How Does the Osmo RAW Compare to Professional Cinema Cameras?

osmo-raw-lab-test

The DJI Osmo RAW has finally arrived at the cinema5D office, as you may have seen in Nino’s Osmo review. But the focus of our attention is the DJI Zenmuse X5R Camera, which can be attached to the DJI Osmo with the Osmo X5 Adapter and produce powerful cinematic RAW footage in 4K with Osmo stabilization in an ultra-compact form factor.

[UPDATE]: We have now also compared the X5R to the X5 and X3: LINK

We were very curious to find out what the camera quality was really like. Here are our lab test results when compared to professional cinema cameras.

dji-osmo-raw-vs-ursa-mini-2

DJI Osmo RAW Compared to Professional Cinema Cameras

The DJI Osmo RAW version is quite a bold little camera. Very small, light and with a promise of 4K RAW that can also take to the skies when attached to the DJI Inspire 1. The integrated stabilizer makes this an extremely convenient tool, and the fact that there’s a potential for high quality footage with its RAW recording makes it compete with much more professional and expensive cameras. We’ve tested the ergonomics of the device, now let’s see how the little Osmo RAW compares in terms of image quality.

Dynamic Range

Dynamic range is difficult to measure properly, especially on a RAW camera where processing is done by the user and not by the camera. We’ve gone through DaVinci resolve to create a flat image of our recording and measured it with our software.

A good dynamic range rating allows us to capture more shadows and highlights in high-contrast scenes. The X3 camera on the original DJI Osmo suffers from a very poor dynamic range, which is especially problematic when capturing landscapes with the DJI Inspire 1 drone.

We’re testing usable dynamic range with a DSC labs XYLA-21 transmissive test chart. Unfortunately, our test lens, the Zeiss 50mm Cp2 macro, is not compatible with the Osmo, so we used the 15mm F/1.7 MFT lens that came with the Osmo X5R.

Our software measured about 12 stops of usable dynamic range on the DJI Osmo RAW (Zenmuse X5R camera). This is similar to the rating of the Sony a7S II, 3 stops more than the Blackmagic URSA Mini 4K and 2 stops less than the Arri ALEXA.

dynamic-range-osmo

DJI advertises the dynamic range of the Zenmuse X5R camera at 12.8 stops. It is rare that manufacturers are upfront when it comes to usable dynamic range.

Note that a RAW camera doesn’t necessarily produce more dynamic range. Doing numerous comparisons with test charts, we have learned this over time. It is likely that the original Zenmuse X5 gives you more or less the same dynamic range. RAW on the other hand gives you finer gradations, more possibilities in post and an image that is more solid and can be graded further without destroying the image quality. When we pull up that information in the blacks there is noise, but a lot more steps are visible behind it, just like on the Blackmagic URSA Mini 4K that displays a lot of noise early on. 12 stops of usable dynamic range is a good rating for a cinema camera. It is also what the C300 mark II achieved. As always, the Arri ALEXA stays untouched with 14 stops of usable range.

The Blackmagic Micro Studio Camera 4K is also small and shoots RAW, but is not an alternative as this one was limited to about 5-6 stops of dynamic range in our tests.

Image Quality

We determine image quality by looking at sharpness (resolution) and aliasing with the help of stars and tubes. Usually they reveal where an image breaks down in terms of resolution.

The shots below were taken at the best ISO speed of each respective camera.

DJI Osmo RAW Image Quality - 100% crop of 4K image

DJI Osmo RAW Image Quality – 100% crop of 4K image.

As you can see, despite any lens cushioning and lens softness, the DJI Osmo RAW is quite close to the Canon C300 mark II and more or less on par with the Varicam 35 / LT. The Sony a6300 is the only camera that leaves the others behind in terms of resolution. Who would have thought.

Looking at the a6300, it has become clear that size is not the deciding factor when it comes to quality resolution. The Osmo DJI RAW proves that yet again, with an impressive image quality that comes close to the bigger cinema cameras. Here are a few more comparison shots of other objects on the test chart:

IQ_dji-osmo-amira

IQ_dji-osmo_ursa

IQ_dji-osmo_fs7

IQ_dji-osmo_varicam-lt

IQ_dji-osmo_c300-ii

In summary, I would say that in terms of image quality the DJI Osmo RAW can hold up remarkably well in comparison to professional cinema cameras currently on the market.

