Will the Scarlet have noticeable rolling shutter skew?

[nick and]

I posted this topic on another forum a week ago, but I thought I'd post it here now that I'm a member and there are quite a lot of people here.

We are currently shooting with a Panasonic HMC150. We selected this camera because it has solid state memory, a low price, and CCD sensors rather than CMOS.

We're now looking for a camera that can produce richer images, preferably something with a larger sensor or at least better sensitivity. In any case, our budget is around $10,000.

Some type of DSLR setup seems to be out of the question, because all of these cameras use CMOS sensors and exhibit significant rolling shutter effects. We shoot material with a lot of high speed motion so this is not an acceptable characteristic.

We're quite interested in the Red Scarlet, but I'm a bit concerned about the rolling shutter issue. I understand the Red One has this problem to some degree, tho not nearly as bad as most CMOS cameras.

I read the One has a read/reset time of 9µs, whereas the new sensor cameras will have a 5µs speed. It's my understanding film cameras have a 4µs time, so at first glance it looks like the Scarlet and Epic will be very close to film cameras. However, a film camera works somewhat differently than a line-by-line CMOS sensor, so I'm not sure if this is directly comparable.

In any case, I'd love to hear what those 'in the know' have to say.

 

[Kenneth Elkington]

It's up to you whether it's acceptable, but from what I've seen, the RED one only exhibits noticeable skew in the most extreme circumstances, like super quick panning. If Mysterium-x is just over half of the skew time as has been said, then I think you'd be hard pressed to ever see a variance from film.

 

[Justin O'Neill]

The mechanical shutter on a film camera leads to 4 microseconds of skew. The RED One had 9µs which led to issues in certain situations. The Mysterium-X in the Scarlet and Epic will have 5µs so it will be very close to film.

Here is jim's quote:

Sorry to slightly disagree with Mitch. Spinning mirror is 4 microsecs and Epic is 5. One won't help Epic at all. It already matches (virtually) what a spinning mirror does on a film camera for skew.

Jim

 

[Charles Angus]

It's too bad one of the things that changed was user-controllable skew... At one point, global shutter was on the table as a menu option, IIRC.

 

[Stephen Gentle]

It's too bad one of the things that changed was user-controllable skew... At one point, global shutter was on the table as a menu option, IIRC.

I only saw speculation about that from people on this forum, not actually from the RED team. The only official word I remember hearing was that it's going to be faster than RED ONE's sensor.

 

[Lawrence Bansbach]

It's too bad one of the things that changed was user-controllable skew... At one point, global shutter was on the table as a menu option, IIRC.

I only saw speculation about that from people on this forum, not actually from the RED team. The only official word I remember hearing was that it's going to be faster than RED ONE's sensor.

Actually, Jim Jannard himself mentioned the possibility more than a year ago, well before the redesign:

I have resigned myself to having to de-bunk the "rolling shutter myth". People tend to hang on to soundbites. We could easily do a global shutter. It just isn't the right solution. If we made a global shutter sensor, everyone would wonder why the footage looked sterile.

One of the options we have is to give the user adjustable read-reset times... just so we don't hear that we are "marketing". From zero (global) to something (what we recommend and looks right).

 

[Pierre T.]

There is a simple criterion : a camera that can do n fps cannot have a delay of more than 1/n between the top and the bottom of the image. It seems that the 2/3 Scarlet should do more than 120fps at full resolution (and that was last year and the frame rates for the Epics have since been increased, so maybe ...), so the delay between the uppermost and lowermost line of the sensor would be less than 1/120s. You can also say it differently, as a still camera, these allow at least 1/120s sync speed compared to 1/250s or 1/300s for high end DSLRs. Actually, I believe it will be faster than that.

For comparison, the 1080p image extracted from a 5D II has a delay of 1/40s (so at least three times slower).

Conclusion, I just would not worry about that one. Note, that slow motion should amplify the perceptible skew.

 

[Zakaree Sandberg]

no more than what the red one has..

 

[Deanan]

There is a simple criterion : a camera that can do n fps cannot have a delay of more than 1/n between the top and the bottom of the image. It seems that the 2/3 Scarlet should do more than 120fps at full resolution (and that was last year and the frame rates for the Epics have since been increased, so maybe ...), so the delay between the uppermost and lowermost line of the sensor would be less than 1/120s. You can also say it differently, as a still camera, these allow at least 1/120s sync speed compared to 1/250s or 1/300s for high end DSLRs. Actually, I believe it will be faster than that.

For comparison, the 1080p image extracted from a 5D II has a delay of 1/40s (so at least three times slower).

Conclusion, I just would not worry about that one. Note, that slow motion should amplify the perceptible skew.

A couple of clarifications... skew is tied to the read/reset rate per row and not tied to fps. Also, max sensor fps on scarlet 2/3rds is 180fps.

In the case of S35 or 2/3rd, skew is way way less than RED One.

 

[alex lemon]

A couple of clarifications... skew is tied to the read/reset rate per row and not tied to fps. Also, max sensor fps on scarlet 2/3rds is 180fps.

In the case of S35 or 2/3rd, skew is way way less than RED One.

180 fps!! :shocked: I take it that is in "burst" mode, but still...awesome!

