Splitting the buffer between CF and SD on 5D3?

Started by Dannington, June 09, 2013, 06:47:02 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Dannington

Has anyone thought of this? (of course they have)  If the bottleneck is in waiting for the CF card to complete it's write cycle could you split the image say 70/30 between the CF and SD then merge them on the desktop?  Is it just a case of diverting the stream from one device to the other?

My other idea - which may turn out to be just as valid - is a large plastic bucket just underneath the cf card door, catching all of the data as it spills out.

a1ex

Is it worth the extra hassle (programming complexity + workflow) for a 20% improvement in write speed?

Dannington

I very much know you're the expert.  20 percent seems quite a bit to me.  It'd be interesting to know if it's technically possible.  If the split point was fixed rather than dynamic I could see how recombining the frames on the desktop would be pretty straightforward - I don't know much about the in-camera side though.

jjpk976

I think this would be extremely useful especially since it might allow higher resolution in 60fps mode. Or maybe higher then 1920x1088 continuous on the 5d3, or is the camera just not capable of that? Anyway this would be an amazing feature to work on maybe after normal raw video is more stable and has an official beta.

jjpk976

Or a mode when the buffer gets like 75% full it start writing to both cards would be a highly valuable feature as well! but it might be hard to process in post unless ML made some kind of mark to when it started writing to the sd. Anyway this is all very exciting to me, thanks a lot ML team for all your hard work! :)

1%


Stedda

IMO until it is found that it's just not possible to have a CF to SSD/HD Device I dont think it's worth the Devs effort.

RAW recording is just about a month old... I think people expect too much....
5D Mark III -- 7D   SOLD -- EOS M 22mm 18-55mm STM -- Fuji X-T1 18-55 F2.8-F4 & 35 F1.4
Canon Glass   100L F2.8 IS -- 70-200L F4 -- 135L F2 -- 85 F1.8 -- 17-40L --  40 F2.8 -- 35 F2 IS  Sigma Glass  120-300 F2.8 OS -- 50 F1.4 -- 85 F1.4  Tamron Glass   24-70 2.8 VC   600EX-RT X3

Dannington

1% - In terms of audio - i'm actually interested in LTC generation in camera, sent out of the headphone jack into a recorder (along with all of your other sound of course).  I'm a video editor and this kind of stuff on professional cameras is dead handy for syncing audio to video. (sorry to switch topics)

Stedda - Not SSD, the built in SD and CF.  There are a lot of posts on this site from people trying to get the very best performance from different cards/camera settings etc.  This might be a way to beat the limitations of both controllers by combining them, whilst not a true raid system you might be able to lever extra performance providing you can switch the data between the cards quick enough.

Stedda

I understood your post I don't think you understood mine.

If someone manages to create a device that we can hook a SSD up to there's no point to using both controllers and the DEVs time has been wasted.
5D Mark III -- 7D   SOLD -- EOS M 22mm 18-55mm STM -- Fuji X-T1 18-55 F2.8-F4 & 35 F1.4
Canon Glass   100L F2.8 IS -- 70-200L F4 -- 135L F2 -- 85 F1.8 -- 17-40L --  40 F2.8 -- 35 F2 IS  Sigma Glass  120-300 F2.8 OS -- 50 F1.4 -- 85 F1.4  Tamron Glass   24-70 2.8 VC   600EX-RT X3

Dannington

Sorry, misread your post.  I'm really interested in the SSD option too.  It'd be the absolute perfect solution.

jjpk976

The ssd option would require an external power source and to be mounted of rails though right? I feel like it would greatly hinder the potability of the camera, but i guess we wont know until there is a working prototype. I feel like writing to two cards would be a great thing to look into after all the crucial components of raw are stable :) I'm sorry, I don't mean to seam like I am asking for to much, I would happy with any direction the Devs decide to go with ML, i'm still dumbfounded with the possibility of raw.

Joachim Buambeki

Quote from: a1ex on June 09, 2013, 06:50:39 PM
Is it worth the extra hassle (programming complexity + workflow) for a 20% improvement in write speed?
If those 20% make the difference between beeing able to get the desired framerate (1080p30 for example) or the desired resolution (1920x1280@ 24fps for example) and not beeing able to get it, then yes it's totally worth it. Beeing able to overcrank to maybe 33fps for a 24p base shot would be a welcome addition.

Using a SSD (that hasn't even surfaced as a prototype) would not only hinder portability but almost double the cost of a ready to shoot rig (5D3+2cards+lens vs 5D3+cage/rails+battery-mount+battery+SSD-mount+SSD), using card splitting in ML  costs nothing extra (buying an extra SD card maybe...) so I don't see the time spent to develop it beeing wasted.

jjpk976

I think our point has been stated well enough at this point, its just a matter of if the devs deam it worth the time and effort to make it happen, I which I understand it they don't, I imagine it would be a huge pain to implement but on the other hand it could be a huge gain too :) Thanks again ML team

Joachim Buambeki

Another idea would be to have the data moved to the SD card in the background while not recording.
This way one could use a lower capacity (=cheaper) superfast CF card for shooting and cheap 64Gb SD cards for internal offloading between takes.
Additionally the splitted files could be merged during the offloading process (though this might be an even more programming intensive task than just splitting them up).

jjpk976

That can already be done, with the file manager module

linyingyen

120+MB/s would enable 48fps HFR on 2K 2.39.

a1ex

95+20 is smaller than 120 if I'm not mistaken...

(that's the theoretical speed, according to benchmarks)

jjpk976

Well i usually get around 99-105mb/s when recording of my cf card so 99 + 20 would allow 48fps for at leased a few seconds right? here's my benchmark if it makes any diffrence :)
http://i.imgur.com/DpUTeMx.png

I've only tested out crop mode too if that also makes a diffrence

a1ex

Nice!

For 48fps 1920x672 (max resolution) you only need 103.5 MB/s. According to my simulator, you should already get 260 frames at 80MB/s and 700 at 95MB/s.

Try it here: https://bitbucket.org/hudson/magic-lantern/src/tip/modules/raw_rec/speedsim.py

linyingyen

Quote from: a1ex on June 10, 2013, 02:37:49 PM
95+20 is smaller than 120 if I'm not mistaken...

(that's the theoretical speed, according to benchmarks)

I thought it is suppose to be 100+ on the 5D3? As stated in the spread sheet that is.
(100+)+20 would be 120+, if my elementary math skill does not fail me that is.  :)

atarijedi

What about streaming the data? or is that already done? or is it even possible to get that low level?

The ATAPI-7/UDMA7 standard included a new streaming command set specifically to benefit audio/video devices that use hard drives.

Shield

Quote from: 1% on June 09, 2013, 09:24:12 PM
Leave the SD for audio :)

I would much rather see the audio fixed too, first. 

tihon

1920x492 with 60 fps - is the only thing i need. it give me a 1920x 820 after an  upscale and make me so happy ))))
Cinema, cinema, cinema

sletts02

Just stumbled upon this - any roadmap for development?

SDX

It'll be a feature that comes with the new format and has already been partly implemented. Follow progress here: http://www.magiclantern.fm/forum/index.php?topic=7122