Ambarella h264 codec chip 15Mb/s+, 1080/720p60 9Mb/s, $1K pro versions and cheap cam. at DVinfo.net

Wayne Morellini · January 16th, 2006, 07:43 AM

Found some more news articles on the range of ambarella full h264 codec chips that they also want to use on $799 cameras. They are planning pro versions of the chip at $1K, data rates of over 15Mb/s and their technology pages quotes 4GB for an hours of 720p/1080i at 60 frames a second, which works out around 9Mb/s (I'm guessing some sort of minimum).

http://www.eetimes.com/news/latest/s...leID=174906908
http://www.edn.com/blog/400000040/post/1080002308.html
http://news.com.com/Upstart+aims+to+...tml?tag=st.num
http://www.soccentral.com/results.as...&EntryID=17247

http://www.ambarella.com/technology/

On their front page there is an image of a camera, my guess is that it might be a production partners camera. Does anybody knows who's camera that is, or who's styling? The only thing I can see at the moment is that the case shares some design cues with the Panasonic SDR-S100 SD camcorder (rectangular centrepiece in the front end etc) but that might only mean that the same designer or 3rd party case manufacturer did it for them.

http://www.ambarella.com
http://reviews.cnet.com/Panasonic_SD...-31481096.html

Somebody just posted this Samsung h264 camera, I don't know if it's related or not:

http://www.dvinfo.net/conf/showthread.php?t=58261
http://www.newswire.ca/en/releases/a.../05/c4989.html

Ron Evans · January 16th, 2006, 09:52 AM

That is a Sony PC1000 Mini DV camcorder. I expect the camcorders using this chip will be the Sanyo and Samsung. The Sanyo is the one that has been announced at $800 and looks to be a really interesting camcorder for the price. Not much different to the Samsung at more than twice the price!!!

Ron Evans

Wayne Morellini · January 16th, 2006, 11:57 AM

We have been discussing it in Samsung/Sanyo threads, and it seems to be around 17Mb/s+ for the Samsung and around 9Mb/s for the Sanyo, and it looks like the Sanyo doesn't use an h264 chip, only a mpeg4, so the Samsung maybe twice as efficient also (if it uses an full h264 chip). Camcorder companies don't always advertise that they are using third party chips, so it might be possible (don't put much on it) Sanyo is actually using a lesser version, as h264 could be called Mpeg4, as it is part of Mpeg4.

Dan Euritt · January 16th, 2006, 01:49 PM

that is a great series of articles, wayne: "Company officials claim that H.264 has, on average, 2.25 to 2.5 times better compression efficiency than MPEG-2 at the same frame rate, frame size and quality level"

that's exactly right, i've seen it myself in the encoding that i do... this thread should be a sticky for everyone who is confused enuf to believe that mpeg2 is "efficient".

if you are wondering why hdv uses mpeg2, it's due in part to sony owning nearly 150 mpeg2 patents... that expensive patent pool is also why sony is backing mpeg2 as a blu-ray hd dvd standard... they are trying to make money off of their intellectual property, regardless of whether or not it's the best choice.

Ron Evans · January 16th, 2006, 02:20 PM

Wayne having read the articles you pointed to the chip comes in three types with a close match for the Sanyo using the second and the Samsung using the highest power version. The givaway seems to be that image stabilization is part of the first two and not the highest power ( Samsung is OIS, Sanyo DIS). I look forward to getting more info as time pass but both these look interesting.

Ron Evans

Wayne Morellini · January 17th, 2006, 08:50 PM

Ron, I would confirm that before buying it though. Maybe video in formation in one of the players can tell if it's a h264 stream or an mpeg4 stream. If it isn't you might be able to buy a h264 camera for the same price shortly after it is released.

Ron Evans · January 17th, 2006, 09:49 PM

Wayne, I will wait until I have seen both. I am in no real hurry as I have a few cameras but they both look very interesting. This technology in a " real" camera would be very attractive though.

Ron Evans

Wayne Morellini · January 18th, 2006, 01:35 AM

I have come across this transcoding article on another board. It shows comparisons between Z1 HDV originals and mpeg2, h264, and wmv at the same bit rates.

http://www.sundancemediagroup.com/ar...frameserve.htm
http://www.sundancemediagroup.com/ar...ing_Part_2.htm

The results are interesting. The Z1 HDV shows extensive blocking, but the chaotic water hides it, and this is even though there is not much a range of different values colour or intensity across large areas of the screen, which should reduce compression load. But viewing. the footage should only result in the waves looking funny, and the foam looking worse.

Their pictures looked a bit strange, they weren't always the same frame, and some of the mpeg4 ones had an increased blur in them compared to the sample frame of less quality. So I think it is possible that the frames in question might have had some blur induced by movement/timing, or something else.

