M-X, read-reset, & shutter speed

[Steven Caesare]

I'm looking forward to seeing what the shutter-speed possavilities with the DSMC's will be.

With no mechanical shutter, the shutter speed should be determined by the read/reset time of the sensor, correct?

Assuming the 5ms M-X r/r time, that would seem to imply a maximum of 1/200th of a sec. Doesn anybody know if the 5ms time Jim has mentioned for the M-X is true on the Epic platform as well as on the R1?

-Steve

 

[Terry VerHaar]

When you say 1/200th "maximum," do you mean it can't go to a shorter exposure (i.e. 1/300th, a larger denominator) which is actually a smaller number, or do you mean it can't go to a "bigger" number like 1/100th or 1/8th?? I think I take you to mean the former - which really should be referred to as a "minimum" exposure, no?

If so, am not sure why a 5 ms read-reset time would limit the exposure. Even if the read-reset time is 5 ms (1/200th of a second), why couldn't the sensor be "exposed" (read) for 1/1000th of a second or less (just as an example)?

If I am not mistaken, read-reset time is just the difference in exposure from the top to the bottom of the sensor but does not limit how long the sensor is "on" for exposure. Someone please correct me if I am wrong in this perception. :confused1:

 

[Steven Caesare]

You are correct with your first indication... it's the "maximum" speed the sensor can be read at, and thus the shortest duration shutter. IOW: You can't have a FASTER shutter speed than 1/200th sec, if my logic holds true.

Now, I could be ignorant of CMOS sensor design, but I am unaware of if it's possible to turn the sensor "off" to stop exposure duration as long as it's beeing exposed to light. Most of the high-speed stuff that can do short durations seem to implment a global shutter, which Jim has indicated the M-X is not.

But I don't even play a sensor engineer on TV, so perhaps there's something else in play.

An interestintg thought is that the EPIC is capable of 120-150FPS. If it can maintain a 180-degree shutter angle for those frame rates, then it would HAVE to be capable of faster-than-1/200th speeds. So I dunno.

What's the fasted shutter speed the R1 is capable of with M-X?

-Steve

 

[Terry VerHaar]

A quick glance at the RED ONE manual suggests that the RED ONE can shoot at up to 1/2000th of a second? How would you explain this if your assumption about the read-reset time is true? Or maybe I am missing your question altogether.

 

[Andrew M.]

I use 1/1000 sometimes on R1.
Normally, (RED is not normal the time you dwell on the photocell reading it and how fast you can read all photocells is loosely connected.

 

[Steven Caesare]

Clearly the "shutter" time is far faster than what the R/R would imply then.

And that's a good thing... thanks for chiming in. I was wondering if we were going to be able to get in to the same shutterspeed neighborhoods as DSLRs with mechanical shitters, and it sounds like we are.

-Steve

 

[J. Bernard Vallon]

the shutter most likely operates by 'closing' the top lines first, like a curtain, on down the sensor in time. This gives each line the necessary 5ms to r/r, but keeps the exposure time for each line to much smaller durations.

 

[Terry VerHaar]

the shutter most likely operates by 'closing' the top lines first, like a curtain, on down the sensor in time. This gives each line the necessary 5ms to r/r, but keeps the exposure time for each line to much smaller durations.

That makes a lot of sense. I'd love to get an "official" explanation from RED - or at least from a camera sensor guru who can verify how this works. :thumbsup:

 

[Rick Burnett]

The read/reset number is the time delay reading down the entire sensor, from the tests I have done, I have verified this. The thing is, you don't have to wait for the the whole sensor to be read before resting each line that was read, however, how long that line takes to be read and reset itself I have not seen published. That is the limiting factor.

 

[Rick Burnett]

That makes a lot of sense. I'd love to get an "official" explanation from RED - or at least from a camera sensor guru who can verify how this works. :thumbsup:

This is correct, this is why they call it a rolling shutter. Each line is read out as you move down the sensor horizontally, then reset in the same fashion when the next frame needs to start.

 

[Terry VerHaar]

This is correct, this is why they call it a rolling shutter. Each line is read out as you move down the sensor horizontally, then reset in the same fashion when the next frame needs to start.

Thanks, Rick.

I hate to be thick about this, but just to make sure I understand the details...

Each entire line, all pixels from left to right, is read at the same time, correct? Kind of "flashes" on (as opposed to one pixel at a time from the left to right)?

Then the next and the next... one line at a time, in very rapid succession?

