All posts by alisterchapman

Atomos Samurai Review.

It’s taken me a while to finish this review as I’ve been pretty busy. While preparing it I posted various frame grabs that I pulled from the Samurai and the Gemini on a couple of blogs and forums and these have created a bit of a stir that was never intended.

Anyway, what about the Samurai? Well to be honest I was impressed before I had even turned it on. When you buy a Samurai, not only do you get a neat little video recorder, but you also get a flight case full of everything you need to use the recorder (with the exception of a hard drive or SSD). The flight case contains the Samurai itself, a couple of adapter cables to go from the mini BNC’s to full size BNC sockets. A couple of batteries, a neat dual channel charger, a pair of drive caddies and a docking station for off loading your material. Everything is well made.

SIMPLICITY

One of best bits about the Samurai is that it is extremely simple to use. A push button on the side turns it on and your quickly greeted with the home screen. From this one screen you can record clips, enter playback mode, go in to monitor mode, format a drive, turn on timecode triggering and choose whether your progressive files get tagged as interlace or progressive (more about that in a bit).

MEDIA:

For my tests I used a 250Gb 2.5″ hard drive that I had kicking around. I’m not a fan of using spinning discs to record on, but in terms of cost per GB they are very affordable. The discs spin at between 5,400 and 7,200 rpm and this means there are some quite high gyroscopic forces in play. Twist the unit too fast or drop it and you may damage the disc leading to the loss of valuable material. The Samurai though, can also use SSD’s (Solid State Drives) which are very fast, cost effective memory devices with no moving parts. If your going to be bumping your Samurai around an SSD is the way to go, but if you need lots of cost effective storage a hard drive may work better for you, provided your sensible with it. One recommendation of my own is not to be tempted to use one big 750GB hard drive, but instead to use several 128GB or 250GB drives. This way if a drive is lost or does fail you don’t loose quite so much. While you should, as I did, be able to use pretty much any drive, there are a few that Atomos do not recommend. Atomos regularly test new drives as they become available and they provide a list of recommended hard drives and SSD’s. It’s almost certainly best if you choose your drives from the list, which can be found here.

SETUP:

There are very few menu settings on the Samurai and this makes it simple to use and fast to set up. If your using a camera with “record run” timecode the Samurai can be set to watch for rolling timecode and as soon as it does go in to record. I found a 2 or 3 frame delay between the camera going into record and the Samurai triggering, which is noting at all and in most cases you’d probably never realise there was this minute delay. Of course you can also press the big red record button on the touch screen LCD or if you have a camera with LANC you can plug a LANC controller into the Samurai and then plug the Samurai into the camera. The remote would then trigger both units together.

PsF:

One of the buttons on the home screen turns on and off PsF recording. PsF is the method used to send 1920×1080 progressive footage via HDSDi. PsF is used by just about every 1080P camera. It’s clever because it allows you to plug a progressive camera into an interlace monitor, and everything still works just fine. It does this by taking the progressive image and splitting it into 2 fields before passing it down the cable. So to the receiving device (monitor, tv, recorder etc) it looks like an interlace signal, but as there is no temporal (time) difference between the two fields, when shown on an interlace monitor what you end up with is a progressive image. The issue with this is that when you feed something like the Samurai with a PsF signal,it can’t tell whether it’s getting a progressive image split into fields or a real interlace image with two fields. This is where the Samurai’s PsF button is useful as it allows you to tell the Samurai whether the signal is progressive or interlace so that the recorded clips are correctly flagged. A word of CAUTION about this. If you are at all unsure, it is safer to not flag the clips as PsF. Why? well once a clip is flagged as PsF your edit software will always treat the odd and even lines within the frame as occurring at the same moment in time, so any effects or renders will also be progressive and this may result in motion artefacts. If you have a PsF clip flagged as interlace this problem does not occur as generally if you render from interlace to progressive the fields get combined anyway.

POWER AND BUILD:

Moving on… On the back of the Samurai are locations for 2 Sony NP series batteries. These are really common batteries that can be purchased very cheaply. Atomos will also provide you with a free D-tap adapter if you wish to take power from a 12v battery. To get the adapter all you need to do is to register your Samurai on the Atomos web site. You can also by additional adapters and a power supply from you Atomos dealer. The batteries can be hot swapped so in theory you could shoot forever, or at least until you drive is full.

