Is there any confirmation on the uncompressed analogue out?

[Wayne Morellini]

I have heard it claimed the new single chip Sony HDV cameras will have uncompressed analogue out, has anybody got confirmation of this, any links?

[Radek Svoboda]

It will, chesk the thread on the Japan brochure.

Radek

[Wayne Morellini]

Thanks, I have seen it, but I need firm proof of this capability before I get anymore excited. That is why I am posting to see if anybody knows of any other forums that may have confirmation, or projects, like the FX1 has.

[Radek Svoboda]

http://hdvforever.com/hdv/hvra1j/default.htm

Check the bottom of the brochure:
3. Analog component Output terminal Making use of the attachment cable, analog component output is possible. (Input is not possible)

Radek

[Wayne Morellini]

Hi Radek

Thanks, but I am meaning proof that the component's output is in a high quality uncompressed format.

I have looked up the threads on uncompressed output on the FX1 and Z1, and all is far from completely clear. For one, I don't trust these guys to test it properly, many of there tests lack some logical refinement. There is split opinion on wherever the Z1 does have it, and the FX1 doesn't, some saying neither, some saying both as well. Some people might be testing output from tape, they also are chucking it through cineform, but cineform compression is playing up because their computers have inadequate power. I don't think any of them have even tried to put fast complex detail movement is front of the camera to cause compression to break down, and compared the results simultaneously through Firewire and component output. The other issue is that output canbe sabotaged so as not to compete. Such as low Signal to Noise ratio (less than 8-bit 48db, but pref more to get around digitisation problems. Note: most cameras are such poor quality they will probably not get this much from any output, let alone 10 or 12 bits). Colour balance/shift etc are also other problems. There are reports that levels are coming out incorrect, too low, too hot, not situated (compared to cameras likely to be over saturated, so not good comparison) but to have true levels and nice latitude in the colours you would expect them to be more neutral. All that stuff about lack of delay proving that there is no compression, take it with a pinch of salt. The truth is the delay is purely a function of processing power, and you can get a high end Pentium PC's worth of HDV compression power on a chip that will fit on your fingertip. So the delay canbe very short, they could even partly process the image for the component output. They don't even have to compress, they can simply nuke levels on the fly (no discernible delay needed, maybe one frame). There is reports that the Output is 3:1:1, 2:1:1, 1.6:0.8:0.8 etc, so we don't know precisely what format is being outputted. The problem is that for people with the right equipment, and the right expertise, that this is a really easy to test this. If you were careful you should be able to find most of this out from live monitor images (that is, if your digitiser is upto scratch and doesn't have poor Signal to Noise ratio, even if it is the same signal to noise ratio (SN) it could add to the outputs SN to bring the recorded SN down, that is one of the reasons you need SN overhead on both output signal and digitizer). Remember also, don't trust data sheets for SN figures, measure it at the output, it maybe vastly different, the weather, and temperature, cables/connects could also effect it. So it is best to look up proper electrically bench tested reviews. On the CMOS Altasens camera (the best on the market) they report something like upto 96db or nearly 100db SN on some settings, so there is hope that the new Sony CMOS will do at least 48db on the sensor chip. So what do you get out of it, more detail, less movement artifacts, hopefully more correct color levels and balance (though they can easily adjust those on the way out and reduce the colour range).

That is why I prefer not to work with component out, but if it is done right it canbe good enough. The other reason I am interested in this camera is for easy use in documentary making. You will notice that none of the RAW camera projects are setup for easy on the fly shooting like a camcorder, but for serious shooting (not ENG but for cinema) the HC1 with adaptor canbe setup for both. So it might even turn out to be a second camera (still looking at the Sumix).

I understand the component to HDSDi capture devices, that everybody is so excited about, costs around $1500, this is the price of the camera, without even the capture computer. So, I have heard there are cheap component input devices (forget the HDSDi stuff for now) so I will have to google that. I have followed a link to a component over CAT% Ethernet cable device ($180) I am currently sending off an email, to find out if it converts it into an Ethernet electrical signal that an Ethernet port can read, but then you still need recording software. Also you remember that VIA chipset (CN400) that has direct video input port (probably only 8-bit, but the latest processors will be 2Ghz, the main boards 17cm, or maybe 12cm square, and if we are luck there is a HD recording software ability might be inbuilt. I was advised on this chipset solution by a VIA rep last year, but it didn't have the features we were looking for in the RAW camera projects. I think that rigging the computer capture software to read component and control recording through the Firewire camera control signalling might be the best way. It might take me months to get around to researching those things.

