Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

The Update Thread

Status
Not open for further replies.
egon touste said:
regardless of the technology .. unless red have in camera special noise reduction built in, that may benefit from smaller pixels .. larger photosites are always going to collect more light than smaller ones .. more photons hit them
Click to expand...

Correct. Why would this be considered a Weapon upgrade as well if it didn't perform better across the board or at least equal to with 8k at S35?
 
egon touste said:
you may be missing my point..

if the performance is equal .. its not an upgrade .
if its a small improvement .. its probably not good value

make the photosites better, and also larger.. because epic-x dragon lowlight is poor .. that's what im saying

arri haven't upgraded their sensor..in how many years ?.. its still 2.8k .. and has better lowlight than dragon imo

now imagine an awesome helium pixel scaled to arri resolution , or 4k since that is the foreseeable future ..

red are missing a trick imo
Click to expand...

I'm going to say this again...Jarred's photo of the White Helium shows the camera set at 3200 ISO.

That can't be random.

Yes, smaller photosites collect less light, but that's only one part of the equation.

Once you pack THAT MUCH resolution into a sensor, the noise can also get so fine as to not be noticeable. Especially when printed down to 4K or 2K.

And in regards lowlight, it's already been shown in numerous tests that a WEAPON (not Epic), a WEAPON Dragon with Lowlight or Standard OLPF is cleaner in lowlight than Alexa.

Highlights? That's another matter.

But lowlight? Yes.
 
..maybe far fetched, but it would be cool to see an 8K sensor that could perform some sort of pixel-binning in order to output lower resolutions, faster frame rates, and/or lower compression rates, while maintaining its full VV or S35 sensor size. That MX and Dragon sensors have to crop in for those things makes a lot of sense, but does introduce a little pause now and then. It would be cool to have access to a wider range of frame rates, resolution, & file sizes, without the compromise of sensor cropping.

Totally different on a stills camera, I imagine, but I always dug the "Sensor+" feature on Phase One cameras that somehow combines 4 pixels into 1 big one, lowering resolution and file size, but upping sensitivity. Canon's "small RAW" settings on DSLRs are a similar concept, I think, but with less usefulness (I've never found full-size 5D RAW files to be a burden).

Is there anything like that in the world of motion capture?
 
Im pretty happy with my Weapon ISO levels. I think its a good, not the best, but a good performer in lower light. Tho, I don't really try and shoot at night without lighting but... Either way I don't think the 8K resolution alone is going to drop the Noise down enough to be that significant, but we'll see. I certainly wouldn't cast an idea from a Red touch screen shot.
 
Matt Beardsley said:
..maybe far fetched, but it would be cool to see an 8K sensor that could perform some sort of pixel-binning in order to output lower resolutions, faster frame rates, and/or lower compression rates, while maintaining its full VV or S35 sensor size. That MX and Dragon sensors have to crop in for those things makes a lot of sense, but does introduce a little pause now and then. It would be cool to have access to a wider range of frame rates, resolution, & file sizes, without the compromise of sensor cropping.

Totally different on a stills camera, I imagine, but I always dug the "Sensor+" feature on Phase One cameras that somehow combines 4 pixels into 1 big one, lowering resolution and file size, but upping sensitivity. Canon's "small RAW" settings on DSLRs are a similar concept, I think, but with less usefulness (I've never found full-size 5D RAW files to be a burden).

Is there anything like that in the world of motion capture?
Click to expand...

http://www.kinefinity.com/shop/terra_preorder/?lang=en
It's one of the features of their 6K camera system.
 
Nick Morrison said:
I'm going to say this again...Jarred's photo of the White Helium shows the camera set at 3200 ISO.

That can't be random.

Yes, smaller photosites collect less light, but that's only one part of the equation.

Once you pack THAT MUCH resolution into a sensor, the noise can also get so fine as to not be noticeable. Especially when printed down to 4K or 2K.

And in regards lowlight, it's already been shown in numerous tests that a WEAPON (not Epic), a WEAPON Dragon with Lowlight or Standard OLPF is cleaner in lowlight than Alexa.

Highlights? That's another matter.

But lowlight? Yes.
Click to expand...

