R5 recording limits temp/sensor correlations (continued)

[Horshack]

This is a continuation of the thread here.

Recent Recap

A lot has transpired since that thread reached its conclusion - an owner in China was able to reset the thermal timer and bypass the cooling-off period by removing the clock battery. A few weeks prior to that I had proposed an experiment to do the same by inhibiting the camera's ability to save its thermal management to NVRAM by removing the main battery. I repeated proposed the same experiment here with more details. Once word of the China clock-battery spread (thanks to EOSHD) I was finally able to get a volunteer for the main-battery experiment, which was successfully performed by J Marus ,then later by Matt Granger, Electroholic Anonymous, and Andrew Reid, the latter two of which demonstrated recording 8K for 50 minutes over multiple battery-removal sessions without any special cooling/fridge tests.

I'm going to keep this thread focused on the technical investigation and not spend too much on theories centered around marketing/segmentation/crippling. Those are well handled in other forums such as EOSHD and YouTube.

New Revelations - workaround for corrupt video file


While the hack was important in revealing new details about Canon's thermal management it's not meant to serve as an actual workaround since it requires pulling the battery while the video is still be recording (otherwise the camera saves the thermal state to NVRAM when the user stops a recording). This renders the video unplayable/unusable. Instead of a fully-formed .MP4 file the technique leaves a .DAT file, which represents the mdat atom of the video file (video+audio frames) but is missing all the MP4 container metadata to play or use the fie.

While working on tools to reconstruct the missing MP4 container metadata, an EOSHD forum poster the name of J- proposed a brilliant workaround to use FAT32-formatted cards, which would force the camera to split up a continuous recording into 4GB splits (due to FAT32's 4GB per-file limitation). It wasn't clear if this technique would avoid the NRAM-saving that's seen when stopping a recording. After experimenting on my Canon RP I found it does avoid the NVRAM when writing the 4GB splits, and more importantly, it was later verified on a R5 by Electroholic Anonymous here.

So far this FAT32 trick only works on SD cards - neither Electroholic Anonymous or Hoka Key have been able to get the camera to accept a FAT32-formatted CFE card. Normally the camera will only format cards <= 32GB as FAT32, using exFAT for larger capacity cards. This limitation has been worked around by formatting the cards under Linux - Windows 10 wont let your format cards > 32GB as FAT32 (but will mount them) and the camera wont accept FAT32 cards formatted under OSX. More details here.

So there is now a plausible workaround that includes saving the entire recorded video sans the last 4GB lost after a battery pull (which is only about 45 seconds for an 8K video).

 

[Horshack]

Electroholic Anonymous (EA) performed two successful long-run video recording experiments today, first a 40 minute 8K30P that resulted in a final EXIF temperature of 72C and a second experiment for 8K24P that ran for 145 minutes at 65C, which is graphed below from data provided by EA:

8K24P for 145 minutes, final temperature of 62C

Both experiments used the FAT32 workaround to retain the video files, which were the source of the EXIF temperature samples.

EA is planning a second 8K30P experiment to see if the temperature still reaches 72C vs only 62C for 8K24P. While the difference is likely the result of the more frequent sensor-readout and DIGIC processing from handling 25% more frames, there were other differences as well. The camera was on the table for the 8K30P experiment and on a tripod for the 8K24P. That might have affected the thermal conduction at the bottom of the camera.

 

[Bernie_King]

You were called out on the Fro's latest Photo News Fix. Starts at 4:16



--
http://www.pbase.com/bernarrking;
http://www.bkingphoto.com

 

[JakeJY]

So far this FAT32 trick only works on SD cards - neither Electroholic Anonymous or Hoka Key have been able to get the camera to accept a FAT32-formatted CFE card. Normally the camera will only format cards <= 32GB as FAT32, using exFAT for larger capacity cards. This limitation has been worked around by formatting the cards under Linux - Windows 10 wont let your format cards > 32GB as FAT32 (but will mount them) and the camera wont accept FAT32 cards formatted under OSX. More details here.

There is a third party Windows utility "FAT32 Format" that will allow you to format cards (and USB drives) larger than 32GB as FAT32. Here's a guide on how to do so (but with USB drives):

https://www.howtogeek.com/316977/how-to-format-usb-drives-larger-than-32gb-with-fat32-on-windows/

I have used this tool personally to format some larger cards for security cameras that only take FAT32.

Also from the same guide supposedly using format command in powershell works also (although it seems to do a full format that takes a long time).

Officially for SD cards any card 64GB or larger is considered SDXC and is supposed to be formatted as exFAT by default, so cameras typically do so.

https://kb.sandisk.com/app/answers/detail/a_id/2520/~/sd/sdhc/sdxc-specifications-and-compatibility

 

[Horshack]

So far this FAT32 trick only works on SD cards - neither Electroholic Anonymous or Hoka Key have been able to get the camera to accept a FAT32-formatted CFE card. Normally the camera will only format cards <= 32GB as FAT32, using exFAT for larger capacity cards. This limitation has been worked around by formatting the cards under Linux - Windows 10 wont let your format cards > 32GB as FAT32 (but will mount them) and the camera wont accept FAT32 cards formatted under OSX. More details here.

