Panasonic Removable Lens Four Thirds HD AVCHD Camera

LizaWitz said:
Frankly, this camera delivers what the scarlet promised-- a 4k camera with interchangeable lenses great for shooting HD. Course, not in raw mode, etc, but its a whole lot cheaper....and it will be shipping in months.
Click to expand...
Scarlet doesn't promise 4K in their lower-cost models, they promise 3K. This GH1 doesn't promise 3K, it promises HD (which is nigh unto 2K).

Scarlet claims use of cinema lenses, modular/expandable design, and RedCode wavelet compression at 42 megabits, plus slow motion at up to 150fps, for $3,000. The GH1 claims 1080p and 720p using AVCHD, interchangeable lenses, and a body-only price probably around $1,000. The GH1 will have a sensor about 3x as large as the Scarlet 2/3. You could get a Scarlet with a bigger sensor than the GH1, but those start at $7,000.
 
What's the deal on no release date? It's basically a timestamp on how soon we get to see real footage, not the tradeshow stuff.

I'd be happy to replace my D90 with this...
 
I saw Summer mentioned for release. Wish it were more specific or even sooner. As we all know, a lot can happen in five months. I'm sure this isn't the only new interchangeable DSLR hybrid coming out this year.
 
Just for clarification, the new GH1 isn't technically a DSLR right. Since it doesn't have the mirror etc. it's just an ILC (interchangeable lens camera) or at least that's how I'm told Panasonic refers to it.
 
Barry_Green said:
Do you guys realize what you're asking? Not even a class 6 card could support AVCIntra @ 50mbps. Class 6 is supposed to support a sustained speed of 6 megabytes per second, or 48 megabits. So even under the absolute perfect conditions at all times, 50mbps still wouldn't be possible. And if it were, it would be on the absolute edge. There would be no margin of error, no room for redundancy, no safety net whatsoever.
Click to expand...

They can always RAID 2 cards. That would be a nice cheap solution for higher quality video.
 
One feature that I'm curious about is that aut follow focus. If it works fine and precise, doors open up to a different shooting style: Run & Gun with shallow DOF.
 
LizaWitz said:
H.264 is a very flexible codec, in fact its not even necessarily fixed bitrate. It can be variable bitrate. I think many people are still hung up on the days when the format was DV and the encoding was "fixed compression" and the bitrate determined quality.

This is not the case anymore. The quality of an H.264 stream is dependant on several factors, including bitrate, but also including the complexity of the encoding being done, the horsepower of the CPU doing the encoding and the time available to he CPU to do that encoding.

The 5D drops much of the sensor information and barely encodes the video and so it needs a massive bitrate to get by--- because it doesn't apparently have the horsepower to encode the video in real time at higher quality, it makes up for it with bitrate. The only info I've found so far shows CAVLC to be 10% less efficient than CABAC (http://escher.elis.ugent.be/publ/Edocs/PPT/P106_246.pdf)

Its quite possible that the GH1 will encode video at a higher quality than the 5D, even though the bitrate is much lower. Two reasons: Superior input: The GH1, by all accounts, reads the full width of the frame-- a 4k image, and reduces that down to the 1920x1080 video frame. The 5D just drops all that info. So, better input quality. Secondly, the GH1 has a dual core CPU, and probably one of those cores is a dedicated H.264 encoder. I believe the 5D just had video added on late in development as a response to nikon-- its probably encoding in software, and thus relying all on bitrate. Finally, the GH1 has %33 longer time between each frame to do the encoding. That goes directly to the quality bottom line.

It remains to be seen what the GH1 video looks like, but we can't draw many conclusions at all from the 17Mbps bitrate. This isnt' tape.

Frankly, this camera delivers what the scarlet promised-- a 4k camera with interchangeable lenses great for shooting HD. Course, not in raw mode, etc, but its a whole lot cheaper....and it will be shipping in months.
Click to expand...

I've seen quite a bit of footage from different cameras, and if the Canon is dropping something...it must be pretty irrelevant. The images have as much picture edge sharpness (low spacial frequency MTF) as any camcorder output I've seen (including SDX900 DVCPro50) and more (much more usually) texture detail (mid-range spacial frequency MTF).

AVCHD specifies this chart:

OverviewChart1.GIF


Wikipedia says:

Canon and Panasonic camcorders use High-Profile@Level-4.1, up to the AVCHD format's maximum bitrate of 24 Mbit/s. To date, Sony camcorders have only used Main-Profile@Level-4.0, at a maximum bitrate of 17 Mbit/s.
Click to expand...
4.0 specfies a 20Mb/s video data bit rate for Main Profile and 25Mb/s for High Profile. and 4.1 specified a max data rate for video of 50Mb/s for Main Profile and 62.5Mb/s for High Profile.