The two problems we face is that most cheap equipment is consumer grade, and thus likely only to offer 8-bit digitisation. The camera might not even be up to 8 bits through the component.

I will give you a word of advise on component. Human vision starts to integrate levels around 4-bits (16 levels) so that transitions look smoother (first RCA digital TV worked at 4bit in early 80's) looks reasonable at 5-6bits (30db), it maxes out around 7 bit (128 levels) on a certain shade of green (or was that yellow) but because human vision levels are not linear like video, we use 8bits to compensate. 4-bits = 24db, any camera below this you should consider throwing away. I think you can even spot leveling on 8 bits. I think the record industry uses 20db overhead to guarantee 96db quality (so that's 126db) so without trying to calculate it, lets assume it is similar for us. So ideally, if you had 68db camera to capture process, you are doing well. You will also note the extra overhead comes in handy for low light solution, 48db should give you near optimum results (I usually factor in 6db more from digitisation theory) the extra 20db would allow you to put the gain up in low light situations (maybe 20db) and maintain a optimal image, but you also have another 18db before you hit 5-bit, and 24db before you hit 4-bit to play with. If the sensor does even 14 bits, but only has the SN for 8-bits (or the component DA converter has less) most of that extra in the 14bits will be wasted. And remember that the bits have nothing to do with the latitude the camera can reach, only detail. For a comparison the micron cameras (elsewhere on the forums) have SN around the high thirties to mid/low forties (depending on model), 10 or 12-bits, and latitude might be around 7-8 stops. I have heard it said that the FX1 has around 5.5 stops. as a comparison. By the looks of the sample images from the Sony brochure (Glass skyscraper, maybe) and the sample shots (with rapid leveling of bright skin showing some 8- bit to tape clipping) I think it is possible that the sensor has a stop range (in optimal lighting) of somewhere between 10-13 stops. Actually I think we should check the images again, to find out if there is any noise in the black areas of that village shot. No discernible grain (monitor adjusted to have true back level and true white level) might indicate the sensor can do at least 48db S/N. I think camcorderinfo might be in for a shock when they test this in low light.

[Radek Svoboda]

Wayne,

I would stay away from new CMOS cameras. If you want save money get FX1 from states for 2,850 USD or something like that, just shoot 60i, use DV Film to create 24p. Good deinterlacing really works. You really get 720p quality from 1080i. The pros are using 730 HDCAM, deinterlace and it's Varicam quality. That you get better progressive by shooting 720p is BS.

FX1(E) has uncompressed analog. There is no delay. There was guy who looked at Service Manual. You have ADC and then tees off to uncompressed output and to processor. CineFrame happens in MPEG2 processor too. Screen only stimulates progressive effect. Delay in FW is caused by 12-15 GOP.

FX/Z1 is only one F-stop slower than F900, 100,000 USD camera. Pixels are 1/2 height, same width of 900. That's as good as you can get. HAD CCD's are used, world's best.

Sony really went out their way to create superb product. It stopped Matsushita in their tracks. JVC's new HDV products were delayed by year. They must seen the quality and could not afford another embaresment.

HD100 was demonstrated in Vegas via uncompressed component. People who seen compressed images said thet they were significantly worse.

The people to ask about quality of component out are Frederic Haubrich, DSE, David Newman and guy who writes HD for Indies. There is link to other sites on his blogg, at top is something like HD24. It is Michael Brennen's site. That guy is excellent DP, owns 900 and Z1, has tested Z1 w/component out. Send him email. Also ask HD for Indoes guy how to get in touch of guy who brought his Z1 for testing to him. HD for Indies guy had some people with their cameras, tested Varicam, F900 and Z1 uncompressed and compressed but will not release his report, probably would embarace one of his sponsors. Anyone has sponsors on his site or does work for some manufacturer, I don't trust his opinion, although it seems DV Info is neutral.