But, will the first line "shut off" after the set exposure time (say 1/48th of a second), and then the next and the next, again in "rolling" sequence - each shutting off a little after the previous one; each after having been correctly exposed at 1/48th?

So, then the read-reset time would be how quickly this process could start again with line 1? Somewhere, with the Epic and MX around 9-14 ms, I believe? (Although I think the target is to get it down to 4-5 ms, eventually.) In other words, line one could fire again just 9-14 (or so) milliseconds after it shut off, to start the process over again?

In this manner, each line of pixels could be "on" for as little as 1/2000th of a second (or even less), and not be limited by or effected by the read-reset time. Yes?

Again, thanks for the info and sorry if I am obsessing over understanding this completely. :beer:

 

[Rick Burnett]

I cannot say specifically for the Red One or other Red cameras, but typically, all of the sensors I have studied perform similar to what you think, although you do not need to wait for all the lines to read out to start again.

Each line, from what I see, reads out all the bits in parallel. This is why there is no gain in FPS if you choose to shoot a narrower image, since it takes the same amount of time to read 1 or 4096 pixels. After that line is done reading out, I am not sure exactly what is happening. If it were me, I'd leave it alone till the end of the shutter time and then reset it. The 9-14ms is to go across the whole set of rows. If you have a very long shutter time, chances are, you are starting the next set of captures on a row before you've read out all the frame before it.

With a fast shutter time, you are resetting and reading very quickly, and then, waiting the rest of the frame time before you do it again.

So, if it takes 10ms to read/reset say 4096 rows, each row then probably takes 2.44us which means theoretically it could do 1/409600 if it could reset/read immediately. I am sure there are other timing considerations and bottlenecks that prevent this fast, but, it just shows that right now, the rolling shutter hasn't been limiting the shutter time itself.

That said, to me, the big issue is the fact that the rolling shutter takes that long, you are going to get a lot of slant in movement at really fast shutters because the complete lack of motion blur is going to not hide that. But we know this.

 

[Fredrik Harreschou]

This is why there is no gain in FPS if you choose to shoot a narrower image, since it takes the same amount of time to read 1 or 4096 pixels

Not sure if I am following you on this one.. The R1 is capable of doing 30 fps at 4K and 120 fps at 2K. Because they crop the sensor for faster read outs at lower pixel counts. The shiXXy DSLRs skip full lines. Please correct me if I am wrong...

 

[Rick Burnett]

Not sure if I am following you on this one.. The R1 is capable of doing 30 fps at 4K and 120 fps at 2K. Because they crop the sensor for faster read outs at lower pixel counts. The shiXXy DSLRs skip full lines. Please correct me if I am wrong...

We are not talking about DSLRs here so it'd be nice if you could keep your inflammatory comment to yourself. Thanks. It has NO relevance to this discussion.

The reason for the FPS difference is that the sensor is cropping out rows, not just columns. Removing columns does not affect read/reset since when a row is read, each column in that row is read out at the same time (in parallel). So, the shorter you make the crop, the faster you can get through the serial operation. But the width of the crop has no difference (other than if you had a data throughput issue in the processing which doesn't seem to be the case with the Red cameras. )

 

[Fredrik Harreschou]

OK, see what you mean. Didn´t mean to offend anyone.

 

[Steve Wake]

Thanks, Rick. I learned a lot from that last post.

Now I'm wondering if EPIC has a special stills sensor mode and or ASIC that would operate at extremely fast shutter speeds, compressing the time between the line by line readouts.

Also (maybe) gives some insights into HDRx.

 

[ericyoung]

We are not talking about DSLRs here so it'd be nice if you could keep your inflammatory comment to yourself. Thanks. It has NO relevance to this discussion.

The reason for the FPS difference is that the sensor is cropping out rows, not just columns. Removing columns does not affect read/reset since when a row is read, each column in that row is read out at the same time (in parallel). So, the shorter you make the crop, the faster you can get through the serial operation. But the width of the crop has no difference (other than if you had a data throughput issue in the processing which doesn't seem to be the case with the Red cameras. )

I may be wrong, but I'm pretty sure the reason for cropping is not to change reading out time. It's to reduce the amount of pixel data at the higher frame rate to fit into the Redcode processing pipeline bandwidth and the transfer bandwidth to recording media.

It must be a processing data throughput issue, otherwise we'd be able to get 5K at 100fps on the Red One M-X which has the same sensor as the new Epic-X.