New CMOS sensors used in low cost popular camcorders (Sony HC-3/5/7 and Canon HV-10/20) can produce exellent quality in resolution 1920x1080p60.
We only need solution like HYDRA or ANDROMEDA http://reel-stream.com/
and DVR "santa rosa" based with Cineform HDlink to capture into CineformRAW.

With LOW COST GIGe no need HDMI capture cards, short cables and expresscard adapters for notebooks!

I'm of the same opinion, if GigE can handle the high data throughput of uncompressed HD, then it's a great choice, as the interface is already incorporated into most new laptops.

I'd like to see an HDMI -> GigE adapter, obviously the quality wouldn't be quite as good as a direct feed off the CMOS sensor, bit it'd make it far easier to install, (eliminating the need to open up the camera), which should also bring the costs down.

No need HDMI-to-GIGe! Straight from CMOS sensor in same way as SI2K.
Please, read about advantage of Cineform RAW: http://www.cineform.com/technology/CineForm_RAW.htm

David Newman said year ago about possibility to implement CFRAW into FPGA design. In this case no need a bulky dvr on PC based. Just as Elphel but with good Canon CMOS sensor http://dvinfo.net/conf/showthread.php?t=63677

P.S. Simon, but if you want converter... http://www.svideo.com/ext-hdtv-cat5.html

1920 x 1080 30p at 10 bit on the hv20 is 593 megabits per second. this would barely be doable on gigabit ethernet ONLY if it was pci express based (many new controllers are, most old ones are not). pci based gig e has to share the bus with other devices like usb (and often hard disks on notebooks) and would likely drop frames. 60p would not be possible on any single gig-e connection.

1920 x 1080 12 bit 30p would stress the ethernet connection and probably have dropped frames every so often.

then you have the issue of cmos sensors and their affinity for total integration. many of the newer sensors (unsure about the canon) do not output raw at all. they output a processed image right off the chip. not that this is particularly bad, but it would provide no advantage quality wise over hdmi. it will also be a much larger signal, unable to stream over gigabit ethernet.

an ethernet solution would be very usefull overall, but we will need to wait for 10g to proliferate before it becomes really viable for quality HD. especially for mods to consumer cameras where you have little to no control over output.

terence.

This is most interesting and worth looking into, especially with some speculation as to whether HDMI outputs 1920 after
it has already been downrezzed to 1440 and back up to 1920.

1920 x 1080 30p at 10 bit on the hv20 is 593 megabits per second. this would barely be doable on gigabit ethernet ONLY if it was pci express based (many new controllers are, most old ones are not). pci based gig e has to share the bus with other devices like usb (and often hard disks on notebooks) and would likely drop frames. 60p would not be possible on any single gig-e connection.

1920 x 1080 12 bit 30p would stress the ethernet connection and probably have dropped frames every so often.

then you have the issue of cmos sensors and their affinity for total integration. many of the newer sensors (unsure about the canon) do not output raw at all. they output a processed image right off the chip. not that this is particularly bad, but it would provide no advantage quality wise over hdmi. it will also be a much larger signal, unable to stream over gigabit ethernet.

an ethernet solution would be very usefull overall, but we will need to wait for 10g to proliferate before it becomes really viable for quality HD. especially for mods to consumer cameras where you have little to no control over output.

terence.

Terence, thank you for your opinion. Another solution instead of RAW via GIGe
come from Rai Orz, developer of Drake Camera "RAW direct to HDD":

Take your brain free from all PC stuffs. The A/D Unit, inside (or outside the CMOS Sensor) have 10, 12 or more output pins (each pin is one BIT). The output rate is 33-66MHz. FPGAs can handle those data speeds in realtime. Shift or translation to 16Bit words are simple works. Next part is the Harddisk. Not a controller or interface, you can write direct to disk. A HDD needs words, not bytes. There are 16Pin (one pin = one byte) and you can connect it also direct to the FPGA or MC.
Also you can split the datas to more than one hdd. (First word to first HDD, second to second...)
But some sensor chip manufactors do the work for you. Most chips (2M pixel, or more) have two, or more output ports. Your FPGA can handle this and you can write to multiple HDD, without software logic. (First chip output go to first HDD, and so on)http://dvinfo.net/conf/showpost.php?p=242054&postcount=2071

direct to hard disk is potentially a simpler idea.
you can read off the sensor to an fpga, and then out from fpga from an onboard sata/sas controller, or to an offboard controller. in either case youll need off board serializer to interface the drive as the fpga's dont have that kind of speed. drawbacks are that youll need multiple disks to capture uncompressed. 10 bit raw from the hv20 would need 3 7200rpm drives, 10 bit 4:2:2 would need 5 drives. so your are effectively needing a raid array, whether you split outputs(if possible) or actually stripe the disks. there is a transparent external sata raid card which would do the trick on one e-sata port.

