Can the GH5 Capture 4K 24p and 60p Simultaneously!?
Photo Moment - April 20, 2017
Is it possible to record slow motion 60p and regular speed 24p UHD/4K footage at the same time? How about 1080p 24 and 1080p 180fps? With the GH5 and an Atomos recorder… yes :-)
6:50 — Skip to the actual show start
PRODUCTS MENTIONED IN TODAY'S PHOTO MOMENT (MANY ARE AFFILIATE LINKS… affiliate links make me a little bit of $ but cost you nothing more!):
•• Atomos Ninja Inferno ••
B&H - https://bhpho.to/2otybDV
Amazon - http://amzn.to/2opnaS8
Adorama - http://jal.bz/2o73z7M
•• Manfrotto 496RC2 Compact Ball Head w/ Quick Release ••
B&H - https://bhpho.to/2n45tJf
Amazon - http://amzn.to/2mNWycS
Adorama - http://jal.bz/2n3Z5ls
Apologies, Joseph, for using your forum for this message, but it is important enough for anybody interested in or using GH5.
Panasonic is planning to add 400mbps ALL-I video encoding for GH5 this summer, but that does not raise the bar of image quality for the videos recorded by the camera as ALL-I has much higher requirements for the bitrate (compared to IPB) for an equivalent image quality. I've created a petition for Panasonic to add support for high bitrate IPB modes into GH5.
If you like that idea, please sign the petition: https://www.change.org/p/panasonic-introduce-4k-4-2-2-ipb-400mbps-and-fhd-4-2-2-ipb-200mbps-modes-for-panasonic-gh5?recruiter=44141647&utm_source=share_petition&utm_medium=copylink
Panasonic is planning to add 400mbps ALL-I video encoding for GH5 this summer, but that does not raise the bar of image quality for the videos recorded by the camera as ALL-I has much higher requirements for the bitrate (compared to IPB) for an equivalent image quality. I've created a petition for Panasonic to add support for high bitrate IPB modes into GH5.
If you like that idea, please sign the petition: https://www.change.org/p/panasonic-introduce-4k-4-2-2-ipb-400mbps-and-fhd-4-2-2-ipb-200mbps-modes-for-panasonic-gh5?recruiter=44141647&utm_source=share_petition&utm_medium=copylink
Thank you so much on clearing that up! It makes much more sense now and the way you explained it is very clear specially on how the IPB can be better suited than ALL-I. This information is very good to have and I hope your petition gets noticed and approved!
TL;DR version
--------------------
Yes, IPB handles fast movements poorly (compared to ALL-I), but we need both of the highest bitrates that can be handled by the camera within the thermal envelope. Give both and let the operator decide which one suits the subject more.
Long version
------------------
ALL-I codec compresses the bitmap of each frame separately. If we take 4K 24p 10bit 4:2:2 for example, each "I" frame would roughly be 3840 x 2160 x 24 x (10 [Y'] + 5 [Cb] + 5 [Cr]) = 3796.875 Mbps. Thats the raw stream with chroma subsampling (4:2:2) applied.
Now we have to compress it further using two methods - intraframe (ALL-I) or intra+interframe (LongGOP/ IPB).
For ALL-I case the only way to compress it is to apply DCT and entropy encoding, which requires quite a bit of bitrate for a decent quality. The higher is the compression rate here, the more prominent the macro blocks are. The two obvious benefits of this compression are the ability to capture very quick movements (those do not require any extra bandwidth/processing compared to static ones) and lower computational complexity during video editing in post (no need to recreate frames using information from the neighboring frames, like it happens for IPB encoding).
For LongGOP/IPB case, the same approach is used for "I" frames, but those are emitted only once in a while (typically - half the framerate, every 12-th frame in our case). Another approach is used for the remaining 11 frames of each 12-frame period. The encoder divides the frame into macroblocks and for each of the macroblocks tries to find its previous location in the preceding frame. If it succeeds, it records only the information about how the macroblock has moved (plus some bitmap error correction data as moving objects ca slightly change their shape/color/luminosity). If it fails, it encodes the macroblock as bitmap (just like for "I" frame case).
