7D Tester13 and 7D

Kholi said:
Why? There's nothing wrong with the compression as is.

There WAS plenty wrong with the GH1 compression.

All of a sudden the 7D isn't adequate?

MJPEG on the other hand IS topping out at 80 thanks to Michael Pappas' dedicated testing. And that's far beyond what either of the Canon's are doing... and that's just SDHC card and controller. It's not that hard to tell, even in a small vimeo box: the AVCHD is startlingly clear, and at 60+ inches of high def real estate looks the part.
Click to expand...

So with that said Kholi, are you opposed to someone trying to make the 7D compression better?

The 7D compression is fine and I'm not saying it's inadequate, I'm just saying that (as we have seen from the GH1 hack) it can be improved on. There are number of H264 flavors, and I'm sure the current 7D flavor could be improved.
 
Last edited:
I dont know if 80 is a realistic number being tested. I might be wrong but I believe they are AVERAGING 40-50.
 
gene_can_sing said:
So with that said Kholi, are you opposed to someone trying to make the 7D compression better?

The 7D compression is fine and I'm not saying it's inadequate, I'm just saying that (as we have seen from the GH1 hack) it can be improved on. There are number of H264 flavors, and I'm sure the current 7D flavor could be improved.
Click to expand...

I think the GH1 is a better camera to focus on.

It's cheaper and offers more out of the gate. The 7D Compression was already pretty decent, for keying it's a pain but for everything else it looks fine.

I don't think it's a worthwhile effort.
 
Barry pretty much has my list for what I'd want in the 7D, all the things they added to the 5D mostly.

I'd also like to make it so the wheel above the shutter button adjusted ISO and NOT shutter speed. I have searched through all the menus and it cannot switch to that. I rarely EVER switch shutter speed, but I do mess with ISO.
 
With the 7D you have a 12 minute recording time limit...if you increase the bitrate on the camera you'd be cutting that down even further. It's a non-issue on the Gh1 because it has no recording time limit and could already hold tons on a card...now it takes up a bit more space, more comparable to the Canons, but has no limit still...so you can just use a larger card.
 
I could deal with less recording time, I don't do much live events, but aren't there flavors of h.264 that are 4:2:2 that would be a better compression than 4:2:0
 
Well... 4:2:2 vs. 4:2:0 is a tradeoff of color resolution versus bitrate. At the same bitrate (i.e., a fixed 50mbps) the 4:2:0 version will have lower color sampling, but fewer compression artifacts too. The 4:2:2 encoding adds about 1/3 to the total data stream, so while it gives you something (double the color sampling) it also takes away from your data stream.

So if you wanted 4:2:2 while not giving anything else up, you'd have to up the data rate too.

h.264 can do 4:2:2 in higher-level profiles, yes. Whether the 7D (or any particular camera) has the ability to code that, remains to be seen. But yes, if there was a firmware hacker working on the 7D, that's something I'd want them to explore.

Frankly if a firmware hacker wanted to tackle the 7D, I think its most egregious problem is the purple/orange chroma aliasing. I'd want to see them expend some efforts to see if they could come up with a cleaner method for that.
 
Barry_Green said:
Well... 4:2:2 vs. 4:2:0 is a tradeoff of color resolution versus bitrate. At the same bitrate (i.e., a fixed 50mbps) the 4:2:0 version will have lower color sampling, but fewer compression artifacts too. The 4:2:2 encoding adds about 1/3 to the total data stream, so while it gives you something (double the color sampling) it also takes away from your data stream.

So if you wanted 4:2:2 while not giving anything else up, you'd have to up the data rate too.

h.264 can do 4:2:2 in higher-level profiles, yes. Whether the 7D (or any particular camera) has the ability to code that, remains to be seen. But yes, if there was a firmware hacker working on the 7D, that's something I'd want them to explore.

Frankly if a firmware hacker wanted to tackle the 7D, I think its most egregious problem is the purple/orange chroma aliasing. I'd want to see them expend some efforts to see if they could come up with a cleaner method for that.
Click to expand...
It's times like this I wish I would have just bitten the bullet and made computer science my major..
 
brockraum said:
It's times like this I wish I would have just bitten the bullet and made computer science my major..
Click to expand...

If it makes you feel any better, most computer science degrees don't get into firmware programming but focus more on the theory of programming than the actual programming itself. While I majored in electrical engineering, my minor is in advanced computer graphics (3D).

Most of the firmware developers I know are some of the most amazing coders I know. I can program pretty well on a variety of platforms (Linux,Sun,Mac,Win) and in VERY MANY languages and they programs circles around me. They are who I go to when I can't do something :)
 
grimepoch said:
If it makes you feel any better, most computer science degrees don't get into firmware programming but focus more on the theory of programming than the actual programming itself. While I majored in electrical engineering, my minor is in advanced computer graphics (3D).

