have zero fringing. Maybe it has something to do with the lens thar were used. but i think they were Leica C lens so that would not be great for the price they cost

 

[Wayne Morellini]

@Mike Nagel,

The piece was graded with our LUTs. We'll have an interview with Jonny about the workflow and the grading process soon over on our blog.

Thanks.

Goran from Lutify.me

Goran, any chance that you could get some ungraded stills that show the cameras performance for comparison against the graded ones?

Thanks.

 

[Bobby Farokhzad]

Is the Helium sensor a finished product or is it a work in progress still being tweaked? Has there been a side by side comparison between the Helium and the Dragon (paging Dr. Phil)? I would love to know more about the DR and the noise floor of the sensor... 8K resolution is absolutely amazing but I hope it doesn't come at the expense of other factors.

 

[Nathan DuMoulin]

Can someone explain, like your talking to a 8 year old, how a sensor with smaller pixels and with no in-camera gain, can have less noise? Smaller photosites would mean less light reaching each of them thereby resulting in more noise. What's the magical maths I'm missing here?

Back illuminated sensor tech perhaps?

 

[Jacek Zakowicz]

Can someone explain, like your talking to a 8 year old, how a sensor with smaller pixels and with no in-camera gain, can have less noise? Smaller photosites would mean less light reaching each of them thereby resulting in more noise. What's the magical maths I'm missing here?

It comes down to the technology.
Noise is in most part caused by heat. Less heat- less noise- better signal to noise ratio- cleaner image.
As the technology advances the power requirement is lower- there fore the amount of heat generated is less and the level of noise is less. On top of that the black shading algorithms are being improved so they are more efficient and allow more detail to be retained in the image while filtering out the base noise.
All these advancements contribute to cleaner image from smaller pixels as the tech marches on...
Hope that helps...

 

[Adam Montville]

It comes down to the technology.
Noise is in most part caused by heat. Less heat- less noise- better signal to noise ratio- cleaner image.
As the technology advances the power requirement is lower- there fore the amount of heat generated is less and the level of noise is less. On top of that the black shading algorithms are being improved so they are more efficient and allow more detail to be retained in the image while filtering out the base noise.
All these advancements contribute to cleaner image from smaller pixels as the tech marches on...
Hope that helps...

According to the official specs, posted by RED, it has pretty much the same specs as Dragon, even Signal to Noise ratio.
http://nofilmschool.com/sites/defau...erwide/public/red_wle-specs.jpg?itok=le0OuXuo

So I wonder if these are official specs, being a 'pre-release'?

 

[Wayne Morellini]

Can someone explain, like your talking to a 8 year old, how a sensor with smaller pixels and with no in-camera gain, can have less noise? Smaller photosites would mean less light reaching each of them thereby resulting in more noise. What's the magical maths I'm missing here?

8 yr old ae!

Well, its like you have a remote control car, but then, you ste gifted with a smaller remote control car, but it is faster and more powerful and goes longer, due to now having Tate earth motors, higher gearing and lithium batteries little Zeb. The insides were upgraded. Its like your fathers 4 cylinder big car compared to your uncles turbo 4 cylinder sports var.

Back illuminated sensor tech perhaps?

They poked, poked that as rubbish technology when I suggested it a number of years SGI, so of course, it could not possibly be that, could it?

It comes down to the technology.
Noise is in most part caused by heat. Less heat- less noise- better signal to noise ratio- cleaner image.
As the technology advances the power requirement is lower- there fore the amount of heat generated is less and the level of noise is less. On top of that the black shading algorithms are being improved so they are more efficient and allow more detail to be retained in the image while filtering out the base noise.
All these advancements contribute to cleaner image from smaller pixels as the tech marches on...
Hope that helps...

Black.body heat etc, ussually accounts for a small part of noise doesn't it, but at these levels when the rest of the electrical related noise is dealt with maybe it is the most significant noise source left? However, that means that a hot day should be a lot kote noisy than a cool night, something we can test with black footage at two different temperature. But the 80db SNR on the sensor sheet I have here from nofilm school, is rather high. 10 years ago the altasens was supposed to get something like 96db under certain condition I think, used in the SI 2k. Whatever the real case with that, for a 16 bit pixel I would want at least 102db, if not 144-150db SNR. So, it is all relative to what you are expecting. However, if you use electron multiplying tech, which I think was one of the onse looked down upon here, you can multiply the image above the niose, so that the noise floor does not matter so much, meaning 102dn would be very acceptable.

So, this 80 db SNR would look rather move in low light under those conditions (as putting up the gain normally would increase noise, but in am electronic multiplier you would not need to most likely.

 

[Wayne Morellini]

Now, black algorithms, I'm unfamiliar with that term, but I can guess it has to do with noise removal. Which is something I was publically pooed, pooed, by a child, as usual (the only ones that illogically poo poo), from another manufacturer years ago. I think I shut them up when I pointed out, unlike what they were saying in opposition, noise is not an authentic part of the scene, and you could actually restore lost scene data using a 3D noise algorithm to a more closely authentic form than with the noise, so it was in effect the opposite of what was claimed. Of course they could have shut up because they were sulking, like many children who poo poo around forums (mainly narcissistic people that point the finger at others). Definition of narcissism includes lower intelligence, which nearly 100% correlates with my experience on forums of people trying to put one over on me as to their superior grasp of everything rather than being serious about discussing a line of intellectual inquiry.

