If the image looks like crap there's no point in investing money in the system. By saying EFFECTIVE you're saying that that the image quality is better than or as good as the MKII. You're pretty sure about that, and if it isn't then you're going to lose a lot of credibility.
Panasonic Removable Lens Four Thirds HD AVCHD Camera
- Thread starter DavidNJ
- Start date
If the image looks like crap there's no point in investing money in the system. By saying EFFECTIVE you're saying that that the image quality is better than or as good as the MKII. You're pretty sure about that, and if it isn't then you're going to lose a lot of credibility.
DavidNJ
Carbonite Member
Lisa, we are back to where we have seen, and stills are posted here from various cameras using the same H.264 encoding. Panasonic uses it already in existing products; there H.264 encoding is no better than anyone elses.
The Canon, the SR11, the HF S10, the HMC150 all use it. And they all show about the same thing.
It is very good compared the MPEG2...BUT ALL THE CAMERAS WE ARE TALKING ABOUT HAVE THE SAME ENCODING.
So, unless you can produce some EVIDENCE that this is magical H.264, you claim holds no water.
There have been NO good clips posted, including by Panasonic, taken by this camera. The few that have been posted are all mediocre...just like you would expect from the recording. Note that the two 25Mb/s cameras, the HMC150 and the HF S10/100, are all pretty impressive. The 45Mb/s 5DM2 and even small sensor SX1 are pretty impressive; they also point out the differences of the 5DM2's large sensor and good glass. On the SD780 (recording 720p30@25Mb/s it still looks bad, probably really showing that cameras small sensor and lens issues.
Unfortunately, the effecrtive equivalent of 17Mb/s 1080p30 never lets the equivalent depth come through. It is barely appropriate for a 1/2.5" sensor and tiny cheap glass.
The encoding quality would have to be a SWAG of a 4 sigma outlier in quality. None of the clips posted support that.
The Canon, the SR11, the HF S10, the HMC150 all use it. And they all show about the same thing.
It is very good compared the MPEG2...BUT ALL THE CAMERAS WE ARE TALKING ABOUT HAVE THE SAME ENCODING.
So, unless you can produce some EVIDENCE that this is magical H.264, you claim holds no water.
There have been NO good clips posted, including by Panasonic, taken by this camera. The few that have been posted are all mediocre...just like you would expect from the recording. Note that the two 25Mb/s cameras, the HMC150 and the HF S10/100, are all pretty impressive. The 45Mb/s 5DM2 and even small sensor SX1 are pretty impressive; they also point out the differences of the 5DM2's large sensor and good glass. On the SD780 (recording 720p30@25Mb/s it still looks bad, probably really showing that cameras small sensor and lens issues.
Unfortunately, the effecrtive equivalent of 17Mb/s 1080p30 never lets the equivalent depth come through. It is barely appropriate for a 1/2.5" sensor and tiny cheap glass.
The encoding quality would have to be a SWAG of a 4 sigma outlier in quality. None of the clips posted support that.
Last edited:
Park Edwards
Veteran
it's not when you consider putting a 35mm adapter on the front and the amount of light loss you get. plus, 17mbps vs 24mbps may be an issue for guys who have clients and want the best possible picture there is, but not really an issue for me. it doesn't seem significant enough. I think the d90's footage looks good and it's bitrate blows. I'll admit to that. The canon mK II looks great, but it lacks manual control over shutter and no 24p. These two things are essential in my eyes. Not 7mb's. Can someone really see the difference in 7mb's?
Gordon Prince
Well-known member
Which cameras?
John Caballero
Veteran
Originally Posted by DavidNJ
There have been NO good clips posted, including by Panasonic, taken by this camera. The few that have been posted are all mediocre...
That is exactly the reason that we can not analyze and discuss this camera logically. Until better clips start showing up and somebody has one in their very own hands and reports with authority about it we are wasting our time trying to get to any conclusions.