what i belive the ideal solution is, is to have a board (inside the cam ideally, tripod bolt mounted for a modded cam) that has an fpga for capture and format, and another dsp or fpga that has programmable hardware codecs. then an E-sata port and a gigabit ethernet port and probably usb for updating. this will allow you to choose between uncompressed, and whatever codec you happen to have in hardware form. codecs will be upgradable too.

so what you end up with is nothing all that different from what we already have, but instead of a one trick pony (hdv mpeg2) youll have full flexibility on what format you want.

this isnt all that complicated either, the hardware and software exists, its only the codecs that youll need to aquire to make this a viable project.

that all said. when talking about off the shelf hdv cams like the hv20.. theres little to no point in doing this. the hv20 for example has 1920 x 1080 8 bit 4:2:2 uncompressed output via hdmi. would raw be better? is it even possible to capture raw from it? debatable with this sensor as its pretty grainy with any amount of gain. files would be smaller, sure. but if you captured raw, you would need to see the footage somehow for exposure, because it would not be the same as what you see in the cams viewfinder. which means youd need to tether to a notebook, or actually add a display interface to your hack. thats a difficult task at best. so.. if you are tethering to a notebook/monitor, and need an external small box to do the capture and stream.. this is no better a solution than using the hdmi to a little black magic card in a box to a notebook express card slot. its functionally identical, and physically similar too. in fact, the only real difference would be that the hdmi solution doesnt void camera warranty and costs 1/10 the price.

terence


terence.

Terence, you are right. Your arguments are very convincing. Thank you.

Is there anything definitive that has told us whether or not the HV20 downrezzes 1920 to 1440 and back for live HDMI out?

the only thing i can say is that it doesnt look like it downresses then upresses. its very sharp and has alot of detail.

what it actually does inside the cam, youll really never know. its a single chip sensor so theres alot of interpolation already, and its hard to look at the footage and say "yes, its really 1920". only someone with inside info at canon will know for sure as its not mentiened in any literature. they dont say much of anything other than "it has hdmi".

terence

the only thing i can say is that it doesnt look like it downresses then upresses. its very sharp and has alot of detail.

what it actually does inside the cam, youll really never know. its a single chip sensor so theres alot of interpolation already, and its hard to look at the footage and say "yes, its really 1920". only someone with inside info at canon will know for sure as its not mentiened in any literature. they dont say much of anything other than "it has hdmi".

terence

I agree with Terence. It's a lot sharper than the HDV but there's no way to know if it's really true 1920. Looks great though.

Regarding true 1920x1080, take a look at my comparisons of your footage at http://dvinfo.net/conf/showthread.php?t=94079&page=2

I was trying to talk an bunch into developing an HDMI to USB, or UWB wireless solution. They already had most of the technology, 100mb/s codec, up to 200mb/s, with dual chips up to 400mb/s wavelet.

Maximum of GigE, is apparently 100MB/s, or 800Mb/s, some simple compression would be best. Ethernet cable is capable of much more, but not in GigE format. I think we Will eventually see an GigE solution, it would be an good solution to send HDMI from PC to remote panel, or remote player to an PC.

I suspect then parts are out there to be combined into an Alternative imaging project that writes directly to hard disk. Opencores, and the usual chip manufacturers suspect, Analogue devices already has an Jpeg2000 codec.

I recently communicated with somebody about HDMI capture, as you know many cameras are 8 bits, and this is cutting the stops of latitude down compared to reel stream. If only an reel stream camera was much less, i would be much more interested.

Just for information:
Shooting to RAW with Canon PowerShot (DIGIC II - based only) (http://digicanon.narod.ru/rawenabler.htm)

Now need to find tech specification for HV10/20 and possible these hackers can get raw from cmos of Canon camcorders :)

Kind of interesting discussion given that I've been recording 2K down GigE the last couple of days. It's a squeeze, you do need the right NICs and drivers but it can be done.

Bob, you talk about SI2K camera and Cineform RAW?

Interesting, what sort of solution are you using?

Another solution is to take an GigE from the camera, and route through embedded solution to GigE disk solution. If directly from sensor, then somebody posted an GigE sensor board that interfaces to sensor of choice. It is over at the New Cinema camera project thread in alternative imaging.

http://www.dvinfo.net/conf/showthread.php?t=96349

Open cores may have FPGA disk solutions.

I recorded uncompressed 8bit 422 1080 60i from an HV20 to Mac Pro last week - incredible step up from HDV, and very useful. I use the nanoConnect HDMI to HD-SDI convertor to my Decklink HD card (digital capture with a 150 foot coax cable!)

