View Full Version : FS5 - Macro blocking codec issue in 4k??
Cliff Totten December 19th, 2015, 02:07 PM So it looks like NewsShooter.com has introduced the idea that the FS5 is suffering some codec breakdown?
New tests show Sony FS5 macro-blocking and noise issues when recording 4K internally (http://www.newsshooter.com/2015/12/18/new-tests-show-sony-fs5-macro-blocking-and-noise-issues-when-recording-4k-internally/)
Has anybody else seen this yet on their FS5?. It seems that NewsShooter has reported this to Sony.
Is it possible that Sony is only using a one pass XAVC-L encoder on the FS5 instead of the usual two pass encoding process that is supposed to exist on other Sony XAVC-L/S cameras?
I hope we can all dive deeper into this and either debunk and disprove it or get Sony to investigate a solution to it.
CT
Andy Wilkinson December 19th, 2015, 03:46 PM One of their senior design engineers is on record (and on camera) stating clearly that FS5 uses two-pass encoding.
To me this is simply that the XAVC codec, when used at a lowly 100Mbps and in Long GoP format is bound to stress as soon as things "get challenging" (low light plus fast movement). Why the big surprise?
To me this is simply the price we pay with having it record at a lowly bit-rate so it can use relatively cheap media, SDXC. Sure, I would have loved it if the FS5 offered a higher bit-rate option (and better All I encoding etc...) but it is what it is.
I have always regarded the FS5 as having what I call "entry level 4K" (actually, technically QFHD until a future firmware update). We all know its only 4.2.0 8-bit so you are not going to be doing serious 4K work for National Geographica channel with a FS5, especially at 100Mbps!
I bought it primarily to use as as a highly flexible, small and lightweight yet highly ergonomic 10-Bit 4.2.2 FullHD cam with some lovely features including HFR and its variable ND. It is also ideal for manageable hand-held gimbal work. So far, it seems that the FullHD 50Mbps two-pass encoding is "sufficient" as image quality is comparable (or better) than my Canon C100 with its lowly 4.2.0 8-bit AVCHD codec at 24Mbps.
Sure, it will be interesting to see how this story pans out but I for one won't be loosing any sleep over it.
Olof Ekbergh December 19th, 2015, 05:33 PM I can show you some absolutely horrendous macro blocking from the FS7 as well, this was before I figured out how to shoot with it.
All cameras are a compromise one way or another. Any cam that uses CF cards are stuck with a low bit rate. I loved my C100 but sold it when I got my A7s. Yes I missed it, but the A7s is so nice in low light. The FS7 blow them both out of the water.
I see the FS5as a replacement for my C100 and IMHO it produces better quality then the C100. But the C100 PPs included were much better than the Sony included PPs. However that is the beauty of these new cams and modern NLEs, you can make them all look good. These days all the cams $5,000.00 and up (some less even) can produce wonderful results, but it is up to the operator and editor/colorist to make it really shine.
So far I like Cine1 and Cine4 with a few tweaks to sharpening (reduce) and black point (lower) with a bit of sat (increase), and I also use Cine 2 and 3 with some tweaks in custom mode in the 5. This works well for me and I think iso 3200 is the max I want to shoot at with the 5. Low light I use my A7s it blows everything out of the water. Mostly I use iso 1000 in the 5 and in Slog I over expose so my LumaPro is set to iso1200 and exposure is perfect for what I do.
Know what I need to keep my clients happy and myself as well and the 5 makes my life easy hiking or skiing around the back woods as I am getting older. It is also nice to be able to shoot some UHD internally, though I did not buy the 5 to be my 4K main cam, that is why I have the 7.
All this said if Sony can improve on the higher iso shaky cam style of shooting great (not part of my style), but so far I have not found anything objectionable in my style of shooting from the FS5. I believe in the 7 second rule and good lighting even if it is just good use of bouncing and flags.
Alex Leith December 20th, 2015, 04:05 PM If the A7Rii can record (consumer-grade) 4:2:0 8bit XAVC-S 100Mbit 4K *without* any noticeable macro blocking (even at higher ISOs) then the FS5 *should* be able to record (theoretically superior) 4:2:0 8bit XAVC-L 100Mbit 4K without it becoming unusable.
