Scarlet-W 5k is actually Scarlet-W 6k in disguise

[AndreeMarkefors]

From the coordinated article/interview with Jarred on Nofilmschool we can read the following nuggget:

NoFilmSchool: You’ve mentioned that RAVEN is a physically smaller (4.5K) cut of the DRAGON sensor, is that the same with SCARLET-W, or is it similar to the original SCARLET that had 5K M-X sensors that weren’t up to the highest spec but worked for 4K and lower resolutions?

Jarred Land: SCARLET-W uses the same sensor as the 6K DRAGON, but only has an active 5K image area. We are the only company other than Sony in this market that designs both our own ASICs and sensors, so we can leverage yields in the manufacturing process to accommodate an array of products in different price points that all share the same DNA.

So, when asked specifically if Scarlet-W uses a cut of the sensor, Jarred say it's the same sensor as in the 6k, but with a smaller active area. Make no mistake: the Scarlet-W already has a 6k sensor in there!

Let this post then, officially be the first in a large number of nagging posts, always polite and constructive, but never yielding, in an effort to wear team RED down until they activate the whole sensor for non-competitive fps designated for STILLS work. I'd honestly take 5fps, but why not bump it to 12 fps, same as what my Canon 1Dc can do in raw?

I'm NOT interested in people asking for 6k 24 fps at high compression ratios. Or any type of wider ratios extending film shooting above 5k. RED has made the "5k active area" decision for a reason. But I REALLY feel that Weapon owners would not mind if we could shoot 6k stills with Scarlet-W.

Let the nagging begin!

 

[sam karr]

Yeah it's weird that they let the 6k sensor in the scarlet-w without activating it entirely. But the biggest mystery for me is why limiting 5k at 8:1 48fps? With the new datarate and higher compression we could easily go up to 60fps @5k FF 10:1 or so I guess, or even higher with 14:1 compression. They let the full power @4.5k, and limit the 5k framerates. It's not like 60fps would destroy the epic or the weapon. But it would feel like having a real 5k slowmo camera with full frame 5K 60fps which is the first real step for slowmotion. We're jumping from 48fps 5k to 100fps 4.5k. That means always going down in resolution for regular slowmo FF.
But well, they did an unexpected gift to Raven owners offering 4.5k. Red is unpredictable.

 

[Subhadip Sen]

Jarred is pretty clear that this is done to increase yields, something semiconductor manufacturers have been doing for decades, of course. Yes, it is a 6K sensor, but it goes into a Scarlet W instead of a Weapon because the 1K lookaround has some defect or poor frequency/power characteristics. Instead of throwing away these semi-defective chips, they can simply be reused in a Scarlet W, lowering costs and increasing yields. This is the same reason, for example, a GTX 970 has 1664 cores @ 1.05 GHz; while a GTX 980 is 2048 @ 1.15. It's the same silicon.

I'm not saying every Scarlet W sensor is like this, so it may partially be a product differentiation strategy too. Either way, it is safe to say that Red cannot just "activate" the extra resolution, not for every Scarlet W. The 6th K is simply broken.

 

[Anton Shavlik]

edit: nvm

 

[Michael Tiemann]

Yeah it's weird that they let the 6k sensor in the scarlet-w without activating it entirely. But the biggest mystery for me is why limiting 5k at 8:1 48fps? With the new datarate and higher compression we could easily go up to 60fps @5k FF 10:1 or so I guess, or even higher with 14:1 compression.[...]

Sam, you don't need to guess. Guessing leads to error. The mathematical fact of the issue is very, very simple: SCARLET-D could do 5K 24fps at 8:1 with a 72MB/sec data rate. SCARLET-W has double the data rate, and it can do 5K 48fps at 8:1. 48fps is twice the frame rate as 24fps, which means SCARLET-W is using all its horsepower to deliver its result. You can call it a limitation, but the limitation is due to mathematics. Moreover, the datasheets clearly state that it can do 5K-WS 60fps at 8:1, which is brilliant. SCARLET-W really is 2x the camera at about 1/2 the cost, which is very Moore's Law of RED in this case. That is very sweet.

