Stop Worrying About Compression With an On-Camera Video Recorder


Full Raster High Definition Video has six times the number of pixels as SD video. Theoretically it should take up six times as much space as SD. However, AVCHD has a slightly lower bitrate (24 Mb/s) than the DV codec used on most of the SD tape cameras of yore (25 Mb/s). How is this possible? Compression, and lots of it.

In most situations, compression is your friend, but in some situations it can lead to frustration. Unfortunately, (with the exception of the Panasonic HPX Line) all cameras below $15,000 record heavily compressed footage. Because of this limitation a number of external video recorders have arisen to free videographers from the shackles of compression. But the feature set, price and workflow of these recorders vary wildly, so choosing one can be tough.

This article will explain what advantages there are to using an external recorder and how to choose the best one for a particular camera. It will also briefly go into the capabilities of the various on-camera recorders B&H sells. It is important to note that we are only going to look at recorders that are designed to record in a higher-quality format than what is recorded in camera. There is another subset of on-camera recorders that record the same video codec as is recorded in camera to a disk or card in order to bypass recording to tape or to record for much longer.

Before diving into the capabilities of each recorder, it is important to understand what methods are used to compress video, and what detrimental effects they can have on the image.

Chroma Sub-Sampling

One of the oldest ways to cut down on the amount of information in a frame is through chroma sub-sampling, which has been around since the analog days. It reduces the bandwidth of video by skipping the color information on some pixels, depending on the level of sub-sampling. A luminance value (brightness/darkness ratio) is still recorded for every pixel.  A 4:2:0 image, which is the chroma sampling method used in AVCHD and XD CAM EX, only changes color every fourth pixel. Where 4:2:2 sub-sampling changes color every other pixel.  4:4:4 color sampling retains all color information. While only having color information for every fourth pixel may sound extreme, it is actually not as noticeable as you may think, as is demonstrated in these images:

Chroma Sub-Sampling
500% blowup of this image in two colorspaces:
4:2:2 Color (Canon XF Codec) 4:2:0 Color (AVCHD)

The eye is much more susceptible to differences in luminance, so the loss of color information is very hard to spot, but it is there. This is why almost everything is distributed in 4:2:0. However, if you plan on doing heavy post work such as green screening and color grading, then it is much better to have footage that is 4:2:2 or higher, because the more you push footage the more sub-sampled color shows.

If you have a camera that shoots AVCHD, XDCAM EX or HDV then your camera records internally at 4:2:0. However, if your camera outputs uncompressed video from its HDMI or SDI connector, then it will output in 4:2:2. Also, all of the recorders in this article record in at least 4:2:2, so if 4:2:2 is what you are after then don’t let that narrow down your search. If you are looking for 4:4:4 then either the Convergent Design Gemini Recorder or the Sony SR-R1 is what you are after, but make sure your camera can output it.

Color Bit Depth

Color Bit Depth is another element that influences bandwidth. Most cameras record internally with 8-bit color. This means that each variation in color is defined as an 8-bit value, or 8 numbers in binary. This leads to 256 possible variations in color hue. 10-bit color, as you might expect, uses a 10-bit value which leads to 1,024 possible variations in color hue with just a 20% increase in bandwidth. Similar to 4:2:0 color sampling, 8-bit is common for distribution and doesn’t look much different from 10-bit color, to the eye. The large amount of extra color information in 10-bit color can make a huge difference in color correction. However, a camera’s image sensor has to be able to resolve the difference in color hue or the extra color bit depth will just show an increased amount of noise. It is for this reason that most cameras do not record at or output 10-bit color, with a few exceptions. All of Sony’s XDCAM EX line output 10-Bit Color through their SDI ports. The 2/3”-chip models and the F3 will benefit the most since the large sensors will produce less noise. It is also worth noting that all of Panasonics P2 Card cameras record 10-bit color internally. As far as recorders go, every recorder except for the Convergent Design NanoFlash record in 10-Bit Color. But currently there aren’t many cameras that output 10-bit color. Here’s looking to the future, though.


A 1920 x 1080 video signal with 4:2:2 sub-sampling and 10-bit color depth will take up a huge amount of space, about 10GB per minute. And there are a few recorders available, such as the Blackmagic Design Hyperdeck Shuttle 2 that will record fully uncompressed video. But if you are looking for something that takes up less space than 600GB an hour, then you probably want some form of compression applied to your image. The question is, what type and how aggressive do you want it to be?

