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 :)

Pilot shoot in Chicago

David, on your test images you don't mention "MX."

Is this the old sensor or the new?

"I just did some rough tests, nothing scientific, comparing the Red (downconverted from 4K HD to 1080P) to the EX3 and the Canon 550D T2i. The Red had a Zeiss Super Speed on it, the EX3 had the stock lens, and the Canon had to the stock 18-55mm zoom, which I acknowledge is not a good lens, but I did shoot it at f/5.6."
​

If it's the old sensor, I can understand.

If it's the new, then I'm in disbelief.

These much cheaper cameras are holding their own. I see 11 stops across the board. What justifies the extra $20-30K?
 
It was the new M-X Red.

The new cameras in general are pretty sensitive, Red M-X benefits from those new developments that make the sensitive, low-noise Canon image possible. However, the Red M-X is still less noisy than the Canon in HD mode -- I don't know if the way the Canon gets HD is making the noise more obvious or it simply is a magnification issue since the full Red frame and full Canon still frame are simply larger overall.

I was a bit surprised that the EX3 was pretty decent in low-light once you pushed the gain a bit. But even then, the Red M-X is less noisy, but you are right that all three aren't that far apart in low light.

But obviously you are paying more for a lot more than noise level. Resolution for one thing -- the Red is a 4K RAW camera, the EX3 is a 3-sensor HD camera, and the Canon barely shoots HD resolution.

The fact that all three cameras can register the 11-steps in a SMPTE chart is no big accomplishment. There's a difference between seeing brightness changes and being able to use that information when color-correcting, I wouldn't say that the bottom end of the Canon or EX3 is quite as usable as the bottom end of the Red M-X and you've got more than 11-stops to work with on the Red.

You have to shoot a lot more tests to really determine the workable dynamic range in terms of ability to color-correct.

Besides, it's always the last 10% of quality that accounts for 90% of the cost increase, whether it's a lens or a digital camera.

You also can't see the 1/2" sensor DoF that the EX3 has when pointed at a chart. You can't see chromatic aberrations from the prism-block or ENG lens. You can't see the chroma aliasing of the Canon (I turned off the color) either but it's definitely there.

I mean, if you are shooting for theatrical, you have to look at the images BIG to see where they fall apart. Look at "Public Enemies" for example -- I'm sure that the F23 and EX3 both matched pretty well when pointed at charts but on the big screen, you can tell that the EX3 is crapping out sooner in extreme situations (too bright, too dark, etc.) than the F23 is.

And ultimately remember that I'm dumbing everything down to HD -- the Red is a 4K camera that can achieve 3.5K resolution.

I can't believe we've already moved from wondering how the Red One manages to be more than 1/4 the cost of most pro HD cameras like the $150,000 Sony F35... to implying that it's too expensive for the quality you get out of it. The Red One is a BARGAIN.

You are paying, among other things, for a processor capable of handling a full sensor high-rez output (10MP or so) in RAW mode at movie frame rates.

All I can say is go out and shoot a feature on all three cameras and you'll see by the end of the first day why the extra $20-30K is justified!
 
This is sort of why I am loathe to post tests online, people taking them too seriously while looking at cropped & compressed JPEGS. This is why I take pains to say that they are unscientific and not particularly accurate (my own tests) and why I take anyone else's tests with a grain of salt.

I just posted them for the rough impressions that make, not the fine details. I figured that in this case, since what I am shooting is for TV distribution, a rough impression is useful. For theatrical projection, I'd have to do all of this work at full resolution of each camera and project the results on a cinema-sized screen to form an opinion.

And my rough impressions are that the Red has more resolution than the EX3 and then Canon has even less than either. The Canon has larger softer more chromatic noise patterns (actually reminds me a bit of the old Red sensor in 3200K at 800+ ASA levels), whereas the Red M-X and EX3 has the tighter noise that you typically see in high-end HD and digital still cameras.

As far as contrast, the Red has more workable DR in color-correction, the Canon has almost none, you can't drift too far from what you shot. The EX3 is closer to the Red in this regards but I wouldn't draw conclusions about DR from my SMPTE chart -- for one things, you're looking at RedGamma for one, Cine Gamma 4 for the other, and whatever Neutral means for the Canon (I assume it falls within Rec 709's world.) I took all three stills into Photoshop and chopped off the ends in Levels so that the blacks would be black and the whites would be white. I didn't take much time to match them, but I had to add contrast to the Red and EX3 chart to get closer to the Canon image.
 
David Mullen ASC said:
...I can't believe we've already moved from wondering how the Red One manages to be more than 1/4 the cost of most pro HD cameras like the $150,000 Sony F35... to implying that it's too expensive for the quality you get out of it. The Red One is a BARGAIN....
Click to expand...