Lowlight

IQ_dji-osmo_lowlight

The DJI Osmo RAW X5R camera shines at ISO 100, but it quickly loses its power when you crank up the ISO. For the shots above I used minimal post processing. You can see that at ISO 800 there is already a considerable amount of noise in comparison to ISO 100. That said, look at the shot of the cat earlier on where the blacks were pulled a little and some noise reduction was applied. This has to be kept in mind when shooting RAW. You can still treat the image much better than from cameras that use heavy compression.

It’s clear that the Zenmuse X5R is not a lowlight wonder and the DJI Osmo RAW should be used at low ISO speeds. I would recommend not to go beyond 800 for high quality shots. Ideally you should stay low, because the Osmo RAW records at ISO 100 at all times. Any other ISO speeds you set in the app are just a “preview” and have to be processed (pushed) in post production.

Rolling Shutter

The famous rolling shutter effect that haunts CMOS sensors can be especially troubling on shots where a lot of movement is involved. The Zenmuse X5R that rides the DJI Osmo RAW and flies on the DJI Inspire 1 is certainly meant to be involved in a lot of moving shots.

Unfortunately, the Zenmuse X5R doesn’t shine here. With 25ms of readout time it is at the worse side of the spectrum, and most proper cinema cameras have much lower rolling shutter ratings.

rolling-shutter_osmo-x5r

DJI Osmo RAW Conclusion

DSC09182

Look at that tiny 4K camera and stabilizer. The DJI Osmo RAW certainly has a niche of its own. A small, stabilized camera that delivers 4K RAW images at 24p. There’s nothing like it right now that delivers RAW, is so easy to use and can be mounted onto an affordable and powerful drone. The question is whether or not the RAW can keep up with cinema standards and truly deliver high quality footage.

It is quite difficult to draw a definitive conclusion from what I’ve seen. On the one hand, the footage we get is really powerful, has beautiful deep gradations, a high resolution, organic look and can really be played with in post production. If you’re used to Inspire 1 drone footage, this upgrade will simply blow you away and the difference to a professional cinema camera like a VariCam, Alexa or C300 mark II will be hard to spot when properly post processed. Of course the Zenmuse X5R wouldn’t simply replace a cinema camera, as such a camera is about more than the final end result of a picture you can achieve. An Arri ALEXA is certainly still in a different class altogether, but I think the Osmo RAW will become relevant in the cinema sector for certain applications.

osmo-raw-lab-test-back

There are also a few shortcomings you should consider. In terms of a “true cinema camera”, rolling shutter is absolutely terrible. The Canon 1D C from Canon’s cinema line rightfully has a place in the same ranks here. But especially because the DJI Osmo RAW is used in situations with a lot of motion, the strong rolling shutter can be a problem that might put many professionals off. Another downside is the mediocre lowlight performance. If you’re coming from DSLR’s this will be limiting, but if you’re used to Blackmagic or film cameras you probably won’t notice it.

osmo-raw-lab-test-lens

All in all, the DJI Osmo RAW is a remarkable piece of gear. In the right hands, when the RAW is exploited with some time in post production, this can be an extremely powerful tool. The fact that this is not just a camera, but a solution as a miniature gimbal camera for on the go or in the air is intriguing. Is the price of $4000 a lot? Yes. Is it worth it? You decide. Let us know your thoughts in the comments.

 
BUY IT
€3,000.27 exc. VATNot in stock - Order now and we'll deliver when available - Normally ships in 4-6 weeks -...
 
BUY IT
$99 exc. VATIn Stock
 
BUY IT
€77.92 exc. VATIn stock for immediate dispatch
 
BUY IT
 
BUY IT
€470.39 exc. VATIn stock for immediate dispatch
 
BUY IT

Leave a Comment

Login to comment

You are not subscribed to this post. Follow new comments

Mironi Ghebregergis Reply
Mironi Ghebregergis June 8, 2016

Too expensive even it has a big features. …I think when it comes to filmmakers they would love it but will not be able to use it due to the price

Josh Miller Reply
Josh Miller June 8, 2016

you sir have absolutely no idea… look at a steadycamsuite and a red camera how expensive they are and u realize that this is absolutely a best price deal..