The build quality is very good. The bulk of the unit is machined from aluminium, so it should take the general bumps an knocks of professional use. There is a 1/4″ threaded mount on the top and bottom of the chassis for mounting. HDSDi Input and output is via mini BNC connectors and Atomos provide you with a couple of adapter cables that go from a mini BNC plug to a standard BNC socket. While this works OK, on the review sample the connectors were straight and as a result these rather thin connectors stick out quite a long way. I don’t like this, they look a bit vulnerable to me. Atomos tell me that are looking into right angle connectors and this would be a big improvement in my opinion. The touch screen also acts as a monitor so you can check that you have not got the camera overlays turned on by mistake. Some external recorders don’t have a monitor and I’ve heard horror stories of entire shoots being ruined because the operator didn’t realise that the camera data was getting recorded burnt into the video clips. It’s not the brightest of screens and perhaps lacks a little contrast, I think some of this is due to the membrane used for the touch screen, but it’s very useable. I could probably benefit from some kind of hood in bright conditions.

BENEFITS AND IMAGE QUALITY:

Crop from frame of leaves blowing in wind, recorded using F3's internal 35Mb/s codec.

So why do you need an external recorder, most cameras are capable of producing a good image without an external recorder aren’t they? Well in general, yes, most cameras will produce an acceptable, first generation image. But when you start making multiple copies or doing extensive post production work on the internal recordings, very often they simply aren’t up to the task. Take a look at the frame grabs presented here (click on them to view full size), one recorded internally on the F3 at 35Mb/s and the other on the Samurai using ProRes HQ at approx 185Mb/s.

The same frame as above but recorded using ProResHQ on the Samurai

When you look at the unprocessed, 1st generation frames, it is perhaps hard to see a difference, but when you do just a little bit of post production grading then the difference becomes much more noticeable. The crop is of the top right part of the frame. The tree branches were blowing in the wind and this motion makes the codec have to work quite hard. When there is only 35Mb/s of data to play with the codec really struggles, while the extra data available to the ProRes codec of the Samurai means that the codec isn’t struggling at all. The other thing in this example is that the F3’s internal recordings (like most AVCHD or Mpeg2 cameras) use only 8 bits of data and the colour is sampled 4:2:0. The Samurai is 10 bit, 4:2:2.

MORE BITS ARE BETTER:

With an 8 bit recording you only have 235 brightness levels. That’s fine if you don’t manipulate the image in any way, but if you modify the image contrast or brightness you will create gaps and jumps between those steps and this will appear as banding in your pictures. 10 bit recordings have 940 brightness levels, so any steps will be much, much smaller so banding after adjusting brightness and contrast is significantly reduced and is not normally visible.

When you compare 4:2:0 against 4:2:2, the later has twice as much colour detail than the former. Now I have to confess that I often to struggle to actually see this difference in progressive material, but if you shoot interlace you can see a difference. In either case, having more colour detail will once again improve the flexibility of your material in post production and can give a small improvement in the appearance of chroma noise.

NOISE AND MACRO BLOCKING:

Talking of noise…. if you choose to look closely at the frame grabs I have provided then you may think that the recordings from the F3 have less noise than the Samurai. Well, they do and they don’t. You won’t see as much fine grain noise in the 35Mb/s frame grabs as the Samurai frame grabs because the F3’s internal codec has to discard a lot of picture information to fit it into 35Mb/s. One of the things that gets discarded is noise, but as well as noise there is also a loss of some subtle fine textures. So be careful when you look at the images, the F3 material may look cleaner, but it’s also lacking the fine detail and texture that has been captured by the Samurai. Check out the video below. First you see the full frame clips from the F3 and Samurai, ungraded and then graded. Finally you’ll see crops from the clips. Check out the ugly, blocky artefacts from the F3’s internal recordings on the first clip, then the breakup of the leaves in the second clip and finally the strange wobble of the hedge and bricks in the last shot. These are all typical motion artefacts from the long GoP Mpeg 2 used by the F3 internally.

Please watch the clip full frame 1920×1080. Please also consider that youtube adds a lot of artefacts of it’s own, but the side by side clips will be affected equally. Try pausing the clip during the side by side branches clip.

These may seem trivial in a highly compressed web clip, but let me assure you they are plainly visible on my 42″ TV. When I did compare the Samurai to the Convergent Design Gemini which is a fully uncompressed recorder there was not a lot between them. The Gemini does produce a slightly better image as it has the capability of recording a pristine, mirror image of what comes out of the camera, there is no compression, no loss, no artefacts and if quality is your absolute priority this should be the type of recorder to consider. But the Gemini produces some very, very big files that will need to be backed up and stored. For example an hour of material from the Gemini requires 750GBs worth of the very fastest SSD drives. A 750GB hard drive in the Samurai costs less than 1/10th of the price and will record over 16 hours of material. Not only that but you can backup 16 hours of Samurai material in around the same time as it would take to backup 1.5 hours of Gemini material. In terms of image quality I would say that the Samurai is more than adequate for most types of production, so you do have to ask yourself whether the small quality improvement you can get from the Gemini justifies the more expensive, more complex and slower workflow? Me personally, I’m seriously considering the Samurai as my recorder of choice for all my day to day production work. I’ve looked at the Pix-240 and it’s just too big and bulky, the Ki-Pro mini doesn’t have a monitor and the NanoFlash is only 8 bit. The Samurai ticks all the right boxes. However, if I have a big screen, high end production where the more complex Gemini workflow is not an issue, then for those productions I would use the Gemini. It’s horses for courses, the right tool for the job.

