Noise Levels of Various Codecs

[Dan Keaton]

Dear Friends,

This is a subject that we have not publically discussed before.

While we have run numerous tests in our lab, and have had confirmation from well-respected expects, we have previously avoided this subject.

However, this subject has come up on some very well-respected blogs, and we thought it was time to comment.

Our nanoFlash uses the Sony XDCam 422 and XDCam Ex codec, as built into a codec module that we purchase directly from Sony.

It is no accident that we choose this codec module, as opposed to choosing ProRes.

Our internal testing has shown:

Sony XDCam 422 Codec is lower in noise than the various flavors/bit-rates of ProRes.

The true Apple ProRes Codec is higher in noise than the above Codec.
By "true" I mean the actual ProRes Codec as provided by Apple in Final Cut Pro.

The third-party implementations of the Prores Codec are higher in noise still.

And one should be aware that not all third-party ProRes Codecs have the same inherent levels of noise.

Background:

For codecs, like ProRes, the decoding of ProRes is rigidly defined.

However, the encode process is not specified. This is up to the company that is writing the encoder.

The quality of the encoder code does vary, with some being lower in noise than others.


We choose the Sony XDCam 422 codec, as we could purchase a very sophisticated, very mature, implementation. We did not have to write an encoder, as both the encoder and decoder came ready made, and thoroughly tested, with very rigid quality control.

One of the side benefits of using our Sony XDCam 4:2:2 codec, is that very efficient, high quality modes are available. And, at the same time, very high quality modes are available, such as our 280 Mbps mode.

Being able to record, at 50 Mbps Long-GOP or 100 Mbps-Long-GOP allows one to have very low-noise footage, and still have file sizes that are smaller in size than other recorders.

At least one very well-respected blog, Alister Chapman's XDCAM-USER.COM blog, is now discussing the inherent noise that is added when using the non-Apple implementations of ProRes.

He has a limited number of files that one can download for close examination.

It is our intention to run, or have an independent party test the various recorders that are on the market today.

We will not be able to run these tests this month, but we will run them as soon as possible.

Of course, we like, whenever possible, for our customers and potential customers to run their own tests and draw their own conclusions.

I welcome your thoughts and suggestions.

[Ed David]

and unfortunately as I found in a rival external recorder found - recording to DnxHD had macro blocking issues.