Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Unravelling HDRx

Á

Álex Montoya

Well-known member
Joined
Mar 25, 2007
Messages
1,413
Reaction score
0
Points
0
Website
www.imdb.com
That was shot with HDRx+2, I think. Click here to watch it in high resolution since I believe there's a lot to see.


a006c0270324320000119b.jpg



I'm loving Magic Motion, by the way.

---------------------
Álex Montoya
www.morituri.es
 
Interesting to see that the X frames are using a faster shutter as they alternate. If that's indeed the case.

I wonder what other changes happen between the two frames. I've only done still shots with HDRx, but this is the first I'm seeing something with moving objects.

So maybe A-frame is 1/48th, and X-frame is 1/96th, or something like that?
 
Someone posted these estimates for the shutter degree change for the x track. They seem pretty accurate.

degrees HDR-X fraction
180• off 1/48
90• +1 1/96p
45• +2 1/192
22.5•* *+3* 1/384 *
11.25•* *+4* 1/768 *
*5.625• *+5* 1/1,536 *
2.8125• +6 1/3,072

So if you set a 1/48 degree shutter and +3 HDRx the X track will have a 1/384th of a shutter. Keep in mind these are just estimates.
 
Joseph S. said:
Interesting to see that the X frames are using a faster shutter as they alternate. If that's indeed the case.

I wonder what other changes happen between the two frames. I've only done still shots with HDRx, but this is the first I'm seeing something with moving objects.

So maybe A-frame is 1/48th, and X-frame is 1/96th, or something like that?
Click to expand...

Well, if you are using +2 stops of HDRx, the shutter speed for the x track would be 1/192nd if the a track is at 1/48th. That would give you two stops.

According to everything we've been told about HDRx, the two exposures start at the same time and the x track is read out earlier. I don't believe they alternate.

EDIT: Justin got here first with the whole rundown.
 
Justin O'Neill said:
Someone posted these estimates for the shutter degree change for the x track. They seem pretty accurate.

degrees HDR-X fraction
180• off 1/48
90• +1 1/96p
45• +2 1/192
22.5•* *+3* 1/384 *
11.25•* *+4* 1/768 *
*5.625• *+5* 1/1,536 *
2.8125• +6 1/3,072

So if you set a 1/48 degree shutter and +3 HDRx the X track will have a 1/384th of a shutter. Keep in mind these are just estimates.
Click to expand...

Mathematically speaking those shouldn't be estimates.

HDRx track *should* be:

Exposure * 2^HDRx #

So:
48th * 2^+3 HDR = 384
 
Well, by measuring the blur streaks of both streams it looks like the X one is half the length, which seems to indicate HDRx+1. That's counterintuitive since Metadata in Redcine indicate "HDR stops over 2"

overlapw.jpg


But more importantly, if I haven't done anything wrong, I believe the test shows clearly that both exposures don't start at the exact same time although they overlap.

If main exposure was 1/50th of a second and X exposure was 1/100th of a second, the décalage between them both seems to be 1/200th of a second which sounds about right, since the car at 60Km/h would advance 8,5cm in this time lapse (1/200th of a second, i. e. 0.005 seconds), which seems to be the distance between the start of the X exposure and the one of the A exposure.

Anyway, HDRx works perfectly since, as I said, both exposures overlap. The only 'problem' I find is that you end with a slightly longer blur streak than you would get at 1/50th of a second (i. e. 0.02 sec). At the end you have a streak that is 0.025 seconds long.

hombrepy.jpg


---------------------
Álex Montoya
www.morituri.es
 
This easily became my favorite thread in a long time, as I really like the technical aspect of HDRx.

So we know at 5K, we can do a maximum of 48fps in HDRx mode, which is 96fps (A+X) as it records.

I also noticed the shutter selection maxes out at 1/96th no matter the stops (2 thru 6).

In 2KWS it goes to 120fps and 1/288th shutter max. So it's alternating/overlapping at 240fps max, in HDRx. Wow.

Now I'm trying to learn how this happens. Seems the Epic brain has to switch back and forth between (for example) 1/48th and 1/384th, 48 times per second? I wonder if it really alternates between the two or if it does it at the same time, or does it split the difference a fraction of the way through and overlap a bit.

This is stunning when you think about it. I think we take for granted the amazement factor of the Epic.

If you said this 6 years ago, people would laugh at you (and they apparently did).

And only a few know what Mr. Jannard has around the corner.
 
Edited after paying closer attention:

Out of curiosity, I did a little test to see something faster than the car movement, and recorded it at 1/96th shutter, and 2 stops, moving the top handle really fast to see where the positions are.

I guess it records the X frame, then the A frame (notice the two positions X and A with red pillow). Note, I made the X frame brighter so it's easier to see.

My curiosity is whether it's taking one shot and splitting it DURING capture, like at the start of the frames, or whether it's actually doing two sparate captures. Like X frame at 1/96th, then A frame totally and completely AFTERWARDS.

None of this matters anyway. Just being a nerd.

X.jpg


A.jpg
 
Last edited:
Joseph S. said:
Out of curiosity, I did a little test to see something faster than the car movement, and recorded it at 1/96th shutter, and 2 stops by moving the top handle really fast to see where the positions are.

I know that we all know it records the A frame, then the X frame (notice the two positions A and X with red pillow). Note, I made the X frame brighter so it's easier to see.

My curiosity is whether it's taking one shot and splitting it DURING the capture, like at the end of the frames, or whether it's actually doing two sparate captures. Like A frame at 1/96th, then X frame totally and completely AFTERWARDS. Notice the layer of both A and X in the 3rd picture.

None of this matters anyway. Just being a nerd.
Click to expand...

This is weird. This is backwards from every other HDRx test I've seen. Interesting... now I want to go back and look at some of my HDRx footage.

EDIT: It's also backwards of the footage above. Look at the car and the man the X Track is at the beginning not the end of the exposure. You sure you didn't make a mistake and pull the X Exposure from the next frame?
 
I'll read that article now.

Gavin, I think I made a big mistake. After reviewing the footage, the position where I captured the frame, it ends up the direction was in the other direction.

Which actually makes the X frame first and the A frame second. Same as Alex's test.

So it grabs a fast exposed X frame first, and then continue with the A frame with a longer exposure. Sharing some overlapping data.

I'm just guessing here, not stating any facts :)
 
Joseph S. said:
I'll read that article now.

Gavin, I think I made a big mistake. After reviewing the footage, the position where I captured the frame, it ends up the direction was in the other direction.

Which actually makes the X frame first and the A frame second. Same as Alex's test.

So it grabs a fast exposed X frame first, and then continue with the A frame with a longer exposure. Sharing some overlapping data.

I'm just guessing here, not stating any facts :)
Click to expand...

Ok, good you're back in line with the reset of the universe. Order restored. :D

Yeah it essentially makes a copy of the frame buffer a few milliseconds into the regular exposure and saves that copy out as your X-Track. So it should be at the beginning and it should be identical to the first few ms of your normal exposure.*

*This explanation matches every piece of HDRx footage I've ever seen and every test I've since done... but Jim Jannard keeps telling me it's completely wrong. I'm not sure on which point and he won't tell me. But I'm relatively certain that it's correct.
 
Yes, that's the only way you can make a temporal HDR with rolling shutter. The camera saves the readout for the shorter period and then continues for the extended time and saves again. Taking two frames one after the other would be a complete mess with CMOS – as long as we don't have global shutter, maybe the Dragon will get one ?

The only other approach I know of is spatial. It has probably been used in the Genesis/F35, maybe Alexa and is definitely used in C300. Canon even shows it in their graphics: They have an array of four photo sites, one for blue and red and two for green, one lighter and one darker (the Genesis/F35 did something similar with stripe filters on CCD). This approach has its problems, since very small point lights might flicker when moving. Plus, I doubt you can get to very high ratios, since there's always passing some light from neighboring photo sites.

RED's approach is not without problems too, but the best for moving images I've seen until now.
 
"order restored" - haha.

Got it, so it's overlapping at one point obviously.

I'll go back to more reading and less typing now.
 
Last edited:
You must log in or register to reply here.
Back
Top