This popped up on the Sony Professional Europe Facebook page this evening.

Anyone care to guess??

There is something a little curious about the specs for the sensor in the FS700:

Why create a sensor with 11.6 million pixels and then use only 8.3M? It is normal to have some extra pixels that are used for setting black levels etc, but this is a massive difference between the number of actual pixels on the sensor and the number that are used to create the pictures. Where are all the un-used pixels? Given that this is a Super 35mm sensor, the active area used for video is APS-C ish sized, so it’s quite a big sensor already. What would you put a near full frame 35mm  11.6 MP sensor in these days? That’s a low pixel count for a modern large sensor DSLR or stills camera, even the compact NEX7 stills camera has 24MP. If (and this is just random speculation) the FS700 is taking an 8.3MP window out of the middle of the sensor, that makes it a pretty big chip. Another thought is that the FS700 does read the full height and width of the sensor but then uses some pixel skipping only actually reading 8.3MP, but why do that? In stills mode the camera only uses 8.4MP yet with so many extra pixels you could get higher resolution stills. So… why 11.6 MP and what else was this sensor designed for?

OK, not really very scientific, but I’m busy on some paying projects at the moment and the weather is very changeable so I only had a short window to do this. I’m still exploring the image quality of the FS700. It is very good, of that there is no doubt, but my benchmark right now is the F3. So I just did a very simple side by side test to look at noise and dynamic range. The scene has about an 11 to 12 stop range if you include the specular highlights and reflections off the silver car bonnet and the brightest clouds. If you take the S-Log clip as the reference the clouds are at about +5.5 stops over nominal middle grey and the darkest part of the image, the black stand holding the chart, is about 6 stops under. I wasn’t looking to actually measure anything here, just get a feel for the differences between the cameras.

The cameras were set to 800 ISO for all the clips, so the FS700 had +6db gain applied for all clips while the F3 had +6db applied for the Cinegamma and standard gamma clips and no additional gain in S-Log. Frame rate was 25fps with a 180 degree – 1/50th shutter. The HDSDi out from the cameras was recorded using ProRes 4:2:2 on an Atomos Samurai.

For the S-Log sample I exposed using the DSC Labs S-Log exposure reference chart (which you can see in all the frames) by placing the cameras centre spot meter over the middle grey and aiming for 38%, however my waveform monitors are telling me the mid grey exposure was actually 35% so I’m about a 1/4 stop under (and need to check why I didn’t get 38%). For the Cinegamma 4 tests I used the histogram to keep peak white at about 95% with similar mid range exposure. In fact the mid grey patch on the DSC S-Log chart is around 38-40% on both cameras which is just a touch low for the cinegammas (I normally aim for 42%-45% with cinegammas). Exposure for the standard 709 gamma was established with the histograms trying to get a reasonable balance between clipped highlights and a reasonable mid range. The result is that the standard gamma shots are under exposed by around 1.5 stops, mid grey is only 33% on the F3 and 30% on the FS700. I would normally aim to put middle grey around 45-48% for Sony’s REC-709 compliant gammas. If this shot included a person or face then I would have been forced to either over expose the sky still further or use some fill lighting or a reflector to bring up the foreground. This is typical of the dilemma you get when trying to expose a scene with a greater range than the camera can deal with, do you overexpose the sky to preserve the mid range or underexpose the mid range to keep the sky. Either way something has to suffer.

While not very scientific I think the clips highlight some interesting differences between the two cameras. The most striking difference is the colour. Both cameras were set to preset white at 5600k with their standard colour matrices. White does appear to be white, but the F3 when not in S-Log is clearly more saturated and has a touch more red and a lot less blue that the FS700, so clearly I’m going to have to do some work on the matrix to get these two cameras to match better. Next thing to note is that the Cinegamma curves are quite different. The FS700 curve has more gain in the mid range which results in brighter upper mid range compared to the F3. The dynamic ranges are very similar, I could have exposed the FS700 about a stop lower to gain a little more highlight room, but this would have resulted in some quite dark mid tones and a little loss of shadow detail. In both cases the cinegammas give a quite appreciable increase in dynamic range over REC-709. I would guess at about a 1.5 to 2 stop improvement in dynamic range gained from using Cinegamma 4 over the Sony REC-709 compliant curves + knee. The S-Log clip from the F3 shows the marked increase in dynamic range that you get when using log. The brightest clouds are about 5.5 stops over middle grey with the peak recording level reaching about 89% which means there are around a further 1.5 stops of unused headroom available.