Intraframe Compression

Intraframe Compression was the first type of digital compression to appear. It compresses each video frame independently. DV, DVCPRO, AVC Inta, ProRes, DNxHD and MJPEG are all intraframe compression codecs. Modern intraframe codecs can achieve virtually lossless results at relatively low bitrates (compared to uncompressed video). For example, ProRes 4:2:2  is a 10-bit 4:2:2 codec that records at 220Mb/s which equates to about 1.65GB per minute. This is considerable in size, but still way less than the 10GB needed to record a minute of uncompressed 10-bit 4:2:2 video. The problem arises when you’re recording at a much lower bitrate. Intraframe doesn’t hold up well, which is why another compression type exists.

Interframe Compression

What do AVCHD, XDCAM, H.264, Canon’s MXF codec and HDV all have in common? Interframe, or Long-GOP compression. Interframe takes all the benefits of intraframe compression and takes it even further by combining several frames together into a group and recording only the differences between the frames. This leads to a huge gain in quality at lower bitrates, compared to intraframe codecs. Interframe codecs have a few disadvantages, though. They take a lot more computer muscle to edit, which is why many editors choose to transcode them into a high quality intraframe codec such as ProRes to edit with, which takes time. The second reason is that interframe codecs can degrade in quality when there is a high amount of variability between different frames. In some instances, such as shaky handheld footage or scenes with large crowds, interframe codecs will “break” and become pixelated. The effect can easily be seen on sites such as YouTube or Vimeo. Luckily, the codecs used in professional video cameras are much less susceptible to breaking than what you usually see online, and when they do it isn’t as noticeable. For an example, look at these frame grabs from a shot with very fast motion:

Interframe Compression
Frame Grab Comparison
Interframe (AVC-Intra) Interframe (AVCHD)

As you can see, the AVCHD footage begins to block up into macro-blocks because the speed at which this scene is changing between frames exceeds the capability of AVCHD’s bitrate; the intraframe footage is frame independent so it doesn’t matter how much motion there is. Using an external recorder will allow you to always record in an intraframe codec so you don’t have to worry about your codec breaking. Also, you can skip transcoding your footage to a more editing-friendly codec to save time.

What Recorder Should I Get?

Before picking out an external recorder it is important to consider how your camera outputs uncompressed video, if it does at all. If your camera outputs 10-bit color then you probably want to steer away from any 8-bit recorders, but if it only outputs 8-bit color then that isn’t an issue. If your camera only outputs 4:2:2, then getting a 4:4:4 recorder will just make your recording files larger. If you plan on shooting progressive footage, it is important to check how your camera outputs progressive footage, especially if you plan on recording through HDMI.

Many cameras output progressive footage in a 60i video signal. Over SDI there is a standard for alerting the recorder or monitor that the incoming footage is really progressive and that pulldown removal should be done to the footage. Therefore, if you are recording via SDI the chances of a pulldown mishap are vastly lower (though in some instances you may still run into an issue). Over HDMI there is no standard. Sony has started implementing its own way of flagging progressive footage in its NXCAM cameras, such as the FS100, and there are now a few recorders that work with them. But in most cases progressive footage will be recorded to an external recorder as 60i over HDMI, and so you must go through an extra step in post to de-interlace the footage.  

Also consider what editing platform you are using. The two most common codecs you will see on recorders are ProRes and DNxHD. Both are 4:2:2, 10-bit, Intraframe codecs that can record up to 220 Mb/s. They are very similar codecs, except ProRes is owned by Apple and DNxHD is owned by Avid. If you are a Final Cut editor, then ProRes will lead to an easier workflow and if you are an Avid Editor, DNxHD will. Keep these things in mind as you choose which recorder will work best for you.

AJA Ki-PRO Mini Compact Field Recorder

The AJA Ki-PRO Mini is a small external recorder that records to CF cards. It is important  to use approved CF cards or your footage may have dropped frames. It records variants of both ProRes and DNxHD via either HDMI or HD-SDI inputs. You can add additional audio tracks via its XLR or RCA audio inputs. The Ki-PRO Mini is powered via 4-pin XLR, and a number of powering options are available. AJA has updated the firmware of the Ki-PRO Mini to work with various new cameras, such as the Sony FS100U, so it’s a good purchase if you plan on keeping the recorder longer than your camera.

AJA Ki-PRO Portable ProRes File Recorder

The AJA Ki-PRO Portable stretches the definition of camera mountable, but if you are willing to deal with the extra heft, the original Ki-PRO can double as a VTR for studio use as well. It has just about every form of video input you can think of, including analog. The Ki-PRO records ProsRes variants to disk drives, SSDs or SxS cards. It is powered via a 4-pin XLR input.