 

[Wayne Morellini]

Jacek. Can you tell me more about the functioning of black algiritjms? I've been trying to look them up with no success.

Thanks.

 

[Tuomas Hakala]

My understanding as well. So question remains: how can you reduce noise with a much smaller photosite, with no in-camera gain or NR, when the laws of physics remain constant? Not trying to start anything here - just trying to understand.

How does one solar panel produce more output than other one while they are physically the same size? Technology goes forward... if you look at the digital photo cameras, the megapixels have gone up every year so by that logic the cameras now should be like 30-40 times noisier than they were earlier.. but actually it is quite the opposite, pictures seem to get cleaner. While many things contribute to this, sensor designs also get better year by year

 

[Dominik Muench]

Noise is in most part caused by heat. Less heat- less noise- better signal to noise ratio- cleaner image.

so simplified, the lower I set the target temperature on the camera the better ?

 

[Tuomas Hakala]

If the two solar panels have the same stated Wattage Output . . . then they would output the same Wattage
If two photo sensors have the same stated Signal to Noise ratio . . . then they would have the same Noise level

It's not magic - it's physics that I am trying to understand. Without in-camera gain / noise reduction and with the same stated SN ration, how can a sensor, with much smaller photosites, have less noise?

So what I was trying to say that in two solar panels with same size, the new ones are more sensitive to light so therefore they output more. To be honest I have no idea how signal to noise ratios are counted. If RED is saying it is same the same, while users say there's less noise then something is wrong either in that number, user reports or the SN ratio is only part of the equation and can't be directly used to estimate how much noise you have in practice. Under what kind of conditions is this ratio measured? How much it changes in practical use for example because of the temperature? All cameras do image processing, even if it is raw so maybe something got better there or the other hardware is just able to make better use of that sensor. Also if you master to lower resolution and resize then the noise (which is now smaller) becomes less visible than if you stayed in the original resolution. Lots of the people are still monitoring in HD. I guess we have to wait for more tests. Anyways I think we have already seen on still cameras that the pixels/photosites do not just get smaller, they better and i'm sure there are tons of questions that go into sensor design based on what you want to optimize it for.

 

[Mike Nagel]

no users have done real stress tests yet, as the OG Helium squad just got their stuff today... so we gotta wait...

but obviously:

everybody is looking at the downsampled image, which the downsampling algorithm cleans up... the higher res the source has, the more improvement u'll see (on the lower res output)...

so - even if the specs are 1:1 to that of Dragon - since u're now downsampling from 8K compared to 6K u'll see an improvement in your HD/2K/UHD/4K delivery output...

that would not be the case if u compare on a native source pixel to pixel level... although the noise texture itself also plays a role and that may/will be different since this is a new sensor with different characteristics...

I would not be concerned that much about noise when shooting 8K - that will be cleaned up one way or another - question is if all the other stuff holds up or improves compared to Dragon, since we have quite a smaller pixel pitch...

 

[Mike Nagel]

8K down sampled to 4K compared to 6K down sampled to 4K is a different conversation. Is there actually less noise or is it just perceptual?

that is the only conversation there is and a clear calculated advantage of an oversampled high res capture provides, which has always been REDs approach.

but you wouldn't know that.... ;-))))

 

[Antony Newman]

+) As each sensor pixel is struck by light (photons), there is a chance that 'electric charge' is moved.

+) If 1 in 10 photons are converted to the movement of charge we could describe this as the sensor having a 'quantum efficiency' of 10%.

+) If on AVERAGE the sensor only sees 10% of the light - we end up with NOISE as sometimes more than 10% is converted, and sometimes less.

+) Red, Green and Blue pixels quantum efficiencies are generally different - and dependent on sensor design for that colour.

+) The noise is as BAD as the performance of the WORST colour.

+) If HELIUM sensor has (purely for example purposes only) increased the quantum efficieny of BLUE pixels compared with the Dragon sensor from 10% to 17%, And the RED & GREEN pixels were already > 17% QE, then the noise of the pixel could be reduced dramatically (all other quantities constant).

+) IF the signal to noise ratio was improved by 1.7 times in this way when capturing light over the area of one Dragon pixel, it would be possible to reduce the area (by around 1.7) times and end up at the SAME signal to noise ratio.

Or to put it another way, a pixel with a higher QE needs less light (photons) to convert into the same number of charge carriers (which are added up to create sensor readout for a pixel).

+) Other aspects that effect noise are the absorption characteristics of the colour filter array (RED, GREEN And BLUE) over each pixel. The MX pixels' filter designed to only let Blue light in, would NOT completely block out other colours. This is normal. The more extreme the rejection of Out of Band (ie unwanted) colour ... the WORSE the QE of conversion of the Colour desired.

+) It is likely therefore that the Helium sensor has an improved Quantum Efficency of RGB, with less wasted (dead area) around each actively collecting area. And from the red look of Helium sensor that Infrared radiation (lots of out of band noise) is being far more effectively reflected.

AJ.

Can someone explain, like your talking to a 8 year old, how a sensor with smaller pixels and with no in-camera gain, can have less noise? Smaller photosites would mean less light reaching each of them thereby resulting in more noise. What's the magical maths I'm missing here?