More pixels less noise!
Clean ISO 3200!

You're full of it, Nick. :D
 
PS heres the ACS test of Weapon Dragon vs Epic Dragon vs Alexa.

https://vimeo.com/151445749

It clearly shows Weapon is cleaner in shadows than Alexa, and the ACS commentators make note of that. And this is in spite of Dragon having MORE and SMALLER pixels than Alexa.

The test also clearly shows Weapon is a bit cleaner than Epic Dragon, suggesting the new boards are also contributing to improved imagery.

All this suggests its certainly possible Helium could be cleaner than Dragon, much as Dragon was cleaner than MX...regardless of smaller pixels...
 
Last edited:
Going from 5.4 to 5 microns is much easier than going from 5 microns to 3.65.
The improvement over Dragon would have to be 3 times greater than MX > Dragon. BSI might help a little, and perhaps a smaller process node. But still... i find it very hard to believe it'll be better than Dragon.

Just compare the Sony A7s and Sony A7rii. The A7rii has a BSI sensor, but it still can't match the non-BSI A7s in low light.

Alexa has a different CFA than Red which lets through less light. But you get better colors in return.
 
Last edited:
Patrick Elliott said:
Correct. Why would this be considered a Weapon upgrade as well if it didn't perform better across the board or at least equal to with 8k at S35?
Click to expand...

If it upgraded an 8k weapon, the sensor is bigger and the pixels would be the same size, bigger than on s35 helium, and probably better. But when is Red going to start ro use that 27 stop omron hdr car camera technology width color constancy (the Quality HDR) from 12 years ago.

Now Matt, if they add pixel binning it affects the mix of the pixels and affects the performance. I know what you mean, but there is limited binning because per 4 group of pixels only two greens can be properly binned, but the binning circuit might reduce performance. It is why rolling shutter of tennis used instead of global etc, and probably natural HDR rather than active versions. However, I prefer pixel counters that can count even a single photon. Instead of a pixel well collection, it counts up hits, but if there is a flood of photons, how do you reliably count them (which is not so needed, as photon counting is really a low light deep feild astronomy photography technique I think). But, accuracy at higher light levels is not as much a issue, and naybe if they could combine a well and a counter system they could get the best of both world's. Put it this wat, a electron multiplication system out does Red, Arri, whatever in the cinema market at the moment for low light, likely by a looong way. The images produced are pristine for low light noise, as the electrons from incoming photons is multiplied far above the noise floor. If they can figure out how to do this properly on a cinema camera, it makes everything so far look like a low performance toy cameras.
 
Nick Morrison said:
PS heres the ACS test of Weapon Dragon vs Epic Dragon vs Alexa.

https://vimeo.com/151445749

It clearly shows Weapon is cleaner in shadows than Alexa, and the ACS commentators make note of that. And this is in spite of Dragon having MORE and SMALLER pixels than Alexa.

The test also clearly shows Weapon is a bit cleaner than Epic Dragon, suggesting the new boards are also contributing to improved imagery.

All this suggests its certainly possible Helium could be cleaner than Dragon, much as Dragon was cleaner than MX...regardless of smaller pixels...
Click to expand...

It's interesting that they downplay how good Weapon is with shadows by saying "this is way under what is normal", but when looking at over exposure, that's normal, since Alexa handles it well.
I didn't realize Weapon was this good with shadows though, a good video to show the difference and improvements of Weapon over Epic Dragon even if they share the same sensor.
 
If increasing resolution does not reduce apparent noise ... then colour dithering doesn't work, and DSD also does not work.

bitmap_diffus_gif.gif


If however moving the above image from further away transitions the 'noisy grain' into a hue ... then smaller pixels will likely reduce the apparent noise in an image.

This could mean that if Dragon pixels were reduced in size, and had an equivalent Noise / Unit area, the increase in 'sample' points in the image could be used to 'better average out' each displayed pixel and reduce noise in the final image.

Couple this with the likely hood that a Helium is likely to have better Noise-per-unit-area than a Dragon pixel, and it looks like Helium 8k could have a stop (or more!) recommended ISO range compared to Dragon 6k.