Atomos ATOMNJA001 Ninja Video Hard Disk Recorder 

The Atomos Ninja is an HDMI-only recorder that records ProRes to user supplied 2.5-inch hard drives or SSD drives. For handheld use, SSD drives are recommended to eliminate the chance of dropping frames due to vibration. The Ninja is powered by Sony L-style batteries and comes with two. It is controlled via touch screen on its built-in confidence monitor. The Ninja makes a good companion to the Nikon D800 with its most recent firmware update. One thing to remember when using the Ninja is that if you plan on recording progressive footage you will need to manually do pulldown removal in post because the Ninja is unable to recognize progressive footage that is delivered in a 60i stream. If you do not do the correct type of pulldown removal, the image resolution will be severely affected.

Atomis Ninja Atomis Samurai

Atomos Samurai HD-SDI Hard Disk Recorder

The Atomos Samurai is a newer addition to the Atomos lineup. Similar to the Ninja, it records ProRes to user supplied 2.5-inch drives. A paid upgrade to allow recording to DNxHD has also been announced. The Samurai records via HD-SDI only and has a brighter and higher resolution screen than the Ninja. It is also able to do pull-down removal as long the camera flags it correctly.

Blackmagic Design Hyperdeck Shuttle 2   

The Hyperdeck Shuttle 2 is the least expensive recorder and is able to record fully uncompressed video as well as DNxHD. The Hyperdeck records to conventional SSD drives but be sure to get one of the approved drives if you plan on recording uncompressed footage. The Hyperdeck accepts both HD-SDI and HDMI inputs. It is powered via a built-in battery, but other powering options are available. It is also important to note that if you plan on recording progressive footage via an HDMI input, you will have to manually do pull-down removal in post.

Convergent Design Gemini 4:4:4 Uncompressed Video Recorder

The Convergent Design Gemini also records uncompressed video, but can do so at 10-bit 4:4:4. It is a great match for the Sony F3 with S-Log. The Gemini is great for recording Log footage from any camera because it is able to output footage to a monitor with a LUT applied for easier monitoring while it records the flat Log footage internally. The Gemini records to high speed 1.8-inch SSD drives and with a paid upgrade, is able to record from two cameras simultaneously, which is great for 3D work.

Convergent Design Gemini Convergent Design nanoFlash

Convergent Design nanoFlash

The Convergent Design nanoFlash is a unique recorder. It offers the flexibility of recording either interframe or intraframe compression onto CF cards. An interframe codec at 100 Mb/s is much less susceptible to motion artifacts than the 24 Mb/s of AVCHD. The nanoFlash is also able to record intraframe at up to 160 Mb/s. All of the nanoFlash’s codecs are variants of Sony’s XDCAM. The nanoFlash is also very power efficient and comes with a battery. The one downside of the nanoFlash is that it only records 8-bit color. However, if your camera only outputs 8-bit color, this isn’t an issue.

Sound Devices Pix 220 and Pix 240

Sound Devices has a history of making professional recorders for the audio industry and has just entered the video industry with two on-camera video recorders, the Pix 220 and Pix 240. Both recorders record either ProRes or DNxHD to either 2.5-inch SSD drives or high-speed CF cards. They are powered by two Sony L-style batteries or an external AC or DC source. Two additional audio tracks can be added to the embedded audio from the camera via two XLR input jacks. The Pix 220 is HDMI only, while the Pix 240 is HDMI or HD-SDI. Both recorders are able to do pulldown removal for recording progressive footage.

Sony SR-R1

The Sony SR-R1 is a high-end recorder that records Sony’s HDCAM-SR codec to SR memory cards. HDCAM SR has previously been a tape-only codec because of the massive amount of bandwidth (up to 880 Mb/s) it takes up. But with memory cards now up to 1 Terabyte in size, users of Sony’s high end cinema cameras can now skip recording to tape. HDCAM-SR is 10-bit, 4:4:4, and intraframe. The SR-R1 is also able to record two cameras simultaneously to one card, for easily organized 3D post work.


Can i use a Ninja Inferno with my canon eos mark III and record video in fullhd at 120p ?


I read the information that the new fuji XT2 records 4k 30fps internally onto the SD card and outputs in F-Log through HDMI 8bit 4:2:2 onto external hdmi recorder.

The question is would for example Atomos 4.3" Ninja 2 recorder be able to record this 4k 30fps footage from XT2? Or is this recorder only able to record 1080p footage from XT2, and if I want the 4k to be recorded externally I would need external hdmi recorder capable of recording 4k?

