Webcast cameras

Dear layzweb, what's the state of the art in cheap, low-light video cameras and/or switchers?

If you've been watching the DNA Lounge webcasts over the last decade, you may have noticed that the picture is kinda dark and grainy, nightclubs being what they are.

I'm wondering whether the available tech has advanced to the point where I could improve low-light performance and/or resolution without spending a fortune.

I strongly suspect the answer is no, but I figured I'd throw it out there.

About half of the shots you typically see on the stream are coming from that same batch of camcorders that I bought back in 1999, all of which are Sony TRS17 or similar. They are indestructible. They're "nightshot" camcorders, meaning they have IR, but we don't use that. They just happened to be pretty good in low light without IR. The rest of the shots, especially for shots of live acts on stage, come from a pair of Panasonic WV-NS324 pan-tilt-zoom cameras.

All of them feed analog SD NTSC to a video switcher, and from there to the webcast. Details.

So, one option would be to replace them all with whatever the lowest-end HD camcorder is, re-cable everything for HDMI-over-Cat5, and get an HDMI switcher. This would mean getting rid of the panning cameras and replacing those with fixed-position fixed-zoom shots, which is probably fine. But, even if the camcorders are only a couple hundred bucks each, that would still probably come out to over $5k, which is kind of steep for something that makes us no money whatsoever. It also couldn't easily be done incrementally, due to the switch from composite coax to HDMI.

You'd think there would be an easy way to deliver video from the cameras to the switcher as MPEG streams over Ethernet, instead of going through uncompressed HDMI and a bunch of Cat5 converters, but if there is, I'm unaware of it.

Please note: before you suggest a camera or camera system, bear in mind that most "security systems" are designed to be used in environments that are as bright as the surface of the Sun. Most non-camcorder video cameras eat shit in less than 7 lux or so. What I have now are lower than 1 lux. If something says "0 lux" that's a lie (that means "it comes with an IR spotlight, and will give you a goofy-looking black-and-white image.")

Previously.

Tags: , , ,

19 Responses:

  1. Dave says:

    Have you looked into the Axis PTZ network cameras? I use quite a few of the Axis P5534 for webcasting live events. My system ingests the MJPEG stream that the camera provides (there's also an h.264 stream available as well), but they do produce a secondary box for getting "standard" video out of the camera and into analog equipment.

    Camera is very well-built, single cable install (PoE), and operates in low-light (0.74 lux for color, 0.04 lux b/w).

    Only issue is cost - they aren't cheap at all (~$1500 last purchase I made), but with proper placement and good presets set up, I imagine you could get away with replacing your current cameras with 3 of these. Plus, there's a published API for controlling the camera over IP using HTTP GET requests (including changing presets, etc).

    Email if you want more info or want to see some of the webcasts using the camera.

    And no, I don't work for Axis, or any company related to Axis. I'm just a happy consumer.

    • jwz says:

      Those look pretty nice, but that's a lot of money. Looks like even Axis's non-PTZ cameras that are < 1 lux are > $800. Assuming they even work with the same system.

      So if you had several of these cameras, and you wanted to switch among them under software control, resulting in a single video stream being handed to Flash Media Live Encoder (presumably meaning, the video has to come in to a Mac on a Firewire cable), how would you go about that?

      Meanwhile, I assume (but have no proof) that you can get an HD camcorder with 0.7ish-lux performance and an HDMI output for under $200. Do seven stationary cams trump one panning cam? Probably, assuming the real cost isn't overly bloated by the required cables and converters.

      • Adolf Osborne says:

        No, I can't answer your question.

        But I can say that the better Axis cameras do better than I, myself, can do in low light -- and my night vision's really fairly good.

        They're noisy, and grainy, and the exposure time can go up (along with a decrease in framerate) to the point of being very blurry with fast movement, but they really do OK in the dim light that most folks would call "dark" using the stock lens arrangement.

        Panasonic's offerings are similar in performance, though very different in implementation (Axis likes to use CMOS sensors, while Panasonic likes CCDs). And Axis's HTTP interfaces are -far- less braindead, and more likely to be supported by $random_app than Panasonic, whose interfaces and protocols are designed by Japanese sadists.