It certainly is..that's why some 6,000 or so have been sold.
 
I finished the Fox TV pilot last week, a 14-day shoot. We shot something like 60 hours of footage for a 42 minute pilot, 97% of it on the M-X Red One, a couple of shots on the Sony EX3, maybe one shot on a Canon T2i is in the final cut, and there was an opening car chase sequence where the 2nd unit portion was shot with Genesis cameras, just because another Fox pilot had just wrapped and we could use their cameras and crews to shoot this day of second unit work. It was easier than trying to find another three Red cameras for that day, in addition to our own package – plus I didn’t have to figure out how they were going to download their data during the day since they were on the other side of Chicago from the main unit.

The M-X Red One cameras performed very well overall, very few bugs or other problems, in all sorts of weather.

The best thing, of course, was using the new M-X sensor with the new color science – I believe it’s a major improvement, eliminating most of its weaknesses compared to other digital cameras and surpassing the competition in other areas. I was able to routinely rate the camera at 1000 ASA for most interiors; this was a godsend because most of the show was shot on 24-290mm Ang. Optimos mostly at the extreme telephoto end. We often didn’t even cut the camera after we shot the wide shot, we kept zooming in tighter and tighter until we were grabbing tight coverage at 290mm on two cameras as actors crisscrossed the rooms. So focus-pulling was near impossible and the only help I could give the focus-pullers was as much stop as possible, hence the 1000 ASA indoors. Outside in daytime, I stuck to 500 ASA, usually working at T/16 – and believe me, there is not much depth of field at T/16 when you’re in a close-up at 290mm; the background is still a complete blur, but at least you have some wiggle room for the actors to move, focus-wise. But night exteriors were often shot at 1600 ASA, wide-open at T/2.8, and focusing was a nightmare.

I’ve just started to grade the material at Encore in their D.I. theater, a real treat for a TV show, and the colorist is amazed at the low noise of the 1000 ASA footage, even at 3200K balance.

Because of the shooting style, a lot of this shoot involved available light whenever possible, and what lighting I could do had to work for most of the coverage; I rarely was allowed the time to relight a close-up totally, not when we had to average 50 to 60 set-ups a day. Luckily, being a police drama, a natural look to the lighting was appropriate. I was also fortunate that our main actress, Jennifer Beals, looked great and didn’t much special lighting attention.

We had a couple of scenes on the streets under the elevated train tracks in Chicago (the “El”) and the contrast was very challenging – luckily the new M-X sensor handles that sort of situation very well. The TIFF frames I grabbed look very much like film scans, a very natural-looking contrast.

I had one scene where the rolling shutter was a bit of a problem – a night exterior outside a hospital with dozens and dozens of police cars in front. Chicago police cars use a rack of blue strobes on the roof; it never seemed to be much of an issue before, but this scene was shot at 1600 ASA, 270 degree shutter, because it was such a huge area to light, and the blue strobes were a major source of illumination. With so many going off at once, you could see these faint half-lit patterns on the white police cars, almost looking like the roll bar of a TV monitor. Anyway, there was no way to avoid using the strobes so I lived with the effect, figuring that once the scene was cut-up, it would almost be like those blue flares you get with anamorphic lenses, sort of a stylistic trick of the scene.

Because so many locations had sodium streetlamps or industrial fluorescents and metal halide fixtures, I ended up gelling almost every light I had with some amount of Plus Green, knowing I’d have to pull down some green in post overall. In camera, we turned down the Green Gain from 1.0 to 0.8 or 0.7 so that the monitor image wasn’t too green.

We rented the M-X cameras and lenses from Fletcher of Chicago, including their Cosmos cart which has RedCine-X and is set-up for back-ups to RAID and LTO. It can even generate dailies using a Red Rocket, but we were shooting so much material a day that our digital loader often worked an hour or two after wrap just doing back-ups. We sent a copy on a shuttle drive to Encore in Los Angeles; we got DVD dailies back a few days later. Encore is using a Red Rocket to transcode to 1080P; everything is being stored on their servers, including the R3D originals should we need to go back to them.

Maybe eventually I’ll be able to post some selective frames.
 
David- thank you so much for sharing these real world examples of the sensor- People like me get dizzy reading all the technical saber rattling (i.e Alexa vs MX) et al.Just want to say thank you for sharing all this incredible practical experience, no-nonsense threads like these are what make this site great. Thanks also for sharing the onset and post workflow.

MY brother lives in Chicago, and Im eager to see the EL shots and especially the blue police light effect you spoke of.

