Should I buy an FS700 right now?

[Ryan Douthit]

Thanks Chris, some good info there. Nice samples, too. I am very familiar with excellent work out of the FS700. I own a network that runs a number of shows filmed with it, I've also edited a bunch of footage shot with that camera. I just don't own one personally and that's my conundrum. After spending several years with AVCHD on my NX5U and then the FS100, getting XAVCS on the A7S has really appreciate the greater grading flexibility of a more modern internal codec. (As nice as ProRes is, I wouldn't use it for everything or my investment in NAS backups will be through the roof.)

CRC 2014 Series, shot on FS700:
CRC 2014: Rocky Mountain Rally - Driving Sports TV

[Steven Reid]

Originally Posted by Christopher Young

In the meantime a little performance tip. If you want full 10-bit out of the FS700 we've found taking the Raw Component video output to a Decklink Analogue to SDI Mini Converter then to a 10-bit external recorder, Samurai Blade, gives us a true 10-bit 1080 HD video that stands up to much more pulling and pushing in the grade. The FS700's SDI out is 8-bit so you end up with two padded bits when recording to a 10-bit recorder.

Chris, where did you learn that component out is 10 bit raw? And must the FS700 have the 4k firmware upgrade to enable this?

[Gary Huff]

Originally Posted by Steven Reid

Chris, where did you learn that component out is 10 bit raw?

Component Out is not 10-bit raw. He's getting a padded signal as well, just doesn't realize it (and it's undoubtedly worse quality than just plugging the Blade into the FS700 directly).

[Steven Reid]

Originally Posted by Gary Huff

Component Out is not 10-bit raw. He's getting a padded signal as well, just doesn't realize it (and it's undoubtedly worse quality than just plugging the Blade into the FS700 directly).

And how do you know that (the complement to my question to Chris)? I've not seen any documentation, official or otherwise, discussing bit depth from the component out.

[Thomas Ruberto]

I too am agonizing over a similar decision. Within the next month, I will require UHD capabilities for a number of customers. I've only had the FS700 for a short time - I own a small travel magazine and shoot videos for advertisers. It paid for itself within hours of use. To jump to UHD, I either need to buy the Odyssey 7Q+ (between $6-7,000 with two 1TB drives, case, cables, battery plate, Zacuto mounts, etc.), or use that money plus the proceeds from selling the FS700 (it has less than 10 hours on it, plus the newer Sony 18-200mm, and the Zacuto Shooter) to upgrade to the FS7. I cannot justify (for now) the FS7 plus the adapter and an Odyssey 7Q+. So many options with the FS700 + the 7Q+ versus the newer FS7 and it's ergonomics. For the travel and attractions videos I shoot, I'm almost always on a tripod and/or slider and can drive my equipment to the site. However, my mid-life crisis has me trying to document climate change in Arizona, which I would also like to shoot in UHD. I've never shot with an external monitor - is it practical to rig the FS700 plus a 7Q+ and hike and shoot with it? Uuuugghh... wish someone would just tell me what to do.

[Gary Huff]

Originally Posted by Steven Reid

I've not seen any documentation, official or otherwise, discussing bit depth from the component out.

Component out is an analog signal, bit-depth basically has no meaning except as an analogy. I highly doubt the Component Out signal is being processed in the chain before the SDI and HDMI outs. That's what would be required, otherwise it's taking the same signal it's putting out of the SDI and HDMI ports and turning it analog (for compatibility's sake) out of the Component output.

[Steven Reid]

Originally Posted by Gary Huff

Component out is an analog signal, bit-depth basically has no meaning except as an analogy. I highly doubt the Component Out signal is being processed in the chain before the SDI and HDMI outs. That's what would be required, otherwise it's taking the same signal it's putting out of the SDI and HDMI ports and turning it analog (for compatibility's sake) out of the Component output.

Thanks. That is what I suspected and, hence, why I pressed Chris to elaborate. When I owned a Canon XH-A1, similar discussions arose from time to time, and the conclusion always was that you just couldn't get a better image from any analog out. At best, it was the same (quality) as the digital out.

[Christopher Young]

Originally Posted by Steven Reid

Chris, where did you learn that component out is 10 bit raw? And must the FS700 have the 4k firmware upgrade to enable this?