Under magnification the 10Mb/s Mpeg4/h264 does look a little worse, the Mpeg4 does try to cover up the blocking caused by HDV but losses some definition and ads some noise (look around the edges of the froth and look at the grey patch of the riders pants). The Mpeg4 looks like it needs twice the bandwidth, but this could be because the constraints of the mpeg2 encoder are passed onto it, where as if it was used in filming it would be able to make better choices to start with. So maybe 10Mb/s is a lot closer than it looks. But the problem with HDV is that it needs double the bandwidth in the first place. I think the choice for 30 minutes to 4GB (around 17-19Mb/s) by Samsung was a wise one, but I would have higher-quality/sports modes that go closer to 25Mb/s+ if I were them. In comparison, an high end pro camera that had 4:4:4/4:2:2 10bit, and true 1920*1080p24/25, 50-100Mb/s H264 would be wonderful.

I would also like to add something. A lot of people might confuse Mpeg4 and h264 (because it is mpeg4 Part 10) and not all h264 codecs do a good job, because they only support part of h264. I don't not know what the case is here, where ever this is a very good h264 transcoder.

Comparison between mpeg4 and H264 coders:
http://compression.ru/video/codec_co...c_h264_en.html
http://compression.ru/video/codec_co...4_2005_en.html

Forgot to add, is that the transcode article states that the tables in mpeg4 are optimised or smaller data-rates then mpeg2. Obviously we hope they re-optimise them for higher data-rates in the gear we buy.

Ron Evans · January 18th, 2006, 12:26 PM

Wayne I think the test might be more of the capabilities of Vegas and Nero recode to extract a still image from the various data streams!!! I think that we are now entering a time when things may not be as they seem. Certainly I can say that HDV played back from my FX1 component out to my Sony HiScan CRT TV looks wonderful. Played back through the same component cable to my Dell 2405 looks pretty poor in comparison. The same video played back from my PC to the same Dell 2405 ( DVI input) looks good again !!!! Maybe a still frame of the material taken from a Canopus Edius timeline may be better or maybe Liquid could do a better job. They likely all construct the single frame from the MPEG stream differently so one would expect a different result. Interesting times.

Ron Evans

Dan Euritt · January 18th, 2006, 12:43 PM

wayne, that silly "test" has been around for awhile, and it has more holes in it than a piece of swiss cheese... it's little more than an advertisement for vasst and sony vegas.

when spot first announced it on the vegas board, i was all over him about the errors that had been made, but nobody bothered to correct any of 'em.

don't jump to any conclusions about codec quality, based on what you read there.

Wayne Morellini · January 18th, 2006, 10:18 PM

Yes, I thought some of the things were a bit strange, like suggesting you couldn't objectively say much and got different frames one obviously with a lot of blur in it. I ignored most of it and did my own analysis, to see how good the other codecs reproduced the HDV picture (which again is a poor thing to use in a situation like that) or changed it. So, on that point, we can see that it did not quiet keep up with it, but might have made twice as good decisions in the first place if it had been the cameras codec. Can we at least say that a h264 camera should do better than it? What do you guys think of their performance in this way, would 16Mb/s+ be enough?

Wayne Morellini · April 5th, 2006, 01:00 PM

Well, here is the news, there is no news.

March has come and gone, but nothing has been revealed. This is probably not as bad as it sounds, it could be for a number of reasons, and is not too uncommon in the chip industry. Chips from smaller companies tend to be manufactured at another companies chip plant. If a manufacturer misses a submission date for his product, or his slot is moved, then it has to wait. Then again, they could have spotted something they wanted to change, or one, or more, of the camera manufactures might be holding off on announcements.

Joseph Aurili · July 12th, 2006, 04:52 PM

And still no news...

Wayne Morellini · July 16th, 2006, 10:49 AM

I got some news from one of the honchos there, can't say what it is but, I said no news and wait before I found this new information, so I have nothing more to add. Have the spec sheet on the next fov X--3 chip, also private, but it has been posted publicly on a website before and pulled. I hope there will be a camera that uses both these things together (haven't heard any information in this direction, only a wish).

AVCHD seems to be a good bit of news though. The ambarella has spec over 15mb/s, and is the first cheap h264 chip that enables cheap h264 cameras and way ahead of competition). So, I would guess it is possible they might go together.

Speaking of something not so obvious. The Samsung 720p H264 camera is due August, I do not know which chip it uses, maybe not an ambarella one I guess.