But it can get so bad in low light that at 6400ISO it's like a mush of blocky noise in the shadows, that stays fixed on the screen and doesn't move with the camera!
So REGARDLESS of the technical limitations of the camera relative to the FS7... If the A7Sii and A7Rii can do it, why can't the (theoretically superior) FS5 do it?
Ray Lee December 21st, 2015, 01:10 PM I haven't got the blocky mushy stuff on my fs5 in UHD, do you have samples? I would like to see if I can recreate it.
If the A7Rii can record (consumer-grade) 4:2:0 8bit XAVC-S 100Mbit 4K *without* any noticeable macro blocking (even at higher ISOs) then the FS5 *should* be able to record (theoretically superior) 4:2:0 8bit XAVC-L 100Mbit 4K without it becoming unusable.
But it can get so bad in low light that at 6400ISO it's like a mush of blocky noise in the shadows, that stays fixed on the screen and doesn't move with the camera!
So REGARDLESS of the technical limitations of the camera relative to the FS7... If the A7Sii and A7Rii can do it, why can't the (theoretically superior) FS5 do it?
Alex Leith December 23rd, 2015, 09:50 AM So this is one example of the sort of breakup you get with the codec. This is ISO 3200 @100bit UHD. As you can see there's barely any movement in the frame... I really don't think it should be showing signs of struggling to handle the image. But edges show weird blocky nonsense.
This was with Firmware 1.1.
FS5 Codec Issues on Vimeo
Andy Wilkinson December 23rd, 2015, 09:57 AM Damn! This is much worse than I'd feared - but you've posted a very good example to demo the problem (Note: I had to go to Vimeo to get it in HD but it's still obvious in low res).
Sony, we have a problem...
Jack Zhang December 23rd, 2015, 10:51 AM That looks like bad CABAC encoding. Might be similar to how Sony AVC in Sony Vegas almost never looks as good as Handbrake. Edge smearing is usually prominent in misconfigured H.264 encoders.
If the chips really are from the A7s II, it just might not be advanced enough for the XAVC extensions in the pro XAVC-L vs XAVC-S. This also does not bode well for the PXW-X70 100mbps update.
Ray Lee December 23rd, 2015, 12:21 PM I just can't get mine to do this, could it be the way you import the footage?... I use EditReady to ProRes and bring that into FCPX... also the sample seems underexposed a little maybe, could that cause it? Could certain PP gamma setting be better or worse?, I want to replicate it so I can avoid it... its a small 8 bit codec, I never expected it would cut into F65 sales :)
This is one of maybe the several attempts at getting mine to do it
first 15 sec out of camera
next 15 corrected to what I feel looked real
next raised the levels like crazy, the exact opposite of what you should ever do
last 15 seconds raised levels a little and tossed a bunch of looks on it then tweaked color again
Once on YouTube it has the issue
FS5 test 9 UHD - YouTube
Download a compressed H264 version but without YouTube artifacts here (514MB)
http://TheSBimage.com/2015/514mbFS5UHDh264mov.zip
Ray Lee December 23rd, 2015, 12:24 PM So this is one example of the sort of breakup you get with the codec. This is ISO 3200 @100bit UHD. As you can see there's barely any movement in the frame... I really don't think it should be showing signs of struggling to handle the image. But edges show weird blocky nonsense.
This was with Firmware 1.1.
FS5 Codec Issues on Vimeo (https://vimeo.com/149878142)
Is this your footage? would be helpful to know all your settings, thanks
Alex Leith December 23rd, 2015, 12:29 PM Yes indeedy that's my footage.
PP5, UHD, 25p, 100Mbit, ISO3200, F8.0, No ND, 35mm e-mount lens (also tried with Canon 24-105mm and metabones speed booster )
It shows up when there's (even a little) movement in the frame, and it's especially visible on edges. Can you try again, just gently moving the camera?
I know my shot was a little underexposed (although not terribly) but I wouldn't expect that to cause that kind of break-up in the image.
I tried EditReady too... Colours looked very slightly more vibrant, but the result was more or less the same.
