Retrofit Panny DVX-100A to uncompressed HD?
 

I just received an email inviting me to tune into a live streaming program featuring Reel Stream's Andromeda data acquisition system. When mated to the SG-DVX-100a it is said to provide uncompressed HD via a USB 2.0 output.
Anyone here have any expereince with this product? Here is the site: http://www.reel-stream.com/index.php
Reg

I hear you can fit a V8 in a Volkswagon beetle.

Mmmmm, but I think I'd rather just buy a Camaro.

- ShannonRawls.com

Re: Retrofit Panny DVX-100A to uncompressed HD?
 

Hey! That was cute...
Perhaps you missed the claim of "10 bit 4:4:4 output." For someone who's already made an investment in a DVX-100A - why not. I am more curious about the veracity of such a claim from an SD camera.
Reg

For what it's worth, Reel-Stream was born here on DV Info Net. That entire concept came from one of our discussion threads in our Alternative Imaging Concepts forum. Just like an 18-year-old teenager, Reel-Stream has moved out of their parent's house at DV Info and have rented their own apartment and have begun earning their own living. They're a great bunch with an amazing service. Check out their message boards, which were inspired by ours. Tell Juan and Jeremy I said hello and remind them to take care of me when I'm old and feeble, just like good kids always do for their folks.

Chris & Barry,

This is interesting.

I just went through the site and I have a question....

How is it that the DVX100a glass is good enough for 720p 4:4:4 uncompressed HD footage on the Andromeda, but I can't slap a Canon 3x lens on my XL-H1?

I read your posts @ http://www.dvinfo.net/conf/showthrea...hlight=3x+lens but it seems that didn't apply to the DVX. Why not?

Is Panasonic DVX Glass just flat out better then Canon XL2 Glass?

- Shan

The 3x wide angle lens isn't XL2 glass. It's XL1 glass. Sure it can be used on an XL2, but it was designed back in 1998 and it was the very first additional XL lens you could buy for the old XL1. Some folks thought it was a little soft even for the XL1. More people thought it was rather soft on the XL2. And everybody pretty much agrees that it's way too soft for the XL H1 when shooting in HD.

Shannon,

Also, there is the added fact that the number of pixels on an XLH1 chip vastly outnumber the pixels on the DVX. The DVX w/andromeda is pixelshifting it's way up to an HD image. The lens only has to resolve enough for the pixels on the chip. The same holds in reverse...if you've got more pixels, they're just wasted if your lens can't resolve enough detail.

So, the DVX lens was made and optimized for the chips in the DVX. How good their HD image is with the andromeda actually has less to do with the lens itself, and more to do with how good the processing is in SculptorHD (their capture software).

With the XLH1, if you throw an old XL1 lens on it, you're essentially wasting a good chunk of the pixel power on those HD chips, because the lens won't resolve enough detail.

Well with that logic, why haven't Canon/JVC gone with the extreem pixel shifting so they could use cheaper glass? Something doesn't add up. Andromeda also wants you to buy an Apple laptop to make it work. A poor marketing decision in my opinion, should be dual PC/Mac at the least.

Ken,

I think if you do a little research on the Reel Stream web page you'll find the choice for the mac platform was because of data stream usb requirements. Not a mac vs pc preference.

The mac mini's and laptops were the most cost effective means to an end.

FWIW, I've been following Andromeda since Juan started out here a few years ago, and...

IMHO the images coming out from the beta tests are some of the best I have seen from a sub 10k recording system.

I also believe that we all owe Juan/Reel Stream our deepest gratitude for.....

1. Challenging the mainstream manufactures to give us more, Now! And....

2. Helping to accelerate to market, new products (like HVX200/Red) that we all desire.

Amen....

I'm sure from the Andromeda and on Dvxuser.com Jan from Panisonic heard of the competition and lowered the P2 card price dramatically.