 

[sam karr]

Sam, you don't need to guess. Guessing leads to error. The mathematical fact of the issue is very, very simple: SCARLET-D could do 5K 24fps at 8:1 with a 72MB/sec data rate. SCARLET-W has double the data rate, and it can do 5K 48fps at 8:1. 48fps is twice the frame rate as 24fps, which means SCARLET-W is using all its horsepower to deliver its result. You can call it a limitation, but the limitation is due to mathematics. Moreover, the datasheets clearly state that it can do 5K-WS 60fps at 8:1, which is brilliant. SCARLET-W really is 2x the camera at about 1/2 the cost, which is very Moore's Law of RED in this case. That is very sweet.

http://www.red.com/tools/recording-time

But scarlet-D can also do 5k 48fps 16:1. Which will give 5k 96fps 16:1 with double datarate if I follow your condescending lecture. Well the link will help you.

 

[Kwan Khan]

Ufff...

 

[Christopher-lee dos Santo]

Please help me understand this. I am a Scarlet Dragon owner... So, this new Scarlet-W sounds really great, but it's 13.8 megapixels compared to the Scarlet Dragon's 19 megapixel sensor. Does the Scarlet Dragon not have more resolution compared to the Scarlet-W due to this, or does the Scarlet Dragon only use 13.8 megapixels like the Scarlet-W at 5k resolution.

If not, and the Scarlet Dragon does indeed have more pixels than the Scarlet-W, does this not mean the Scarlet-W is a downgrade? Sure, the newer body is great, and shooting 5k at 4:1 is amazing, but i shoot 5k WS at 7:1 and having more resolution (pixels) would give a cleaner 4k finish would it not?

Just wanting to understand the differences here before I dive into this new Weaponized Scarlet. It sounds amazing, but want to be sure.

 

[sam karr]

Please help me understand this. I am a Scarlet Dragon owner... So, this new Scarlet-W sounds really great, but it's 13.8 megapixels compared to the Scarlet Dragon's 19 megapixel sensor. Does the Scarlet Dragon not have more resolution compared to the Scarlet-W due to this, or does the Scarlet Dragon only use 13.8 megapixels like the Scarlet-W at 5k resolution.

If not, and the Scarlet Dragon does indeed have more pixels than the Scarlet-W, does this not mean the Scarlet-W is a downgrade? Sure, the newer body is great, and shooting 5k at 4:1 is amazing, but i shoot 5k WS at 7:1 and having more resolution (pixels) would give a cleaner 4k finish would it not?

Just wanting to understand the differences here before I dive into this new Weaponized Scarlet. It sounds amazing, but want to be sure.

19 megapixel is when you shoot 6k. When you shoot @5k with the scarlet dragon you use "only" 13.8 megapixels, like with the scarlet-W. For motion the scarlet-w is a better camera, twice better power. But you can't use anymore 6k for stills. 5k on scarlet-w and 5k on scarlet-dragon have the same amount of pixels, 5120x2700, and the same pixel size (0,005). The only difference is compression (and framerates @4.5k and lower). You will get a cleaner finish because you have a lower compression.

 

[Michael Tiemann]

If you look at the Media Operations Guide (version N) you will see that across the WEAPON line, the trend is toward lower compression rates, not higher frame rates. The Carbon Fiber Weapon (RED's highest spec model to date) doesn't offer higher framerates than EPIC DRAGON (except maybe in one corner case or two). What it does do is to reduce all the compression ratios to (low) single digits. The SCARLET-W follows its bigger sibling's pattern very closely here.

Now I will make a guess (which could be wrong): RED has decided that the amounts of compression required for the frame rates they were advertising were, ultimately, a compromise that hurt both users and the brand. A further guess is that when it came time to design the WEAPON ASICs, they made an implementation choice to reduce the maximum amount of compression available vs. other things the ASIC might be tasked to do. If the compression limitation is wired into the ASIC, not some line of software code, it's not a decision that can be undone by a firmware upgrade.