One more thing – can you try focus mag at 8.0x in UHD... then switch to HD and try focus mag again. The view seems to be the same size, but on mine it looks really blurry in UHD but relatively sharp in HD.
Nate Haustein December 23rd, 2015, 12:33 PM This doesn't have anything to do with AUTO knee settings or something like that, does it?
I haven't seen this same behavior yet in my UHD footage. It definitely breaks up on movement, and creates sort of strange horizontal 'lines' in the picture, but at 100mbps, I was prepared for that when I bought it. Attached photo is mid whip-pan, obviously terrible in my living room, but shows how the XAVC-L breaks up when stressed.
Alex Leith December 23rd, 2015, 12:37 PM No change with manual knee... but interesting everything gets a lot better if I switch to SLOG3... Edge artefacts disappear completely. Unfortunately, SLOG3 grades HORRIBLY in 8bit... :-(
Nate Haustein December 23rd, 2015, 12:42 PM Alister Chapman said at his seminar to use S-Log2 with S-Gamut3.cine in UHD - just throwing out ideas right now but it couldn't be a color thing?
Ray Lee December 23rd, 2015, 12:48 PM "One more thing – can you try focus mag at 8.0x in UHD... then switch to HD and try focus mag again. The view seems to be the same size, but on mine it looks really blurry in UHD but relatively sharp in HD"
I noticed that, I actually switched to 4x because it did not really help me at 8x with it so soft... OT but this is the first camera I have ever owned or used that I did not turn the sharpness down, I think I may actually raise it up a click or two in UHD? (turned it down in HD like normal like every other camera aI have owned)
Did you download my clip? I had a good deal of underexposed out of focus background and did not have the issues.
Ray Lee December 23rd, 2015, 12:50 PM This doesn't have anything to do with AUTO knee settings or something like that, does it?
I haven't seen this same behavior yet in my UHD footage. It definitely breaks up on movement, and creates sort of strange horizontal 'lines' in the picture, but at 100mbps, I was prepared for that when I bought it. Attached photo is mid whip-pan, obviously terrible in my living room, but shows how the XAVC-L breaks up when stressed.
I watched another sample with banding in the blown highlights and I think the auto knee was 100% to blame for that. Not sure it would affect the shadow areas like in this sample would it? Wish I understood all the options and setting available in the PP menus... feel like I know just enough to be dangerous
Ray Lee December 23rd, 2015, 12:57 PM No change with manual knee... but interesting everything gets a lot better if I switch to SLOG3... Edge artefacts disappear completely. Unfortunately, SLOG3 grades HORRIBLY in 8bit... :-(
hmmm, totally guessing maybe one of you would have more info.
If its ok in SLOG but bad in other gamma settings could it be an issue with noise removal ? the camera does remove noise except in LOG correct? I have seen some really messed up footage from bad noise removal attempts in post.
Also there have been people saying they have no issue with an external recorder? I know you really shouldn't get much if any improvement (still compressed 8 bit isn't it?) I really couldn't see any difference external to ProRes or internal on my FS100
Alex Leith December 23rd, 2015, 01:26 PM Did you download my clip? I had a good deal of underexposed out of focus background and did not have the issues.
I did – that you for sharing. To see the edge-smearing problem there needs to be at least a little bit of movement in frame. Would you mind trying again with a little movement?
I'm also seeing the same effect (although the noise is less) when viewed externally (PIX-E5H)... I can't find the speed drive for it, so I haven't tried recording, but I don't see why it would get any different when I hit record on the PIX-E5H
:-( :-( :-(
Ray Lee December 23rd, 2015, 03:38 PM I used a slider for this one... should have cleaned the track but it will work 10000iso (I think its more like 5000-6400iso according to my Sekonic light meter but thats a topic for latter :) )
Download the h264 file here 320MB
http://TheSBimage.com/2015/UHDfs5test13.zip
first 30 seconds is direct from camera
next 30 edited to look like what I feel the scene looked like
the last 30 is the lifted shadow look I see on video all the time
Alex Leith December 23rd, 2015, 06:05 PM Thanks Ray,
I'm downloading the file now.
So here's my (very quick and rough) test... it does demonstrate the problem.