Doesn't the DVX100 have a 4x3 SD sensor? Given that's the case, the best you could get out of it is a really nice 4x3 SD image, which will still only have SD-level detail if you upsample to HD resolution. I can't see going to all this trouble when you could just buy a Canon XL-H1 and tap the HD-SDI outputs for a nice widescreen, 1.5 Gbps HD image.

This is all great, but why would you pay $6000 panny included or $3000 for the conv instead of spending that money on an already existing HD camera? 24P maybe?

I don't know guys, not meant in a bashful way or anything I just can see myself spending my money there, I am sure a couple of people will buy it as it looks like a great product, but for example; why not buy an Z1 and a Micro 35mm?

Or if you are strictly panny then get the new HD camcorder!

I am just trying to make sense out of this, someone just help me!~

When the Andromeda started out, the new HD cameras were not out yet.

When the new HD cameras do come out, the Andromeda has a small advantage in that it gets the most exposure latitude possible from the CCDs. At SD resolution, this is probably the biggest advantage the Andromeda would produce.

Hey guys,

You all have some good questions regarding Andromeda.

I could try to answer them....

But I think the best thing would be to have the MAN himself explain everything.

What do say Juan?

Could you please chime in here?

Thanks

Hello everyone,

I've read through the posts and it seems that most of the questions are directed at how one can extract resolution higher than SD from the DVX, so I will start there.

You are correct that the sensors in the DVX are approximately standard definition size...they are 770x492 (NTSC pixels) to be exact. However, do note that there are 3 sensors, and they are not aligned, but shifted from each other at a sub-pixel distance. There are about 1.1MP total on the DVX.

For example, picture this: The blue sensor is shifted diagonally 1/2 pixel from the red sensor. What this means is that each blue pixel picks up a detail that lies in between red pixels. So the blue sensor records a completely different set of details than what the red sensor picks up. Thus, even though each sensor is 770x492, the actual resolution of the 3CCD imager block is higher than the resolution of each individual sensor.

This is nothing new, it has been used for many years. There are some cameras out there which actually use this technique to take high-resolution stills from lower resolution sensors. This technique is also similar to how single color-sensors work, such as the sensors in the Arri D-20 or Kinetta. Each pixel location only records one color.

Now this is not a miracle solution to gaining resolution, it has its advantages and disadvantages. One of the advantages as Glenn has pointed out, is that since the pixels on each 1/3" sensor has a larger area than the pixels of a 1/3" sensor with say, 1280x720 pixel count, this tends to yield better dynamic range(latitude) among other things. We have measured the dynamic range of the Andromeda output from a DVX100A to be around 9.5 stops. This is more dynamic range than just about any prosumer and even many proffessional camcorders in the market today.

To be fair, the technique does have down sides, such as that a camera with native high-resolution chips will still have an edge, usually in vertical resolution, although at the cost of dynamic range. Look at the output from the XL-H1, and take a look at the samples in our site. You will note that the 1080 output from the XL-H1 is sharper, but it looks like absolute video because of the very limited dynamic range and resulting tonality of the image.

We've shot side-by-side comparisons with still 35mm film, and the main difference in the look of the image is the depth of field. We've had people send us their own color-corrections which are almost indistinguishable as far as tonality, and almost resolution!

Our goal is to provide a unique tool to filmmakers, one that hasn't been available before with unique advantages. It all depends on what the filmmaker wants and the compromises he/she is willing to make.

I think there is a lot of fascination with pixel counts, in general people seem to think that more pixels will immediately make the footage more film-like. This is not necessarily true, all you have to do is put your favorite movie DVD on your TV and then watch some DV footage. There's a lot of differences other than resolution, and the main difference in my opinion is the huge amount of latitude that film captures. There are ways to address these problems in camera designs although nobody seems to be doing it, except Panavision.