In both, similar encoding is used. The Canon is clearly display more detail. That extra detail seems to be present, within the limits of the lens and sensor. which are astronomical on the 5DM2 with good glass. However, looking at even insexpensive SR11 and HV20 footage captured through HDMI, tremendous extra detail (at least in picture edge sharpness, the textures still seem a bit hammered which I think is a lens issue). This is further evidence that the storage bit-rate is the limiting factor.

That said, the advanced protocols of AVCHD seem to make a difference. Comparing HMC150 25Mb/s AVCHD to HPX170 DVCProHD, I find the footage remarkably similar. That protocol has 1.75Mb per frame. At 24fps, the AVCHD average 1Mb per frame, however the long GOP makes up the difference. I imagine 720p60 or 1080p30 which would take advantage of the higher 100Mb/s DVCProHD bandwidth would look better in DVCProHD.

My knock on AVCIntra is that I don't see the Intra giving and an advantage over the otherwise identical long GOP version of the otherwise identical protocol. However, the inter-frame protocol would allow more efficient compression, allowing more data to be captured.

If Panasonic is supporting HDMI out on the GH1, it may exceed the 5DM2, IF bit-rate is still the limiting factor. If not, the Canon's superior glass and probably superior sensor (minimally 4x the area) should give it the edge.
 
Last edited:
Sorry for even mentioning red. I'd like to see Red be competitive, but looking at their store, a CF module, and one 16GB CF card is $1,000. So, I think I'm not their target market.

DavidNJ said:
I've seen quite a bit of footage from different cameras, and if the Canon is dropping something...it must be pretty irrelevant.
Click to expand...

The canon is dropping lines of pixels to quickly get a 1920x1080 frame out of their full frame sensor. They aren't dithering or otherwise downrezing the full frame, they're just dropping the data. This is, at least, what I've read in numerous forums. The result of this is a moire pattern that can appear in footage and be problematic. I don't deny the 5D produces wonderful images-- I'm just saying if the panasonic is making full use of their 4k wide sensor, and not just dropping data to get an HD frame, then that will help them be competitive with the 5D in image quality.

In both, similar encoding is used. The Canon is clearly display more detail.
Click to expand...

If you're talking about the 5D vs. the GH1, neither of these statements are valid. The second cannot be claimed until we've seen video from the cameras. The first is simply an assumption on your part. Just because two cameras use an H.264 codec does not mean they are encoding to the same quality level or even with the same methods. H.264 provides a huge array of choices in the way you're going to encode your video, how many frames forward and back you look for similarities, what algorithm you use to compare frames to find similarities, etc. In fact there are so many choices one can make in encoding that I can't keep track of them all.

Here, look at the man page for the x264 open source H.264 encoder. All of these options have an effect on the CPU load necessary to encode a given piece of data, and the quality of the resulting bitstream.
http://www.linuxcertif.com/man/1/x264/

This is further evidence that the storage bit-rate is the limiting factor.
Click to expand...

That was a non sequitur. Bitrate is a factor, but that does not mean its the only factor.

If not, the Canon's superior glass and probably superior sensor (minimally 4x the area) should give it the edge.
Click to expand...

The first thing the Canon does is throw out %75 of the pixels in the image. Then it compresses the rest, and it does so at a very high bitrate, requiring minimal CPU horsepower. The high bitrate is to compensate for the lack of encoding horsepower on the camera.

Given the GH1 has 33% more time to encode the information, probably doesn't throw out %75 of the pixels, has a duel core CPU, one of which is probably dedicated H.264 encoding hardware, and can use Canons "superior" glass with an adapter, there's no reason to believe, at this point, that the video will be inferior to the 5D.

Just looking at the encoding pipeline, the GH1 should probably deliver about %266 more quality per megabit than the canon, more than making up for the differences in bitrate. (Figuring the dual CPUs give twice the horsepower and the %33 longer time to encode being multiplicative.)

We'll see how good a job Panasonic did here. But you can't judge quality on bitrate alone.
 
Last edited:
I honestly think the video quality should be as good as hdv or better but not to the level of the Canon 5D. The compression and lenses and codec are pretty amazing. I think we are asking for to much from the GH1. Canon codec is around 40 Mbit/s but it is 30p so if it was 24 frames it would be around 32 Mbit/s. Still almost having double the bit rate of the GH1. Canon just has a fatter straw period. It seems that if you were ok with hdv and you just wanted some killer dof this is the camera for you. No 35mm adapters. Cheap media and long record times. So nice.
 