SLOG3 / SLOG2 / ITU work fine. All other profiles show the blocking. Noise is better when recorded externally, but the issue is not fixed by recording to an external recorder.
I've uploaded the prores file to Vimeo. Needs to be watched at very least in HD to see the problem properly.
Sony FS5 - UHD Blocky Noise Test (External Recording) on Vimeo
Alex Leith December 23rd, 2015, 06:23 PM Wow! Ray, yours looks pretty good. I can see a little hint of blocking in the top edge of the keyboard as the slider does a little grind ;-) (44 seconds). I think it might be vertical movement that the camera has a problem with.
It reminds me of the quality you used to get when mismatching the interlaced version of AVCHD on a progressive signal (or vice-versa).
Ray Lee December 23rd, 2015, 07:06 PM Thanks Ray,
I'm downloading the file now.
So here's my (very quick and rough) test... it does demonstrate the problem.
SLOG3 / SLOG2 / ITU work fine. All other profiles show the blocking. Noise is better when recorded externally, but the issue is not fixed by recording to an external recorder.
I've uploaded the prores file to Vimeo. Needs to be watched at very least in HD to see the problem properly.
Sony FS5 - UHD Blocky Noise Test (External Recording) on Vimeo (https://vimeo.com/149915169)
The best example of the issue I have seen, I went back and just used a fluid head panning nice and smooth up and down and yep, I have it clear as day. Now at lower iso I don't see it so I figured it is what it is.... the fact that it is not showing up on the LOG footage gives me hope it can be fixed via firmware.
Someone on the FS5 Facebook page asked about the lag in UHD, I assumed they had OSS on or confused rolling shutter with lag... your video makes me think its this same issue.
**The last video I posted was a REC709 profile, maybe that the difference
Alex Leith December 24th, 2015, 04:53 AM The best example of the issue I have seen, I went back and just used a fluid head panning nice and smooth up and down and yep, I have it clear as day. Now at lower iso I don't see it so I figured it is what it is.... the fact that it is not showing up on the LOG footage gives me hope it can be fixed via firmware.
I really hope it can be fixed.
I think I feel pretty disappointed that I can see this at 3200ISO. And if I'm looking hard it does show up at 1600ISO too... And I can even see hints of it at 800ISO if I'm pixel peeping.
It seems to be when recording UHD (both internally and externally over HDMI) in any of the cine profiles, vertical movements on (even slightly) underexposed edges with some contrast (worse on out-of-focus areas). The problem does not show up in LOG (although there are other issues with dealing with LOG in an 8bit signal).
Cliff Totten December 24th, 2015, 07:56 AM Uggg,...I'm three weeks away from buying my FS5 and this problem is making me feel nervous and queezy about it.
I'm hoping to see more samples out there soon. If this was an internal codec only problem, maybe I could deal with that as I intend to shoot with a PIX E5 allot. However, if this problem exists before the codec gets it than we are out of luck.
Oh man......
Olof Ekbergh December 24th, 2015, 01:24 PM I have both the FS5 and the FS7. I love both of them. The 5 is the perfect Bcam to the FS7.
I have made a few tests with the FS5 and FS7 side by side in UHD, and if exposed correctly they look identical.
I have not been able to reproduce the noisy edges others are reporting, I am not saying it is not real. I have just not seen it in my footage yet. So I think it is limited to high ISO somewhat shaky cam, possible a bit under exposed. I like so far to rate the FS5 at 1200 to 1600 in Slog (I use a light meter or an Odyssey with scopes to be exact in exposure when getting used to a camera, then I just use the histogram and zebra once I know how to expose) and 1000 in CINEgammas, it gets noisy from 3200 and up, but this can be toned down in post. For real low light I just use my A7s.
On slight glitch I have found in UHD FS5 footage is that in FCPX and in CatalystBrowse the footage on my Macs gets clipped at 100ire. This does not happen with the HD footage, so I hope this is just a meta tag in the footage that needs to be tweaked by Sony, I noticed this after I updated to 1.10 FW. It may have been there in 1.00 as well. But I am only now getting to testing the UHD and comparing to the FS7.