This logic Could mean that it would be in our interest to have a 70k (11k) Helium VV sensor over a 35k (8k) Helium VV, as there would be proportionally less noise when displayed at 4k UHD.

AJ
 
egon touste said:
wow, you are totally missing the point .. but making mine ..
so what you're saying is.. when arri release a new sensor, benefiting from new advances like weapon that may have happened over the last 5-6 years..and if they stick with their big/better pixel design pjilosophy.. itsll wipe the floor with dragon .. that's my predicition.. arri highlights, and improved shadows..

arri have large slow moving fans that are quiet also .. red have small fast moving fans.. these are design choices.. of course epic is much smaller..

the fact were comparing old and new techonology , and the old is doing so well...

earlier in this thread i wrote "special" noise reduction.. that was a typo.. i meant spacial noise reduction..
more pixels may help with that .. but traditionally that will increase the sensors fill factor .. reducing overall available light catching wafer goodness..

its good that weapon is holding up so well in the shadows ..
Click to expand...

Not sure what your point is (exactly), but you do know Arri is (supposedly) making a 6K s35 sensor right? Thats been the rumor for 2 years. Meaning they're going smaller pixel as well. How they do that and maintain the fantastic peformace of their current sensor...is perhaps why it's been so delayed to market?

They can't stay at their pixel size if they want to make a real 4K s35 solution. Their workaround so far has been a bigger physical sensor, aka Alexa 65.
 
Last edited:
I've just realised a new way of noise reduction, scrap that, reverting to just an idea, as the way I was thinking could confuse real data with noise, but then again a lot of pixel values are fairly static or slow moving during sample. As the pixel well is charging up, fluctuations can be indicative of a form of niose that can be accessed and filtered, but that is only one form. In ADC fluctuations might more regularly imply noise. So all through the chain of sampling after well finishes sampling (even in a rest phase, which might indicate a persistent noise event) unexpected behaviour can be examined and isolated, as innthesw behaviours non noise behavour should be predictable. And wlhile writing this I realised one of my old new circuit innovations might be able to be used to sample pixels.

Now the issue, new circuits potentially increase heat and noise, but can there be a gain? Now, can this be applied to photon counting? That might produce maximum benefit ratio.
 
egon touste said:
simple, the test is comparing a 6year old sensor with the latest sensor from red ..

as for arri making a real 4k s35 camera .. they don't seem that bothered currently, and a couple of my clients bought Amira because they aren't either it seems.. and none of their clients are grumbling either .. go figure

I'm guessing denser pixels may help with debayering somewhat (the colour samples that need to be guesstimated are nearer to known colour samples), but other than that ..as you say, they are only necessary to reach some perceived need for a certain resolution, for a certain sized sensor .. for reasons that aren't always clear to alexa users .. but I'm guessing at the expense of iso performance / depth of field/ optical flexibility (i won't say creative flexibility, because higher resolutions gives you c. options also )

I actually like red highlights . ive also seem tests that seem to indicate the sony f55 has more highlight latitude than alexa (they may be flawed tests) .. but overall image quality is not as organic as alexa, or dragon. the sony has the global shutter though , which i would consider a massive plus
Click to expand...

Agreed on all points. Arri's only impetus to have a valid 4K solution is many productions are now mandating a true 4K capture (Netflix, Amazon, apparantly HBO, etc).
 
Boy Jarred is a smart guy. Shifted the whole convo from "where's my camera?" to "what do you think this new camera is?" ;)
 
If Red wants to bring back the $5,000 off plus $5,000 in Red credit, I would be willing to switch my Scarlet W upgrade to a Weapon upgrade. Just saying. I'm sure that would jump start the que.
 
Brian Pascale said:
If Red wants to bring back the $5,000 off plus $5,000 in Red credit, I would be willing to switch my Scarlet W upgrade to a Weapon upgrade. Just saying. I'm sure that would jump start the que.
Click to expand...
I would upgrade my epic dragon to weapon if this happened again. I didn't know about the credit and stuff, just heard the day one announcement of $19,000 for the upgrade and checked out. But now at $15,500 and a $5k credit, I would have been all over that had I known about it.
 
Status
Not open for further replies.
Back
Top