Wayne Morellini · July 17th, 2006, 01:06 AM

More news, Sony showing new AVCHD camera next week, and AVCHD being upgraded to 24mbs and hard disk etc here:

http://www.dvinfo.net/conf/showthrea...stpost&t=67127

http://www.dvinfo.net/conf/showthread.php?t=71571

January 16th, 2006, 07:43 AM	#1
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Ambarella h264 codec chip 15Mb/s+, 1080/720p60 9Mb/s, $1K pro versions and cheap cam. Found some more news articles on the range of ambarella full h264 codec chips that they also want to use on $799 cameras. They are planning pro versions of the chip at $1K, data rates of over 15Mb/s and their technology pages quotes 4GB for an hours of 720p/1080i at 60 frames a second, which works out around 9Mb/s (I'm guessing some sort of minimum). http://www.eetimes.com/news/latest/s...leID=174906908 http://www.edn.com/blog/400000040/post/1080002308.html http://news.com.com/Upstart+aims+to+...tml?tag=st.num http://www.soccentral.com/results.as...&EntryID=17247 http://www.ambarella.com/technology/ On their front page there is an image of a camera, my guess is that it might be a production partners camera. Does anybody knows who's camera that is, or who's styling? The only thing I can see at the moment is that the case shares some design cues with the Panasonic SDR-S100 SD camcorder (rectangular centrepiece in the front end etc) but that might only mean that the same designer or 3rd party case manufacturer did it for them. http://www.ambarella.com http://reviews.cnet.com/Panasonic_SD...-31481096.html Somebody just posted this Samsung h264 camera, I don't know if it's related or not: http://www.dvinfo.net/conf/showthread.php?t=58261 http://www.newswire.ca/en/releases/a.../05/c4989.html

January 18th, 2006, 01:35 AM	#8
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I have come across this transcoding article on another board. It shows comparisons between Z1 HDV originals and mpeg2, h264, and wmv at the same bit rates. http://www.sundancemediagroup.com/ar...frameserve.htm http://www.sundancemediagroup.com/ar...ing_Part_2.htm The results are interesting. The Z1 HDV shows extensive blocking, but the chaotic water hides it, and this is even though there is not much a range of different values colour or intensity across large areas of the screen, which should reduce compression load. But viewing. the footage should only result in the waves looking funny, and the foam looking worse. Their pictures looked a bit strange, they weren't always the same frame, and some of the mpeg4 ones had an increased blur in them compared to the sample frame of less quality. So I think it is possible that the frames in question might have had some blur induced by movement/timing, or something else. Under magnification the 10Mb/s Mpeg4/h264 does look a little worse, the Mpeg4 does try to cover up the blocking caused by HDV but losses some definition and ads some noise (look around the edges of the froth and look at the grey patch of the riders pants). The Mpeg4 looks like it needs twice the bandwidth, but this could be because the constraints of the mpeg2 encoder are passed onto it, where as if it was used in filming it would be able to make better choices to start with. So maybe 10Mb/s is a lot closer than it looks. But the problem with HDV is that it needs double the bandwidth in the first place. I think the choice for 30 minutes to 4GB (around 17-19Mb/s) by Samsung was a wise one, but I would have higher-quality/sports modes that go closer to 25Mb/s+ if I were them. In comparison, an high end pro camera that had 4:4:4/4:2:2 10bit, and true 19201080p24/25, 50-100Mb/s H264 would be wonderful. I would also like to add something. A lot of people might confuse Mpeg4 and h264 (because it is mpeg4 Part 10) and not all h264 codecs do a good job, because they only support part of h264. I don't not know what the case is here, where ever this is a very good h264 transcoder. Comparison between mpeg4 and H264 coders: http://compression.ru/video/codec_co...c_h264_en.html http://compression.ru/video/codec_co...4_2005_en.html Forgot to add, is that the transcode article states that the tables in mpeg4 are optimised or smaller data-rates then mpeg2. Obviously we hope they re-optimise them for higher data-rates in the gear we buy. Last edited by Wayne Morellini; January 18th, 2006 at 07:57 AM.*

January 16th, 2006, 09:52 AM	#2
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	That is a Sony PC1000 Mini DV camcorder. I expect the camcorders using this chip will be the Sanyo and Samsung. The Sanyo is the one that has been announced at $800 and looks to be a really interesting camcorder for the price. Not much different to the Samsung at more than twice the price!!! Ron Evans

January 16th, 2006, 11:57 AM	#3
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	We have been discussing it in Samsung/Sanyo threads, and it seems to be around 17Mb/s+ for the Samsung and around 9Mb/s for the Sanyo, and it looks like the Sanyo doesn't use an h264 chip, only a mpeg4, so the Samsung maybe twice as efficient also (if it uses an full h264 chip). Camcorder companies don't always advertise that they are using third party chips, so it might be possible (don't put much on it) Sanyo is actually using a lesser version, as h264 could be called Mpeg4, as it is part of Mpeg4.