So I can just now say that everything I have shot so far in HD and UHD looks great and grades well for me anyway. I loved my C100 but IMHO this tiny Sony is far more usable and the codecs are far superior. The Vari ND and DigitalZoom are outstanding features.
Merry Christmas to everyone, don't eat too much, now I am off to drink some Glögg with the family.
Jack Zhang December 24th, 2015, 02:58 PM Okay, this is likely not a codec issue now that I've seen the latest video but an internal processing issue. Remember, the processors that process the incoming data from the sensor are based on DSLR designs, (low power ASICs) so any image processing will not have enough power compared to a specialized high power FPGA that the FS7 has. This is likely down to the processor is just too weak to do image processing at standard gammas, but somehow has enough power to process S-Log and pass that through no problem?
Might want to wait for the RAW update before getting any hopes up. That might be the saving grace of the camera.
Edit: Got a message from a friend saying that it's likely a poorly implemented noise reduction algorithm in the processors. Try turning Crispening to -99 or turning the Detail section of a picture profile off. (if the option is there in the picture profiles, or turn off the equivalent that controls noise reduction)
Ray Lee December 24th, 2015, 03:39 PM REC709 also has no issues, and I really tried to find them
Alex Leith December 24th, 2015, 05:03 PM Got a message from a friend saying that it's likely a poorly implemented noise reduction algorithm in the processors. Try turning Crispening to -99 or turning the Detail section of a picture profile off. (if the option is there in the picture profiles, or turn off the equivalent that controls noise reduction)
Unfortunately the sharpening controls aren't as detailed as the FS7... Turning off the detail or turning the crispening down to 0 (that's as low as it goes) doesn't seem to make any noticeable difference.
Jack Zhang December 24th, 2015, 05:22 PM Crippled firmware options... not sounding good. Looks like the only way to bypass the noise reduction is to use S-Log.
Unfortunately this is a bug in the hardware that the firmware cannot fix unless noise reduction is entirely turned off. A new revision of the ASICs with better noise reduction algorithms is required to solve the problem on standard gamma.
Alex Leith December 24th, 2015, 05:26 PM I haven't tested this myself, but someone just commented on my Vimeo clip to say that they were seeing the problem at all of the following (and above) ISO settings:
Gamma - ISO
Standard - 2500
Still - 2000
Cine1 - 2000
Cine2 - 1600
Cine3 - 2500
Cine4 - 2500
ITU709 - 2500
ITU709 (800%) - 8000
SLOG2 - 8000
SLOG3 - 8000
Alex Leith December 24th, 2015, 05:47 PM I have not been able to reproduce the noisy edges others are reporting, I am not saying it is not real. I have just not seen it in my footage yet. So I think it is limited to high ISO somewhat shaky cam, possible a bit under exposed.
Interesting... I can see it really clearly just using the focus magnification. It is most apparent on underexposed edges – although EVERY shot has areas that are underexposed (they're called shadows ;-D)... I wonder if this effects different units to different degrees.
I'm finding it frustrating now - I can see it making edges around things (like people in interviews) look a little bit more messy as they move. I didn't initially notice it... Now I can't stop seeing it. :-(
Merry Christmas to everyone, don't eat too much, now I am off to drink some Glögg with the family.
And Merry Christmas to you too. Enjoy the season and time with family and friends :-)
Jack Zhang December 24th, 2015, 06:01 PM I haven't tested this myself, but someone just commented on my Vimeo clip to say that they were seeing the problem at all of the following (and above) ISO settings:
Gamma - ISO
Standard - 2500
Still - 2000
Cine1 - 2000
Cine2 - 1600
Cine3 - 2500
Cine4 - 2500
ITU709 - 2500
ITU709 (800%) - 8000
SLOG2 - 8000
SLOG3 - 8000
This is signs this is noise reduction related. Again, the algorithms are hard coded into the hardware ASICs so the image processing ASICs have to be revised and replaced before this issue is fixed. A FPGA is re-programmable with firmware, and the FS7 has a FPGA.
Cliff Totten December 24th, 2015, 06:50 PM Crippled firmware options... not sounding good. Looks like the only way to bypass the noise reduction is to use S-Log.