The fact remains that SCARLET-W is 2x the SCARLET-D at about 1/2 the price (if you add the cost of WiFi, LUTs, external recorders for prores, etc to the SCARLET-D). I think it is disingenuous to look at a datasheet that clearly offers 2x the performance of the previous model and complain that it's been deliberately crippled to be not 2x better. You may have wished that they had allocated the 2x performance push to a different figure of merit (framerate instead of REDCODE compression rates), but it's not like it's just not there.

 

[Scott Crawley]

Jarred is pretty clear that this is done to increase yields, something semiconductor manufacturers have been doing for decades, of course. Yes, it is a 6K sensor, but it goes into a Scarlet W instead of a Weapon because the 1K lookaround has some defect or poor frequency/power characteristics. Instead of throwing away these semi-defective chips, they can simply be reused in a Scarlet W, lowering costs and increasing yields. This is the same reason, for example, a GTX 970 has 1664 cores @ 1.05 GHz; while a GTX 980 is 2048 @ 1.15. It's the same silicon.

I'm not saying every Scarlet W sensor is like this, so it may partially be a product differentiation strategy too. Either way, it is safe to say that Red cannot just "activate" the extra resolution, not for every Scarlet W. The 6th K is simply broken.

Those who want to learn more about this can google Product Binning or follow this Wiki link: https://en.m.wikipedia.org/wiki/Product_binning

 

[sam karr]

If you look at the Media Operations Guide (version N) you will see that across the WEAPON line, the trend is toward lower compression rates, not higher frame rates. The Carbon Fiber Weapon (RED's highest spec model to date) doesn't offer higher framerates than EPIC DRAGON (except maybe in one corner case or two). What it does do is to reduce all the compression ratios to (low) single digits. The SCARLET-W follows its bigger sibling's pattern very closely here.

Now I will make a guess (which could be wrong): RED has decided that the amounts of compression required for the frame rates they were advertising were, ultimately, a compromise that hurt both users and the brand. A further guess is that when it came time to design the WEAPON ASICs, they made an implementation choice to reduce the maximum amount of compression available vs. other things the ASIC might be tasked to do. If the compression limitation is wired into the ASIC, not some line of software code, it's not a decision that can be undone by a firmware upgrade.

The fact remains that SCARLET-W is 2x the SCARLET-D at about 1/2 the price (if you add the cost of WiFi, LUTs, external recorders for prores, etc to the SCARLET-D). I think it is disingenuous to look at a datasheet that clearly offers 2x the performance of the previous model and complain that it's been deliberately crippled to be not 2x better. You may have wished that they had allocated the 2x performance push to a different figure of merit (framerate instead of REDCODE compression rates), but it's not like it's just not there.

I dont' see any compromise that would hurt both users and the brand regarding compression and datarate. We had 5k 48fps @16:1 on Scarlet-D, it's a high compression, and it was ok. So I don't see how 5k 60fps 10:1 would be a shame for red image. It would be fucking clean. We are just discussing courteously and wondering why, as Red spirit had always been about not limiting cameras on purpose. They even launched a 4k camera and make it 4.5k a few days later because they found a way to do so, even if now it's so close to scarlet-W specs.

 

[Christopher-lee dos Santo]

Thanks Sam... That clears up a lot of questions. Appreciate it.

 

[Nathan DuMoulin]

http://www.red.com/tools/recording-time

But scarlet-D can also do 5k 48fps 16:1. Which will give 5k 96fps 16:1 with double datarate if I follow your condescending lecture. Well the link will help you.

It's quite probable that it's not a question of compression or the 140MB/s datarate limit, but more so the limitations of actual sensor readout prior to compression. At 5k, there is probably just too much bandwidth necessary to read the full 5k at a higher frame rate. Hence why you can get higher frames rates on cropped modes, as the smaller image area means less required bandwidth, and avoids the bottleneck created by trying to process too much data in real time.