Hey Barry about 5 months ago you said...
2) it's a lot more efficient. AVC-HD at 8 megabits can about match the picture quality of HDV at 25 megabits. AVC-HD at 24 megabits should be able to match XDCAM-HD at 35 megabits.
Click to expand...

Do you feel that way right now? If not what do you think avchd is at compared to hdv now?
Do you think the 17 megabit avchd codec is as good as sony's 35 megabit mpeg2 codec?


Thanks
 
According to Sony, 9 megabits of AVCHD matches 25 megabits of HDV. So when I said 8, I might have been a tad optimistic, but not all that much.

I would say that yes, definitely, 17 megabits of AVCHD should match XDCAM-HD or XDCAM-EX at 35 megabits. Thing is, you can only compare 60i to 60i, becaue AVCHD @ 17mbps is only 60i. EX @ 24p is going to look noticeably better than EX @ 60i, because progressive is inherently so much easier to compress than interlaced.

AVCHD @ 21-24mbps is easily a match for XDCAM-EX at 35mbps.
 
Nik Manning said:
I honestly think the video quality should be as good as hdv or better but not to the level of the Canon 5D. The compression and lenses and codec are pretty amazing. I think we are asking for to much from the GH1. Canon codec is around 40 Mbit/s but it is 30p so if it was 24 frames it would be around 32 Mbit/s. Still almost having double the bit rate of the GH1. Canon just has a fatter straw period. It seems that if you were ok with hdv and you just wanted some killer dof this is the camera for you. No 35mm adapters. Cheap media and long record times. So nice.
Click to expand...
I can't agree more. I think this cam's picture should match up perfectly with my HV20. The thing is if I wanted some of this cam's capability on the HV20 I would have to slap on an adapter further degrading the picture. If the codec holds up nicely like I think it will then I believe I would be one satisfied customer.......until Scarlet that is.
 
I don't think the Canon is throwing out 75% of the pixels. The interpixel gaps would be so huge as to have other artifacts. Also, it would lose its low light advantage, which is clearly not the case.



It does throw away 19% of the pixels, since 16x9 can only use 3159 rows.

Yes, there are different algorithms used in H.264/MPEG4 Part 10. However, when you define it as High Profile Level 4.1 many are specified. I doubt Canon is giving up much of an advantage to Panasonic here, other than CABAC vs. CAVLC.

The down loaded Canon clips I have viewed have had bursts to 45Mb/s, and sometimes sustained periods in the 44Mb/s range. At least according the VLC.

I thought the moire pattern were because of reflections north of the Nyquist limit. Possibly because the optical low pass filter is set for 5600x3700 resolution?
 
Barry_Green said:
According to Sony, 9 megabits of AVCHD matches 25 megabits of HDV.
Click to expand...

So, H.264 is giving AVCHD a better than 2:1 efficiency advantage over the MPEG-2 video in HDV according to Sony. I'm sure they have a good basis for saying that. As a rule of thumb, 2:1 is the ratio I use, but of course video compression is an art.

DavidNJ said:
I don't think the Canon is throwing out 75% of the pixels. The interpixel gaps would be so huge as to have other artifacts. Also, it would lose its low light advantage, which is clearly not the case.
Click to expand...

Dropping every other line horizontally and vertically means removing %75 of the information, but would only leave a single pixel gap between the pixels... this would be the same as producing the same sized sensor, and same sized pixels, but with lower pixel density. If you can find details on the method canon uses to get the 1920x1080 frame, I'd be happy to see it as its possible I'm wrong. I've seen it asserted many times that this is the case and have not seen the contrary, though not from authoritative sources.

The DIGIC 4 image processor used in the 5DmkII is the same one canon uses in its SX1, and is Powershot Line, specifically the G10 and the SX10 which have been out for while. These two cameras both capture H.264 video, however they do so at a maximum resolution of 640x480! The 5D has the same CPU, and thus is woefully underpowered for compressing HD-- a huge step up from 640/30p to 1080/30p. The only way to make it work is to do the bare minimum possible compression and have a very high bitrate. And this is what they did.

This may be the source of the video time limits-- I expect the CPU is overheating. I don't think its unreasonable to believe that people are right when they say the pixels are dropped to downrez-- this CPU is very busy trying to encode H.264 at HD resolutions, and all the other things it needs to do, and resolution reduction would just add to that load.