Unfortunately this is a bug in the hardware that the firmware cannot fix unless noise reduction is entirely turned off. A new revision of the ASICs with better noise reduction algorithms is required to solve the problem on standard gamma.
I have done many 100% crops on UHD recordings on my RX10-II, A7s-II and X70. None of them have ever displayed these artifacts.
I don't know but I strongly suspect that the FS5 uses a similar motherboard or chipset as these guys.
I bet it's fixable in firmware
Jack Zhang December 24th, 2015, 07:57 PM ASICs are fixed function. Unless the image processor is an FPGA it cannot be modified cause it's been optimized and engineered for a specific setup. ASICs once configured for one configuration have to be re-engineered for a different configuration from the hardware level. I doubt a simple firmware fix will do. It'll be like the EX1 going to Firmware V1.20 where you send the camera in for a firmware and hardware upgrade that fixes the issue.
Hate to admit it, but the engineers at Sony cut corners in this cam to meet market demand.
Cliff Totten December 24th, 2015, 09:53 PM If Sony already has circuits and logic (A7s-II, RX10-II and X70) that can already do what the FS5 does, it would make financial sense to re purpose them into a new FS5 design. Just change the sensor and keep a motherboard design that you already buy and manufacture chips for and are well tested on other cameras.
It could be a coincidence but notice how the FS5's (firmware 1.1) monitor, HDMI and internal recording limitations are exactly identical to the RX10-II and A7S-II.
If they don't have this FS5 artifact problem on any other model, than I suspect the fundemental hardware is fine.
Again, I don't know if the FS5 is a repackaged A7s-II with different sensor or not.....just guessing here.
Jack Zhang December 25th, 2015, 05:09 AM Those were all actual DSLR sensors with sometimes better SNR ratios. But when the noise reduction algorithm was "made" for the FS5, they just figured the same ones from the DSLR would work equally well on a Super35 sensor. They're not the same. The X70 takes the sensor from the RX100, a 1'' type point and shoot stills camera, so the ASICs don't need to be changed much.
Going from a DSLR sensor to a Super35 sensor is probably what caused the fixed function ASIC to behave differently. It's like the X200's lens issue. They assumed the optics were close to the same as the old PMW-200 for a higher zoom ratio, but it's not the same calibrations.
Cliff Totten December 25th, 2015, 11:02 AM Wow.....well? Its going to be very interesting to see how this plays out:
Will more samples come forward with this same problem?
Will Sony become aware of this at some point?
If they can't fix it, how will it affect future FS5 sales?
Could Sony issue a FS5 recall to save the product?
Oh man....I'm still hoping for the best: Either no more reports of the problem or a Sony firmware fix.
Maybe you are right....I just hope not.
Noa Put December 25th, 2015, 12:18 PM Could Sony issue a FS5 recall to save the product?
The turkey on my plate has as much hope not being eaten as your hope the fs5 might be recalled. With a bit of luck Sony might provide a firmware fix, if this is possible at all and otherwise will make you happy with a more expensive mark 2 version in a year that doesn't have this problem. I could swear the turkey in front of me just nodded so he agrees. Happy Christmas all :)
Olof Ekbergh December 25th, 2015, 12:39 PM Shot using Cine4 Gamma iso 3200 UHD factory settings for detail.
This is on a HD timeline, but I blow it up first 200% then 400%.
Interesting is that viewed on my iMac screen I see all kinds of strange aliasing etc when played back full screen, but on my eval monitor SDI out it is very clean. I think maybe the problem some are seeing is how they are viewing it. You need to see it on on a Pro evaluation monitor. On my MBP I get some strange aliasing as well but I think that is the GPU coping badly with the 100mb/s UHD Long GOP.
Or maybe I got lucky with my FS5. Bottom line is I am not seeing anything terrible. Remember this is 4X HD in other words it is like a 420 8bit HD stream at 25mb/s how good can it possible be, something has to give. And I think of iso 3200 as pretty high iso. EX1 and EX3 cams are only good to about 1600 before getting noisy and the is at 35mb/s.
This is with the kit lens, no image stabilize and all the way wide, slightly out of focus.