Most of the firmware developers I know are some of the most amazing coders I know. I can program pretty well on a variety of platforms (Linux,Sun,Mac,Win) and in VERY MANY languages and they programs circles around me. They are who I go to when I can't do something :)
Click to expand...

Hah, well I think you've found something you can "go to them" with :laugh:
 
gene_can_sing said:
Hey Barry,

Do you think that they can get rid of the purple/orange aliasing? Or is it too baked into the hardware?
Click to expand...

That has to do with the way the sensor does line-skipping...not really fixable. The GH1 uses pixel-binning which is why it doesn't have quite the same level of aliasing, and may also be why it avoids color contamination.
 
I don't see why the 12 minute limit on the Canons can't be hacked out. And of course a higher bitrate would be nice.

But yeah the aliasing, rolling shutter, and moire are pretty much unfixable with the Canons.

But I would definitely love something like the Magic Lantern for 7D and the T2i.
 
Angry Leprechaun said:
I don't see why the 12 minute limit on the Canons can't be hacked out. And of course a higher bitrate would be nice.

But yeah the aliasing, rolling shutter, and moire are pretty much unfixable with the Canons.

But I would definitely love something like the Magic Lantern for 7D and the T2i.
Click to expand...

Because it's fundamental to the file system used by the cards....the GH1 doesn't have it because AVCHD is designed to bridge clips and has a whole metadata structure...I kind of doubt that quicktime wrapped h.264 can do the same.
 
gene_can_sing said:
Do you think that they can get rid of the purple/orange aliasing? Or is it too baked into the hardware?
Click to expand...
Nobody outside of Canon knows how their sensor works, or where that purple/orange stuff is coming from.

However, something bizarre I've seen on the GH1's MJPG 1080p mode ... it actually *adds* the purple/orange moire, just like a Canon! When you leave MJPG in 720p mode, there's no color moire, but if you upscale it to 1080p (in-camera), it starts appearing.

So ... would it be possible to make a proper 720p in the Canon that doesn't have moire? maybe. Who knows? I just think it's really odd that the GH1 is completely clean of it, yet when Vitaliy patched the firmware for 1080p MJPG, all of a sudden the same kind of chroma moire came around!

Potentially that might mean it's fixable, but it might mean going to 720p. But who knows?
 
Ben_B said:
That has to do with the way the sensor does line-skipping...not really fixable. The GH1 uses pixel-binning which is why it doesn't have quite the same level of aliasing, and may also be why it avoids color contamination.
Click to expand...
Where does this information come from? AFAIK, nobody inside Canon or Panasonic has ever confirmed how their sensors work, and Canon has gone so far as to deny line-skipping and said that they instead use pixel binning.
 
Barry_Green said:
Nobody outside of Canon knows how their sensor works, or where that purple/orange stuff is coming from.

However, something bizarre I've seen on the GH1's MJPG 1080p mode ... it actually *adds* the purple/orange moire, just like a Canon! When you leave MJPG in 720p mode, there's no color moire, but if you upscale it to 1080p (in-camera), it starts appearing.

So ... would it be possible to make a proper 720p in the Canon that doesn't have moire? maybe. Who knows? I just think it's really odd that the GH1 is completely clean of it, yet when Vitaliy patched the firmware for 1080p MJPG, all of a sudden the same kind of chroma moire came around!

Potentially that might mean it's fixable, but it might mean going to 720p. But who knows?
Click to expand...
Definitely noticed this too Barry. We won't really know unless someone with a substantial amount of firmware coding knowledge dives into this thing head first.......
 
When editing in adobe prem cs5 does the color space change to from 4:2:0 to 4:2:2 ? Because it can edit native h.264 codec
 
vx1kOrnot said:
When editing in adobe prem cs5 does the color space change to from 4:2:0 to 4:2:2 ? Because it can edit native h.264 codec
Click to expand...

I'm not sure what you mean. Converting to ProRes allows you to work in 4:2:2 color space, but it doesn't mean that there is now extra chroma information in your original footage. You can't add in what wasn't captured there in the first place.
 
In fact 7D have one big advantage - CompactFlash with UDMA support.
And this cards have controller inside and are very fast.

Problem with 7D is that it is based on ARM. And I have only limited time access to full latest IDA Pro today. And IDA Free is not able to disassembly ARM firmware.

Current status on 7D is as follows (according from my talks with Trammel).
He have disassembled version for long time. But ML loading technique no longer work (bootable modification and auto loading of specific file).
So, you need actual firmware modification. And Trammel even knows all algorithms and encryption keys that are necessary fo this. But he seems afraid to share them due to DCMA (as he must comply to USA laws).

So, to conclude. Even simple modification requires you to solve many problems.
First, obtaining latest IDA version.
Second, checking and adjusting information about firmware file structure.
Third, making dumper (Trammel did one for previous versions).
Use dumped firmware to find encryption routines and keys.
Next, good unpacker and decryptor plus packer and encryptor (plus checksums calculator).
And only in this moment you could make simple modification.
 
You must log in or register to reply here.
Back
Top