EPIC HDR...

[JohnF]

How well do HDR modes deal with transients (Strobes, fireworks etc)?

JohnF

 

[Rob Ruffo]

If they don't put HDRx™ in Scarlet than the next generation of DSLRs (Canon, Nikon, Panasonic, Sony, Fuji, Sigma) could "eat" a lot of

Scarlet's sale with their new generation of sensors that will have the better S/N ratio, higher sensitivity, etc...

HDRx™ is must have for any future RED product Scarlet, Epic, or whatever is called the next camera generation (even with Monstro inside!!!).

That means HDRx™ should be already standard in any future RED product.

I think Sanjin you underestimate just how awful DSLRs are as tools on sets that are not "personal project oriented", and where, if, say, a shot is ruined due to moire or some other problem, it means an expensive re-shoot and likely a lost, valued client. I'd much rather a solid 2/3rds cam than a shaky full frame that lets me down half the time.

 

[Michael Hastings]

Isn't it likely that HDRx is essentially exposing two "fields" (sort of like interlace except they are each full resolution) so call it "48fps-24P/24P+/-3" and using two different shutter speeds to get two different exposures?

Then those two frames are combined, possibly with some interpolation?

And this gives you the FPS limitation since 48fps HDRx is really "96fps-48P/48P+/-3" and some of the remaining processing power has to go to overhead?

 

[Jeff Kilgroe]

I don't think that's how they're doing it. It's been said that there is no temporal offset or sequential frame capture. My guess is they have a way to take two distinct exposure readings from the same read/reset cycle on the sensor. Thus creating two differently exposed frames, but from the same instance in time, therefore no motion issues or interpolation to deal with. However, having two different exposures from the same sensor read-out still creates two images and uses 2X the rate.

 

[Lauri Kettunen]

Isn't it likely that HDRx is essentially exposing two "fields"

Once the cameras are out, we will get a better idea. If the HDRx mode does not work with a 360 degree shutter, then two fields sounds plausible. But, although not familiar with the hardware limitations, would rather guess that they manage to read every line of the sensor twice or something like this. Besides, Jim showed this image with waves on the sea. It's difficult to conceive interpolatation between such fields you suggest would work out in this case.

 

[Daniel Reichenbach]

My bet is that EPIC is the 1st digital camera to beat film in every way.

It is smaller than every 35mm film camera. It has more resolution. More dynamic range. The color science is now perfect. It is cheaper than a film camera. It records to a reusable SSD or CF. The results are NOW.

I have to say that EPIC achieves everything we ever set out to accomplish. Wow.

Jim

Jim, I was at IBC and TED showed me the EPIC and I agree totally with you: There is NO other camera, that will give me all these advantages, from every point. ALEXA is a great product and will find there costumers, no doubt. But my heart beats for revolution AND resolution AND dynamic range AND design AND less money for more camera AND weight AND... You are incredible guys, thanks.

 

[Paul Nordin]

I think Sanjin you underestimate just how awful DSLRs are as tools on sets that are not "personal project oriented", and where, if, say, a shot is ruined due to moire or some other problem, it means an expensive re-shoot and likely a lost, valued client. I'd much rather a solid 2/3rds cam than a shaky full frame that lets me down half the time.

Hi Rob, Not really wanting to enter into another mind numbing theoretical/hypothetical heated debate of Epic vs. Scarlet vs. 5Dm2. But from your comments it sounds like you are having a harder time using current HDDSLR than may be necessary. It is possible to shoot with cams like 5Dm2, and never disappoint your clients, in fact quite the opposite. I've never let any clients down shooting HDSLR and I've used that "brain" in a variety of types of shooting (mostly commercial and narrative). Red has done a masterful job of positioning Epic/Scarlet brains as the sensor technology around which an acquisition technology and workflows need to be applied. Same thing for cams like 5Dm2.

Anyway. This is not a HDSLR forum, but if you would like to know more about my approach to working with the 5Dm2, which generally guarantees client satisfaction, contact me offlist and perhaps it can be of help.

 

[Michael Hastings]

I don't think that's how they're doing it. It's been said that there is no temporal offset or sequential frame capture. My guess is they have a way to take two distinct exposure readings from the same read/reset cycle on the sensor. Thus creating two differently exposed frames, but from the same instance in time, therefore no motion issues or interpolation to deal with. However, having two different exposures from the same sensor read-out still creates two images and uses 2X the rate.

So I think you're suggesting maybe they read the same row of pixels twice before moving to next line?

Do we know if the exposure offset has to go in one direction I.e +3 or minus three from what is set? Or are there 3 exposures? One plus and one minus? Or am I missing something obvious?

 

[Roberto Lequeux]

No promises but we are looking at HDRx™™™ with smaller frame sizes and higher fps.

When we said there was a lot of power inside an EPIC... we weren't kidding.

Jim

Awesome. Flexibility is power! It will be very interesting to find out what kind of small FPS gain we will get from dropping down to an S35 area.

