PXW-X70 Firmware V3.0

Upgrade is free of charge and available in 1Q of FY 2016 (TBC)

New Features & Improvements:
● Support for 100 Mbps QFHD recording with XAVC format*

PXW-X70 V3.0 firmware will be downloadable here in 1Q of FY 2016.

(*)PXW-X70 will need to be 4K capable, which requires a paid upgrade with CBKZ-X70FX, in order to record in 100Mbps QFHD XAVC.

Excellent - look forward to seeing how much the picture improves!

Yay, 100Mbps! :-D

Maybe the new codec will not hard clip at 100IRE in FCPX like the 60Mbps one. :-|

Paul

Excellent - look forward to seeing how much the picture improves!

I don't see any difference between the 60 and 100mb codec on my ax100, unless I"m shooting a scene with small details and with motion which is normal as it stresses the codec more.

I tend to have access to live flames a lot, and they tend to pixelate quite badly on the 60Mbps codec. I also shoot 30-33db a lot, and with the black gamma stretch the 60Mbps codec macro blocks badly as well in dark areas. :)

Paul

That's what I meant with scenes that stress the codec and that's where you should see an improvement but in most cases you really have to pixel peep to see any difference at all.

Much needed to make it up to par with the consumer model.

Now to wait for fast enough ASICs for a low power draw XAVC-S 4K 60p encoding.

I've heard there are some under the hood things about Sony's in camera encoder that separate it from the "consumer" equivalent at the same data. One is that the x70 (and other pro cameras) are doing a two pass encode which improves quality. Two pass "live" encoding means that the buffered shot is being gone through twice to better allocate the bits.

I think some users have noted a very short delay between stopping the camera and the 4K recording finishing the write to card. If that's the case it may be due to the second pass.

BTW that could explain why the AX100 had 100mbps and the X70 60mbps at 4k. The x70 is doing a second pass. This also would make it harder to push the higher data rate. This might also explain why 4k was not ready at the time the x70 was released.

I've had conflicting thoughts whether they based the FS5 off of the X70 or the A7SII's encoder chips. The menu looks like the X70, but other people believe it's the A7SII's encoder modules.

Much needed to make it up to par with the consumer model.

Now to wait for fast enough ASICs for a low power draw XAVC-S 4K 60p encoding.
application-specific integrated circuit - that's hardware right?

Yes. The more specific term for encoder and decoder chips.

Current ones that have high efficiency (The only ones that work in cameras without the need for fan cooling) do not have enough power to encode 4K 60p.

so its nothing to look forward to with my ax100

The menu looks like the X70, but other people believe it's the A7SII's encoder modules.

Spent a day with the FS5 today and would go as far as to say it's like a big X70 but with colorimetry almost identical to the FS700, far more so than the A7S models we have worked on. In fact side by side the FS700 and FS5 were very hard to tell apart.

After running the FS5 up on the charts and scopes we ended up getting the best "in camera look" out of PP4 with the knee set to 85 and the slope to -3. This gave us the best looking highlight roll off. If you don't push peak white exposures above 90IRE it will look quite natural on the way up to just over 102IRE. Much prefer the manual knee adjustment because if you watch the behaviour of the auto knee in rapidly changing lighting conditions such as stage strobe or rapid pulse lighting you can actually observe the auto knee pumping trying to compensate and keep up.

Also dropped the color depth on red to +3 (the adjustment is counter intuitive.) Then finally adjusted color phase to -3. These last two adjustments are totally subjective but for my eyes tone down the rather reddish magenta hues in skin tones that some of the Sony cameras tend to exhibit.

One observation a couple of us made was that the center crop mode is a definitely a little noisier and softer. Overall though quite impressed with the package so far.

Chris Young
CYV Productions
Sydney

Chis, would you say that 85% -3 knee would work well for pushing out 100IRE direct to air clips, without requiring bringing highlights down 8% in post? I am thinking that anything over 102% with your settings would appear rather blown out, and that an on air clip at 100% would not even be a noticeable change to a viewer.

Also, I have a new REC709 PP4 color correction calibration. I have been having the damndest time getting the red as perfect as possible. With an almost 99% subject matter having high intensity red LED and halogen lights, I am still seeing scenes where one red will be orange, while the one next to it is magenta! I would like to message you the settings if you could side-by-side them with my old settings and whatever you are happy with, and maybe provide some feedback or modifications, especially pertaining to color depths.

As for the 4K encoding, my X70 continues to REC for at least a full second after I hit stop. It became noticeable when in post every clip ended with a second of video of the camera swinging away from the shot. And I think Sony did a very good job with the processing in the camera, compared to other models. The simple fact that i can spit out ANY HDMI/HD-SDI resolution despite the recording format is something no other Sony camera i know of can do. This means dedicated processing to accomplish this, which is strange because more ENG specific models don't even have this, much less 4K.

Paul

I've heard there are some under the hood things about Sony's in camera encoder that separate it from the "consumer" equivalent at the same data. One is that the x70 (and other pro cameras) are doing a two pass encode which improves quality. Two pass "live" encoding means that the buffered shot is being gone through twice to better allocate the bits.