EDIT: Quick question. I am certainly familiar with shutter speed, but how does changing the shutter angle effect the shot? I read about that all the time on here
from DP's and would love clarification. Thanks
 
Same here. I would love to see grabs of the locations and lighting you described. Thank you for sharing. Can't wait to hear about post either. I would love to see what you do to the 1600 shots with access to the RAW. However, it sounds like you might not be working with the r3ds?
 
Since this is for television, it was all converted to 1080P in RedGamma; if there are any problematic shots, timing-wise, we have the option to go back to the R3D files. Ultimately, Fox broadcasts in 720P so the 1080P master will get downsampled to that.

The most problematic shots to time will be the car chase under the EL; we had three cameras rolling on two different camera cars (the scene involved two cars racing side-by-side with dialogue shouted between them, so for safety's sake, shots where you could see the actors involved a tow rig.) One car was towed by an SUV with an Ultimate Arm, with me, the focus-puller, the camera operator, the director, the driver, and the remote crane operator, all crammed inside the SUV (very snug). The other car was towed by a camera car. I had one camera on the camera car, the camera on the Ultimate Arm crane, and a third camera on a hostess tray locked-off and triggered remotely.

I could only control the exposure of one camera (the Ultimate Arm crane-mounted one) myself with a handset. The B-camera crew on the other camera car could manually adjust iris. The camera on the hostess tray had to have the iris set before we hit the road.

So the problem was that we shot all day long, from sunrise to sunset, in all directions, and from full sun outside of the EL to being in and out of the shadows under the EL. So exposures varied wildly. I could ride the iris on my camera, the B-camera crew were given a range of f-stops to work in, but the hostess tray camera had to just be set for one stop that split the difference between the brightest and darkest parts of the roadway trip. In retrospect, even under the EL, the actors' faces got hit with more sun than I guessed and some of those shots are a bit clippy.

But the other cameras had the reverse problem, we kept the clipping down but with the cameras panning wildly between two cars, I couldn't throw much light on the faces in the car, especially since one camera was on a remote crane sweeping across the hood, etc. The best I could do was put a 800w Joker on the Ultimate Arm crane right next to the remote head. So in many shots, I'll have to dig out the faces in the dark.

The second unit stuff shot on the Genesis had some of the same clippiness problem -- I suspect that they were just shooting in Panalog and looking at Panalog in the viewfinders and monitors, no LUT box, which unfortunately tends to make one overexpose because white in Panalog is supposed to be at 70 IRE, not 100 IRE. So there is a tendency when operating with a Log image on the viewfinder to make white look closer to 100 IRE, causing you to lose some of that overexposure detail that Log is supposed to give you.

This car chase stuff was not the sort of thing where you could take the time to look at waveforms, zebras, etc. We were driving at 50 mph shooting fast action and whipping the cameras around to full sun and full shade, so exposing was an extremely fast guess on the fly, if any stop adjustment was made at all.
 
David Mullen ASC said:
This car chase stuff was not the sort of thing where you could take the time to look at waveforms, zebras, etc. We were driving at 50 mph shooting fast action and whipping the cameras around to full sun and full shade, so exposing was an extremely fast guess on the fly, if any stop adjustment was made at all.
Click to expand...

This is one reason why I'm so thrilled that Exposure Check was added to be an option for the GPIO ports, so one port can start/stop the camera and the other port can instantly show you exposure even during a take.
 
Thanks for the insights, David.

Is diffraction an issue at T/16?
 
Thanks for the reports David. Practical real world feedback is ALWAYS worth a good read and doubly so when it comes from someone as respected as yourself.

Graeme
 
Awesome! Thanks for the info, Mr. Mullen.

Just curious, what did you record onto? (REDRAM, REDDrives, or CF cards?) I suppose you used RED RAMs for the car chase scene in order to avoid dropped frames.
 
I am watching a news story about a cop drama set in Chicago. Local cops intervened in a store robbery scene not realizing it was a film set. The actors didn't realize the cops were real either.

This wasn't your shoot was it David?
 
We mostly recorded to Red RAM's except for Steadicam shots, where we used the CF cards. I had two Red Drives for C-camera work but I avoided using them in general.

Practically speaking, even though I'm sure diffraction starts to become an issue, I generally don't see a sharpness loss when stopping down big zooms. They don't really seem snappy to me anyway until you get past f/5.6. And this is all for TV broadcast where subtle sharpness differences will not be noticed on most sets.

I did notice, however, that on a day that I rented the smaller, lighter (relative to the 24-290mm) 17-80mm Optimo for remote crane work (and I used it for the last shot of the show, a night shot of a car driving towards the skyline -- at least it opens to T/2.2) that it looked snappier, sharper at the wide-angle end than the 24-290mm. Seems like a better lens overall.