 

[sam karr]

It's quite probable that it's not a question of compression or the 140MB/s datarate limit, but more so the limitations of actual sensor readout prior to compression. At 5k, there is probably just too much bandwidth necessary to read the full 5k at a higher frame rate. Hence why you can get higher frames rates on cropped modes, as the smaller image area means less required bandwidth, and avoids the bottleneck created by trying to process too much data in real time.

You're probably right.

 

[Grant Smucker]

I highly doubt the sensor is maxed as this is the same dragon sensor in the epic which is achieving speeds much higher.

It's quite probable that it's not a question of compression or the 140MB/s datarate limit, but more so the limitations of actual sensor readout prior to compression. At 5k, there is probably just too much bandwidth necessary to read the full 5k at a higher frame rate. Hence why you can get higher frames rates on cropped modes, as the smaller image area means less required bandwidth, and avoids the bottleneck created by trying to process too much data in real time.

 

[Nathan DuMoulin]

I highly doubt the sensor is maxed as this is the same dragon sensor in the epic which is achieving speeds much higher.

Something has to process that data from the sensor, which is likely the internal boards, which vary greatly from version to version. I'd imagine the quality of components effects everything from bandwidth, to cooling, processing power, write speeds, etc.

 

[Grant Smucker]

Something has to process that data from the sensor, which is likely the internal boards, which vary greatly from version to version. I'd imagine the quality of components effects everything from bandwidth, to cooling, processing power, write speeds, etc.

Correct. I thought you were referring to only the sensor.

I think what sam is saying is, if it has the same Dragon 6k sensor and can write 140MB/s, why don't they let you record 6k 24 8:1 (72 MB/sec), or 5k 96 16:1 (127 MB/sec). They are limiting the camera in software.

 

[Scott Crawley]

Correct. I thought you were referring to only the sensor.

I think what sam is saying is, if it has the same Dragon 6k sensor and can write 140MB/s, why don't they let you record 6k 24 8:1 (72 MB/sec), or 5k 96 16:1 (127 MB/sec). They are limiting the camera in software.

It is not as simple as that. My understanding is that Scarlet's sensor is the same Dragon in that it is cut from the same wafer but it does not perform reliably at the same speeds. Rather than toss the lot they divide them into groups based upon performance and rate them according to speed. Those that rate at the higher spec go into Epic Weapon while the slower ones go to Scarlet.

Therefore, if the chips were fast enough to perform at Epic Weapon resolutions and frame rates they would be placed in Epic systems.

No it is not just the sensor. Other proceeors in the system such as the ASIC are manufactured and sorted (or binned) in a similar fashion.

Is software used to throttle the system down and prevent errors? I don't doubt it, but it is not a limitation that is placed on the system arbitrarily in order to cheat you.

 

[Phil Holland]

From the coordinated article/interview with Jarred on Nofilmschool we can read the following nuggget:
So, when asked specifically if Scarlet-W uses a cut of the sensor, Jarred say it's the same sensor as in the 6k, but with a smaller active area. Make no mistake: the Scarlet-W already has a 6k sensor in there!

Having done interviews like this in the past, I can totally see where something "missing" might lead to others "hypothesizing".

However, I'm nearly 100% confident that Jarred meant to say "it's the same sensor technology as in the 6K".

It's clearly listed as 25.6x13.6mm sensor size on the RED product and tech specs page.

It appears that while Scarlets of the past (MX and Dragon) had a slightly larger sensor size that RED has taken new steps to make sure that all of the cameras in the DSMC2 line-up have at a minimum "at speed" and high speed filming capabilities at maximum resolution and format size of each sensor size.

I think there are indeed new cuts of the DSMC2 Dragon Sensor Technology in 4.5K, 5K, 6K, and 8K sizes for Raven, Scarlet-W, and Weapon models.