Mironi Ghebregergis Reply
Mironi Ghebregergis June 8, 2016

To pay for osmo above 3k $ is maybe OK for u…. but some filmmakers won’t be able to pay this….everybody ain’t have money ……gear lover

Reply
Tristan Barrocks June 8, 2016

I have to agree with Josh. Like if your looking at this saying this is too much money than you may not be at the level where your productions can have this as an option. No disrespect, but you’re getting RAW lossless 4K on a 3 axis gimbal for $4000, pretty good to me.

 Mathias Häcki Reply
Mathias Häcki June 8, 2016

Its not lossless raw. It records cinema dng.
Still a great option/value for its askng price.

 Mathias Häcki Reply
Mathias Häcki June 8, 2016

I’ve shot a few aerial projects with the x5r yet.
But I’m still not a real fan of it for many jobs and often use the X5 instead.

Here are my 2cents:

– The file handling is still very annoing and unsafe. The ssd’s must be read out with either DJI’s Cinelight on Mac or a simple DJI file copy tool on PC. These programms convert the files to cinema dng. None of the two options uses a checksum process. So you’ll never know if the files are being copied faultless.
This results in a bigger need of ssd’s as you’ll have to keep all files until you checked them manual frame by frame.
– SSD’s are still overpriced. There are standard Samsung EVO 640 cards build in a custom case and some custom DJI firmware. You can get these ssd cards for about $250. Sadly the firmware blocks non unlocked ssd’s. At least there’s now a 3 for 2 deal from DJI. But its still hard to get hands on spare cards or spare readers out in the market as DJI seems not to have big enough stock.
– You can’t delete single files. The only option is formating the whole card
– The histogram is still adjusted to the X5 h264 file. So it gives you a false reading regarding the x5r raw footage. This can lead to underexposing the shot.
– The created dng files are not premiere compatible, but they work in AfterEffects. Adobe might solve this one soon. Converting with a tool called slim raw does the job to get you files that run in premiere and even shrinks the file a bit.
– The X5r is almost to heavy for DJI Inspire 1 drone. The drone has a hard time to climb for altitude fast. This can be a real problem for certain shots. Flightime goes down about 2 min compared to the X5. Its still around 10 min, which is more than ok with me.
– Photo wise its still the same as the X5 delivers.
– The supplied 15mm lense is ok for starters. Just make sure you stay between f5.6 and f11 to get decent sharpness. The 12 and 25mm Olympus lenses are a much better and way sharper choice.
– Using ND’s is a must. But as you can’t adjust the gimbal you’ll need a few tricks. I use coins and tape to manage a decent balance, depending on the setup. Otherwise you risk burning the motors due of overload.
– If you change lenses, you’ll need to calibrate lenses everytime. Why can’t they integrade lense profiles in the firmware?
– You’ll loose a lot of settings every time you turn the camera off. Sadly the same with the X5

The X5r delivers nice footage. But its still 200-250MB every second. RED manages to work with almost the half of that rate. Hopefully DJI adds some compression option as a update, if its posdible with the current hardware. But my guess is we’ll see it in the next generation camera.
Also many jobs do not really require raw as a need.
Netflix shoots some series with the older X5 and 60 Mbps h264. DJI shoots a lot of their own projects with the X5, even they could just grab some X5r of the shelf when ever they want.
Why? Because often the 60Mbps h264 delivers what they need, if the cameraman configures for a correct image and safes a lot of addtional work in the post.
There are still a lot of things for DJI to sort out to make this camera as good as it could be.
Until then I’ll only use it as a tool for projects that absolutly requires raw becsuse of stuff like heavy planed grading.