Steven ~

Sorry. Just to be clear I meant uncompressed full 4:2:2 bandwidth. I should have explained myself a little bit better. Sorry if I'm teaching Grandma to suck eggs but what I meant was that the analogue component you capture is not coming off at any bit rate. It is a pure analogue HD video stream that has been encoded to full HD 4:2:2 bandwidth meeting the BT.709 HDTV standards of 74.25MHz for the luminance channel and 37.125MHz for the two chroma channels (i.e. the 4:2:2 designation as the chroma channels are sampled at half the rate of the luminance channel). If you measure said component signal this is what you will see, refer to the screen shot measurement. You will see its frequency is measured at 74.25MHz which is the specified maximum.

Now look at table #1 graphic and look to the bottom of it and you will see that for both 60Hz and 50Hz video the maximum frequency for 1920 HD Progressive video is 74.25MHz. This in other words is a full bandwidth analogue stream.

If this full bandwidth 4:2:2 analogue video stream is fed into a high quality (10 to 12-bit path) component to HD-SDI converter, Decklink, AJA etc it will encode that analogue stream to a 10-bit HD-SDI output. Which if fed into an external recorder such as an Atomos ProRes recorder will be recorded as a true 10-bit HD-SDI signal. Alternatively you can take this full bandwidth 4:2:2 analogue stream and feed it straight into an AJA KiPro’s HD analogue inputs and it will transcode it and record it as a 10bit HD ProRes file for you.

Re your question as to whether the camera needs to have the 4k upgrade the answer is no if you are working HD only. That upgrade has no effect on the full bandwidth RGB component streams from the sensor that get encoded to the 4:2:2 analogue output. The sensor analogue RGB streams also run through A to D converters and once they are in the digital domain they get encoded to an 8-bit 4:2:2 SDI stream that runs out of the SDI BNC. The FS700 also outputs an 8-bit 4:2:2 HDMI signal.

If you want a RAW output from the FS700 to encode to either 2K or 4K then yes you have to have the 4K upgrade installed. An overview of what is required and what is happening and how the RAW output gets deBayered and converted is outlined here by Dan Keaton I think. There are other useful links on this page for further info.

https://convergent-design.com/produc...ony-fs700.html

Again a lot of this is moot because it assumes you have a camera with a signal to noise ratio of at least 60dB to be able to get the maximum out of a 10-bit 60dB encode.

The mathematical representation of SNR for an image is defined as the ratio of the light detected on the sensor (signal e) to the sum of the noise in reading the signal through (noise e). SNR is expressed in units of power or decibels (dB).

SNR (dB) = 20log (Signal e- / Noise e-)

Quantified examples: 8-bit with a ratio of 256:1 has a dynamic range of 48dB.
10-bit with a ratio of 1024:1 (a 4x increase over 8-bit) has a dynamic range of 60dB.
Let’s now take this through to something many more of us would be familiar with, the humble CD. The CD is how many bits? It is 16-bit encode with a ratio of 65,536, (a 256 x increase in information over the 8-bit system) and it supports a dynamic range of 96dB.

Ergo. Unless you have a camera giving you better that 60dB SNR there is not a lot to be gained in 10-bit encoding from the noise point of view. In spite of the extra noise I find 10-bit encoded files do handle heavier grading better as they are less likely to start banding given the extra levels to work with. The physiology of the human eye tolerates noise in an image far better than banding artefacts.

To give this a bit of a point of reference let’s look at a few cameras out there today with their listed “typical” SNRs. The PMW-400 is claiming 56dB, The new PDW-850 (CCD) is claiming 62dB, this is getting good! The PXW-FS7 is claiming 57dB.

But there again it cannot all be defined by objective empirical data alone. Take the Sony A7s for example, quote “Signal to noise ratio drops from 45dB at ISO 100 to 31.4dB at ISO 3200—clean images throughout. It dips slightly below 30 at ISO 6400 but can still produce good quality images at that setting.” Any of us who have used the A7s can attest to the fact that the picture noise out of this camera is far from objectionable even at reasonably high ISOs. If this camera was recording 10-bit depth one hell of a lot of it would be noise with these sorts of SNR figures.