CONCLUSIONS:

For the money the Samurai is all but impossible to beat. Lets face it, the price is very good. Sometimes a price can be too low, I often associate cheap prices with sub-standard products. While the price is low, the Samurai however is not sub-standard. It’s well put together, very capable and has good recording media flexibility with the option to choose codecs (ProRes or DNxHD) and media types (SSD or HDD). With the PMW-F3 it can be used with standard gammas or 422 S-Log. It brings a significant improvement to the quality of the images you can record and makes the workflow even easier than when your using SxS (no need to re-wrap from MP4 to .mov with FCP). I think its great as an all-round every day recorder. 9/10

This is why you want an external recorder!

As you may know I’m preparing a review of the Atamos Samurai as well as the Convergent Design Gemini. Both excellent devices, but suited to different applications. If your wondering why you need an external recorder, below are a couple of crops from some footage that I shot today on the F3. High compression codecs can struggle when there is a lot of motion in the frame. What you see in the pictures below is some leaves on a tree blowing in the wind, one frame from the F3’s internal recordings and the other from the ProRes HQ from the Samurai. The clips were graded, as it was shot using S-Log, but this is a quite realistic scenario and with TV’s getting bigger and bigger, this kind of thing is getting much more noticeable.

F3 35Mbps Internal Recording
Samurai ProRes HQ

More Codec and Gamma Tests.

More Gemini, Samurai, AC-Log and S-Log sample frame grabs. See download box at bottom of post.
I had thought, when I first wrote this post that I had discovered a strange issue where the 444 RGB recordings from the Gemini had more dynamic range than 422 recordings. I didn’t think this was right, but it was what my NLE’s (FCP and CS5.5) were telling me. Anyway to cut a long story short, what was happening was that when I dropped the Gemini RGB files into the timeline the levels got mapped to legal levels, i.e. nothing over 100% while the YCbCr 422 clips go into the timeline at their original levels. The end result was that it appeared that the 422 clips were clipping before the 444 clips. Thanks to Waho for suggesting that it may be a conversion issue with the frame grabs, I was able to see that it was simply the way the NLE’s (both CS5.5 and FCP were behaving in the same way) were clipping off anything in the 422 clips above 100% both in the frame grabs and also on the monitor output. As the RGB files were all below 100% they were not clipped so appeared to have greater dynamic range.

Anyway….. below is a new set of frame grabs layered up in a single photoshop file showing how the various codecs and recorders and codecs perform. The levels in these have been normalised at 100% to avoid any dodgy clipping issues. I’ve included F3 Cinegamma 4, plus my AC-Log picture profile, plus Samurai ProRes, Gemini S-Log and F3 Internally recorded S-Log of a very extreme contrast situation. Use the link below to download the photoshop layers file. You’ll need to me a registered user to access the link.

[downloads_box title=”More codec test grabs.”]
Photoshop Layered Frame Grabs v3
[/downloads_box]

Alister’s Product of the Year 2011

Well the new year is less than a month away and at this time every year I start to consider my product of the year. To qualify the product must be available for general purchase within the year, so the C300 can not be included. In the running at the moment:

PMW-F3, FS-100, Convergent Design Gemini, Atamos Ninja/Samurai, Matrox Ultra Studio, F65 (if it ships before year end). Media Composer 6, Zacuto EVF, Cineroid Metal EVF.

Anyone think I should be considering anything else?

Canon C300 open house at Visual Impact. Price £9995!

Crowds gather around the Alexa, F3 and C300

I spent most of the morning at the Visual Impact C300 open house. They had 3 C300’s on show. One on a camera set alongside an F3 and an Alexa, the other was a very simple handheld configuration for people to play with and a 3rd on a stedicam rig.
What impressed me the most was how nice the camera was to hand hold and how good the rear viewfinder is. You certainly don’t need anything extra to shoot with it.
On the camera set I was a little less impressed. At the beginning of the day when I first quickly looked at the 3 monitors I could see no obvious difference between any of the cameras. The monitors were fairly small 20″ Panasonic monitors. However when I went back to take a closer look, for some reason the C300 looked a little soft or not quite in focus. No matter how I played with the focus, the edges of the image looked soft compared to the F3 next to it. Both the F3 and the C300 had Arri 32mm master primes so they should have been similar. I was told all the cameras were set up with standard gammas, but looking at the C300 it looked quite flat, so perhaps it was setup with the Canon Log gamma and the resulting lower contrast was making the image appear softer. There were so many people at the event that it was difficult to really get at the camera to figure out what was going on. I don’t think something was quite right, I’ve seen better from the C300. I asked if I could record some footage but was told that this was not possible on this occasion as these are pre-production cameras. Shame, I had a pocket full of CF cards and SxS cards. Anyway I have been promised a test shoot very soon.

I really hope that the softness was a peculiarity of that particular camera or the way it was set up as the C300 would for me be fantastic for my storm chasing and extreme weather assignments. I’d love to take one up to Norway in January to see if it sensitive enough to shoot the Northern Lights without having to resort to slow shutters or long exposures (Stop Press… I’ve placed an order and hope to have one in time for Norway). I’ve done the math, and according to my calculations it should just about be sensitive enough at 22,000iso with a f1.4 lens to shoot the Aurora in real time. It would be really cool to try and stream the Aurora live from Norway in January. On other assignments I could shoot using Canon L series glass or my B4 to Canon adapter and get broadcast ready material without needing an external recorder. I still think the Sony F3 with S-Log and an external 444 recorder is capable of a better image, but that’s a significantly more expensive package and more cumbersome, power hungry etc. It’s horses for courses. The C300 for me looks to be fantastic for simple, fast, easy shoots where shallow DoF is desirable, while I would continue to use the F3 where the extra bulk of an external recorder and the slightly more complex S-Log workflow will not be an issue as I think the ultimate image quality will be better. I might just have to place an order for one, especially as the price will be less than anticipated. The price I was offered from Visual Impact today was £9995, available from mid January…. Just in time to take to Norway????

S-Log on a non S-Log PMW-F3 and Log on an EX1/EX3

Note: There is something up with the frame grabs. For some reason they are very dark. I’ll look into this in the morning and get some more accurate grabs online.

First of all let me say thanks to Ben Allan on CML list for getting me thinking about this. He has already started experimenting with creating a log style Picture Profile for the EX1. All the setting you’ll find here are my own work and based on tests done with real scenes and some dodgy home made latitude test charts 😉

Ben’s musings on CML made me consider what S-Log is. In essence it is nothing more than a clever gamma curve that allows you to capture a greater dynamic range than is normally possible with conventional gamma curves. The reason why the standard gamma dynamic range is normally constrained is in part simply because if you record too large a dynamic range and then show it on a conventional monitor or TV, it simply does not look right. So to make it look right it must be graded in post production. In order to do a significant grade in post, the quality of the recording has to be good enough to withstand a fair bit of pulling and pushing. As a result 10 bit recording is recommended (however it is still possible to work with lot with top quality low noise 8 bit recordings, not that I would recommend this). Anyway as both the standard PMW-F3 and EX1/EX3 have 10 bit outputs I decided to see if it was possible to come up with a picture profile that would mimic a Log curve and then see if it actually brings any real world advantage.

Genuine S-Log, mid grey @38%

First up I experimented with the F3. I already have the S-Log option, so this gave me a benchmark to work against. To mimic S-Log you need to increase the gamma gain at the lower end of the curve, you can do this with the Black Gamma function. I know that with S-Log the cameras native ISO is 800 as this is the sensitivity at which maximum dynamic range can be realised with the F3’s sensor. So I started my experiments at 800iso. I could bring up the shadow detail with the Black Gamma but I notice that I appeared to be trading off some highlight handling for shadow information, so while the images kind of looked like S-Log, they did not really gain any latitude.

AC-Log v1. Very similar to S-Log, same exposure as S-Log

During this process I realised that my mid range sensitivity was now a lot higher than with genuine S-Log, so I decreased the camera gain so I was now at 400iso and started tweaking again. Now with Black Gamma all the way up at +99 I was seeing around 1 stop further into the shadows, with no impact on highlight handling.

When I tested my new Picture Profile on a real scene, exposing as you would S-Log with mid grey at 38% I was very pleased to find some very similar images that do grade quite well. As well as the Gamma tweaks I also incorporated a few other changes into the profile to increase the overall grade-ability.

CineGamma 4, mid-grey at 38%

There is a definite improvement in shadow reproduction. It’s not as good as real S-Log, but it does give a very useful improvement for those without S-Log. One interesting point is that the exposure between the two log frame grabs posted here is not changed, so even though the camera is set at 400iso, when the picture profile is applied the camera behaves more like an 800iso camera and exposure should be set accordingly.  I think my PP (which you can download at the bottom of the page) brings a little under a one stop improvement in DR, real S-Log is about 2 stops.

If you click on the image captures you can view them full frame. When you compare the AC-Log and Cinegamma 4 images you should be able to see more shadow detail in the tree on the right of frame with the AC-Log yet the sky is further from clipping as well.

So what about the EX1 and EX3, can the same be done for them? Well this is much more of a challenge as the EX cameras are much noisier. Simply bringing up the Black Gamma does help you see into the shadows a bit better but it comes at the cost of a lot of extra noise and really makes it un gradable. Normally I don’t recommend using negative gain as it can reduce the dynamic range of the camera. But I figured if I use negative gain and then increase the gamma gain that should cancel out any dynamic range loss. To then avoid the usual -3db reduction in highlight performance I adjusted the overall gamma gain to return the peak output level to 109IRE. After a bit of fiddling around with my test charts and waveform monitors I could see that it was possible to gain a small amount of dynamic range, a little under 1 stop, however there is an overall increase in the noise level of about +4db. Now that doesn’t sound too terrible, but to gain the extra stop of DR you have to under expose compared to standard gamma’s, typically with S-Log you would put mid grey at 38% (use the centre spot meter on the EX1/EX3 and a grey card). This works reasonable well with this fake log picture profile. The problem however is that when grading you may find that you have to add still further gain to bring skin tones to a normal level and this will accentuate the noise. You could use something like the Neat Video plugging to reduce the noise and in this case I think this sudo Log picture profile could be handy in tricky lighting situations. The EX1R Log picture profile, to work correctly MUST be used in conjunction with -3db gain, any other gain setting and you will loose dynamic range. Again like real S-Log, 10 bit external recording is desirable, but why not play with the picture profile and try it for yourself. It is a bit experimental, I’m not convinced that the extra stop of DR is worth the noise penalty on the EX1R, but then I’m spoilt as I have an S-Log F3.

I have uploaded both the F3 and EX1R picture profiles into a single zip file that you can download below. You will need to have an account on xdcam-user.com to download them, or register for a new account first. Un-zip the package and copy the SONY folder to the root of an SxS card, so you should have both a BPAV folder and a SONY folder in the root directory. The cameras will need the latest firmware versions to load the single profile directly. In the Picture Profile menu choose an empty PP and then in the bottom PP menu chose “load”.

[downloads_box title=”F3 and EX1R Log Picture Profiles”]
F3 and EX1R Log like profiles
[/downloads_box]

Codec and Recorder Tests: Layered Photoshop File

I have created a single photoshop file with multiple layers each containing some of my codec test results. Layers include, Uncompressed DPX, Samurai ProRes, ProRes HQ and DNxHD 185 encodes from the DPX and F3 Internal Mpeg 2 @ 35Mb/s. By zooming in and turning layers on and off in Photoshop you can quickly see the differences in the quality. One thing this highlighted was some duplicated pixels in the DPX files from the Gemini. This issue has now been fixed.

[downloads_box title=”PhotoShop Codec Layers”]
Download the file here
[/downloads_box]

Sony 3D link option for PMW-F3: Quick Look.

I’ve been playing with the Sony 3D link kit for the F3. I’m still discovering some of the fine details of how and what exactly it gives you, but what I will say right now is that it really, really simplifies 3D with a pair of PMW-F3’s.

The kit comprises of a couple of SxS cards with the software keys required to activate the cameras plus a special cable that connects the two cameras together. After loading the software keys on to both cameras you get an additional menu option to activate the 3D link in one of two modes, master or slave. With one camera connected to the other and one set as master, the other as slave the first thing that happens is that the cameras will sync with each other. My initial testing indicates that this is more than just Genlock, it appears to be full shutter sync. I’ve tried a few fast shutter speeds and so far have not seen anything to suggest the shutters are not in sync, but a proper test is still needed to confirm this.

Other things that happen is that the cameras will both go in and out of record in sync and the timecode of the master camera is used by the slave. If you change the shutter on the master camera the slave also changes. I’m still learning more about the way the 3D link system works and I really like it a lot. Anyone doing a lot of 3D work should seriously consider it. It is pricey at around $5k for the kit for a pair of cameras (although it does also include a couple of SxS cards) but the stable sync, dual start/stop and small things like the sync shutter speed changes does remove another set of things to worry about during your 3D shoot.