        Meanwhile, all of these color cameras suck at doing color in low light, and behave much better in black-and-white mode (sometimes by even physically sliding the RGB filter out of the way with little robot servos) when there's very little light.

        So, on the assumption that monochrome is the way that it is: Have you considered using IR to brighten the place up a bit? (We don't need HD to see the goings-on -- we just need more light.)

        I, personally, have some PTZ gear (using not-so-new Sony optics, FWIW) at a facility which has quite good fidelity out to several hundred yards on a very dark night with only a small(ish) narrow-beam IR emitter providing illumination. (Prior to installing the emitters, they were useless at night.)

        • jwz says:

          Everything looks stupid under IR. Fabrics and hair are the wrong color and peoples' eyes glow. It's a funny effect the first time you see it but after that it's just annoying. Black and white is not an option.

  2. Adolf Osborne says:

    Hmm. Well, then.

    I admit that my focus is on security, so having lovely pictures of people is far less important to me than seeing that people are present and that they are doing things: While IR works great for what I'm doing, I can see how it doesn't work for you.

    However, I maintain that you're just lacking light, and that NTSC(ish) is fine for the current state of things. Throwing HDMI at the problem is a lot like trying to fix something with DLNA, but there isn't any other standard for consumer HD stuff, and pro HD stuff is still alarmingly expensive. MPEG over IP can work, but that brings you squarely back into the realm of "security cameras."

    Perhaps use cameras that permit better (bigger) lenses to be mounted which throw more light at the sensor? It ain't rocket surgery. I don't have any ideas for specific suggestions at this instant, but you already know that it's easy with an SLR: Just snap on a "faster" lense.

    I will keep it in the corner of my little pea-sized brain, and ask a few other people who might have a clue. Maybe something will come forth in the next day or so.

    • jwz says:

      It's true that throwing HD at it is kind of a backwards way to go about it, except that A) HD gear is what counts as the "low end" now, and B) when I see video recorded here using cheap HD camcorders, I'm shocked at how good the low light performance is compared to these 10-year-old SDs. This is probably just a factor of the CCD having more pixels on it, and/or the software being better, but... it's a thing.

      Cheap cameras (video or still) don't have removable lenses. It's part of why they're cheap. Sometimes you can screw an additional lens on front (a fisheye or something) but that's only going to make it darker.

      • Adolf Osborne says:

        Still hasn't been a day or so, but: Can you show me an example of shocking low light performance from a cheap HD camcorder at your club so that I can compare to the murky webcam images and glean some idea of a baseline for what you expect?

  3. Michael Dwyer says:

    For what it is worth, SDI runs over the same 75ohm coaxial that your current composite video runs over. So, it is possible that you will be able to reuse your current cabling with some types of HD. But cameras that provide SDI-out start in the $1k range, and the used market probably doesn't really exist, yet. Plus you'd still be back to having to replace your switching systems.

    When you look at it that way, the cost of the wires is probably not a big factor.

    • Adolf Osborne says:

      SDI is cool, but I think it's a dead-end here. It's silly to spend that sort of cash on gear for something that is essentially just for fun.

      It's interesting that SDI uses 75-ohm coax: The curse of NTSC strikes again. First with NTSC itself, then with VGA and S-Video and Component, along with S/PDIF, along with (I'm sure) others, all to let folks use their existing 75-ohm patch panels.

      Even though a characteristic impedance of 50 ohms would've worked just as well, and always results in equivalent wire that is both smaller and less expensive, it's 75-ohm FTW.

      Meanwhile, small signal wire is (almost) always cheap. Installing it is the expensive part.

      • a_0001 says:

        HD-SDI from the cameras to a switcher is definitely the standard for television studio production, but I agree that it probably can't be done within the stated budget.

    • If you try to run SDI over the same 75ohm coaxial you run composite over you'll definitely run into issues (wave collisions for one, dropped frames trying to go to anything higher than 720p).