Thus, for the cases, when there is no rapid movement in the frame (that causes the encoder not to find previous locations for macroblocks and thus requiring more bandwidth due to the emitted bitmaps), IPB encoding gives better image quality for the same bitrate. It has much higher computational complexity in both - encoding and decoding. Raising the quality for IPB encoding would mean dropping less bits in DCT (not adding much for processing) plus looking for moving macroblocks in wider areas (adds a lot and is memory bandwidth bound).
To summarize, ALL-I encoding handles fast movements much better, but requires higher bitrates for equivalent image quality (compared to IPB), while IPB encoding handles less dynamic streams much better compression rate wise, allowing higher image quality for the same bandwidth (compared to ALL-I).
What my petition is about is giving both tools to people and letting them choosing, whether they want higher IQ at the expense of inability to keep the image quality (IQ) at that level in case of rapid movements, or they want to have constant IQ that is lower, but handles rapid movements with no IQ degradation.
IPB at 150 Mbps gives IQ equvalent to 400 Mbps ALL-I (if we have no rapid movement in the frame). By implementing 400 Mbps IPB Panasonic would give us an ability to get maximum IQ out of the camera (for some subjects).
--------------------
Yes, IPB handles fast movements poorly (compared to ALL-I), but we need both of the highest bitrates that can be handled by the camera within the thermal envelope. Give both and let the operator decide which one suits the subject more.
Long version
------------------
ALL-I codec compresses the bitmap of each frame separately. If we take 4K 24p 10bit 4:2:2 for example, each "I" frame would roughly be 3840 x 2160 x 24 x (10 [Y'] + 5 [Cb] + 5 [Cr]) = 3796.875 Mbps. Thats the raw stream with chroma subsampling (4:2:2) applied.
Now we have to compress it further using two methods - intraframe (ALL-I) or intra+interframe (LongGOP/ IPB).
For ALL-I case the only way to compress it is to apply DCT and entropy encoding, which requires quite a bit of bitrate for a decent quality. The higher is the compression rate here, the more prominent the macro blocks are. The two obvious benefits of this compression are the ability to capture very quick movements (those do not require any extra bandwidth/processing compared to static ones) and lower computational complexity during video editing in post (no need to recreate frames using information from the neighboring frames, like it happens for IPB encoding).
For LongGOP/IPB case, the same approach is used for "I" frames, but those are emitted only once in a while (typically - half the framerate, every 12-th frame in our case). Another approach is used for the remaining 11 frames of each 12-frame period. The encoder divides the frame into macroblocks and for each of the macroblocks tries to find its previous location in the preceding frame. If it succeeds, it records only the information about how the macroblock has moved (plus some bitmap error correction data as moving objects ca slightly change their shape/color/luminosity). If it fails, it encodes the macroblock as bitmap (just like for "I" frame case).
Thus, for the cases, when there is no rapid movement in the frame (that causes the encoder not to find previous locations for macroblocks and thus requiring more bandwidth due to the emitted bitmaps), IPB encoding gives better image quality for the same bitrate. It has much higher computational complexity in both - encoding and decoding. Raising the quality for IPB encoding would mean dropping less bits in DCT (not adding much for processing) plus looking for moving macroblocks in wider areas (adds a lot and is memory bandwidth bound).
To summarize, ALL-I encoding handles fast movements much better, but requires higher bitrates for equivalent image quality (compared to IPB), while IPB encoding handles less dynamic streams much better compression rate wise, allowing higher image quality for the same bandwidth (compared to ALL-I).
What my petition is about is giving both tools to people and letting them choosing, whether they want higher IQ at the expense of inability to keep the image quality (IQ) at that level in case of rapid movements, or they want to have constant IQ that is lower, but handles rapid movements with no IQ degradation.
IPB at 150 Mbps gives IQ equvalent to 400 Mbps ALL-I (if we have no rapid movement in the frame). By implementing 400 Mbps IPB Panasonic would give us an ability to get maximum IQ out of the camera (for some subjects).