Even Michael Brennen is biased because he's trying push F900 to clients, naturally and some things he says about Z1 are not accurate, like compression level, but he's very honest and will also confirm you that properly deinterlaced 1080i will give same as native 720p quality. He's done that.

As to CMOS, Sony's making as sensitive as CCD now, I believe, but it is 3 MP in 1/3". The FX/Z1 is 1 MP in 1/3", has 3 sensors. The onechipper will have 1/2 pixels green though. I would expect low light performance be 2x worse than they advertise for the CMOS camera. There is some new processing used to up it, also at low light are likely to be less color saturation than on 3-chipper.

Let me know if you find anything.

Radek

[Wayne Morellini]

I think I am liking these camera more and more.

I just looked at this A1 brochure and realised that it might be some dual slope latitude extension scheme, that I have been very eager on.

I also have found some curious note numbers next to the component output chart in the image cropping feature. But I don't know precisely what these note numbers refer to (they maybe are about the cropping/stretching features above it, but the third option is not listed). What do you think.

http://hdvforever.com/hdv/hvra1j/default.htm

If all this is true, it maybe the greatest camcorder yet.

But the problem is that this is not mentioned in the HC1 camera, that has not even cinema mode mentioned:

http://hdvforever.com/hdv/hdrhc1/default.htm

But realistically, they need to drop the price of the A1 to some more reasonable level, $2500 or $2000. At this rate I will be able to buy a HC1 for $1500, put an lens adaptor on it, or buy the A1 for two thousand dollars, that's $2000, more, and nearly $500 more than the $3086 price that Amazon is currently selling the Sony 3 chip HDV camera, bad idea. They must be trying to boost the recent sagging price of consumer/prosumer equipment.

Thanks

Wayne.

[Radek Svoboda]

Wayne,

I think is what Sony is doing is protecting their broadcast market. They freaked when some European commission decided go with 720p50. They pressured them and they backed down. They'll deciding later. Except for minor US market, whole world is 1080i, I read somewhere that Sony and Panasonic split HD stuff with Sony doing 1080i/p30, Panasonic 720p60/1080i. Now Tandberg developed MPEG4 1080i/720p encoders/decoders for European broadcast. It seems Germans will be using MPEG4.

Sony needed to flood market with quality low cost pro 1080i equipment to help them protect 1080i broadcast. I don't think Sony crippled anything on these cameras. The only thing we don't have true 1080p24-30. I bet new Panasonic 200 will not have either.

BTW, Matsushita, which is JVC and Panasonic will be doing same thing to help sell 720p50-60 countries converting to HDTV. We are benefitting. If not for the high end stuff, these two world's largest electronic manufacturers would feed us HD prosumer products slightly better than HD1/10 level. Now we have excellent new JVC HDV coming, excellent Panasonic P2 HD. I seriously believe Sony has better HDV products ready introduce too.

Radek

[Radek Svoboda]

Originally Posted by Wayne Morellini

I also have found some curious note numbers next to the component output chart in the image cropping feature. But I don't know precisely what these note numbers refer to (they maybe are about the cropping/stretching features above it, but the third option is not listed). What do you think.

I believe camera will use whole 1920 pixels without electronic steady shot, with steady shot resolution will drop 25% vertically and horizontally.

Radek

[Radek Svoboda]

Sooner or later the whole industrial world will go HDTV. There is a lot of money to be made. That market segment is a lot more profitable than the consumer, or even prosumer market, although Z1 and FX1 probably cost the same money to make, plus minus couple dollars, so Z1 is a lot more profitable. You'll get better service but the service will be more expensive.

Sony may make 1,000 USD net on FX1, or 30% or 2,500, or 60% on Z1. On F900 they may make 80,000 or 80%. They sell nearly everything station needs for production, post, including software, switches, monitors. They sell packages of everything. They dominate broadcast industry worldwide. If there really is such agreement with Panasonic to split HDTV market, and most Europe goes 720p, will be big blow to Sony.

Sony's new strategy (to push 1080i) is propaganda that future is 1080p with MPEG4 level 10 compression over the same 19 Mbps channel as 1080i MPEG is now.

As to if there was agreement with Panasonic, we don't know, but these two don't compete these two top HD markets.