 

[Peter Mosiman]

A couple of clarifications... skew is tied to the read/reset rate per row and not tied to fps. Also, max sensor fps on scarlet 2/3rds is 180fps.

Nice little tidbit there Deanan. Last we heard max fps was 150 :devil:

 

[Lawrence Bansbach]

Nice little tidbit there Deanan. Last we heard max fps was 150 :devil:

The theoretical maximum may be 180 fps, and that was indeed the original burst-rate figure, but was ratcheted back to 150 fps, possibly because of heat build-up.

 

[nick and]

A couple of clarifications... skew is tied to the read/reset rate per row and not tied to fps. Also, max sensor fps on scarlet 2/3rds is 180fps.

In the case of S35 or 2/3rd, skew is way way less than RED One.

That's great to hear. I was originally looking for an S35, but I am now considering the possibility of a 2/3rd. These are exiting times.

I plan to use the Scarlet for a car-themed show, with high speed motion and lots of quick pans, plus some shots with a fair bit of vibration (i.e. camera suction mounted to vehicles). With that in mind, rolling shutter can be a big problem. I really hope the Scarlet won't exhibit the typical problems, because it sounds perfect to me in every other way.

 

[Pierre T.]

A couple of clarifications... skew is tied to the read/reset rate per row and not tied to fps. Also, max sensor fps on scarlet 2/3rds is 180fps.

Well the read/reset rade per row limits the framerate, you can agree to that. 180fps means at least (180*numberoflines) lines per second. That is the whole argument.

So, if the framerate can be high, than the read-reset per line must be at least as fast. I know that there are other bottle necks to framerate, e.g. codec, but the read rate is one of them.


I think, camera reviewers should adapt requiring manufacturers to publish that figure (lines per second) and learn how to measure it (not very difficult). I measured the lps for the Canon 5D II and estimate it at 43200lps or 23µs per line.

I am confused:

I read in one of JJ's postings about 5µs (per line?) for Mysterium X. At 180fps with 5µs per line, you cannot have more than 1080p, impossible. At 3K 2:1 format, still, with 5µs, you cannot have more than 100fps, with 4µs, not more than 125fps.

 

[paulherrin]

I wouldn't worry. The Scarlet isn't one of those hack-job DSLR's. You'll be good to go.

 

[Kenneth Elkington]

Well the read/reset rade per row limits the framerate, you can agree to that. 180fps means at least (180*numberoflines) lines per second. That is the whole argument.

So, if the framerate can be high, than the read-reset per line must be at least as fast. I know that there are other bottle necks to framerate, e.g. codec, but the read rate is one of them.


I think, camera reviewers should adapt requiring manufacturers to publish that figure (lines per second) and learn how to measure it (not very difficult). I measured the lps for the Canon 5D II and estimate it at 43200lps or 23µs per line.

I am confused:

I read in one of JJ's postings about 5µs (per line?) for Mysterium X. At 180fps with 5µs per line, you cannot have more than 1080p, impossible. At 3K 2:1 format, still, with 5µs, you cannot have more than 100fps, with 4µs, not more than 125fps.

Lines are read vertically, so it's 1500 lines, not 3000. And the red/refresh happens at the end of the exposure, not continuously. This means that you expose, then read, then refresh. That's why your estimates are off.

 

[Charles Angus]

Well the read/reset rade per row limits the framerate, you can agree to that. 180fps means at least (180*numberoflines) lines per second. That is the whole argument.

So, if the framerate can be high, than the read-reset per line must be at least as fast. I know that there are other bottle necks to framerate, e.g. codec, but the read rate is one of them.


I think, camera reviewers should adapt requiring manufacturers to publish that figure (lines per second) and learn how to measure it (not very difficult). I measured the lps for the Canon 5D II and estimate it at 43200lps or 23µs per line.

I am confused:

I read in one of JJ's postings about 5µs (per line?) for Mysterium X. At 180fps with 5µs per line, you cannot have more than 1080p, impossible. At 3K 2:1 format, still, with 5µs, you cannot have more than 100fps, with 4µs, not more than 125fps.

5 microseconds for the entire sensor, I believe.

 

[Pierre T.]

Lines are read vertically, so it's 1500 lines, not 3000. And the red/refresh happens at the end of the exposure, not continuously. This means that you expose, then read, then refresh. That's why your estimates are off.

Will get it right this time 4µs per line allows for 250000 lines per second or 166,67 frames per second with 1500 lines, or 231,48 frames with 1080 lines.

If it was 5µs, these figures become 200000 lines per second and 133,3 fps for 1500 lines per frame or 185fps for 1080 lines per frame.

 

[Stephen Gentle]

A couple of clarifications... skew is tied to the read/reset rate per row and not tied to fps. Also, max sensor fps on scarlet 2/3rds is 180fps.

In the case of S35 or 2/3rd, skew is way way less than RED One.

So the specs on Red.com (which were changed to 150fps burst the announcement before October 30th if I remember correctly) are out of date then? That would be awesome!

Or is 180fps the max the sensor can do, and 150 is the most you can currently record? Maybe 180fps could come with faster media later down the road?

 

[nick and]

Look at the 5D jello:

http://www.vimeo.com/6435335

This is still a concern I have with the Scarlet when mounting to a car. I'm really hoping the 5µs read/reset makes this a non-issue, but I'm not sure how I'm going to be able to find out definitively.