http://goo.gl/A0UeWC

I can only report my experience. With the FS700 encoding the analogue 4:2:2 through a good quality 10-bit A to D gives us a full 10-bit file that handles grading better than the 8-bit output that gets padded to 10-bit via the Samurai Blade's input. Especially when using Cinegamma 3 as it is by far the quietest on the FS700. You will trade off a little bit of the high end dynamic handling but if it is lower noise you want the Cinegamma 3 analogue output is quieter than Cinegamma 3 out of the SDI port and once encoded through a Decklink mini converter to SDI is still quieter... plus it is now 10-bit. In reality though 95% of my work doesn't require this nature of encoding and mucking around. If I'm not shooting XDCam disc I'm shooting AVCHD and it does a great job especially for anything going to the web in 8-bit 4:2:0 MP4s. Shoot in it, cut in it and carefully encode in it and my clients are happy. To me that is the proof of the pudding and that's what I get paid for.

Chris Young
CYV Productions
Sydney

[Gary Huff]

Originally Posted by Christopher Young

I can only report my experience. With the FS700 encoding the analogue 4:2:2 through a good quality 10-bit A to D gives us a full 10-bit file that handles grading better than the 8-bit output that gets padded to 10-bit via the Samurai Blade's input.

And you do have video examples to support this claim, correct? What kind of grading can you do specifically with this method that falls apart with files recorded straight to the Samurai?

[Steven Reid]

Originally Posted by Gary Huff

And you do have video examples to support this claim, correct? What kind of grading can you do specifically with this method that falls apart with files recorded straight to the Samurai?

Chris, you get props for one the most thoughtful posts I've read in awhile. I took Gary's question as asking for the "proof of the pudding", as you put it, perhaps by comparing shots of a blue sky or 8-bit vs. 10-bit frame grabs from pushing a grade pretty far.

Stepping back for a moment, I'm a little baffled (and perhaps really naive, too) as to why Sony wouldn't have simply included a relatively inexpensive 10-bit AD converter in the camera in the first place. I already record straight to a Samurai, so I'm curious, like Gary, to see a few examples of the benefits from this alternative signal path and processing.

[John Nantz]

+1 for Chris

Originally Posted by Steven Reid

Chris, you get props for one the most thoughtful posts I've read in awhile.

Or maybe Professor Christopher.
All this and no pop quiz or homework assignment?

Talk about information overload - there's a lot of it here. I think if my brain were a computer it would freeze.

Thanks for the great educational post, Chris.

[Christopher Young]

Originally Posted by Steven Reid

or 8-bit vs. 10-bit frame grabs from pushing a grade pretty far.

Points taken re samples. Yes I had a think about this and at the moment I don’t have time to run around to set up specific shots where we are comparing apples with apples so to speak. After a bit of thought though I recall it was quite some years back that I first became aware of certain grading limitations when I was working on a clip that had color bars and pluge ramps at its head.

Trying to work out the quickest way to demonstrate what I was trying to say about the 8-bit digital capture compared to 10-bit analogue capture plus the nature of 8-bit banding and 10-bit analogue banding / noise levels I thought the quickest way to demonstrate this was to use Bars and Ramps as examples. A setup that others can very quickly replicate with their FS700s and Atomos recorders should they wish to. Using the color bars and ramps gives us a control test where there are no external influences like light and exposure so I thought that this would be a fairly acceptable place to start.

Obviously using a static subject like bars and ramps of course gives us no indication as to how movement can also effect grading. That’s a topic for another discussion which covers Spatial and Temporal Resolution.

Anyhow back to our bars and ramps. If you look at the attached pix you will see I have three sets of bars on an 8-bit timeline. All these were acquired at 1080 50i. I could have used a full 32-bit floating point timeline but for this exercise I stuck with 8-bit as this is the level most editors seem to work in.

In pix #1 the cursor is sitting on the AVCHD bars. The waveform and vector below show that the FS700 was putting out 16-235 level bars in 8-bit which equates to 0-100 IRE on the WFM. You will notice the vectors of the bars land in their correct boxes on the Vectorscope as one would hope.

Pix #2 Shows the bars captured by the Samurai Blade via the 8-bit SDI output. The ProRes capture is not 16-235. The Atomos recorders capture everything as full swing 0-255. As you can see from the pix it’s not quite there regarding the levels. The other thing to note is the non linear nature of the slope from black to white. It has a distinct curve which is incorrect. I will put these non compliant level anomalies down to this particular Samurai Blade’s state of adjustment. The output from the FS700 when measured externally on a hardware scope is spot on the correct values like they are on in pix #1. I’ve included no screen grab of the scopes on the 422 analogue bars capture as those levels were also full swing and intrinsically the same levels as pix #2.