Gordon Prince
Well-known member
Why do you infer David is just referring the GH1?
He posted:
"There have been NO good clips posted, including by Panasonic..."
"Including" should mean also others, no? That's why I asked: "which ones?" On the others, you have a lot of clips for analyze.
DavidNJ
Carbonite Member
There were lots of clips taken at PMA. The high speed Casio clips (210fps or 60fps at 6Mp) look very good. They hired the UNLV cheerleaders, and provided a lit area and cameras to take pictures. Panasonic had a dark corner and nothing to shoot. There are lazy pans around the room, or takes of a Panasonic person talking about a feature. However, all are a bit soft and noisy.
Then there are the clips on the Panasonic global site: http://www.panasonic.net/avc/lumix/systemcamera/gms/gh1/movie.html They are all down scaled to postage stamps more or less. Not that Panasonic always does that. The STILL images have a full download:http://www.panasonic.net/avc/lumix/systemcamera/gms/gallery/g1.html (GH-1)
Panasonic did have clips running on some large LCD monitors. People reported they looked good. 'Like TV' However, is 'TV' the goal for raw footage? 5DM2 footage, taken by nearly every posting owner, looks more like it was taken with a Varicam. Correction, a Varicam with more texture detail.
Compared to the Canon the Panasonic would appear to fall between a good 2/3rds camera (say HPX500) and the Canon in specification, with one big exception: recording.]
Another HPX 300 thread goes back and forth comparing AVC Intra 100's 100Mb/s H.264 4:2:2 I-frame (@1080p30, bandwidth salled down for other resolutions and frame rates) to the Flash XDRs 100Mb/s GOP or 160Mb/s I-frame 4:2:2 MPEG 2 (constant bandwidth and pulled down storage for all resolutions and frame rates).
This one is easier...there are lots of 17Mb/s clips. It was just a year or so ago that 17Mb/s then later 21Mb/s were released in consumer camcorders. Now we have new camcorders at 25Mb/s and still cameras with the same or more (the Canon 5DM2 and SX1 are 45Mb/s). All with AVCHD.
So here is a reasonably priced camera with interchangable lenses, auto focus in live video mode, a much bigger sensor than any camcorder short of a FF35, a good quality sensor (at $1500 it won't be a great quality sensor) designed for video with a latest generation video processing engine. This should be a great video camera. And I am pretty sure if you looked at the signal coming from the sensor it would be...rivaling the Canon (the sensor design for video and video processing are probably better on the Panasonic, the glass and photosite noise levels are probably better on the Canon).
However, it appears to thrown away in recording. Like using expired film in a $10,000 still camera. However, if it has HDMI out, we could conceivable find out how good it really is. And capture it, for a price. A price double to triple the cost of the camera.
Everyone has conjecture about the next generation Rebel and Nikon. However, the Rebel is likely to have the same Digic 4 processor as the 5DM2. Nikon's current offering uses MPEG1 MJPEG for recording. Samsung has announced a camera, and Olympus has said their next cameras will have video. However, Panasonic, the company if not the division, has a lot more video experience then they do. Not only that, cameras like the DVX100, SDX900, HVX200 and decendents show they have a good concept of cinematography.
So where would the next camera come from? May a D60 Canon with Digic 5? Or a new line of Canon camcorders with an large, oversampled sensor? They have been doing removable lenses since the XL-1 with many used with 35mm adapters. They need to compete with both the still camera division and the video camera competition. It's been a while since I heard anyone ever mention a HDV Canon as a camera to buy. It usually HMC150 vs HPX170 or HPX170 vs EX1, etc., maybe with another Sony thrown in.