How about a comparisson with the URSA Mini 4.6k instead of the 4k? We all know the 4k sucks ;)

Sebastian Wöber Reply
Sebastian Wöber June 8, 2016

Hi Matthias,

Thank you for chiming in with your experience. Just some remarks as I see some of your points differently:

I would not say file handling checksum is a minor issue.
SSD’s are expensive but one SSD will suffice. You can never have enough batteries to burn 30 minutes worth of RAW footage.
The cinema dng files do work in Premiere.
There’ll be more updates to the firmware and app. I think those are not big issues considering the quality you can achieve. Unfortunately I have not had the chance to test the X5. Is the quality of the shots as bad as the mp4 proxies created on the X5R micro sd card?

Thanks

 Mathias Häcki Reply
Mathias Häcki June 8, 2016

The proxies of the x5r are the same h264 you’ll get on the X5.
The x5r dng files do not work on anyones premiere pro i know off. So currently everyone is either placing them in AfterEffect and import the sequence to premiere, or they convert with programs like slim raw or resolve. Be aware that the PC file exporter does not give you any export option. Its really bad.
One ssd card is a way to high risk, plus offloading can go up to 2h,depending on the system used. So 3 to 6 cards is the least to go for a production shot.

Sebastian Wöber Reply
Sebastian Wöber June 8, 2016

Hi Mathias,
The Mac app lets you either export to normal dng or cinema dng for Premiere. I dont know about the PC app.
If you want two cards you can buy two 256GB cards. For most one 512GB card will be enough and for high value productions a second one is a good recommendation. But if you’re going that route then you can afford a second card. I’m not saying it’s cheap. But for run and gun, affordable shooting I think you don’t need a second card.

 Mathias Häcki Reply
Mathias Häcki June 8, 2016

There are only 512gb cards available, if you’re lucky to find s place that has em in stock, abailable yet. 256GB are not in DJI’s lineup yet.
I would not do any job with a single card or reader.
Anything double is the minimum I go for. But that is just my personal way to work. I also want multiple cards to offload footage while on the job.
I don’t want to risk loosing footage due of a problem with a drone.
Filming on the ground is not as risky. But even there, stuff or cards can go bad.
But thats the same with tool.

Sebastian Wöber Reply
Sebastian Wöber June 9, 2016

Ok. But really, just because you want a second card or tool to be safe is not an argument against the tool itself. It shouldn’t make any problems, like any other expensive SSD media that is made to simply work. I guess the drone is more likely to die on its own than that SSD card and unless you’re on a high budget shoot you will not see someone arrive with two drones.

Reply
AlexAdrian Adrian June 9, 2016

I’m like you, also waiting on the URSA Mini 4.6k dynamic range comparison. But I’ve yet to find a scientific test on its dynamic range on it. Hopefully Cinema5d can do one.

Reply
the SUBVERSIVE June 9, 2016

a lot of the sensor’s specs seems to match what Panasonic uses, even more the specific 12.8 stops – and also the noise. one could assume that aside from some tweaks, this is what you would get out of a GH4 if it had RAW.

besides having no crop or that 1.86x crop of a multi-aspect sensor in 4K, it would be nice if Panasonic could get a cleaner image out of a possible GH5.

rumors tell otherwise, but I wished Panasonic could turn the GH/GX similarly to what Sony does with the A7S/A7R, going with less MP for the GH and more MP for the GX instead of trying to find a difficult middle ground in MP count for video and photography – until 8K comes… hahahaha…

Mase Daniel Reply
Mase Daniel June 9, 2016

This is fail. ;)

Reply
Jonathan Pais June 9, 2016

why is this ‘fail’? If the price were lower, I’d so be on board. I’d use the Osmo for motion, my GH4 for locked-down shots. I don’t see the need for using longer lenses or speed boosters with a gimbal, a lightweight wide angle lens would be just fine. It would be great if it had built-in ND filters, but probably not possible in such a compact device. Meanwhile, looking forward to seeing IBIS on the GH5, but it’s no replacement for a gimbal.