Thank you for the answer.


Hi Maros -

Although the Fujifilm X-T2 offers an HDMI output,  it is not supported by ATOMOS.

Wrong... it's very compatible.

If aquisition is in 8bit, i.e: Canon MXF,  is there any benefit to transcoding in post to 10bit, i.e: ProRes 422, if color correction and grading?

Can I capture 4k hdr in an atomos ninja flame from a sony fs100?

Hi Gabriel -

This camera's maximum recording resolution is 1920 x 1080P. Using an ATOMOS Ninja Flame will not change the recording resolution, but it will record the 1920 x 1080P at up to 120fps.

Please contact us via e-mail if you have additional questions:

Not so sure this is correct "A 1920 x 1080 video signal with 4:2:2 sub-sampling and 10-bit color depth will take up a huge amount of space, about 10GB per minute".

I shoot 1080p 14-bit RAW 4:4:4 (completely uncompressed) and 24fps runs about 3.0GB per minute, 30 fps 3.7GB/m.

You can do the math yourself to confirm.

Hi JD -

Example:  340min  for 1080p Apple Pro Res 4:4:4:4 will take up 1TB of space. Uncompressed is just not a consideration anymore with video. Uncompressed would probably be about 6TB for the same recording.

Please contact us via e-mail if you have additional questions:

Hi, can atomos ninja assassin record 10 bit 422 with sony a7sii? 

Yes, the Atomos Ninja Assassin 4K HDMI Recorder will be able to 10-bit, 4:2:2 from the Sony A7S II. 

Well, as long as I know Sony A7S II doens't record 10-bit. So, it won't be actually 10-bit. Am I wrong?

I have the same exact doubt what Rafael have. Please through some light on it. Thanks in advance.

Do you know if Canon 5D mark iii can output 422 10bit via hdmi on to Ninja Star?

The Atomos Ninja Star will record 4:2:2 8-bit in a 10-bit wrapper from the 5D III, as the camera ouputs 4:2:2 8-bit via HDMI.

Thanks Christina,

... which is still ok, I guess for blue/green screen work.  Is there a camera priced less than $3k that can output 10-bit via HDMI? Just curious.



Hi Alkesh -

Sony's PXW-X70 Professional XDCAM Compact Camcorder records HD video using the XAVC Long-GOP codec with 4:2:2, 10-bit color sampling at 50 Mb/second, providing footage with a high data rate that is suitable for broadcast applications.

Please contact us via e-mail if you have additional questions:

can you please list which sub $2000 camera has hdmi out in 1080p 4:2:2 10 bits ?

Hi Amanieux -

We do not offer a camera that meets your stated criteria.  The camcorders offer a 4:2:2 8 bit stream via HDMI.

Please contact us via e-mail if you have additional questions:

Great article, thanks. I own a Sony NXCAM5U that captures AVCHD 4:2:0. internally, to dual cards and its portable harddrive.  However the camera has a SDI jack and a HDMI jack as well.  Other NXcam users and the Sony manual say the camera can output 4:2:2 via its SDI jack.  I am thinking of investing in the Ninja 2 or the new Samuri Blade in order to obtain the uncompressed footage. But there is some argument on the web ( that the difference between 4:2:0 and 4:2:2 is negligible, that is, it's not worth the expense of buying an external recorder especially if the delivery platform for the final footage will be Vimeo or Youtube, because of the h.264 crunch.  What's your opinion? Would capturing via the SDI or HDMI jacks thus avoiding the internal avchd crunch on the data be of a significant enough benefit to warrant the purchase of a external recorder? kind regards ...

There will be two advantages here even for videos heading to Youtube.

1: It is always better to start with the best (least compressed) footage available so you are using that to compress for web from. It will not always be very noticeable but it will technically be better starting with a lower compression.

2: The more important item in this case will be the color space. Recording in a 4:2:2 color space will not give you a visible change but it will give you more information to work with in post. This will be important when color correcting and or grading the video before compression for the web.

I would like to record 1080/50p from my sony A7.
As far as I know Atomos ninja2 only records 1080/50i
Same goes for shuttle2
Is there an affordable field recorder for 1080/50p (or 60p for the US)

What do you think of using the Atomos Ninja with the Nikon D7100 (cropped sensor) as opposed to the full frame D800?

TR wrote:
What do you think of using the Atomos Ninja with the Nikon D7100 (cropped sensor) as opposed to the full frame D800?

on dvxuser a test of 7100+ninja showed the 7100 output was only 720p 8 bits