All of these are reasonable compromises for Canon to make and they still result in good image quality. Panasonic didn't put video in the G1, they waited, and they produced a camera with a new, dual core CPU and a new sensor. They stated their intention to not "make the same mistakes" of other DSLR manufacturers with regard to video. So, they clearly intend to produce a fabulous video image... and they have the advantage of time and focus on video.

Yes, there are different algorithms used in H.264/MPEG4 Part 10. However, when you define it as High Profile Level 4.1 many are specified.
Click to expand...

Like most standards, MPEG4 is focused on interoperability, without tying people down to a technology that may become less ideal in the future. Thus these profiles focus on coding of data, not the algorithms used to come up with the data. Sometimes the coding affects algorithm choices (CABAC vs. CAVLC) but often it does not.

The profiles set the way you can store the image data, but they do not determine how much horespower or what methods one uses to come up with that data.
 
Last edited:
In any system, the limiting factor is bottleneck limiting the others. If the limiting factor is the recording quality, then all the lenses and sensors in the world won't make a difference. One comparison of an HDMI capture vs. AVCHD on an HV20 or SR11 (downloads of both available on vimeo) will show AVCHD does limit ability. HDV much more. I would say slightly less than 2x...maybe a range of 1.5-2x, which is why I don't like the XDCAM EX...I think it is probably the limiting factor on a great camcorder (the EX1/EX3).

Those small camcorders still have limited texture detail...a probably limitation of their sensor size and lens quality.

The 5D has excellent texture detail and picture edge sharpness. That means it was captured by the lens, through low pass and Bayer filters, the sensor, and the recording format. Since it doesn't have a direct HDMI output, we will never be able to compare it directly.

The HMC150 does 25Mb/s and looks comparable (IMHO) to the HPX170. However, while both have good contrast neither blows you away with edge sharpness or texture detail.

So, we know 17Mb/s doesn't cut it from cameras with far inferior optics and sensors. We know whatever Canon does, it works. However, without manual adjustments, and limited frame rate and other choices it is the unsuffixed DVX100.

If the GH-1 could record the image, it may be a suffixed (DVX100a/b or even SDX900) camera, setting new standards.

That Panasonic only posted those very low resolution clips 3 months before projected release date implies that not be the case. And that is very disappointing.
 
LizaWitz said:
H.264 is a very flexible codec, in fact its not even necessarily fixed bitrate. It can be variable bitrate. I think many people are still hung up on the days when the format was DV and the encoding was "fixed compression" and the bitrate determined quality.

This is not the case anymore. The quality of an H.264 stream is dependant on several factors, including bitrate, but also including the complexity of the encoding being done, the horsepower of the CPU doing the encoding and the time available to he CPU to do that encoding.

The 5D drops much of the sensor information and barely encodes the video and so it needs a massive bitrate to get by--- because it doesn't apparently have the horsepower to encode the video in real time at higher quality, it makes up for it with bitrate.

Its quite possible that the GH1 will encode video at a higher quality than the 5D, even though the bitrate is much lower. Two reasons: Superior input: The GH1, by all accounts, reads the full width of the frame-- a 4k image, and reduces that down to the 1920x1080 video frame. The 5D just drops all that info. So, better input quality. Secondly, the GH1 has a dual core CPU, and probably one of those cores is a dedicated H.264 encoder. I believe the 5D just had video added on late in development as a response to nikon-- its probably encoding in software, and thus relying all on bitrate. Finally, the GH1 has %33 longer time between each frame to do the encoding. That goes directly to the quality bottom line.
Click to expand...
very well.. how do you know all this?.. your crystal ball?.. or any secret Panny information from inside? lol

LizaWitz said:
It remains to be seen what the GH1 video looks like, but we can't draw many conclusions at all from the 17Mbps bitrate. This isnt' tape.

Frankly, this camera delivers what the scarlet promised-- a 4k camera with interchangeable lenses great for shooting HD. Course, not in raw mode, etc, but its a whole lot cheaper....and it will be shipping in months.
Click to expand...
where did we hear it?.. :D great 4k camera slogan btw :beer:
 
joe 1008 said:
One feature that I'm curious about is that aut follow focus. If it works fine and precise, doors open up to a different shooting style: Run & Gun with shallow DOF.
Click to expand...

... But it's been this way for decades? Real cameras...
 
filmmaker's gang said:
very well.. how do you know all this?.. your crystal ball?.. or any secret Panny information from inside? lol

where did we hear it?.. :D great 4k camera slogan btw :beer:
Click to expand...

Yeah I'm wondering who Liza is as well, that info sounds like it's insider source. LoL

Keep it coming, though!
 
You must log in or register to reply here.
Back
Top