Hope this helps,
Juan

Thanks for the detailed reply Jaun. A few questions if I may:
How do you get around the 4:3 of the DVX? Does one need an anamorphic adaptor, or does it end up being a HD 4:3 image?
What is the reason that you have chosen a Mac laptop only, why no PC? One last one, why so long to market? I have been following you guys periodically for quite a while, when are you going to ship, if not now, what is holding things up?

Hello Ken,

To answer your questions:

1.)You are correct that the sensors on the DVX are approximately 4:3(actually they are 1.4:1). However since there is a significant increase in resolution, to get a wide image you can letterbox or use the anamorphic. Do note that because we are recording areas of the CCD that are normally cut off by the camera, the Anamorphic yields a 1.85:1 image, and 1.4:1 without it. The maximum frame size without up-rezzing is 1540x984.

It's very similar to shooting full-frame film. The Sculptor software has reticules similar to the video-assist on high-end film cameras, such that you can frame correctly even while recording the entire frame. This also allows the freedom to pan and scan later.

2.)The reason why we chose the mac is because the Powerbook and Mac Mini can do something that few(if any) PC's can do, and that is sustain the required ~30MB/sec over USB2 while being portable and/or tiny package. PC's have a very wide variety of hardware configurations, and even if there are (just) now a few machines that can do this over USB, we would need to write low-level drivers for each machine, and test each PC model independently. With the mac, there are only a few models and implementation was a lot simpler since there is no low-level drivers required to get the maximum performance out of the hardware. Plus, for about $400 you can buy a mac-mini that can do the job.

I know that in the consumer world "common-knowledge" is that USB2 can do 480Mbps, but that is not accurate as far as transfer rates. The common USB2 card you can buy now days does about 30MByte/sec max. Many do less.

3.)If you were following our original discussion here on DVinfo, then yes, it has been a long time. The main reason is that this really only became a commerical endeavour about a year ago. Given that, we had our first beta units out in about 6-7 months from the beginning of product design and development.

Considering that I single-handedly did all the design and development of both the hardware and software that's not too bad is it? :)

We are currently taking pre-orders for our first production batch which just arrived. Sales are scheduled to begin early January.

Hope this helps, let me know if you have any other questions.
Juan

Juan, let me throw a question here...
 

...This is out of the DVX arena but bare with me....

Considering the Cost of a new/used DVX out there and the cost of a new/used FX1, would you consider doing something similar for the FX1?

I don't think is far fetched but then again you are definetly smarter than most of us in terms I would not even dream of making something like you did and wouldn't know how to start.

Given the amazing quality of the FX1/Z1, could you in theory be able to tap into the system and grab the raw image data to be able to record the shot in 24p/30p on the laptop? somehow I remember seeing an article stating that the FX1 recorded progressive and interlaced the image, I'll try to find the post.

I am dreaming here?

Please don't shoot me guys, its just a question.

Hello Tomas,

The concept applies to any digital video camera, however it has to be re-designed based on sensor size, etc.

That said, it is very unlikely that we will do a version for the Sony Z1/FX1 for several reasons.

First, the DVX100+Andromeda already yields more resolution and a lot more dynamic range than the Z1/FX1 in any CineFrame mode.

Second, the Sony has interlaced chips, and we've had almost 0 requests for uncompressed interlaced video. The grand mayority of the input we have received is for Uncompressed 24P which can be easily molded for a particular look without having artifacts, decimation or low dynamic range in the way of filmic color timing and fx work.

Hope this helps,
Juan

Any particular reason why only USB2 and not firewire? I am still a little puzzled about the claims of Apple powerbooks having (more usb bandwidth) is that relevant to only certain PC chipsets, or all? And if so is the limitation the same with firewire? What about firewire 800?
I am also puzzled by the claims of such increased latitude from the SD cam (DVX100a) as I just do not see that dramatic an increase in latitude over the top HDV cams. If the advantages were that marked, why didn't HDV cam makers simply use their SD cams and pixel shift as you have done? Sorry for all the questions but so many things are puzzling to me in regards to this.

Hello Ken!

There are several reasons why we chose USB. First of all, when we started talking to people and showing them the prototypes, one of them was a box mounted under the camera which would have a firewire interface. However, this box design was met with rejection because among other things, it prevented some accessories from being mounted(such as some 35mm adapters), required it's own battery, and would be overall more expensive to the costumer and for us to produce. Every single person we talked to wanted to keep the ergonomics of the camera intact, and wanted just a port added that you could connect to if you wanted uncompressed. They also wanted the lowest price possible and little if no effect on the battery life of the camera.

USB allowed all this. The board is small enough to fit inside the camera, and be invisible except for the extra port. The power is taken from the USB host, so the battery life(as well as all other camera functions) are unaffected. You can mount every single accessory in the same way you would with a regular DVX. And it is a lot cheaper to the end user than an external device with it's own battery power and firewire interface. Another point largely overlooked is that using firewire would require users to have a fast internal drive, as USB drives wouldn't be fast enough. Using a cheap external drive is a big advantage. A 100GB hard disk which can be had for under $200 will store about 1 hour of Uncompressed 4:4:4 footage. Compare this to a $2k P2 card holding 8 minutes.

About the latitude, you don't have to take my word for it. We have done quantitative tests using a backlit ND chart, using an Andromeda DVX-100A. The two frames are the exact same frame(same moment in time), but one is captured at the A/D stage with Andromeda, and the other is what the tape records after the camera processes the A/D data. We have shown these results to knowledgeable people such as Adam Wilt and Barry Green, and they both agree with our assessment of about 2-2.5 stop gain on the highlights. The charts are here:

http://www.reel-stream.com/gallery_t...gallery_type=2

The light coming through each step was carefully measured with a light meter. The Andromeda clearly conserves considerably more dynamic range than the DV output captures.

As far as the HD/HDV cams, the HVX has been reported by Panasonic to have been tested at about 7 stops which is about the same latitude as the DV output from the DVX. The XL-H1 hasn't been tested AFAIK, but everybody can see from the footage posted on here that it clearly has a lot less dynamic range than the DVX or HVX, putting it conservatively at 6 stops, probably less. Same thing goes for the JVC, although it seems to lie somewhere between the DVX and Canon.

Remember that I said pixel-shift has advantages and disadvantages, not advantages across the board. A possible reason why camera companies would go for higher-pixel count sensors and sacrifice other performance aspects is simple:

PIXEL COUNTS SELL

Try a search in all the forums and see how many people are talking about resolution/pixel counts and how many mention, say, dynamic range. You'll see that when asked what they want from a new camera, they immediately say 2K! 4K! If you ask me, give me 16+ stops latitude uncompressed, at 720P and as much color precision and frame speed as possible. The studies done on cinema resolution show that at the best, a standard cinematic print has about 600-800 equivalent lines of vertical resolution. We've been exceeding film for a LONG time as far as resolution goes. All the current and vapor camera designs seem to completely miss these other aspects that nobody is addressing, except Panavision.

And then people wonder why it still looks like video. It's not impossible to do. Everybody seems to be just barking off the wrong tree. :)

BTW, it is likely that Panasonic is using pixel-shift in a similar way to produce 1080P on the HVX. They are also refusing to release pixel counts for the CCD's. I think this illustrates my point.

Hope this helps,
Juan

Juan: any predictions on the dynamic range for video from the HD-SDI outputs on the Canon camera? If this data is coming straight off the sensors, shouldn't it yield roughly the same result as tapping the sensor on the DVX?

SDI and HD-SDI is formatted digital video. This means that it goes through all the in-camera processing except the final HDV compression layer that would normally be put to tape. The output from the HD-SDI port is not raw, so it exhibits decimation along with all the other byproducts of the in-camera processing except compression. The difference with raw output is that it is RGB direct from the sensors(A/D) without any processing done to it whatsoever. It is the most pristine data available inside the camera.

I think it's a great step forward for them to add an HD-SDI output, unfortunately the camera exhibits such limited dynamic range. Even if we did tap the sensors on the XL-H1, I doubt it would yield nearly as much dynamic range as the sensors on the DVX. The sensor element area is so small.

BTW, Area 51???

Cheers,
Juan

Hi Juan. Thanks for taking the time to answer questions. Good luck with the release.
Ken

Juan - thanks for all your hard work -
I am torn - it seems your solution gives a higher res/ less artifacting solution than the HVX -
Is there any possibility/ advantage to making your products compatable with an HVX
?
Thanks,
J

Hello John,

I've been hearing that a lot lately :)

That's a hard question to answer. It might seem like making an Andromeda for the HVX would be a.)as simple as the DVX and b.)yield advantages across the board.

The truth is that, assuming we want to capture up to 60 uncompressed frames a second, the bandwidth is such that the design of the device would have to be completely different, since the bandwidth is not manageable with existing standards such as firewire. We would need to build something like a portable RAID array connected to a box under the camera, which gets complicated.

Also, because the pixel elements on the HVX are smaller, it is very possible that we might get worse dynamic range and sensitivity with the HVX chips.

However, it is too early to say, we will have a better projection when we get a chance to work on one.

It is really up in the air if we will get an advantage in resolution:

http://forum.reel-stream.com/viewtopic.php?t=363

Anybody know why we are in the Area51 section?

Cheers,
Juan

thanks, Juan for your quick response -
if I was already a DVX owner, I wouldn't hesitate

I wish we could compare the Andromeda & HVX200.
The charts are interesting it's hard to copare Tiff's & Jpegs - the Sony looks better than the HVX by far with the JVC coming in second

John,

If you are interested in seeing how the image format affects it, just convert the tif file to jpg in photoshop. Tif is lossless so it will be the same as if it was exported directly from Sculptor.

When we get the chance we'll do a side-by-side with the HVX. From the specs we know however, Andromeda on the DVX100 gets wider dynamic range(9.5 stops), more resolution, no artifacts, no decimation, higher color precision, etc.

Because the data is recorded direct from the A/D converters uncompressed, the filmmaker is free to color-correct the images and give them any desired look without destroying the footage by enhancing artifacts.

Cheers,
Juan

It sounds very promising Juan. How does the DVX lens hold us back with your setup? It must not in at least some respect as it was not designed for the specs your eqipment provides. Both Canon and JVC have been criticised for their HD lens (I personally disagree given cost/performance ratio (esp. the JVC at its price point)) how does the SD Panna lens hold up under zoom and or shallow depth of focus situations? How much CA? How does the sharpness hold up across the image? Any other problems you have noticed?

Hmm, this really doesn't belong in Area 51.

Should I file it under HD Acquisition, or Alternative Imaging Methods?

General HD/HDV Acquisition is my suggestion. It isn't from a Co. as large as Sony/JVC/Panna but it is of a similar vein.

I did that and it introduced artifacts similar to ones in the other JPG files, so posting these images in different formats wasn't a fair comparison. Also, it looks like the Z1U chart was recorded on a different background than the other images, which introduces uncertainties about the effects of lighting on the image. I think we can all glean enough from the images as posted to learn something useful, but it's more useful when identical procedures are followed to generate such comparisons.

Also, to be thorough we should compare images from Andromeda to uncompressed capture options for other cameras. Until someone does that, I don't know how Andromeda compares to, say, connecting a Z1U via component cables to a Kona2 capture card.

It actually is not the weak link in the chain. In fact, if you push detail up in the Sculptor processing for a EIA resolution chart one can see the vertical trumpets beyond 800 and horizontal past 900. The nominal output as shown on the rez chart at:

http://forum.reel-stream.com/viewtopic.php?t=363

doesn't jump up that high because my algorithm doesn't sharpen the image in its default setting, but it's obvious that the lens is more than capable of resolving HD resolution comparable to the other cams. As much resolution as we are already pumping out of it, the practical limit of the imaging block is reached before that of the lens.

The chromatic aberration across the entire range is also far below the aberration visible from the factory lens that comes with the Canon and HD100. You don't need Andromeda to verify this, most current DVX users already know how well it performs. This is not surprising, as the internal lenses in cameras such as the DVX and HVX have a large number of elements in them, and as such it is easier for the optical engineer to zero out aberration. Cheaply making good removable lenses with fewer elements for small chips is a different story, because it relies more on the actual quality of the glass($) and not the optical composition of elements.

Kevin: I posted the jpg of the HVX chart, because:
a.)That was the only format it was available at the time and
b.)it was the chart that showed the highest resolution for the HVX of all the tests done.(+7 detail level)

Barry has just sent me a re-processed BMP which I have replaced in the post.

I try to be as fair as I can. Like everybody else, we don't have every single camera available to us to shoot charts, so the best I can do is pick the best results for each camera and using the same chart.

Cheers,
Juan

Dang.
Juan - you are really tempting me to look for an older DVX.
I would love to mount on a steadicam with a 35mm adapter
(and feed into a 12" powerbook that I could use also as a monitor)

But please explain the ratio is 4:3 > 16:9 ?
I don't quite understand...

Also - I am in NYC - is there anyway to see a model in operation
Thanks again
J

Will a PC laptop enter the equation at some point in the future?

John:

The CCD's on the DVX are 770x492 at NTSC pixels. This yields an image 1.4 ratio. Andromeda allows you to capture all the information from the CCD's, so you endup with an image 1.4 in ratio (1540x984 NTSC pix). However, from there you can target just about anything.

Our goal was to impose no limitations that are not actual hardware limitations. Thus, Andromeda gives you all the information the camera really captures, and the filmmaker is free to use it how he/she sees fit.

Just one example of how this can be exploited: normally the anamorphic adapter yields a 16:9 image. However, since Andromeda records areas of the CCD normally cutoff by the camera, with the anamorphic adapter you actually get an image at the cinematic 1.85:1 aspect ratio.

If you are recording the full 1.4:1 frame, it is much like shooting film. Sculptor provides reticules within the 1.4:1 frame, and you can later apply pan-and-scan techniques if required, etc.

Ken:

We have plans to support PC in the future as long as the available hardware supports the required data rates. The main issue is that Andromeda uses up more USB bandwidth than just about any other USB 2.0 device out there. Thus not many systems are designed such that a high throughput can be mantained from the USB port through the bus and to the drive.

The Apple Powerbook and Mac Mini are very fast in this respect, but for example, the powerful G5 tower actually has slower USB performance. However, with macs we only have a few machines to test for compatibility. With PC's, there's such a large number of variations, that it's a larger task to find out on what models the motherboards are configured in the required manner.

Cheers,
Juan

Thanks again Juan. Keep us posted on the PC developments. Please.
Your product becomes so much more appealing to the 95% of us that own PC gear. When you find a platform that works let us know so we can make sure our laptops comply. Then it is Ebay for a used 100a!

Juan,
Thanks for all the info -
I hope you get your hands on an HVX - I would love to see Andromeda get around this cameras limitations too
BTW: you are now taking preorders, when will they be coming round.... I'll sign up in your forum.
Thanks
John

Hello,

Will make sure to post on any developments. One of the reasons why mac was a good choice(besides the technical stuff i explained) was the existence of the mac mini. For a few hundred dollars you can get a very compact recording system which can be setup in many different ways. We have people who have made their own portable setup with mac minis, or built one into a monitor, etc.

Also, many of our users do their work on PC, they just connect the external hard disk to their PC and use the media from there.

John: We are filling orders like crazy as I type. :) If you are sending in your DVX, it normally takes 5-7 business days before we send it out with the system installed.

Cheers,
Juan

Wow,
More tempting than ever !!!