Correct me if I'm wrong but, this is confusing. The point of IPB Long GOP is higher compression. Why would you raise the IPB Bitrate? What benefit would that be if you could just shoot ALL-I at 400mbps? Yeah you'll get better quality Long GOP but it wont be as effecient.
400mbps at 24p ALL-I = 16mbit per frame
In IPB GOP You're breaking down the 400mbps to where every 3 frames the compression ratio of each frame roughly translates to:
I = 30mbits per frame
P = 15mbits per frame
B = 5 mbits per frame
multiply by 8 to get 24p IPB 400mbit. So you're resolving 8 I frames at 30mbit each but the rest are diluted to 8 P frames at 15 mbit each and 8 B frames at 5mbits each if my calculation and ratios are correct. (Even if it isnt, the breakdown for each I P B frame bitrate would be similar)
Wouldnt an equal spread of 16mbit frames throughout the entire 1 second be better than that compression ratio?
ALL-I is better for motion anyways as IPB Long GOP is preferred for less motion sequences due to the compression and calculation for each P and B frame.
400mbps at 24p ALL-I = 16mbit per frame
In IPB GOP You're breaking down the 400mbps to where every 3 frames the compression ratio of each frame roughly translates to:
I = 30mbits per frame
P = 15mbits per frame
B = 5 mbits per frame
multiply by 8 to get 24p IPB 400mbit. So you're resolving 8 I frames at 30mbit each but the rest are diluted to 8 P frames at 15 mbit each and 8 B frames at 5mbits each if my calculation and ratios are correct. (Even if it isnt, the breakdown for each I P B frame bitrate would be similar)
Wouldnt an equal spread of 16mbit frames throughout the entire 1 second be better than that compression ratio?
ALL-I is better for motion anyways as IPB Long GOP is preferred for less motion sequences due to the compression and calculation for each P and B frame.
Certainly would like to see a comparison of with and without XLR. Also, the best settings. Thanks for all your informative posts.
Great video :)
Just got my GH5 and I love it.... wanted to ask you what SD card will I need when they 400Mbps firmware update will come?
Will this card be enough?
https://www.bhphotovideo.com/c/product/1275616-REG/sandisk_sdsdxxg_128g_ancin_extremepro_sdhc_128gb.html?sts=pi
Or will I need a faster one?
Just got my GH5 and I love it.... wanted to ask you what SD card will I need when they 400Mbps firmware update will come?
Will this card be enough?
https://www.bhphotovideo.com/c/product/1275616-REG/sandisk_sdsdxxg_128g_ancin_extremepro_sdhc_128gb.html?sts=pi
Or will I need a faster one?
+Olivier Braine Going by the numbers, 400 mbps is 50 MBps https://www.google.com/search?q=400+mbps+to+MBps — so by the numbers either should be fine. I simply don't know enough about these new cards to understand everything on them yet sorry.
Thanks for getting back, I appreciate it. The link is supposed to show "SanDisk 128GB Extreme PRO UHS-I"
Max Read Speed: 95 MB/s
Max Write Speed: 90 MB/s
From what I understand it is the write speed that is important. I'm asking because most UHS-II (2 rows of pins) actually have slower writing speed than 90 MB/s. Lexar UHS-II card has for example only 80 MB/S write speed (and 150 MB/s read speed - which actually does not matter as much)
Basically what I would like to know, what write speed is required after the firmare update for 400 mbps 4K modes? Would appreciate if you could check it up if possible.... I don't want to buy wrong cards.... Thanks
Max Read Speed: 95 MB/s
Max Write Speed: 90 MB/s
From what I understand it is the write speed that is important. I'm asking because most UHS-II (2 rows of pins) actually have slower writing speed than 90 MB/s. Lexar UHS-II card has for example only 80 MB/S write speed (and 150 MB/s read speed - which actually does not matter as much)
Basically what I would like to know, what write speed is required after the firmare update for 400 mbps 4K modes? Would appreciate if you could check it up if possible.... I don't want to buy wrong cards.... Thanks
+Olivier Braine I can't actually click that link on my phone (youtube comment system leaves a lot to be desired) but as I understand, anything with the second row of pins will be fast enough.