Radek

[Wayne Morellini]

Originally Posted by Radek Svoboda

Wayne,

I would stay away from new CMOS cameras. If you want save money get FX1 from states for 2,850 USD or something like that, just shoot 60i, use DV Film to create 24p. Good deinterlacing really works. You really get 720p quality from 1080i. The pros are using 730 HDCAM, deinterlace and it's Varicam quality. That you get better progressive by shooting 720p is BS.

Radek

Well that's the issue, it is not better than 1080p, so I might as well by 720p and save myself the extra processing time de-interlacing.

I read all the stuff about delays, and people were reporting delay (confusing isn't it, who do you trust) but the issue is what is the quality compared to the Raw cinema camera we are developing, and the cost. Forgot to mention went through Scott Billups and Mikes Curtis's stuff (HD for Indies). If it can beat a Micron 720p solution (SN and latitude) with uncompressed, and we can find a cheap way to record, then it will do most people. At the moment I think it is likely both cameras have it, but the issue is what will the HC1 give us.

CCD and CMOS:

The F900/F950 is I regret to say, terrible enough to me, CCD is it's problem, seen it used for films. New CMOS gets much better (Altasens for instance). Most CCD cameras probably get around 6-7 stops, high pixel counts knock down latitude, so 5.5 stops for FX1 is possible. The CMOS can head out to a number of F-Stops more (dual slope a lot more) so good CMOS is better than bad CMOS or most good CCD. The dual slope feature I mentioned might be on A1 could give you 13 stops or more, it is used on some CMOS.

So the extra pixels on the HC1 chip (and with A1 dual slope enhancement, if that is what they are talking about) don't necessarily disadvantage it compared to the FX1. There are a few factors the new CMOS chips are excelling in, well capacity, light to electric charge conversion efficiency (QE), and SN (the reasons that Altasens does so well, except it may not have the best well capacity, I don't remember. The Latitude and SN effective the latitude, SN and QE effects the Sensitivity. Do we know who designed this chip, Altasens has a 1/3rd inch coming out about now. CCD's are a mature technology, which means there isn't much advancement they can make on their design. CMOS are only reaching maturity (with Altasens, but double the efficiency might be possible) so they have a bit further to go before there performance increase slows down. But I must admit, I believe low light will be worse. I used to think the filter would effect the low light performance myself. But each pixel is receiving the same amount of light as the same sized pixel in a three chip system, just that the three chip system has many more pixels and they can combine the outputs of neighbouring pixels properly to do pixel binning mode to increase low light sensitivity. But the three chip does have sensitivity advantages too that increase the low light ability, and the filter should loss a bit more light than the 3 chip's prism (but you can't use as wide a aperture on the 3 chip prism). So there might be only a %20-%50 (only a rough guess here) loss with the same size pixel. But the trick that the FX1 uses, is pixel shift, where they can use bigger pixels to achieve the same resolution. Lens aperture is a big factor in light gathering too. So those factors probably account for the 200-300% better low light ability of the FX1, nothing really to do with the CCD chip design. If you replaced the FX1's CCD's with good CMOS you would probably get twice the low light performance. Just that Sony, I guess for intellectual property/market political reasons didn't have CMOS ready last year. But it is interesting to note that you can buy the FX1 at a cheaper price than the A1's retail at the moment. So when are they replacing the FX1? I also agree that the color on the FX1 should be better than the single chip HC1. But I have a possible explanation for the slight washout in the bright outdoor frame grabs. I felt a bit uneasy about my previous explanation then it occurred to me that this is built to accept infrared light for night vision, which means there might not be a IR filter. Without a IR filter (or UV/extreme blue) filter, all the light we don't see reflected off objects outside would go into the camera. To get the max definition each filter colour would have to be dual bandpass to allow the IR through. In daylight this light would wash out the image, but in rooms (like in the can shot) it wouldn't. The problem with image saturation, is that camera companies can increase saturation to mask the true quality of the image. People see bright colours and they say "wow", but it also removes some of the colour latitude/detail that you want for pro editing/effects. So you become able to less push the colours around. They also may unbalance the colour as well.