I'm just not a fan in general of the big zooms, I'd rather use primes. But in this case, it's all part of the style of shooting, primes would have been a nightmare due to all the quick size adjustments we were making, sometimes during a take.

The 24-290mm is so heavy (24 lbs) that I felt at times the back-focus was slipping even with all the heavy support rods and bracketry. We had a night of shooting wide-open when I asked the operator to grab the zoom ring and do some manual snap zooms and the focus went out slightly as he grabbed the ring and tugged. This is why I decided that the remote crane stuff on a camera car should use a lighter zoom, either an 18-100mm Cooke or the 17-80 Ang. Optimo.
 
Here's something I hope happens with the EPIC: I wish the monitor-out signal was 1080P instead of 720P. With these long-lensed shots, the focus-pullers are using monitors to judge focus and I feel that the monitor image on set is not quite as sharp as what I get out of a Genesis, maybe because of the 720P issue. Or maybe it's just the nature of the quick debayering to 720P.
 
That was announced David! (pretty confident in my recollection)

It sounds like it will be the broadcast quality signal too. HD-SDI to a 25.5" Panny would be nice. :) But I'll gladly take that same image compressed with h.264 over Wi-Fi. The on-camera monitors however should take full advantage, and get a tack-sharp 1080p, from full debayer. Since 1080 for distribution from full frame was announced, then I am assuming this is an always-on function handled by the ASIC. For such telephoto work, and since you already have a monster-sized rig going, I suppose you wouldn't find a big 9" on the side to be bad idea. And little 5" for the operator to frame with. So I am assuming you request is already baking.
 
Thanks David. If its not too much would you mind going over your process to prep for a film. (perhaps this should go in your main thread?)
 
David Mullen ASC said:
Here's something I hope happens with the EPIC: I wish the monitor-out signal was 1080P instead of 720P. With these long-lensed shots, the focus-pullers are using monitors to judge focus and I feel that the monitor image on set is not quite as sharp as what I get out of a Genesis, maybe because of the 720P issue. Or maybe it's just the nature of the quick debayering to 720P.
Click to expand...

Amen to that, David.
The best and fastest way to pull focus on these digital cinema cameras (in my opinion) is to have a wireless, position yourself away from the camera and take advantage of triangulation of sight. Make marks, pull the tape (or laser) but also have a monitor that you can trust for verification.
So use the traditional tools of measurement, judgment, and anticipation that any AC would use. But the "new age" tool that really helps is a high-resolution tap (whether it is an HD-IVS on a film camera or preview from a Red or D-21) and a good monitor.
Unfortunately, as you point out, the 720p output of the current Red is a quick and dirty debayer, and not as high resolution as you would hope for to confirm whether it is the actress' eyes or eyelash tips that are sharp.
Perhaps Epic's 1080 output will help in this regard.
Cheers,
Harry
 
Just finished an epic color-correction session to deliver this pilot. There were SO many cuts in this show.

I just wanted people to know that we got to color-correct in a new D.I. theater at Encore and the Red footage looks fantastic on the big screen even at 1080P, which is what we are finishing at. The Encore people and the Fox post people were all amazed at how clean the Red footage was, even though I shot most of it at 1000 ASA, some at 1600 ASA. Hoping it changes some minds at the studios about using the Red for TV, some had some less-than-great experiences dealing with Red for pilots in the past but it's hard to argue now considering how good the image is, and how easily it color-corrects now using RedGamma.

Even my past experiences color-correcting Red two years ago in a D.I. situation were a bit problematic, gamma and color-wise, but now it goes through the system like a dream, i.e. quickly, more like how Panalog is to color-correct. In fact, on the big screen, I felt that our Genesis footage, though it intercut fine for the chase scene (especially since the cuts were so quick) was a bit more "electronic" or video-ish in texture. Hard to describe, but Red has a smooth-but-detailed film-like look while the Genesis had a touch of that overly-crispened look even though edge enhancement was not applied. But I always have felt that Genesis was easier to color-correct and had better color reproduction and dynamic range. Now I'd say that they are the same to color-correct, the dynamic range is the same, maybe the color from the Genesis is a tiny bit more accurate compared to reality, but that the Red looks more detailed yet smoother from the downsampling.

And whereas in the past, the Genesis was less noisy, now the Red M-X image is the one that is less noisy in low-light.

For now, it's still the simplicity of HD tape workflow that makes the Genesis/F35 more amenable to TV post schedules. Red is a bit more like a film-shot TV show's workflow, an extra step in there compared to tape, for transcoding. But as places like Encore develop a data-centric post system, it makes less of a difference.
 
You must log in or register to reply here.
Back
Top