However, I agree with the question - is it worth it to go raw?

I think it'd only be truly useful if the output promised something over what can be captured on this already amazing camera with it's evidently very useful HDMI out. I'd need to see:

a) 10bit option (as I believe that the camera only outputs 8bit now)

or/and

b) 2K resolutions from the sensor

A 1920x1080 4:4:4 8bit option alone wouldn't be worth it over the 4:2:2 8bit that HDMI capture already provides. Worth investigating. True 2k would be killer - in the STILL camera mode, this camera certainly uses enough pixels that it's got! Looks feasable...

Bob, you talk about SI2K camera and Cineform RAW?
Yes.
If it can be done with this camera it can be done with any sensor / lens or I guess a HDMI -> GigE adaptor could also be built. Only possible issue is I think the HDMI ports are only 8 bit and a fair amount of processing is done before the vision reaches the port on the camera so you certainly wouldn't have all the advantages of recording RAW straight off the imager.

4:2:2/ 8 bit is downscaled output from HDMI.
Straight from sensor possible get at least 10(12) bit.
Canon CMOS is 1/2,7" with full resolution 2,96M (2048x1440 ?)
and possible create resolution at 2048x864p24 or capture 720p60.

If for raw data from cmos use smart compression like CineformRAW then stream possible to write onto 2,5" 7200rpm HDD or SDD!

As said David Newman, encoding to CineformRAW need less computation than encode from HDMI to NeoHD or Prospect.

Wayne Morellini also provided link to project "Dirak codec for FPGA". I dont know much about Dirak solution,
but believe it wavelet based as Cineform. If this codec implement to FPGA with ability write to disk...

CMOS(RAW)->FPGA(compressed RAW)->DISK ;)

4:2:2/ 8 bit is downscaled output from HDMI.
Straight from sensor possible get at least 10(12) bit.
Canon CMOS is 1/2,7" with full resolution 2,96M (2048x1440 ?)
and possible create resolution at 2048x864p24 or capture 720p60.


Well, yes. Then it'd be cool and worth investigating ;) I've dropped a lot of money into capture to hard disk from an HV20 - so far, the cable, nano-Connect and capture card are near twice as much as the HV20 itself ;) So, if the price was right, an HV20 upgrade would be cool. It's all about value - will the price be right 10bit at 2k sounds yummy.

Project a Wearable computer (http://dvinfo.net/conf/showthread.php?t=99632) very nice and good for our days.
Compressed RAW from CMOS is our tapeless FUTURE:)
SI2K and RED on this way.
To have a less expensive camcorder need to remember about the wonderful Elphel camera. (http://dvinfo.net/conf/showthread.php?t=63677)
This cam with good CMOS sensor and codec as CineformRAW (or Dirak) implemented into FPGA
can be the best choice in the near future.

I own a 35mm lens adapter and Nikon lenses.. the HV20 is simply the sensor capture medium (which is outputted live via HDMI to hard disk). Honestly, I am using the HV20 mosly for it's HDMI out, it's large CMOS sensor, it's latitude and it's 24P. Great combo for the price!

Soo...

If someone takes a good CMOS sensor and mounts it in place (with magnification) of the ground glass on say, a Cinevate Brevis or Red Rock M2, and adds good computer control support, I'd go for that!

Basically, I'm suggesting a SI 2k system, but a cheaper version with Nikon mounts :D

I wouldn't say it's worth it to add all this support to the HV20 hack because I bet there'll be some start up that's already working on a newer option now :) I challenge Red Rock, SI and RED to make a consumer 2k 10bit option :D

But since this is all chat, sure, someone make a hack for the HV20 :D Thanks for starting this fun thread Serge.

However, I agree with the question - is it worth it to go raw?

I think, like quality difference between DVX100 and Andromeda conversion.

We can only hope that reel-stream could do an HV20 conversion, but an camera with 50fps is preferable. Look at the Andromeda thread and see what they did.

There are an number of people on the DIY cinema camera thread combining to do an FPGA solution, and GigE. join forces (if you haven't already) make it also attachable to camera hack of HV20 and other cameras etc.

What I am going to say is in future, but if you create such an circuit what about build on it to make it configurable to most major low end camera and sensor interfaces (GigE, Industrial etc)? This way, anybody in future can extend the circuit and use in many cameras, but being open also be available to commercial companies to extend and use.

By worth, I meant, affordable and, long-enough lasting (before a cheaper/better 2k DIY system arises). An indepdant system that is simply a 2k chip, lens mount and simple circuits and GigE out w/computer control. I don't think it'd be long now...

Wayne, yes I'd love to see someone make a DIY 2k kit for the HV20 :D Or any CMOS camera. No, I've not read those threads yet.. I'll give them a peek soon.

any progress on this?