The FS700 is a little noisier than the F3, no surprises here. It’s not hugely noisier and the noise levels at 400 and 800 ISO are perfectly reasonable for a broadcast production. At a push I would use 1600 (+12db) if I had to, but I think for me at least the comfort zone is 400 – 800 ISO (0db and +6db).

The FS700 images appear to lack a little of the crispness of the F3. This may just be because the standard F3 is a little over sharpened (in my opinion). The FS700 pictures look more like the S-Log F3 which has no added image sharpening and in fact in some respects the Cinegamma FS700 looks more like the S-Log F3 than the Cinegamma 4 F3. From what I’ve seen you probably could figure out a very flat log type picture profile for the FS700, but I’m not sure that this would bring any significant benefit over Cinegamma 4. The extra noise in the shadows that you would get if you bring up the low end with some black stretch (black gamma) would likely limit the usefulness of any slight extra latitude gained. However you look at it the FS700 does appear to be able to cope with around 12 stops (maybe a little more) just by using Cinegamma 4, which is about as much as you want with a 8 bit camera anyway.

Interestingly I spotted some moire and aliasing from both cameras. The dreaded roof tiles of the houses opposite (a tough test for many cameras) strike again and the F3 is showing a little coloured moire across the roof tops while the FS700 is showing some aliasing on the grill of the silver car. I’m not concerned by either. Yes it would be nice if it wasn’t there, but I’ve shot hundreds of hours with my F3 and it’s very rare to find any shots that are unusable or problematic due to aliasing.

If you want to download the actual footage please use the link below. The file contains a single clip made up of about 5 seconds each of the FS700 with Cinegamma 4 and standard settings, the F3 with cinegamma 4, standard settings and S-log. The clip is a direct copy of the original ProRes 422 files recorded from the HDSDi outputs of the cameras. 10 bit for the F3 and 8 bit for the FS700. The file size is 150 MB and 10 seconds long in total. Lets hope the file sharing service works as advertised!

DOWNLOAD THE SAMPLE CLIPS – FREE!

If you find the clips useful please consider making a small donation using the button below. You can choose the amount you wish to donate. Even small donations of £1 are greatly appreciated and go towards the costs of writing articles and providing further sample clips.

 Here are some jpeg frame grabs. Remember you can click on each image to see it larger and once you’ve clicked though to the larger image there should be a link just above it for the full size 1920×1080 original.

Here’s a few clips shot with the FS700 from last weekends Royal International Air Tattoo. Check out the DHL 767 just hanging in the sky, also notice the almost complete lack of skew on the propellors.

Many of you will have heard about my involvement in the recent production of a film about one of Duran Duran’s concerts last December. Well it’s out on Blu-Ray and DVD now and I think it looks might fine. Directed by Gavin Elder, Produced by James Tonkin it was a great pleasure to work with Den Lennie to help create a special picture profile for the F3’s used to shoot the concert. Here is a track from the Blu-Ray to give you a taste of the look of the video. Please click through and play it full screen in HD for the full effect.

Hello all.

I really need to get an ultra light tripod for my travels. I love my Vinten 5AS and 100AS with CF legs, these are great, but they are around 6kg/8kg and possibly more of a bind is the length of the legs when collapsed. I really need something a bit easier to travel with, especially for my arctic expeditions and overseas workshops. On my recent Asia trip I was typically around 5kg over my baggage allowance on many flights, partly because of the weight of the tripod but also because of the size of the case required to transport it. In addition a big broadcast tripod does tend to catch the attention of customs officials. So I need something that will work with a bare bones FS700 or F3 (3-4kg payload). I realise that cutting back to the minimum on the tripod will effect my ability to achieve stable shots and smooth pans, but most of the shots I do are wide angle.

