Scarlet and READ-RESET Speed

[Terry VerHaar]

It has occurred to me that the overall size of the Scarlet sensor, as well as the smaller size of its pixels (3.3 microns vs. 5.4 in the Epic), might give it an advantage in its READ-REST times. Is this a logical and likely conclusion to draw?

At one point, I believe Jim said the goal was to get the READ-RESET time on Epic to approach the 4 ms or so equivalent of film. Last I remember, they had achieved something like 8-10 ms in the pre-production Epics but had decided to scale that back a bit while they worked out the kinks. In any case, might it be easier to get the smaller Scarlet sensor close to that "holy grail" of 4 ms and thus, virtually eliminate the "wobble" and "skew" issues of the rolling shutter?

Any thoughts? :beer:

 

[Michael J. Underwood]

Ooooh I hope so. I shoot a lot of red carpet, and the biggest issue is the flashes from the gang of photogs at every event. CCDs can handle it, but CMOS sensors usually display flash banding, i.e. the shutter is slower than the strobe, so only a portion of it gets recorded. Unfortunately it's an issue that is hard to search for, "flash" being a super common search term. Can anyone recommend a previous thread that may have covered this?

 

[Chuck Z]

It has occurred to me that the overall size of the Scarlet sensor, as well as the smaller size of its pixels (3.3 microns vs. 5.4 in the Epic), might give it an advantage in its READ-REST times. Is this a logical and likely conclusion to draw?

No and no.

It's tied to the speed of the read-out electronics in the camera.

 

[Clint Lealos]

Epic read reset time is 5ms. I believe Scarlet is the same, but not sure.

 

[Kalani Prince]

5ms, that is damn close ; )

 

[Ben Coleman]

It has occurred to me that the overall size of the Scarlet sensor, as well as the smaller size of its pixels (3.3 microns vs. 5.4 in the Epic), might give it an advantage in its READ-REST times. Is this a logical and likely conclusion to draw?

At one point, I believe Jim said the goal was to get the READ-RESET time on Epic to approach the 4 ms or so equivalent of film. Last I remember, they had achieved something like 8-10 ms in the pre-production Epics but had decided to scale that back a bit while they worked out the kinks. In any case, might it be easier to get the smaller Scarlet sensor close to that "holy grail" of 4 ms and thus, virtually eliminate the "wobble" and "skew" issues of the rolling shutter?

Any thoughts? :beer:

I'm so glad that somebody brought this up because it has been very heavy on my mind as well. The fact that I hate flash banding so much is one of the reasons that I demand ccd chips on live events.

 

[Clint Lealos]

Not sure on this, but I believe you will still have problems with flash units. Its my understanding that the way CMOS sensors "fire" is one line at a time starting from the top. I'm not certain that the increased speed of the read reset time will take care of the problem of flash banding. The reason film never gets flash banding has more to do with the way a frame of film is exposed all at the same time and reset all at the same time, than it does with the time it takes to reset each frame. You may still have to use long decay flash bulbs, but I guess that doesn't help you during live events. Hopefully, someone with more knowledge on the topic can weigh in here, cause I might be way off on this one. Fellas?

 

[Eki Halkka]

The reason film never gets flash banding has more to do with the way a frame of film is exposed all at the same time and reset all at the same time,

Film does get flash banding, and it's not read/reset at the same time. The term rolling shutter actually comes from the physical rolling shutters of film cameras.

 

[Terry VerHaar]

No and no.

It's tied to the speed of the read-out electronics in the camera.

Of course, this is true. I guess what I was trying to articulate was whether, within the limits of these electronics, performance might be aided by the smaller sensor. In other words, is it inherently easier to get read-out from a sensor that is physically smaller?

As an analogy, if I have a 300 horsepower engine (the read-out electronics), will I get better performance in this smaller, lighter chassis (the Scarlet 2/3") or is there some other limitation? Does that make sense? :huh:

 

[Terry VerHaar]

Epic read reset time is 5ms. I believe Scarlet is the same, but not sure.

I don't believe they have made it that far yet on the Epic. I remember Jim saying they were closing in on their 4 ms goal but hadn't reached it yet. Last I remember (but, of course, could be wrong) was they were in the 10-14 ms range after backing off a little from the 8-9 they had previously reached in their testing. I'll go looking for the discussion... (wish me luck)...hmy:

EDIT: OK - So far, I was only able to find this quote from Jim on 3-22-2010:

Many CMOS are well above 30ms (jello). The RED ONE is about 9ms. Mysterium-X (including Scarlet 2/3") is 5ms. Film is 4ms.