Even if just barely higher, it will be a phenomenal tool to have.

 

[Shawn Nelson]

Hmm. Yea, looks like. Sorry Shawn.

I can follow that sentiment with HDRx™™, yes. If Scarlet only had the easy HDR, I'd be more than happy - I'm unsure enough as it is on how to handle all that data, the render times. Twice the amount... I'm actually hoping that if one or the other is problematic for Scarlet, HDRx™™ is the one which loses out.

Correct, I was saying "it'd be nice if Scarlet does have it"

Though it may be a matter of computer power. If it is, then it may be that only Epic is capable of crunching it.

 

[Jordi Figueras]

About all this discussion about HDR in Scarlet I wouldn't worry too much. It's not 'cause I'm going the EPIC way (stage 2) :thumbsup: but following RED's history, if the Scarlet hardware allows it, for sure it will have it. Also I'm quite sure that if it can't do it, they won't change it, 'cause I think the whole point of Scarlet is making the DSMC concept available to people who can't or are not willing to afford an EPIC system, so if it implies a huge raise in the price it's not gonna have it.

Anyway, in the near future almost all camera systems will have a sort of HDR option, 'cause thats what RED does for the industry and the users: they keep rising the bar!

Keep on RED Team! You are doing great!:violent5:

 

[Todd M.]

I don't understand then.

If the single-stream HDR mode is still RAW, with all the usual adjustment parameters available, with no particular "look" or "tone-mapping" applied, then it sounds like it's just a clip with way more exposure

I would imagine this is correct.

And the final product wouldn't even have to look "HDR-ish" if you didn't want it to.

You can always crush down the blacks or blow out the highlights to reduce the effective latitude.

I'm not sure what's missing from that, for HDRx™™ mode to add.

I think there will probably be tradeoffs where certain scenes may not respond as well to the HDR processing that RED is implementing and going for a full 6 stops may introduce some undesirable elements to the image. Whereas, maybe in that same scene adding 3 stops would have had no negative effect. With Magic mode you are limited to only using that +6 image you captured, while in HDRx™ mode you can finesse the additional latitude back down to the +3 range that was the sweet spot.

Make sense?

edit: obviously I'm making some assumptions here since I haven't seen or used the feature.

 

[Jeff Kilgroe]

So I think you're suggesting maybe they read the same row of pixels twice before moving to next line?

Nope.

Do we know if the exposure offset has to go in one direction I.e +3 or minus three from what is set? Or are there 3 exposures? One plus and one minus? Or am I missing something obvious?

We don't know... Only that when enabled it can be dialed from 3 to 6 stops over.

 

[Rob Ruffo]

Hi Rob, Not really wanting to enter into another mind numbing theoretical/hypothetical heated debate of Epic vs. Scarlet vs. 5Dm2. But from your comments it sounds like you are having a harder time using current HDDSLR than may be necessary. It is possible to shoot with cams like 5Dm2, and never disappoint your clients, in fact quite the opposite. I've never let any clients down shooting HDSLR and I've used that "brain" in a variety of types of shooting (mostly commercial and narrative). Red has done a masterful job of positioning Epic/Scarlet brains as the sensor technology around which an acquisition technology and workflows need to be applied. Same thing for cams like 5Dm2.

Anyway. This is not a HDSLR forum, but if you would like to know more about my approach to working with the 5Dm2, which generally guarantees client satisfaction, contact me offlist and perhaps it can be of help.

I know the ins and outs of the cameras I use (of the sensors and settings, I'm not saying I know everything about cinematography), and all cameras except the Alexa which I have not shot with yet. At one time I was even involved in the design of an unreleased firmware for a very popular camera. I think my clients simply have higher standards than yours, and that oversharpened skin pores, less-than-720p actual resolution and crappy dynamic range are not good enough for the people who hire me, who pixel peep even more than the people on this forum. I also can't hand in something that only has shallow depth of field, because that is just bad filmmaking.

The cheapest sets I work on run at $10 000 a day, and that's the rare lowest end of our work here. We don't do wedding videos or low-end local commercials (I'm not saying you do, but your clients are obviously way easier to satistfy than mine.)

I'm sorry if I sound arrogant, but really, I know the ALL the tricks for "helping" a 5D, and I am not impressed, and not interested in working with a substandard tool, or learning kludgy tricks for it that are little more than a series of limitations as to what kind of shots you can film, when we have our own Red MX right here.

ON A MORE INTERESTING NOTE: The best use for tonemapping, I think, is selective, using masks. This can look totally natural, and is not so hard to do, especially if, say, you are masking simple shapes like a window frame. No wirid 90s video game look, just something which looks more like what your eye would see.

 

[Steve Wake]

ON A MORE INTERESTING NOTE: The best use for tonemapping, I think, is selective, using masks. This can look totally natural, and is not so hard to do, especially if, say, you are masking simple shapes like a window frame. No wirid 90s video game look, just something which looks more like what your eye would see.