We know Panasonic will have AVC Intra with 1080p sensors and cinema features. The GH-1 could have been its match or better. *sigh*
Then there are the clips on the Panasonic global site: http://www.panasonic.net/avc/lumix/systemcamera/gms/gh1/movie.html They are all down scaled to postage stamps more or less. Not that Panasonic always does that. The STILL images have a full download:http://www.panasonic.net/avc/lumix/systemcamera/gms/gallery/g1.html (GH-1)
Panasonic did have clips running on some large LCD monitors. People reported they looked good. 'Like TV' However, is 'TV' the goal for raw footage? 5DM2 footage, taken by nearly every posting owner, looks more like it was taken with a Varicam. Correction, a Varicam with more texture detail.
Compared to the Canon the Panasonic would appear to fall between a good 2/3rds camera (say HPX500) and the Canon in specification, with one big exception: recording.]
Another HPX 300 thread goes back and forth comparing AVC Intra 100's 100Mb/s H.264 4:2:2 I-frame (@1080p30, bandwidth salled down for other resolutions and frame rates) to the Flash XDRs 100Mb/s GOP or 160Mb/s I-frame 4:2:2 MPEG 2 (constant bandwidth and pulled down storage for all resolutions and frame rates).
This one is easier...there are lots of 17Mb/s clips. It was just a year or so ago that 17Mb/s then later 21Mb/s were released in consumer camcorders. Now we have new camcorders at 25Mb/s and still cameras with the same or more (the Canon 5DM2 and SX1 are 45Mb/s). All with AVCHD.
So here is a reasonably priced camera with interchangable lenses, auto focus in live video mode, a much bigger sensor than any camcorder short of a FF35, a good quality sensor (at $1500 it won't be a great quality sensor) designed for video with a latest generation video processing engine. This should be a great video camera. And I am pretty sure if you looked at the signal coming from the sensor it would be...rivaling the Canon (the sensor design for video and video processing are probably better on the Panasonic, the glass and photosite noise levels are probably better on the Canon).
However, it appears to thrown away in recording. Like using expired film in a $10,000 still camera. However, if it has HDMI out, we could conceivable find out how good it really is. And capture it, for a price. A price double to triple the cost of the camera.
Everyone has conjecture about the next generation Rebel and Nikon. However, the Rebel is likely to have the same Digic 4 processor as the 5DM2. Nikon's current offering uses MPEG1 MJPEG for recording. Samsung has announced a camera, and Olympus has said their next cameras will have video. However, Panasonic, the company if not the division, has a lot more video experience then they do. Not only that, cameras like the DVX100, SDX900, HVX200 and decendents show they have a good concept of cinematography.
So where would the next camera come from? May a D60 Canon with Digic 5? Or a new line of Canon camcorders with an large, oversampled sensor? They have been doing removable lenses since the XL-1 with many used with 35mm adapters. They need to compete with both the still camera division and the video camera competition. It's been a while since I heard anyone ever mention a HDV Canon as a camera to buy. It usually HMC150 vs HPX170 or HPX170 vs EX1, etc., maybe with another Sony thrown in.
We know Panasonic will have AVC Intra with 1080p sensors and cinema features. The GH-1 could have been its match or better. *sigh*
Last edited:
Why?
DavidNJ
Carbonite Member
New topic...or question: how much information can we find out about the processing engine?
Here is the reason: For our oversampled single CMOS sensor, we first have to up scale each of the four colors (green was sampled 2:4, red and blue 1:4). That requires one of the scaling algorithms in a prior post. Some smoothing or other anti-aliasing may be applied.
Then the image is converted to YUV and downscalled to 1080p or 720p. Similar algorithm has to be used. Some sharpening may be applied.
Then the Cr and Cb need to be created from a 4:1 down scale. It used to be '1/4 to the left, 1/2, plus 1/4 to right' or some variation of adjoining pixels. Now, people are spending a lot of time on the algorithms. Ultimately, it also depends on the algorithm used in final viewing.
So, in addition to the glass, the sensor, and the recording, there is another factor that can make a significant difference.
The question is has it ever been discussed (most likely in research papers or patents) for the commonly available still and video cameras?