Jim

I'll try to find support for my (perhaps feeble?) memory in a later quote but, for now... :embarassed:

ANOTHER EDIT: So, now I found this quote, dated a week earlier than the other...

We have temporarily rolled back the clock speed (read-reset time) for the M-X upgrade cameras to work on several issues... calibration being one. As soon as we fix all the "stuff" we'll roll up the clock speed again. As of today's build, you will get skew "slightly better than the M sensor". Soon, that will go back to where it was when we showed M-X at RED Day. EPIC will be released with 5ms out of the box.

I would have posted this before... but I forgot. Sorry about that.

Jim

Obviously this references a roll back on the MX but, at least at that point in time, says EPic will be 5 ms. So far I have found nothing later (but won't give up looking).

All of this suggests that perhaps my inquiry should be put to rest as Epic and Scarlet will likely be nearly equivalent to film and any non-filmic artifacts will not occur.

I should state that my initial inquiry had nothing to do with concerns about "issues" with any of the RED sensors, just a somewhat academic pursuit to understand the behavior and limits of the CMS sensor and a rolling shutter.

Perhaps now I am...:beatdeadhorse5:

 

[Terry VerHaar]

Film does get flash banding, and it's not read/reset at the same time. The term rolling shutter actually comes from the physical rolling shutters of film cameras.

Yeah - I recall Jim stated that film cameras have the equivalent of about a 4ms read-reset time with their mechanical shutter. Personally, I have no knowledge either way but I'll take this as accurate! :thumbsup:

 

[Ben Coleman]

Is the R-R time of a film camera based on the diameter of the shutter wheel and the speed of which it turns?

I'm just trying to understand the physics of the mechanical beast.

 

[Clint Lealos]

Film does get flash banding, and it's not read/reset at the same time. The term rolling shutter actually comes from the physical rolling shutters of film cameras.

I always thought if a flash went off while you were shooting with a film camera that the flash would appear on the entire frame, unlike most CMOS sensors where the flash only shows up as a white bar horizontally across a portion of the frame. Do I have that wrong?

 

[Gabe Conroy]

I always thought if a flash went off while you were shooting with a film camera that the flash would appear on the entire frame, unlike most CMOS sensors where the flash only shows up as a white bar horizontally across a portion of the frame. Do I have that wrong?

Yeah, because you have to consider that a physical shutter has to cross in front of the film, so there will be times when something like a flash will start and stop while the shutter is en route.

The band on a film frame will however be feathered, which may be more desirable than the harsh cutoff on CMOS sensors.

 

[John Godden]

Of course, this is true. I guess what I was trying to articulate was whether, within the limits of these electronics, performance might be aided by the smaller sensor. In other words, is it inherently easier to get read-out from a sensor that is physically smaller?

As an analogy, if I have a 300 horsepower engine (the read-out electronics), will I get better performance in this smaller, lighter chassis (the Scarlet 2/3") or is there some other limitation? Does that make sense? :huh:

Sorry, but that analogy just isn't valid. A sensors physical size has little/nothing to do with the read time.

JohnG

 

[Steven Veldkamp]

Sorry, but that analogy just isn't valid. A sensors physical size has little/nothing to do with the read time.

JohnG

Actually, it does. If the sensor is bigger, it'll take longer for the entire frame to be read. For example, the canon 5d and the canon 7d scan frames at the same speed but the 5D has more rollingshutter/wobble issues due to the sensor being larger.

 

[Terry VerHaar]

Sorry, but that analogy just isn't valid. A sensors physical size has little/nothing to do with the read time.

JohnG

Actually, it does. If the sensor is bigger, it'll take longer for the entire frame to be read. For example, the canon 5d and the canon 7d scan frames at the same speed but the 5D has more rollingshutter/wobble issues due to the sensor being larger.

Obviously, someone is wrong here, although, John -
1) "little" and "nothing" are definitely NOT the same thing, and,
2) "read time" itself, may not be the same as what I am asking, or the limiting factor

Logically, it does seem that it might take longer - all other things being equal (an important "if)) for the "roll" to proceed down the length of the sensor if it is either physically larger or has more lines. Of course, "logic" may not be the prevailing property here - that's why I was asking. Note, though, I was talking about the time it takes to roll down the sensor, not th time it takes for a single line to read, which is, effectively, the shutter speed. I guess the jury is still out until some sensor guru steps in.

 

[Alvise Tedesco]

Interested as well. I tend to be more forgiving with slow read/reset performance on cheap dslrs than on higher class cameras like R1. Alexa, now, is the reference. Isn't it?
Please Jarred, Graeme, Brent, etc, can you chime in and say if you reached better results than, say, the R1?
Cheers