Accuracy, Brevity and Clarity need to be balanced. Brevity is very important for a YouTube audience, so try to keep the blabbering and needless chatter to a minimum.
That aside, going Live and providing interesting content in an easy to understand way is very much appreciated.
That aside, going Live and providing interesting content in an easy to understand way is very much appreciated.
...but what about shutter variance between HDMI and internal capture? Sounds like you'll get odd motion blur on either or.
I hear ya… in my initial tests I didn't see anything but I will do a proper test on this.
Quick question, is there a cord we can buy from Panasonic that allows us to mooing our new xlr adapter in another placer other than the hotshoe on the gh5? Can't find this info anywhere
No but it's likely that any flash extension cable with the right pins will work. There's nothing unique about the pin configuration
I'm guessing that you're not actually getting normal 24p here. I expect that you have some ~judder effects. Since 24 doesn't divide into 60, the camera can't be skipping a consistent number of frames when generating the HDMI output. My guess is it's probably doing something like alternating between skipping 1 and 2 frames.
Easiest way to test this would be to film something dropping in front of a ruler. Normal 24p should show the object moving the same distance in each frame. Alas, I do not have a GH5 (yet?) to test this on. :)
Easiest way to test this would be to film something dropping in front of a ruler. Normal 24p should show the object moving the same distance in each frame. Alas, I do not have a GH5 (yet?) to test this on. :)
+Quentin Smith Great test! I'll do it.
I was just about to ask if you could do a show about hooking the Gh5 to an atomos as I hope to buy one soon. Amazing! Not watched it yet, do you cover how to hook up a mic? I'm a total video noob
+R Summers Just plug it in :-) I will do a show on the XLR1 soon but if you're a n00b then the regular mic input will do. Check out my mic comparison video: https://youtu.be/WRiEwMWmdOQ
When you're using a 50mm cine lens for example, do you set the "Focal Length Set" on the IBIS to 50mm or 100mm since it's on a 2x crop sensor?
+Eric Zhang I've always set it to what it says on the lens, so 50. Probably worth checking the manual but that's what I've always assumed. The 100mm (in your example) is irrelevant to the camera; it doesn't know what a full frame sensor is.
I can't wait to see what the GH5 can do in the 10 bit color space with the atomos' higher mbps. Also I will totally use this at weddings on the gimbal. Atomos Ninja Inferno will be mounted on the cross bars. This is why I wanted to know if the AF system improves!
If you are set at 180° angle, what is the shutter speed for variable frame rate, example 180 fps? If you have set a shutter speed, say 1/60, what happens as you increase vfr up to 180 fps?
The shutter speed/angle changes the motion blur. therefore keeping the 180 degree rule. 24p 1/48(or 1/50), 30p 1/60, 60p 1/120, 180p 1/360, and so on.
Hey PhotoJoseph...will you be at NAB?
Yep. Photo Moments meetup at 9am on Tuesday. Send me your email so I can add you to the list. Location is probably the Courtyard Marriott across the street but I'll confirm on Sunday.
Isn't it the same just wrapped differently?
Thank you, that´s quite interesting. I just wonder which 24 frames of the 180 captured frames the Ninja will record, because it clearly can not be every 7,5th frame. I don´t own the GH5 nor any recorder, so maybe I make logical mistake somewhere.
But what about the shutter speed/angle? Nomally in SS you would set it to double the frame rate. Ok.. So do you make a choice 1/48 or1/120 and one videos motion blur is off??
Hi Joseph, This works well on the videodevices pix-e recorders too, another thing available is, if you set the GH5 to display a LUT and send the LUT over hdmi you can record the footage with LUT embeded in the recorded stream, looking very good quality straight out of the recorder :)
That's why I fumbled for a moment in the show while my brain processed… yeah that's what I was thinking originally, but that doesn't make sense because the camera is capturing 180fps, and playing back at 24… but the Ninja is capturing only 24 (or whatever it's set to). So the camera is outputting the native while capturing the extras. Weird, but it works.
Comments from YouTube