There is a third party Windows utility "FAT32 Format" that will allow you to format cards (and USB drives) larger than 32GB as FAT32. Here's a guide on how to do so (but with USB drives):

https://www.howtogeek.com/316977/how-to-format-usb-drives-larger-than-32gb-with-fat32-on-windows/

I have used this tool personally to format some larger cards for security cameras that only take FAT32.

The ridgecrop formatter utility mentioned in that article was the first I tried - my RP wont accept cards formatted by it, as discussed here.

Also from the same guide supposedly using format command in powershell works also (although it seems to do a full format that takes a long time).

The command line lets you attempt it but it fails.

 

[traderjay]

Canon has revealed there are 3 temp sensor, 2 internal and 1 external. I think we need to start all the thermal discussion from scratch...

 

[JakeJY]

Canon has revealed there are 3 temp sensor, 2 internal and 1 external. I think we need to start all the thermal discussion from scratch...

Link? Couldn't find it.

I did find they just released a firmware update that was inline with the previous rumor (that there would be some improvement in the recording times):

https://www.newsshooter.com/2020/08/27/canon-r5-firmware-update/

 

[David Hull]

Canon has revealed there are 3 temp sensor, 2 internal and 1 external. I think we need to start all the thermal discussion from scratch...

That makes sense: FWIW: They also said there was a timer. Here is the video where that was referenced (about 50 sec in):

 

[Canochrome]

Being called out by FRO is pretty big time!

 

[Canochrome]

New Firmware is out, at least in the European market. I'm thinking that it may change things regarding "overheating" in a direction favoring the photographer/videographer.

It might be worth checking it out and repeating past testing with the new firmware before carrying out further tests.

 

[rrc1967]

Electroholic Anonymous (EA) performed two successful long-run video recording experiments today, first a 40 minute 8K30P that resulted in a final EXIF temperature of 72C and a second experiment for 8K24P that ran for 145 minutes at 65C, which is graphed below from data provided by EA:

8K24P for 145 minutes, final temperature of 62C

Both experiments used the FAT32 workaround to retain the video files, which were the source of the EXIF temperature samples.

EA is planning a second 8K30P experiment to see if the temperature still reaches 72C vs only 62C for 8K24P. While the difference is likely the result of the more frequent sensor-readout and DIGIC processing from handling 25% more frames, there were other differences as well. The camera was on the table for the 8K30P experiment and on a tripod for the 8K24P. That might have affected the thermal conduction at the bottom of the camera.

it makes sense that the temperature reaches a steady state, but I'm not convinced that 65C internally won't degrade image quality or reliablity.

it should be noted that 65C isn't a good environmental temperature for electronics.

72C certainly isn't.

 

[JakeJY]

Canon has revealed there are 3 temp sensor, 2 internal and 1 external. I think we need to start all the thermal discussion from scratch...

That makes sense: FWIW: They also said there was a timer. Here is the video where that was referenced (about 50 sec in):

Thanks for the link. For those that hate watching video or can't, here's a transcript of the relevant parts from Gordon Laing:

"During a new interview with Canon, I learned the R5 employs three temperature monitors, two internally near the AF and imaging sensor, and one for measuring external ambient temperatures, that's near to the ports. The R5 uses readings from all three, along with a timer, to figure out when to report overheating and how much recording to allow while the camera cools down.

With the firmware update, Canon hasn't actually changed the temperature at which overheating is reported, but will now check the sensors more frequently, giving more accurate feedback and potentially allowing you to record again sooner or for longer. The external ambient temperature also now plays a bigger role, with Canon explaining to me that cooler surroundings may allow the camera to recover more quickly, so a fan or other active cooling could actually now become beneficial. I was also told the new firmware may not make much or indeed any difference to single long clips..."

He also goes on to say with firmware update he could do 4KHQ for 30 minutes and a second 10 minute clip. Then 10 minute rest gives 5 minutes, and 20 minute gives 10 minutes, but it still took an hour to recover to the quoted 25 minutes. For 8K, it improved from 20 minutes to 26 minutes. For more details, further (like his 4K120p tests) watch the video.

 

[traderjay]

USA firmware is up i just upgraded my camera! Based on the review he is able to record many more shorter clips of 4k 120 than before because the recovery time has improved due to better sensor reading.

I really enjoyed all the technical discussions here and the firmware proved there is really no conspiracy theory that all the trolls are alluding to. With what is in the r5 and the form factor unlimited 8k raw is impossible.