I’ve looked at a few options. Maybe a Manfrotto 504 head with the single tube CF legs, or one of the Miller Solo systems. I don’t have a big budget for this so perhaps I’ll just get a set of single tube 75mm bowl legs to use with my Vinten 5AS head. So what do you guys use, what are your recommendations? Please leave a comment with your suggestions.

This is a video of Singapore I shot while at Broadcast Asia. I used a variety of Sony cameras to shoot this. The principle camera was a Sony FS700, used for the slow mo at the beginning and end as well as many of the time-lapse shots which were done by shooting using S&Q mode at 1 frame per second. Also used was a Sony PMW-F3 in frame interval mode and a Sony NEX5 stills camera triggered with a Gentled time-lapse controller.

One of the concepts that’s sometimes hard to understand is why mid range exposure is so critical with most video cameras, even cameras with extended dynamic range. Cameras that use Cinegammas, Hypergammas or even log curves like S-log or Arri Log may give you great dynamic range and extra latitude but it’s still vital that you get your mid range exposed correctly. In many cases, the greater you cameras ability to capture a wide dynamic range the more critical mid range exposure becomes. I’ve often heard comments from users of XDCAM cameras complaining that they find it harder to work with cinegammas and hypergammas than the standard REC-709 gamma.
So why is this, it seams counter intuitive, surely a greater dynamic range makes exposure more forgiving?

First lets take a look at a standard gamma curve. These graphs are not accurate, just thrown together to illustrate the point. The standard gamma for HD, REC-709 can be considered to be near linear. Certainly in terms of “what you see is what you get” the idea behind REC-709 is that if the camera is set to 709 and the TV or monitor is 709 compliant then we will get a linear 1:1 reproduction of the real world. However REC-709 is based on the gamma curves used at the very beginnings of television broadcasting where TV’s and cameras had very limited dynamic range. True REC-709 only allows for about 6 stops of dynamic range and as a result the version of REC-709 used in most video cameras is tweaked somewhat to allow a greater dynamic range in the region of 8 to 10 stops while still producing a pleasing image on most TV’s. Another way of increasing dynamic range is to introduce some form of signal compression. The simplest form of this in common use is the cameras knee circuit. This simply takes anything above a certain brightness level (typically between 80 and 95%) and compresses it. We normal get away with this compression because it’s only affecting highlights like clouds in a bright sky or a bright window or lamp in the shot. Our own visual system is tuned primarily to mid tones, faces, plants and things like that so we don’t tend to find highlight compression overly obtrusive.
When considering your post production workflow and grading in particular, it’s important to remember that in most cases whenever anything is compressed then some of the original data is being discarded. In addition if the amount of compression is non-linear (increases or decreases with amplitude) then when we add a linear function to that, like adjusting the signal gain the non-linearity is also increased.
Based on these assumptions, you should be able to understand that anything exposed in the linear part of a gamma curve will grade very well because there is no extra compression and gain adjustments will behave as expected. Now if you look at the graph of a typical standard gamma curve (as above) you can see that everything below the knee point is pretty linear, so anything exposed in this range will grade easily and well (assuming it isn’t actually overexposed). For this reason standard gamma can be very forgiving to small over exposure problems as a slightly bright face should still be in the linear part of the curve. However overexpose to the point where the face starts to enter the knee area and all is lost, you’ll never make it look natural.