  4. I've been working on a IP video based project for a few years that also needs good light sensitivity and have a large variety of different IP cameras and lenses if you want to coordinate a time for me to bring them over and see how they perform in your actual lighting. My guess is that nothing will cover the intersection of price and performance you're seeking, but who knows.

  5. Stefan Bethke says:

    Can't find the links offhand, but I remember reading at the usual blogs that a few companies announced live streaming boxes as CES, to go with helmet cams as well as regular camcorders, encoding the HDMI signal to h264 as streaming that to u-stream or what have you.
    There's a couple live video mixers for Mac OS that I looked at some time ago that can take live feeds from anything that QT can play, so maybe that could work as a switcher. I'll try to dig up the links.

    • Stefan Bethke says:

      Cerevo LiveShell is the product I was remembering, and I also remember that a similar product was announced at CES as well. I couldn't find any more details (in English, anyway), and no pricing.

  6. O Hai! Someone said I should probably chime in on this thread. I'm sort of in charge of doing this at a certain castle like structure that offers "industrial adult therapy videos." I will let you in that HD is definitely not cheap yet. Add on more features and the price goes way up. Even currently we aren't doing the greatest of jobs but we are totally getting there.

    For our "free" cam offerings we implement a bunch of Axis M1011 that export h264 over to an Amazon ec2 instance of Wowza Streaming Server. Not the most elegant approach but they are free cams.

    For our "paid" offerings, we use Sony EVI-HD1 PTZ cameras running over HD-SDI. This runs into a "meh" Mac Pro running Telestream Wirecast (which tends to be a POS if you push it to its limits (3 different bitrates and a to disk recording)). We are moving to something far more nicer but it wasn't chump change.

    There are a couple of ways I'd suggest trying to do HD en cheap. For your purposes, I'd definitely suggest using Wirecast. Advantages being it will allow you to use IP cameras as an input source and it offers software based switching. You can pretty much script all the functionality of Wirecast using a bit of AppleScript. It'll allow you to publish FMLE streams, however you'll need either need a Wowza Streaming Server, a streaming service (livestream, justin.tv, etc) or a CDN (bitgravity, level3, limelight, akamai). Cons being Wirecast isn't quite a "pro" application, so minimal to no error reporting, bubbly happy Mac OS X interface, and no hardware accelerated encoding.

    The second method is a bit more expensive, you could get a bunch of micro-HDMI cameras (GoPro, Contour+) and do the HDMI to Cat5 and use an HDMI switcher. Unfortunately, you might have issues with switching since there isn't a master clock to run all the signals. You could cause a weird frame drop issue when you switch the signal and it doesn't get the next frame it is expecting. Also, decent HDMI switcher with some sort of external computer control (RS232, Ethernet) come at a hefty price. Then you can run the HDMI signal into a computer with an HDMI capture card (Blackmagic Design or Matrox makes some really nice ones). However, you might have issues trying to capture video with these cards using Linux (if you wanted to go that route). Again, you could use Wirecast on Mac OS X or shoehorn VLC to do it.

  7. Scott says:

    It's about light gathering, so you want whatever camera has the largest lens opening. That means you want an F-number of like f/2 as compared to what the web cams have, which is likely something in the f/8 or f/12 realm. The higher F-number means that less will be in focus (depth of field), but since you have fixed cameras, you can focus it to what you want.

    Much of it is explained here: http://www.axis.com/products/video/camera/about_cameras/lens.htm
    and here on page 12: http://www.axis.com/files/whitepaper/wp_lighting_for_netvid_41222_en_1012_lo.pdf

    Thus if you can buy lenses with more glass in them, you can keep the camera. Or maybe find a 35mm style SLR video-capable digital camera that has interchangeable lenses (like Canon/Nikon style mounts), and then go hunt e-bay for used 35mm film camera lenses.

    • Ben Brockert says:

      Dude has real still cameras, I think he knows what an f/ number is.

      Try finding official published sources for aperture, focal length, and field distortion of low end video cameras. They generally do not publish them, so either you have to wait for someone to figure it out and post it online, reverse engineer it, or test it yourself.