--
Workstations - EPYC 7702 + 3950X || NVIDIA RTX 2080ti || Canon imagePROGRAF Pro-1000

 

[Peak freak]

Anyone know the Exif temps with the camera on / idle / stills for extended period?

Is it then actually too hot for HQ video modes, or is more that the 'clock is always ticking'?

 

[Zeee]

Canon has revealed there are 3 temp sensor, 2 internal and 1 external. I think we need to start all the thermal discussion from scratch...

That makes sense: FWIW: They also said there was a timer. Here is the video where that was referenced (about 50 sec in):

Thanks for the link. For those that hate watching video or can't, here's a transcript of the relevant parts from Gordon Laing:

"During a new interview with Canon, I learned the R5 employs three temperature monitors, two internally near the AF and imaging sensor, and one for measuring external ambient temperatures, that's near to the ports. The R5 uses readings from all three, along with a timer, to figure out when to report overheating and how much recording to allow while the camera cools down.

With the firmware update, Canon hasn't actually changed the temperature at which overheating is reported, but will now check the sensors more frequently, giving more accurate feedback and potentially allowing you to record again sooner or for longer. The external ambient temperature also now plays a bigger role, with Canon explaining to me that cooler surroundings may allow the camera to recover more quickly, so a fan or other active cooling could actually now become beneficial. I was also told the new firmware may not make much or indeed any difference to single long clips..."

That is great. Not surprising that they didn't implement it earlier. They delivered a camera like no other during a world pandemic. I doubt they had the adequate time in the field. There were people testing it out but probably not enough time. It was mentioned it was more than likely on the side of caution because the cool down times were a little absurd.

To those who are hoping you will never see one hour of 8K on this camera.

He also goes on to say with firmware update he could do 4KHQ for 30 minutes and a second 10 minute clip. Then 10 minute rest gives 5 minutes, and 20 minute gives 10 minutes, but it still took an hour to recover to the quoted 25 minutes. For 8K, it improved from 20 minutes to 26 minutes. For more details, further (like his 4K120p tests) watch the video.

--
Another one bites the dust. I feel even more confident that soon things will have a wonderful conclusion

 

[rrc1967]

Anyone know the Exif temps with the camera on / idle / stills for extended period?

Is it then actually too hot for HQ video modes, or is more that the 'clock is always ticking'?

you know .. this bring an interesting point even though it's sort of unrelated or reverse to what you are talking about.

is there a temperature threshold where the camera can no longer reliably perform stills photography without hitting a thermal wall after shooting x minutes of video.

maybe we're looking at this the wrong way. it's not all about video. (surprise).

but Canon is worried that if you shoot too much video, stills may be compromised over time - in other words, they are looking a temperatures, yes, but they are also prioritizing stills over video, so there's margins in there that are meant to protect the ability to shoot stills after video.

 

[traderjay]

In theory they can give user the option to sacrifice still image quality for even longer video via an option with warning. This all depends if that temp threshold is safe for the internal components. Experienced engineers can look at the pcb and figure out the max safe temp, but we won't know the cmos sensor limit because that is born from canon's own foundry.

 

[Zeee]

USA firmware is up i just upgraded my camera! Based on the review he is able to record many more shorter clips of 4k 120 than before because the recovery time has improved due to better sensor reading.

I really enjoyed all the technical discussions here and the firmware proved there is really no conspiracy theory that all the trolls are alluding to. With what is in the r5 and the form factor unlimited 8k raw is impossible.

I agree and I said this on another post. They delivered a camera like no other during a world pandemic. I doubt they had the adequate time in the field. There were people testing it out but probably not enough time to meet sales deadlines. I worked in manufacturing for over 30 years. Not once did we have enough time to train people, set up and test new equipment. Sales would go to the general manager, we were a week behind and all of a sudden it is going live two weeks before the original deadline.

It was mentioned it was more than likely on the side of caution because the cool down times were a little absurd. It would have not taken the conspiracy world to address that. Customer complaints would have done that.

 

[traderjay]

I work in manufacturing (semiconductor and now robotics and 3d vision) as well but on the business marketing side but spent enough years with brilliant engineer to know what is bs and what is fact. I also worked at Canon and understand their engineering culture hence I bought this camera despite all the trolls。

 

[rrc1967]

In theory they can give user the option to sacrifice still image quality for even longer video via an option with warning. This all depends if that temp threshold is safe for the internal components. Experienced engineers can look at the pcb and figure out the max safe temp, but we won't know the cmos sensor limit because that is born from canon's own foundry.

temperatures at or around 60-70C over long durations of time are not considered safe for electronics.

sure they can run at those temperatures, they won't melt, they won't necessarily blow up immediately, but they will suffer an decrease of the MTBF by a significant amount.

(as you are most assuredly aware of from your own experience)

as far as CMOS sensor - it's more a linear noise increase but what it can also cause is drift, not sure but elements like banding may be more prone to appear etc.