FS5 noise test on Vimeo
Jack Zhang December 25th, 2015, 09:31 PM With a bit of luck Sony might provide a firmware fix, if this is possible at all and otherwise will make you happy with a more expensive mark 2 version in a year that doesn't have this problem.
Most likely they'll be fixed with the Mark II version with UHS-II and 4K60p. It'll still be 8-bit though. If Sony really wanted to be greedy, they'd lock the RAW capabilities to the Mark II. I have severe doubts the RAW upgrade will be in the current FS5 if no new hardware is added. (Like the FS700)
Also, Mark II would solve the EVF issue IF THEY BASED IT OFF A CAMCORDER DESIGN, NOT A DSLR DESIGN. (Bad Sony! Bad!)
Cliff Totten December 26th, 2015, 08:03 AM "Also, Mark II would solve the EVF issue IF THEY BASED IT OFF A CAMCORDER DESIGN, NOT A DSLR DESIGN. (Bad Sony! Bad!)"
What do you mean, camcorder design or DSLR design?
The electronic principles are the same. Raw sensor readout, deBayer and gama/color and noise reduction processing and codec encoding. Aside from programmed time recording restrictions, the same circuits will be used for either DSLR or Camcorder.
Yes, sensor noise will be different when you change that out. Noise reduction algorithms will be different but everything else will be the same.
A DSLR is just a software crippled camcorder in a photo camera body. (A photo camera body that is less capable of handling heat too)
Jack Zhang December 26th, 2015, 12:02 PM In the case of the A7s II, they have to do that so as to not overuse power or overheat the chips. It's not simply the case of swap the heatsinks and it should work, the chips were engineered with limitations in mind. Camcorders have more FPGAs than ASICs, and those can handle way more heat and load over time.
Only active cooling in the FS5 is the Sensor. On the FS7 I believe it's both on the main FPGA and the Sensor.
Noa Put December 26th, 2015, 12:15 PM Just curious, how come you know so much about this? Are you an engineer at a cameracompany?
Cliff Totten December 26th, 2015, 05:25 PM I would really hope we can get more FS5 sample uploads here. I'm ready to pull the trigger and buy the FS5 next month and I really want to see if this is a real problem or not.
Could there be some models out there with bad sensors and others that are fine? (A simple quality control issue)
If you purchased your FS5 within the last 30 days, you seriously might want to inspect closely what your camera is recording around the edges. Maybe you have a unit that has this problem?
If anybody does see something, please upload some samples! Sony understandably is not going to lift a finger to do anything if they only see two or three bad videos out there.
CT
Jack Zhang December 26th, 2015, 07:09 PM Just curious, how come you know so much about this? Are you an engineer at a cameracompany?
My friend is working on his own portable high speed camera to rival the Phantom. If you're an engineer, you should check out "tesla500" on YouTube.
Olof Ekbergh December 27th, 2015, 04:46 PM Just did extensive testing on the FS5 trying to reproduce the problems some are seeing.
Here is a long boring video but if you are investigating this you may want to watch. You can download the original UHD file loaded to Vimeo but that is not the same as original it is compressed about ten fold. Contact me directly if you want the direct out of camera file or the simultaneously recorded Odyssey O7Q+ file I will provide those if you like.
This is only the internal UHD file, frankly I see almost no difference in the ProRes version from the 07Q+.
The profile I used is a new one I made, partly because I am having some issues with FCPX clipping above 100ire. So I used CINE2 in this test, I do state in the video it is CINE1 and that is wrong. Though CINE one is very similar it allows up to 109ire.
Anyway this shows iso 800,1600 and 3200 with some custom settings in my PP3
These settings are:
Gamma CINE2
Black Gamma: Middle -4
Sat: +4
Color: Cinema
Detail: -7
Everything else standard PP3 settings.
The interesting thing is the Odyssey file is the same on the scopes and noise vise there is very little difference. If you want to see the O7Q+ file contact me, the vimeo compression will delete any difference.
I am viewing this both on a TOL iMac and an Eval Monitor, not a TV the Eval monitor is a professional piece of equipment and it shows a very true picture and the video is fed by a SDI out from my UltraStudio for Thunderbolt.