January 16th, 2006, 01:49 PM	#4
Dan Euritt Trustee Join Date: Mar 2004 Location: Carlsbad CA Posts: 1,132	that is a great series of articles, wayne: "Company officials claim that H.264 has, on average, 2.25 to 2.5 times better compression efficiency than MPEG-2 at the same frame rate, frame size and quality level" that's exactly right, i've seen it myself in the encoding that i do... this thread should be a sticky for everyone who is confused enuf to believe that mpeg2 is "efficient". if you are wondering why hdv uses mpeg2, it's due in part to sony owning nearly 150 mpeg2 patents... that expensive patent pool is also why sony is backing mpeg2 as a blu-ray hd dvd standard... they are trying to make money off of their intellectual property, regardless of whether or not it's the best choice.

January 16th, 2006, 02:20 PM	#5
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	Wayne having read the articles you pointed to the chip comes in three types with a close match for the Sanyo using the second and the Samsung using the highest power version. The givaway seems to be that image stabilization is part of the first two and not the highest power ( Samsung is OIS, Sanyo DIS). I look forward to getting more info as time pass but both these look interesting. Ron Evans

January 17th, 2006, 08:50 PM	#6
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Ron, I would confirm that before buying it though. Maybe video in formation in one of the players can tell if it's a h264 stream or an mpeg4 stream. If it isn't you might be able to buy a h264 camera for the same price shortly after it is released.

January 17th, 2006, 09:49 PM	#7
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	Wayne, I will wait until I have seen both. I am in no real hurry as I have a few cameras but they both look very interesting. This technology in a " real" camera would be very attractive though. Ron Evans

January 18th, 2006, 12:26 PM	#9
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	Wayne I think the test might be more of the capabilities of Vegas and Nero recode to extract a still image from the various data streams!!! I think that we are now entering a time when things may not be as they seem. Certainly I can say that HDV played back from my FX1 component out to my Sony HiScan CRT TV looks wonderful. Played back through the same component cable to my Dell 2405 looks pretty poor in comparison. The same video played back from my PC to the same Dell 2405 ( DVI input) looks good again !!!! Maybe a still frame of the material taken from a Canopus Edius timeline may be better or maybe Liquid could do a better job. They likely all construct the single frame from the MPEG stream differently so one would expect a different result. Interesting times. Ron Evans

January 18th, 2006, 12:43 PM	#10
Dan Euritt Trustee Join Date: Mar 2004 Location: Carlsbad CA Posts: 1,132	wayne, that silly "test" has been around for awhile, and it has more holes in it than a piece of swiss cheese... it's little more than an advertisement for vasst and sony vegas. when spot first announced it on the vegas board, i was all over him about the errors that had been made, but nobody bothered to correct any of 'em. don't jump to any conclusions about codec quality, based on what you read there.

January 18th, 2006, 10:18 PM	#11
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Yes, I thought some of the things were a bit strange, like suggesting you couldn't objectively say much and got different frames one obviously with a lot of blur in it. I ignored most of it and did my own analysis, to see how good the other codecs reproduced the HDV picture (which again is a poor thing to use in a situation like that) or changed it. So, on that point, we can see that it did not quiet keep up with it, but might have made twice as good decisions in the first place if it had been the cameras codec. Can we at least say that a h264 camera should do better than it? What do you guys think of their performance in this way, would 16Mb/s+ be enough?

April 5th, 2006, 01:00 PM	#12
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Well, here is the news, there is no news. March has come and gone, but nothing has been revealed. This is probably not as bad as it sounds, it could be for a number of reasons, and is not too uncommon in the chip industry. Chips from smaller companies tend to be manufactured at another companies chip plant. If a manufacturer misses a submission date for his product, or his slot is moved, then it has to wait. Then again, they could have spotted something they wanted to change, or one, or more, of the camera manufactures might be holding off on announcements.

July 12th, 2006, 04:52 PM	#13
Joseph Aurili Major Player Join Date: Feb 2006 Posts: 326	And still no news...

July 16th, 2006, 10:49 AM	#14
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I got some news from one of the honchos there, can't say what it is but, I said no news and wait before I found this new information, so I have nothing more to add. Have the spec sheet on the next fov X--3 chip, also private, but it has been posted publicly on a website before and pulled. I hope there will be a camera that uses both these things together (haven't heard any information in this direction, only a wish). AVCHD seems to be a good bit of news though. The ambarella has spec over 15mb/s, and is the first cheap h264 chip that enables cheap h264 cameras and way ahead of competition). So, I would guess it is possible they might go together. Speaking of something not so obvious. The Samsung 720p H264 camera is due August, I do not know which chip it uses, maybe not an ambarella one I guess.

July 17th, 2006, 01:06 AM	#15
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	More news, Sony showing new AVCHD camera next week, and AVCHD being upgraded to 24mbs and hard disk etc here: http://www.dvinfo.net/conf/showthrea...stpost&t=67127 http://www.dvinfo.net/conf/showthread.php?t=71571