Now look at the curve for a typical Cinegamma or Hypergamma. You can see that this curve starts to become more curved and less linear much earlier than a standard gamma. This is how the extra latitude is gained. Compression is used to allow the camera to record a greater brightness range. This extra compression though comes at a price and that is linearity. The further up the exposure range you go the less linear the response (it’s actually becoming logarithmic). The result is that even though you have more dynamic range, if you do overexpose faces and skin tones by even just a small amount they will start to creep into the non linear part of the curve and this makes them harder to grade naturally. You may be less likely to get those ugly blown out highlights on a shiny face typical of video knee compression with cine/hypergammas, but you must still be very careful not to overexpose.

Now if you look at a typical Log curve you will see that compression starts almost immediately. The name of the curve also gives us a clue. It’s Log so it’s not linear to start with, there is almost nowhere on the curve that is actually linear but it is worth noting that the best linearity is lower down the curve. That’s why when you shoot using log your exposure levels are normally considerably lower than with standard gammas. Middle grey should be 38%, faces will be around 45-50% as opposed to the more typical 65-70% of standard gamma. This lower exposure keeps faces and skin tones in the more linear part of the curve where they will grade better. In addition when you do go into the grading suite with your log material do try to use correction tools that will apply log gain instead of linear gain. Most dedicated grading tools should have this option, certainly Resolve does.
So there you have it. Greater dynamic range does not necessarily equate to more exposure tolerance. In fact it’s often the opposite. You might get better highlight handling, but you may find you need to be even more careful with how you expose. As we go forwards (or sideways at least) and linear raw becomes more common place then you will be able to shift you mid tone exposure up and down with a lot more flexibility as with a linear raw camera the last stop of exposure has the same linearity as the first, so in theory your mid tones can sit anywhere in the exposure range. Sony’s F65 is a great example of this. It has 14 stops of linear dynamic range. A face lit with a 3 stop range could be placed in stops 11-14 and would grade down to wherever you want just perfectly.

One of the key benefits of a raw workflow is that normally you will be working at a higher bit depth, at least 12 bit if not 14 bit or 16 bit. This in turn allows the use of linear capture as opposed to the log capture normally associated with conventional video.

Don’t get me wrong, log capture and recording (even things like hypergammas and cinegammas are closer to log than linear) is very good and works remarkably well. But it is limited as it compresses highlight information. Each extra stop of over exposure with a linear recording will contain twice as much data as the previous, while with log each stop contains the same amount of data, so as a result each brighter stop only has half as much information as the previous. Log does allow us great scope when it comes to grading and post production image manipulation, but the higher up the exposure range you go, the less data you have to work with, so how much you can manipulate the image decreases with brightness. As our own visual system is tuned for mid tones this log behaviour goes largely un-noticed. But as modern sensors achieve greater and greater dynamic ranges log starts to struggle while linear copes much better.
It’s not until you try linear raw with a camera like the Alexa or F65 that you realise just how forgiving it is. In log mode the Alexa (and other log cameras like the F3) need to be exposed accurately. Over expose and you risk not only your highlights blowing out but also it becomes harder to get good looking skin tones as these may be up in the more compressed parts of the curve. However with linear, it doesn’t really matter if faces are higher up the range, jus as long as they are not actually at sensor overload.
When you shoot with a log camera it must be treated like any other conventional video camera. Exposure must be correct, you need to watch and protect you highlights, expose to the left etc. A camera shooting raw behaves much more like a film camera, you can afford to push the exposure higher if you want less noise, just like film.

But linear raw comes at a cost, mainly a time and storage cost. We have become very used to the simplicity of working with video. Modern file based workflows are fast and efficient. Raw needs more work, more processing, more storage (compared to compressed at least). But computers are getting faster, storage is getting cheaper. Right now I believe that raw is only going to be used by those that really do need and want the very best flexibility in post production while log will become more and more common for episodic and documentary production. But, the time will come when we can handle raw quickly and easily and then perhaps we will look back at legacy codecs and wonder how we managed. Although saying that, while we still broadcast and distribute programmes using backwards compatible legacy gamma with it restricted dynamic range for display on devices with only 6 stops of display latitude, a general shift to raw with all it’s extra overheads may never happen.