Pix #3 shows a rather dramatic Color Curves plugin. For this test I called it the ‘Torture S Curve’ as it is pretty extreme. The reason it is extreme is to show the effect extreme gamma level adjustments can have. This exact S Curve was applied to all three clips. That was all that was applied. The extended ProRes capture levels would obviously need to be brought back to 16-235 to meet broadcast delivery standards. This could be easily accomplished with a ‘Levels’ plugin or similar.

Now we can pixel peep to our hearts content. Something the average viewer never does. We are our own worst enemies when it comes to these sorts of topics I am sure of it.

Pix #4 This shows a screen grab of the XAVC 24-mbit MTS clip’s greyscale. Note the transition from black to white. It’s not bad considering the extreme nature of the S curve that has been applied.

Pix # 5 This shows a screen grab of the 8-bit SDI to Samurai ProRes HQ clip’s greyscale. Note that in spite all we expect from ProRes HQ it is starting to show quite distinct banding in the greyscale but less noise as one would expect of an 8-bit 48dB dynamic range capture .

Pix #6 This shows the analogue 4:2:2 via an Analogue Component to SDI converter with its 10-bit capture. It also shows banding but less than in the 8-bit capture due to the fact it contains more noise which has dithered the banding. More noise is captured and this is to be expected as the SNR of the camera’s output electronics is incapable of meeting the 60dB SNR required by the 10-bit capture. In subjective general viewing though this increase in noise level and minimised banding is to my eye much more tolerable than the more distinct steps of the banding as is exhibited in the 8-bit ProRes example. I know most of us can tolerate a push up in ISO noise to certain levels and still consider it acceptable. I don’t know anyone though who considers banding even at its lowest levels to be acceptable.

Every case of vision capture is influenced by many factors, detail, movement, dynamic range, codec, bit rate and in more extreme cases vibration to the sensor. This static bar and ramp test can only give one a small indication of the complexities of capturing images but I hope it helps in the understanding of the issues involved. All in all in this example I think the much maligned AVCHD codec handles the extreme Gamma S Curve better than either of the ProRes examples. Go figure.

A while back I came across this clip that some may find of some interest. Done by someone who is pretty through and who was really trying to find out just how good ProRes is in detail retention, noise and color depth compared to the humble AVCHD codec. I think you will be surprised at some of his findings. Apart from his mistake of taking ProRes 422 to be an 8-bit codec. A mistake on his part that he later acknowledges. His experiments were carried out using a Canon C100 but I think one would draw similar conclusions had they been carried out with an FS700 or any number of other cameras for that matter.

AVCHD better than Prores? - YouTube

Chris Young
CYV Productions
Sydney

[Gary Huff]

Why did you not submit an actual frame grab from footage you have shot? I'm sorry but none of this convinces me in the slightest that you are not taking an 8-bit 4:2:2 analogue converted signal in as 10-bit and converting it back to digital. You need to show actual footage, not just a screen grab with this "torture" curve applied from footage shot internally, out to the Blade, and out to your convoluted Component analog capture setup. It needs to have motion and it needs to demonstrate where the Blade captured footage breaks down where the Component footage does not. All you have done so far is given information that I don't think you are interpreting correctly. Your bars at the bottom demonstrate nothing more than banding being hidden by extra color noise introduced by the analog conversion process, which is what I feel is confusing you.

[Steven Reid]

Chris, thanks for taking the time to reply. I'm not eager to lodge a churlish reply. Still, I have to admit that you lost me on the applicability of your color bars and pluge ramp test. Those images aren't coming off the sensor, are they? If not, then I have to join Gary in wondering why you bothered with them.

The video that you linked was excellent and provided the information that, if shot on the FS700, would likely put our questions to rest. I'm sure you're busy like the rest of us, but since you put forth the notion in this thread that a couple hundred bucks for a 10-bit A/D converter could elevate the image quality, I think we're still looking to you to prove up your point.

[Gary Huff]

Originally Posted by Steven Reid

The video that you linked was excellent and provided the information that, if shot on the FS700, would likely put our questions to rest.

Hey Steven, that video was shot on the C100.