This is one of the big RED differences. RED is the only company (to the best of my knowledge) capturing a RAW mosaiced image on the camera and demosiacing in post. By doing it all in post they are able to use the horsepower of a 100+w multi-core processor without time constraints vs. a battery powered processor with a real time constraint of 1/60th or 1/30th of a second.
Correction: the Silicon Imaging SI2K, used for Slumdog Millionaire also records a RAW mosaiced image (wavelet compressed by Cineform) and is subsequently demosiaced on the computer. Cineform's 2007 writeup on the technology says "DeBayering is an extremely compute intensive algorithm that may take as much as much a 1 - 2 seconds per frame ( 24x to 48x) for a 4K image." The GH-1 has a 4k image.
Here is the reason: For our oversampled single CMOS sensor, we first have to up scale each of the four colors (green was sampled 2:4, red and blue 1:4). That requires one of the scaling algorithms in a prior post. Some smoothing or other anti-aliasing may be applied.
Then the image is converted to YUV and downscalled to 1080p or 720p. Similar algorithm has to be used. Some sharpening may be applied.
Then the Cr and Cb need to be created from a 4:1 down scale. It used to be '1/4 to the left, 1/2, plus 1/4 to right' or some variation of adjoining pixels. Now, people are spending a lot of time on the algorithms. Ultimately, it also depends on the algorithm used in final viewing.
So, in addition to the glass, the sensor, and the recording, there is another factor that can make a significant difference.
The question is has it ever been discussed (most likely in research papers or patents) for the commonly available still and video cameras?
This is one of the big RED differences. RED is the only company (to the best of my knowledge) capturing a RAW mosaiced image on the camera and demosiacing in post. By doing it all in post they are able to use the horsepower of a 100+w multi-core processor without time constraints vs. a battery powered processor with a real time constraint of 1/60th or 1/30th of a second.
Correction: the Silicon Imaging SI2K, used for Slumdog Millionaire also records a RAW mosaiced image (wavelet compressed by Cineform) and is subsequently demosiaced on the computer. Cineform's 2007 writeup on the technology says "DeBayering is an extremely compute intensive algorithm that may take as much as much a 1 - 2 seconds per frame ( 24x to 48x) for a 4K image." The GH-1 has a 4k image.
Last edited:
First off, can you answer my question, please?
DavidNJ
Carbonite Member
The Canon has a bigger area for increased DOF, better glass, and the total photosite area is bigger. The 45Mb/s AVC can capture that.
The Panasonic has more photo sites, a better arrangement of the photo sites (my WAG is 3840x2160 on the GH1, 5616x1080 on the 5DM2), multi-channel reads from the sensor, and parallel processing in the video engine. The difference in sensor photo site size and probably quality will show a little...probably not much in good light. The difference in glass will be harder to detect...the glass with and available for the GH1 is not much worse than the Canon: sort of Mercedes vs Infiniti.
The better photos ite layout and more video designed processor should allow a better video bayer demosiacing and subsequent down scaling to 1080p or 720p. That could easily wipe out any advantage the Canon gained. And if you have live uncompressed HDMI in 4:2:2, it would have much more information spatially and in chroma than the Canon.
Panasonic had the G1...video was the big difference in the GH1. Canon added video to their still camera as an afterthought. However, the guy who designed DIGIC 4 designed if for very high capture speeds. The problem is, the guy who engineered the recording on the GH1 (and his/her supervisors) were asleep at the switch.
The Panasonic has more photo sites, a better arrangement of the photo sites (my WAG is 3840x2160 on the GH1, 5616x1080 on the 5DM2), multi-channel reads from the sensor, and parallel processing in the video engine. The difference in sensor photo site size and probably quality will show a little...probably not much in good light. The difference in glass will be harder to detect...the glass with and available for the GH1 is not much worse than the Canon: sort of Mercedes vs Infiniti.