But if the window is blown out, masking and adjusting exposure in post won't get the details back - details which you would have seen in real life. With HDR, there will be some detail to play with when you mask and adjust the window. For those of us who don't have sets, crews, equipment, time, budgets, etc. this kind of flexibility would really be nice to have.

Some people might be going for a 90's video game look. Personally, I hate it. In any case, HDR does not have one "look". HDR tonemapping can be subtle or extreme. Unfortunately, most examples that scream HDR! are extreme and have given HDR a bad name in some circles.

I think there will probably be tradeoffs where certain scenes may not respond as well to the HDR processing that RED is implementing and going for a full 6 stops may introduce some undesirable elements to the image. Whereas, maybe in that same scene adding 3 stops would have had no negative effect. With Magic mode you are limited to only using that +6 image you captured, while in HDRx mode you can finesse the additional latitude back down to the +3 range that was the sweet spot.

Make sense?

[/I]

I agree with the first two sentences, in theory at least. But I've gathered the following from reading Jim and the RED team's posts in this thread:

There is a range of adjustment for EasyHDR or HDRx that is +3 to +6. I assume that means the second data stream would basically be from 3 to 6 stops under your "properly" exposed baseline data stream in order to capture 3 to 6 more stops in highlights that were clipped in the baseline footage. That's why it would be possible in HDRx to just reject the underexposed second stream and go with the non-HDR footage.

"Magic" and "normal" are two settings for HDRx that deal with motion blur. I haven't read anything that would indicate Magic is linked to using +6.

And Jim's mention of the "Magic" and "normal" blur correction modes is why I think some of us are assuming there are two different exposures using two different shutter speeds (the way it's done with stills), or some other temporal difference between the baseline and underexposed frame. That's why there would be a need to correct for motion blur when combining frames. If there was a simultaneous dual ISO readout I wouldn't see a need for blur correction.

 

[Craig Parkes]

Here's a thought - If the HDR values are being generated twice, motion blur wouldn't be much of a problem if you aren't actually generating two lots of RGB images, but instead one RGB image based on the debayer of the first exposure, but with additional luminance values in the second stream based on a second exposure, that is then mapped to the luminance values of the first exposure looking at relative change (discarding any massive changes in luminance generated from temporal differences), transforming the original bayer image based on the differences in value to reach the new exposure level, then debayering again.

Any instance where there is a temporal based discrepancy in luminance, you'd regenerate by averaging the transform of nearby pixels and voila, the majority of HDR detail would be retained. There could be the odd weird artifact but because you have both sets of raw data they could easily either be manually rectified in post given the right tool set...

NB: This is purely pseudo science conjecture - anyone feel free to point out the gaping holes in my electronic imaging know how.

 

[Paul Nordin]

I'm sorry if i sound arrogant

My my. you sound very sorry. And you're also making many assumptions, but its not really important.
Cheers

 

[Jay A. Kelley]

I know the ins and outs of the cameras I use, and all cameras except the Alexa which I have not shot with yet. At one time I was even involved in the design of an unreleased firmware for a very popular camera. I think my clients simply have higher standards than yours, and that oversharpened skin pores, less-than-720p actual resolution and crappy dynamic range are not good enough for the people who hire me, who pixel peep even more than the people on this forum. I also can't hand in something that only has shallow depth of field, because that is just bad filmmaking.

The cheapest sets I work on run at $10 000 a day, and that's the rare lowest end of our work here. We don't do wedding videos or low-end local commercials (I'm not saying you do, but your clients are obviously way easier to satistfy than mine.)

I'm sorry if i sound arrogant, but really, I know the ALL the tricks, and I am not impressed, and not interested in working with a substandard tool, or learning tricks for it, when we have our own Red MX right here.

ON A MORE INTERESTING NOTE: The best use for tonemapping, I think, is selective, using masks. This can look totally natural, and is not so hard to do, especially if, say, you are masking simple shapes like a window frame. No wirid 90s video game look, just something which looks more like what your eye would see.

Yikes... Not a good post. I'd delete the line that says you know ALL the tricks for sure...

Trust me, you don't want people reading that. It makes for a very bad impression.

I'm sure you know this already, but there is not a soul alive in this industry that knows ALL the tricks.


In my experience, the best professionals out there consider themselves students of their craft, and feel the day they know it all, is the day the art betrays them.

I'm quite proud to say I do NOT know ALL the tricks.. But it's been 25 years, I know a few, and I'm learning more every day!


Jay

 

[Art Chong]

In my experience, the best professionals out there consider themselves students of their craft, and feel the day they know it all, is the day the art betrays them.

Gold in text form, that statement is.

 

[Brook Willard]

I will tell you that EPIC is the resolution leader. EPIC is the dynamic range leader. EPIC is the bang for buck leader. And EPIC is the size/performance leader.

Can't wait. This HDR mode is going to be a HUGE deal.