FS5 Testing UHD internal on Vimeo
Cliff Totten December 28th, 2015, 01:45 PM I downloaded this video and didn't notice anything strange around edges when viewing on a 1080 monitor.
However, when I did a zoom/crop and looked at it with a 1:1 pixel view. I DID then notice some of this odd "splotchy" edging.
I really can't say for sure, but could this be a result of some kind of line skipping at the sensor read-out stage? I could be way off but I don't feel like is a compression problem at all. Could it be a strange, improper RGB de-Bayering problem?
Weird. I have never seen this on my A7s-II, RX10-II or X70. Not even at 400% zoom/crop.
What is the FS5 doing that it's cousins are not?
EDIT - I have been looking at this sample a little closer and have noticed that the artifacts seem to occur stronger on horizontal lines much more than vertical lines. I'm uploading a zoomed in clip from your file at 1/4 speed.
What shutter speed did you use for this? Was it 1/30 second? It seems that the shutter speed and motion blur are helping to cover this artifact a bit. Can you try another shot at a very high shutter speed? I think if you take it above 1/200 or something like that, it will make these ugly edges pop right out in a slow motion inspection.
https://drive.google.com/file/d/0B60OfAOTK8ZtdUpHNHkwbXZGZzQ/view?usp=sharing
Check the slow motion video above. again, a faster shutter speed will reveal more.
CT
Jack Zhang December 30th, 2015, 09:10 AM It doesn't have anything to do with line skipping or de-bayering. It has to do with a mis-tuned noise reduction algorithm, which is hard coded to one of the onboard ASICs.
Bayer artifacts and line skipping tend to show the "rainbow effect" but the examples I've seen (primarily Alex's) show less rainbow effects on hard edges and more "ghosting" and loss of detail, more prominent with a poorly tuned noise reduction algorithm.
Also, protip, turn off "Smart Resampling" in Sony Vegas when you slow motion a clip. That eliminates the blur.
Cliff Totten December 30th, 2015, 10:31 AM Thanks for the tip in Vegas. I had completely forgotten about that! ;-)
Do you believe that the code that is hard programmed into these ASICs chips have no ability to be adjusted at all? I mean, there is no code that Sony placed there that says:
"Apply this ASICs FIXED noise reduction pattern or calculation at a strength of X or Y",...lets say 1-10 variable amounts.
Forcing all the video through a completely fixed and hardware locked circuit path would be a very cheesy way of doing it. I have to believe (hope) that there has to be some kind hardware programmed code that allows for some minimal degree of noise reduction variability. (like "color saturation", "shapness" or "contrast" 1 thru 100)
I can see the noise pattern characteristic being permanently "baked" in the chip. If Sony cant change the noise pattern, couldn't they at least "ramp-down" how much of that reduction algorithm is actually applied in final processing? (and possibly reducing the problem?)
That Tesla500 channel is awesome. Really sweet stuff there.
Have you been able to see the inside of the FS5 yet? Or, have you been able to identify the chips the FS5 uses from technical documents?
I really hope we all can shed a larger spotlight on this in the weeks to come.
I might be holding off on my FS5 purchase after all! :-(
CT
Edit: The A7S-II got a firmware update this month. One of the Sony stated fixes was "image quality improvements". I did a few short (original firmware) recordings with the lens cap on to record all black and different ISO settings. I then upgraded to the new firmware and did the same recordings with the lens cap on. I DID notice some slight noise changes between the firmware versions. When I heavily stretched the contrast in both blacks upward, I saw that the original firmware had very faint "horizontal" noise "lines" while the pattern with the new one was more "dot and spot" looking.
Could they have modified the noise reduction process between the two firmware?
Josh Swan December 30th, 2015, 07:22 PM I just did some testing with my fs5 on a large tv and I found three gammas that look excellent in 4K. Rec709 800%, slog2 and slog 3. In my opinion this will cover most any shooting situation with these. They looked completely usable to about 6400 and pretty decent even up to 10,000. The other gammas, the blocky mess on some edges is very apparent, but completely gone on the three mentioned above. Worth taking a look for yourselves, but I feel much better knowing there is a gamma outside of slog that looks great at 4K.
|
|