The better photos ite layout and more video designed processor should allow a better video bayer demosiacing and subsequent down scaling to 1080p or 720p. That could easily wipe out any advantage the Canon gained. And if you have live uncompressed HDMI in 4:2:2, it would have much more information spatially and in chroma than the Canon.
Panasonic had the G1...video was the big difference in the GH1. Canon added video to their still camera as an afterthought. However, the guy who designed DIGIC 4 designed if for very high capture speeds. The problem is, the guy who engineered the recording on the GH1 (and his/her supervisors) were asleep at the switch.
At the moment we're all discussing theory. We've seen footage from the 5DII, but not from a production GH1.
The proof of the pudding, as they say....
FWIW, AP's technical editor, Richard Sibley had a look at the GH1's footage at a Panny event, and reported that it looked pretty good, though not up to the standard of a £2000 HD camcorder, this was watching it on a screen that he said looked 60". I'm willing to concede that this was probably a pre-production model, and the final releasd version will be better, especially for us in PAL land with 25p over 50i.
What's nagging me is that in order to get wide angles with 3rd party glass, we'll have to use seriously wide optics, with the resultant compromises in f-ratio and distortions, and of course, Panny's daft decission only to release the GH1 as a kit with a lump of glass that's of use to neither man nor beast.
Here's the link, Richards comment is near the bottom of the thread www.amateurphotographer.co.uk/forums/showflat.php/Cat/0/Number/768885/an/0/page/0#768885
The proof of the pudding, as they say....
FWIW, AP's technical editor, Richard Sibley had a look at the GH1's footage at a Panny event, and reported that it looked pretty good, though not up to the standard of a £2000 HD camcorder, this was watching it on a screen that he said looked 60". I'm willing to concede that this was probably a pre-production model, and the final releasd version will be better, especially for us in PAL land with 25p over 50i.
What's nagging me is that in order to get wide angles with 3rd party glass, we'll have to use seriously wide optics, with the resultant compromises in f-ratio and distortions, and of course, Panny's daft decission only to release the GH1 as a kit with a lump of glass that's of use to neither man nor beast.
Here's the link, Richards comment is near the bottom of the thread www.amateurphotographer.co.uk/forums/showflat.php/Cat/0/Number/768885/an/0/page/0#768885
Last edited:
DavidNJ
Carbonite Member
His quote is:
A "£2000 HD camera" is a HMC150 or XH-A1. If that were the case, why bother?
More 'bad' news...
First, Steve Mullen's comments on rolling shutter from his PMA clips:
And this from the HPX300 forum here:
Same company, similar techonology...
Richard Sibley on Amateruphotographer.co.uk said:
A "£2000 HD camera" is a HMC150 or XH-A1. If that were the case, why bother?
More 'bad' news...
First, Steve Mullen's comments on rolling shutter from his PMA clips:
Steve Mullen on DVInfo.net said:
And this from the HPX300 forum here:
Same company, similar techonology...
if thats the case thats good for something on a sub 1500$ camera. what the hell are you really expecting on a tiny cam? its doesnt even have 3 sensors to achieve full 3 RGB.
the damn camera is so tiny and you're expecting it to have a processor that can do what pro cams do. Thats just plain idiotic and unrealistic.
by judder im sure he means the judder you get from the 24p. but we cant judge anything yet.
this thread is just getting repetitive with all the technical mumbojumbo that doesnt matter.
The sensor is a damn sight bigger than a 150 / 151's, bigger than 2/3" by quite a way, the fact that the cam itself is small is irrelevant.
If the cam+lens costs £1500+ then I damn well hope it's better than a £5-600 handycam.
As I said earlier, the proof of the pudding....
Until we see footage from a real production model we are all just guessing.
I have the feeling we are hoping for more than the cam is designed to deliver, like the 5DII, and will bellyache when it doesn't. Are we looking for Varicam+ performance for peanuts?
Dave
If the cam+lens costs £1500+ then I damn well hope it's better than a £5-600 handycam.