I think some users have noted a very short delay between stopping the camera and the 4K recording finishing the write to card. If that's the case it may be due to the second pass.

BTW that could explain why the AX100 had 100mbps and the X70 60mbps at 4k. The x70 is doing a second pass. This also would make it harder to push the higher data rate. This might also explain why 4k was not ready at the time the x70 was released.

Although I can't say for sure, I strongly suspect that two pass encoding is a feature that Sony has in all its XAVC flavors, consumer and pro models. The A7s-II also keeps writing to its SD card after you hit stop. It's the same amount of time as the X70 does.

I was told at NAB that the .mxf wrapper creates a little more overhead to pack and unpack versus the .mp4 wrapper. I don't know about this myself but I suppose it's possible.

For me, .mp4 XAVC (long GOP 100 or 60) does playback and edit smoother than .mxf does on my laptop and desk PC.

THANK YOU SONY FOR BRINGING THE X70 UP TO THE CODEC STANDARDS OF THE REST OF YOUR PRODUCT FLEET!!!

Bravo!!!

would you say that 85% -3 knee would work well for pushing out 100IRE direct to air clips7

Not with the X70 Paul. What I was referring to were the specific settings for the FS5 to suit what I find best for my shooting. I am finding all these later model single chip Sony cameras tend to behave in a distinctly different manner to what we have been used to in the previous three chip cameras we have been used to.

One has to remember in Rec 709 that this is only a six stop dynamic range to suit the 99% of displays out there that are only capable of handling a six stop range. Basically 0 IRE is Black and pure white reflectance sits at around 90 IRE. 0 to 90 IRE represents five stops. From 90 to 100 IRE is the last stop in which anything over a white reflectance level has to be dealt with if we have to comply with the broadcasters 100 IRE requirement. The problem is that pure light sources, like you run into in ENG work at night especially when police, rescue and fire guys are lighting emergency scenes, exceed 90 IRE.

Somehow we have to contain these direct light sources and any other source over 90 IRE, such as clouds on a bright sunny day, reflections off water etc that can generate light levels well in excess of the six stop range we have to play with. The trick is to compress these extra stops into the last stop of dynamic range we have left in the Rec 709 range between 90 and 100 IRE.

In the case of the FS5 I found that by starting to run compression on the signal with a knee at 85 IRE just below pure white reflectance of 90 IRE and setting a slope of -3 gave me the most natural looking smoothest roll of into the highlights above 90 IRE. Remember - values are a gentler slope + values are a steeper slope. + 5 on these cameras is NO slope at all so no rolling off the highlights with gentle compression.

Using the 85 and -3 settings the peaks of the signal would then reach around 102 IRE which we can get away with on transmission links that will limit at 100 IRE. The BIG PROVISO here is that you MUST keep your maximum white level reflectance exposure to no more than 85-90 IRE. If you push above 90 IRE on that white car, shirt, building or if there is no white in the scene the brightest part of the scene that is important to capture you will start to blow out the image as it goes into compression.

I think I saw a comment by Jody Eldrige, and I think he was saying that he was not happy with the harsh highlight handling of his FS5. Total agreement here but it's not just with the FS5. I'm finding it with all Sony's bigger sensor cameras. They have to be set up correctly to get the best out of them in Rec 709 and exposed correctly to avoid those nasty highlight issues.

Some of you may have downloaded a test clip I did a year or so ago when trying to squeeze the max out of an X70. If you have that clip stick it on a timeline and look at your waveform monitor. No hard clipping, detail in 99% of the highlights and all looking quite natural. If anyone wants to look at the levels in that clip it is still up there in cyberspace at:

www.sendspace.com/pro/dl/t9yiwl

My suggestions won't work for everyone that's for sure but for the way I shoot and expose the results I am getting are the best that I can squeeze out of Rec 709 in various cameras such as the FS700, X70 and now the FS5. Obviously different settings for all of them.

Chris Young
CYV Productions
Sydney

Chris, REC709 gamma settings in the X70 produce in camera 6 stops, or REC709 when broadcast limits things to 6 stops?

Paul

Generally speaking, a good looking rec709 video can only hold 6 around stops of dynamic range. With the knee setup and lifted backs, you might squeeze another stop to make it 7.

Remember, the X70 does not have an SLOG-2 option the way it's little RX10-II cousin does. With gamma compression, (SLOG-2) that same RX10-II / PXW-X70 1 inch sensor literally matched the same 12 stop dynamic range that my A7s did in my tests!!!

So yeah,...that little 1 inch sensor can do 12 stops but unlike the RX10-II, the X70 is trapped inside rec709's 6-7 stop recording limitation.

If Sony charged $500 for a new SLOG-2 firmware option on the X70, I'd jump on that in a heartbeat. Getting SLOG-II and 12 stops on the X70?....wow! That would be WELL worth it in my eyes.

SLOG-2 eliminates any and all highlight problems on the RX10-II. Anybody else willing to shell out money for SLOG-2 optional upgrade?

Chris, REC709 gamma settings in the X70 produce in camera 6 stops, or REC709 when broadcast limits things to 6 stops?

Paul

As Cliff correctly states Rec 709 is designed around six stops as that was the minimum required for displays to approximate what we could see with our eyes. This standard was laid down years ago but it's the one we still have to work to. Rec709 can go to 109IRE but it's still a six stop range. It's just that we have to compress that back down to 100IRE for broadcasting.

A crude analogy would be that we have a straight linear wound 11" spring and we need to fit it into a 10" tube. Rather than cut 1" off it we compress it to fit it into the 10" tube. But we are going to be a bit tricky here and just compress the last three inches of the spring to get it to fit into the 10" tube. So the 8" point on the spring is equivalent to your video knee and the last three inches of spring compressed into the tube would equate to the Slope of the signal above the video Knee. So basically linear no compression up to 8" and then progressive compression from 8" to 10" with 3" compressed in that range. The spring is Rec 709. We can leave it at 11" or 10". In video terms for broadcast it's still six stops but scaled down to fit into 0-100IRE.

A lot of earlier video cameras had knee points up as high as 95IRE so the signal was good up to that point then suddenly all detail and color got blown out from 95IRE to 100IRE. Hence that horribly clipped look to early video camera highlights.

The other problem is heavy compression above the knee can create strange false colors to appear in the highlights. The reason for this is that the green, red and blue channels do not compress with an identical ratio to one another above the knee point. The higher that knee point is the more savage and unequal that compression becomes in the color channels and the worse the highlights look.

For me I find the lower knee points, mid to high 80s, but well above skin tones, with a softer - negative Slope compression gives me a more pleasant natural looking higher level roll-off to footage shot with most of the modern sensors when shooting Rec 709.

Cliff the problem I have with 8-bit S-LOG is that trying to fit 11-14 stops into the available 50-mbit data rate means you are compressing the signal fairly hard. You still only have the same amount of data storage but you are now trying to fit at least twice the amount of sensor data into the bin. This in tun shows up as more compression noise when you go to grade the footage. Especially as to get that sensor dynamic range captured you are shooting at 2000-3200 ISO on many of the S-LOG Sony cameras. Over exposing by around 1.5 to 2 stops so that your grading is generally downwards which will help compress that noise is really not that great a solution when you only have a 255 level 8-bit bin to start with. Below middle grey in 8-bit LOG you are grading with rapidly diminishing amounts of data per stop.

I've all but given up after nearly four years of squeezing the max out of 8-bit S-LOG. 10 bit with 1023 levels well 64-960 with respect to 0 and 100IRE respectively on the SMPTE 259 standard is a different story. 10-bit S-LOG with four times the data levels is infinitely better to work with. If I need to grade to the max it has to be 10-bit. Most of my client work though is more than happily satisfied with good old well exposed and lightly tweaked 8 or 10-bit 709.

Blow up the attached image and you really start to see the short fall of the 8-bit capture when dealing with fine gradation and this can happen easily in blue skies in 8-bit S-LOG where you only have about 15 shades per stop to grade with in the higher exposure levels. Your 10-bit S-LOG is closer to 65-70 shades per stop and that is a massive difference when grading. Then again to get the best out of a 10-bit grade you are not going to be getting the best out of it if grading in a 16-bit environment. You really need 32-bit depth and that means DaVinci Resolve or even, yes, even Vegas Pro 13 where you can set up your bit depth at the start of a project to be 32-bit video levels or 32-bit full range.

Chris Young
CYV Productions
Sydney

It's interesting. As we all know, 8bit sampling only offers 256 shades of color for each channel and 10bit offers 1024 shades. On paper and by all rights, SLOG should look HORRIBLE after it's stretched back out and "normalized" from an 8bit sample. Let's fact it, 8bit sampling was just not designed to be bent and stretched like this.

However, strangely enough, depending on the type of scene that is being shot, I have actually gotten decent results with SLOG-2 in XAVC-S at 100mbps. Going further, that same SLOG-2 into 8bit ProRes seems to stretch out with even better results.

SLOG-3? Woof,....even that best 8bit codec seems to suffer banding and stepping artifacts. In bright scenes, it's somewhat usable but will show signs of stress quickly.

So, all in all, the 8bit "numbers" look bad on paper but from my tests, SLOG-2 is mysteriously "OK" for allot of scene types in 8bit....and with a good ProRes recorder, it's noticeably better.

looks like second quarter now!

PXW-X70 Firmware V3.0

Upgrade is free of charge and will be available during April-June 2016 (TBC).

New Features & Improvements:
● Support for 100 Mbps QFHD recording with XAVC format*

PXW-X70 V3.0 firmware will be downloadable here during April-June 2016 (TBC).

(*)PXW-X70 will need to be 4K capable, which requires a paid upgrade with CBKZ-X70FX, in order to record in 100Mbps QFHD XAVC.