Reply
Paolo Rudelli June 23, 2016

I dont won’t to be harsh but this is very strange test were you campare: image, camera (lens???), mode, iso and more in nearly random way.
Ok we can get a generic idea of the quality of the camera but is not even enough to make a decision to rent to try out more. Regards

Sebastian Wöber Reply
Sebastian Wöber June 23, 2016

Hi Paolo,
The tests are not random at all. I might have to point that out more clearly in the text. Please let me know how you came to that conclusion. Thanks

Reply
Paolo Rudelli June 23, 2016

Hi Sebastian,
I understand but you test for dynamic range with a alexa, a7s e ursa mini and for image quality with other camera: C300 and Pana varicam with non referenced chart (something like ISO 12233: Ver. QA-76 chart or similar) and you put only low res jpg image of the test with allot of compression or maybe i miss the link to download the tiff or dng file?

Maybe is because i learn, many year a go, to do camera test with 35mm film camera to check and conform all equipment for insurance and maybe i have to high standard but for me this test doesn’t tell me anything really use full to be convinced to rent this camera for a future work.
Also becuse this camera is intersting because is very small with raw recording and can be use in some extreme situation or crazy shot but maybe in a production you need to match the shot with some more high-end gear and can be nice to have a real global idea with what you can match without to many trouble in post.
Regards

Sebastian Wöber Reply
Sebastian Wöber June 23, 2016

Hi Paolo,
Ok I understand. I could certainly post images with a higher compression quality like dng or tif, but they would be less accessible for people reading this blog. The simple fact is that the differences are obvious, so it is enough to post them with a high quality jog codec, as I did. Next, the images are 100% crops, so the resolution you see is the original, full resolution. And regarding your question for an ISO chart. Our chart is custom made and standardized among all our tests. It shows off strengths and versatility in a good way. These tests are made to give our users an observable comparison, not to be studied by scientists. I believe it is the best way to compare the cameras and an ISO chart (we do have a 1.5 meter large ISO chart by the way, for 6K resolution cameras.) will not be beneficial here.
Keep in mind that in a real life environment you are not filming test charts, but objects and structure. Our controlled 3D chart helps you to see some of the important aspects of a camera and the differences are quite obvious and clear. This is the best we came up with to help users make an informed decision.
I hope this helps.
Best

Reply
Paolo Rudelli June 23, 2016

Thanks Sebastian for your explanation.
Yes of course, in real live, you film objects and structure: in the example you give in the image with wire spool the red one by Osmo look terrible but with jpg image i can not really appreciate if is compression problem or the footage is really so crappy (especially if i think of red part in skin tone) and you compare with a fs7 Cine EI native ISO with?? i have to guess is a Cine S-log3 profile, XAVC-I codec ???

I did not say your review is crappy, you guys in cinema5D you do great work but because in this test you compare with high end camera and lens(10k$++) i was expecting a test with a methodology for high end user target.

Sebastian Wöber Reply
Sebastian Wöber June 23, 2016

I understand, yes I agree the spool was a bit oversaturated. That is not due to jpg compression, but the way I processed the image. I can tell you that skin tones are petty impressive on the X5R, much better than on the FS7 as the higher bitrate RAW gives us a lot of clean data.

Reply
Paolo Rudelli June 23, 2016

“I can tell you that skin tones are petty impressive on the X5R” this is really nice to know.

If i understand right what you write the native iso is 100 and is no on camera gain or offset?

What i do not understand is if you apply a LUT (Rec 709?) to the raw or is the native image we see.
A Histogram and/or a waveform with the related image can be also very “speaking” for a DoP ;)

Reply
Kevin Almodovar June 23, 2016

Unfortunately, the one thing that can never be tested is reliability. I have had a DJI Ronin for a couple years, a BMPCC and GoPro’s for even longer. In my opinion, none of these manufacture’s deserve the Pro title.

I have owned Canon, Panasonic and Sony. There is a certain amount of confidence when you buy them and Movi products too. While there is never 100% reliability, these manufacturers have been around a while and you have a certain level of confidence when you are using them on set.

I would never solely rely on a DJI product on a paid shoot without a backup option, been burned. In my opinion DJI and the other companies named, are too concerned with features and getting the product out and never really beating them up to ensure they don’t go down on set.

With all of this said, the price point is way too high for the name going on the side.