As I said earlier, the proof of the pudding....
Until we see footage from a real production model we are all just guessing.
I have the feeling we are hoping for more than the cam is designed to deliver, like the 5DII, and will bellyache when it doesn't. Are we looking for Varicam+ performance for peanuts?
Dave
DavidNJ
Carbonite Member
Actually...we can measure judder...the HPX300 is new Panasonic CMOS camera about to ship and reputable people have said the only 1080 mode in the GH-1, as implemented in that camera, is unusable.
As to single sensors...other than say $10k Nikons and $25k Hasselbads, How about these video cameras: Genesis, RED, Phantom, DALSA, Glenn, ARRI, and Silicon Imaging.
size of the camera does matter. you dont expect a intel atom to perform like a quad core cpu. so dont expect a gh1 perform like a hpx300. atleast, not yet. would it be sharper? sure but nothing can be decided until the production is released. and no, dont expect it to be great. it may have a bigger sensor, but it doesnt have a better processor.
-you're not getting my point. a camera that small will never out perform all those cameras you mentioned. they're all different sensors too. Its not like they all buy on one kind of sensor in one shop!
-GH1 wont reach the level you want it to be. so stop bumping the thread with posts that the answers to are obvious or unknown. wait for the next info release.
-i really dont care what hpx300's performance is. They're completely different cameras with completely different hardware and use. plus it's 1/3 inch and has 3 sensors.
DavidNJ
Carbonite Member
You might not care but I do. I (and lots of other people) would say the 5DM2 records a significantly better image than an EX1/3, HPX300, HPX500, etc...Just a pita with no manual settings and tons of archaic workarounds.
You can buy what you want. I am looking for what I consider dramatic high production quality images. I can live without on camera sound because it is easily (and better) recorded off camera and matching is just a little work. I am not willing to make a $5-8k investment in the 5DM2 given its issues.
This camera has (still using has because we don't know about the HDMI) to deliver the goods. Panasonic is working on the skew...the HPX300 (and future HPX200/HMC160) buyers will demand it.
As far as GH-1 vs HPX300: if it had the 45Mb/s AVC of the Canon the GH-1 would probably take better pictures. You could certainly buy good glass for a lot less.
P.S.
The Canon upset my apple cart. Before I could make a commitment I have to know why. It has led to a rather strange learning curve of image quality measurement (those pesky MTF curves, JImage is my friend), interpolation algorithms (NEDI, Lanczos, etc.), Bayer sensor processing, compression algorithms, sensor design, etc. However, this is a restaurant not the fruit section of the supermarket, you have to pick a complete system. The biggest discovery, which the Canon cameras drove home, was the importance of the compression in recording.
P.P.S
If Barry had created a GH-1 forum, this would be 5-10 different threads.
You can buy what you want. I am looking for what I consider dramatic high production quality images. I can live without on camera sound because it is easily (and better) recorded off camera and matching is just a little work. I am not willing to make a $5-8k investment in the 5DM2 given its issues.
This camera has (still using has because we don't know about the HDMI) to deliver the goods. Panasonic is working on the skew...the HPX300 (and future HPX200/HMC160) buyers will demand it.
As far as GH-1 vs HPX300: if it had the 45Mb/s AVC of the Canon the GH-1 would probably take better pictures. You could certainly buy good glass for a lot less.
P.S.
The Canon upset my apple cart. Before I could make a commitment I have to know why. It has led to a rather strange learning curve of image quality measurement (those pesky MTF curves, JImage is my friend), interpolation algorithms (NEDI, Lanczos, etc.), Bayer sensor processing, compression algorithms, sensor design, etc. However, this is a restaurant not the fruit section of the supermarket, you have to pick a complete system. The biggest discovery, which the Canon cameras drove home, was the importance of the compression in recording.
P.P.S
If Barry had created a GH-1 forum, this would be 5-10 different threads.
Last edited: