HDRx has a problem IMO

[Mark A. Jaeger]

I like the flexibility that HDRx allows. However, the function only behaves as you would expect at 1/48th second shutter. At all other shutter speeds, the addition of HDRx drives the top-bar shutter speed to strange values. I asked Tech Support why this was and I was told that the forced values avoided "non-aesthetically pleasing combinations". I find that answer hard to accept.

Here are a sampling of the forced values:
Stops
HDRx
0. 1/48 1/96 1/24 1 sec
1 1/48 1/60.55. 1/39.69 1/30.33
2 1/48 1/60.55. 1/39.69. 1/15.17
3 1/48 1/60.55. 1/39.69 1/7.58
4 1/48 1/60.55. 1/39.69. 1/3.79
5 1/48 1/60.55. 1/39.69. 1/1.9
6 1/48 1/60.55. 1/39.69. 1/1.02

To my eyes, particularly at 1 second shutter the theory of doubling the shutter speed for 1 stop of HDRx falls apart. It is as though the table of values is up-side-down where 1/1.9 at HDRx=5 should be indexed to HDRx=1.

Do any of you know why the DSMC2 forces these values? Why is it not 1 stop of shutter speed increase for every stop of HDRx no matter what you started with unless you're exceeding speed limits (like more than 1/8000)?

PS: my table looks good until the SAVE. Sorry but I think you can still discern the values.

 

[Mark Phelan]

My guess would be motion blur. HDRx was useful for the one-time I used it in a production. But even with that, there was motion blur that can be seen in the shot. I kept it because the HDRx solution was a better outcome than worrying about the motion blur for the very short clip, but it's still there. The plusses outweighed the minuses.

I'm guessing it was a judgement call of sorts from RED based upon diminishing returns.

 

[Finn Möller]

@Mark can you post the example ? Would be interested. Heard about ghosting with HDRx, but never experienced it by myself. Maybe I don't look close enough.

 

[Michael Winokur]

unloading sensor

unloading sensor

Just a guess but I think your math assumes that the second frame recorded starts at n+0 where n=shutter speed. You probably need n+y where y = the time it takes the i/o and sensor to prepare to start the next frame. Remember we can't unload the entire sensor at 1/300 that's why you crop to a small part of the sensor to get fast frame rates, less lines of sensor sites to unload and reset. So that time is small but not zero.

 

[Mark Phelan]

Just a guess but I think your math assumes that the second frame recorded starts at n+0 where n=shutter speed. You probably need n+y where y = the time it takes the i/o and sensor to prepare to start the next frame. Remember we can't unload the entire sensor at 1/300 that's why you crop to a small part of the sensor to get fast frame rates, less lines of sensor sites to unload and reset. So that time is small but not zero.

Take a look at my comment #25 in this thread where I post a link and a couple images: http://www.reduser.net/forum/showthread.php?173800-Have-we-forgotten-about-HDRx

That scene has motion blur, you can see it in his cap as he moves backwards. It's so fast, it blends into the scene like regular blur, but if you freezeframe you might see it.

 

[Mark A. Jaeger]

Just to be clear. I am not so concerned about motion blur because I use HDRx for landscape scenes where the DR of the Monstro is challenged to cover the scene well. Motion is slight or none.

My main concern is the changes the camera forces on the A-track recording. As I frequently shoot in low light conditions at 1 fps and 1 second shutter I definitely do not need/want the shutter to go to 1/30 at HDRx=1 (see table above). The scene goes dark and there's no way to add light or change other properties to compensate.

I'e been told that the DSMC2 shutter speed forced changes avoid ghosting and other artifacts but, with little or no motion in the scene, this is not an issue for me. I would very much prefer that the ghosting-avoidance would be handled by the DP and not restricted in-camera to values that, for me, do not work at all. A shutter speed of 1/48 is the only shutter speed that I have found that behaves like the documentation describes. For me 1/48, in low light, is not workable.

Your thoughts?????

 

[Mike P.]

Can you add additional shutter options because clicking on the "advanced"/"..." like you can with regular shutter?

Also, aren't the forced shutter options to make the downstream shutter angles less problematic? So it forces you to go with those wonky shutters because the next shutter (1/121.1 for one HDRx stop) results are best? I dunno.

 

[Alan Greene]

I always use shutter angle instead of shutter speed, as I just find the math to be far more intuitive and easier to visualize. I hadn't noticed (or had forgotten) that the shutter angles are restricted in HDRx mode. Of course, it makes sense, as some of the exposure time has to be reserved for capturing the X track.

However, I'm seeing that the shutter angle is much more restricted with HDRx 1 than any other setting (142.9° is as high as it goes). It seems like we should be able to get closer to 240° at HDRx 1, since that would allow for one exposure to be about twice as long as the other. I've probably never used HDRx 1 and have therefore never noticed this limitation.

As soon as I go to HDRx 2, I can open up to 285.9°. This makes sense to me. An X track that's two stops down would be one quarter of the exposure time. Divide the maximum exposure possible (360°) by five (one short exposure, and one that's four times longer), and you get 72°. Multiply by four to get 288°. So it seems that the setting controls the longer of the two exposures?

I obviously don't really know how this works.

Remember we can't unload the entire sensor at 1/300 that's why you crop to a small part of the sensor to get fast frame rates, less lines of sensor sites to unload and reset. So that time is small but not zero.

If that were true, wouldn't it mean that a 360° shutter would not be possible at full resolution?