Component:

As I have never seen Mike (HDFORINDIES) talk about component's correctness or quality, I can only assume that it is high enough that he may not be considering it. He doesn't go too much into things in emails either. Thanks for the links anyway, I will have to get around to them one day. For us we are doing uncompressed untouched RAW, very high quality (compared to all Pro CCD cameras) very correct and balanced levels from digital camera heads. But for all the HDV cameras owners they have no choice, so it doesn't matter how good it is, component is the only thing that is going to give you the detail of uncompressed recording. So, in the main, I don't think people are even measuring these things. It simply doesn't occur to most people, they probably think "uncompressed out, it must be better". But for me it matters (less so now, as the sensor/features appear to be pretty good) compared to investing the money in a Sumix Altasens solution.

[Wayne Morellini]

OK I got through the remaining posts:

Yes that is my thinking, though there maybe some crippling of these cameras (no true progressive, 1440*1080 CCD, slight colour enhancement etc) but I was wondering why they hadn't crippled various things, particularly on these new cameras (notice relatively low saturation, good latitude, good features, component and good balance).

I make no comment on companies splitting things, or on where companies get their brochure photographs from (know a guy locally) as I don't want to get sued. So 720p50 gone, mpeg4, wasn't aware anybody was deviating.

I am interested in the film market as well, so 1080i is what I would use for normal television.

Thanks

Wayne.

[Radek Svoboda]

Wayne,

Don't believe CMOS hype from Altasens, etc. Top CMOS, top CCD have equal quality right now. Sony makes own CMOS, also for Nikon, it's all Sony design. They make CCD's for lot companies. CMOS is not more light sensitive than CCD's right now. FX/Z1 is rated correctly 3 lux. 7 lux rating on new Sony CMOS cameras is with some kind digital processing.

Radek

[Wayne Morellini]

Well if that is the case no electronic camera is worth it. But I don't believe it is, I have seen the shoots, and know how to compare them, and have seen my fair share of sensor (CMOS/CCD) data sheets. If you gave me the specs/data sheet I could even show you what makes the difference.

I have just compared those shoots, and boy it is amazing, not only does this camera see way into the shadows, but way into the highlight. It can't be seen on my monitor at normal setting, even though it is high grade, but I turned the monitor upto max, and went into the black areas of the houses in the village scene at 800% zoom (about 64 times the original size). I could not see obvious grain, though there was stuff there that looked like compression artifactting, but no normal speckle grain. With the extra brightness and magnification, I noticed that the couples (who are getting their photograph taken) that have the strange glowing compressed bright light on there heads, the bright areas are all but fully rendered without burnout. I would not have expected this of a camera of this caliber, the Doco's on this may well be a pleasure to watch compared to PD150 footage.

I have just found a bit of a problem with the lens, it is leading to a slight washout around bright objects, but pretty good. Another, the guy, in the background with the white shirt, it is blown out, but looks natural, compared to how we see things. This is being shoot in the midday sun. I believe that I have found some camera edge enhancement around things too, but not very noticeable. I'm agreeing with you, forget all the CCD stuff, all we got to know is that finally we might be getting a camera with the goods, in the pro-consumer realm.

As you are in the circle I thought I might give you this link for a bit of humour:
http://www.dvinfo.net/conf/showthread.php?t=45971

Done the can shot again, and can see lots of grain in there (apart from discoloration on table). I would like to know the lighting levels where it was shot, there are some strong lighting around as can be told in the reflections on the can. Would anybody like to bet me as to what the SN would be? It must be using gain, but the store could be dim or bright. My guess is the SN at the head of the camera is over 48db, but not too much over. Maybe not so crash hot for light work.

[Régine Weinberg]

As nobody I do guess had hands on this camera and all we do have are some screen shots is this anygood ??
But when it's out, a tech doc make sany sense ?

If we could get raw out as with any proto PCB board with a Cmos on it
we could let's say with a PSP3 convert to AVI...why not.

The PSP3 thing drives me mad.
As you may know Dynebolic comes with any video editing and Audio editing and will nest or dock on any other Linux. Imaging 3 PSP3 networked, all with Dynebolic that's open Mosix and you have the Cinerella render cluster for less than any dual CPU AMD box.

We would be a lot of steps further than with any Silicon-Sumix solution.
A real Camera with a viewfinder, power management, calibration etc.

With the Mini35 crew somebody took the lens of a DV cam of and fitted
an Nikon adapter to it... could be a nice Indie cam poor mans Kinetta