Who Is Buying New CCD IP Cameras? Anyone?

Quite a number of people continue to believe that CCDs are better than CMOS, despite the fact that almost no one (even high end manufacturers) even releases any CCD IP cameras anymore.

I am curious - who of you are buying new CCD IP cameras? What brands?

Also, who is releasing new CCD cameras (in 2012 or 2013)?

The CCD vs. CMOS debate has been a heated one in the photography world for a long time as well, often linked to the Nikon vs. Canon debate as Nikon has long been a CCD proponent... yet that too, is changing now, with more than half of Nikon's lineup using CMOS.

Actually, I have a slightly different question for the collective mind here: does anyone actually base their camera choices on the type of sensor, or even pay attention to it? This seems like an even more esoteric factor than Relationship Between Sensor Size And Megapixel?

Yes, people pay attention to it. Two attendees of the IP camera class brought up the CCD vs CMOS issue and their preference for CCDs. Another actually mentioned that CCDs (specifically global shutter) had advantages for speeddomes.

I am not really against CCDs. It's just that pretty much every IP camera manufacturer has moved away from them. So it's not as much a split but a movement.

I'm certainly not "against" either as a technology either - as with anything else, each has its own strengths and weaknesses, and each will have a different effect on the final image. However, as with pixel count vs. sensor size, there's SO much more to the equation.

When I spec a camera, I'm looking at the final output rather than how it got there. While there is a theoretical advantage to global shutter with PTZs, for example, if I see a CMOS-based camera that does the job better, why would I want to hamper that decision just because it challenges what I already "know"?

Of course, that's the major benefit of IPVM's tests: they provide real-world results to go on, rather than just basing selections on spec sheets.

I agree with you. My practical issue is that even if I wanted to test or try CCD cameras, where I would get them amongst today's new MP IP models?

See if you can get the fanbois to give up some of their stockpiles for testing?

John- we (Hikvision) have them available in 2MP and 5MP versions, but they are large format CCD sensors (2/3" or 1") and are targeted to specific markets like traffic so we have them available for these special projects. They are not mass market.

I remember that. I also remember them being quite expensive. Yes/no?

Also, what benefit do the those large format CCDs have in transit that a traditional 'regular' format CMOS does not?

John- yes- they are expensive (approx. 3x-4x) versus a standard CMOS 5MP and need lenses to match as well. They are priced below many machine vision CCD cameras. Key benefits are better sensitivity and global shutter. They can also do 30fps at 5MP. For traffic, they also support the different snapshot sync modes.

HI John, here an interesting statistic from iSupply website. Somehow I feel on the Surveillance sphere the shift to CMOS is even more pronounced. I think a key reason might be that there is just no need for the CCD as the CMOS quality and variety has really improved over the last 5 years.

Maybe when doing extremely slow motion videos (such as capturing the motion of a water drop, etc) we might find the CCD to still be suitable given that you will find global shutter, high sensitivity (wihtout need of sense up or further image manipulation), etc.

I believe the average cost of CCD vs CMOS on a high megapixel situation is about 4 times more expensive. Also the programing of a CCD is quite more complex.

CMOS sensors have image processing functions within the sensor and also built in functionally like signal amplification, noise reduction, and analog to digital conversion.

With a CCD you have to setup a separate read time, shutter times, and excess charge dump time. And there are many cloks running that needs to be synchronized.

Thus the overall solution might be unnecesary more expensive with a CCD than a CMOS from serveral points of view that you see it.


Oscar, thanks, thorough explanation and data!

Does 'CMOS damage' have any bearing on the subject?

We have talked about lasers damaging / killing cameras before, but it seems hard, rare and possible for both CCD and CMOS.

CCD is hardly immune from burn-in - I have some old Panasonic cameras that will still show you the outline of the backlit door they were pointed at for several years.

Thought I included support information for my question.

If not, here it is again: "Search |cmos damage|"

Adding - Currently listing "About 6,790,000 results".

Apparently Lasers are not the only problems.

Relevance, your honour...

I realize the question is who is buying CCD's? Since Bob from HIK piped up - Who is manufacturing them still besides HIK? > Panasonic is still manufacturing their IP model WV-NW502S and the WV-NP502 as a CCD offering.

There's still a number of older models out there (like the 502s you mentioned which are ~3 years old). I can't recall any MP camera in the last 2 years that was CCD. I am sure there's a few out there but they are definitely few and far between. (btw, that 502 is not close to being competitive with this year's cameras (worse low light, WDR, bandwidth, etc.).

There are certain advantages to CCDs that will make them never go away. CCDs are a natively better technology for low-light applications, and we at Pixon Imaging are about to introduce a new technology that I think will forever make them the winner in this area. It is based on the CCD's ability to bin charge before reaching the readout amplifier. So here's my question to the community. What if you had a wonderful megapixel camera during the day, but at night you simply changed the readout clocks and had a sensor that provided simultaneous multiple images (let's say two to make it simple) which had different effective pixel sizes, single pixels for a high resolution image and a second image with highly binned pixels for a very high sensitivity image? The effect would be that now you could have a high dynamic range image by combining the two images, and you would be able to see into those deep shadows with a low-light sensitivity that was 10x or more than your high megapixel camera could yield under normal operation. Yes, the spatial resolution would be decreased at night, and perhaps you couldn't identify the face of someone in the shadows, but at least you'd know someone was there rather than not detecting them at all. And your would retain higher spatial resolution in the portions of the image that were better lit. Further the tradeoff between sensitivity and spatial resolution would be completely under your control by simply adjusting the clocks. This is not an entirely new idea. WDR binning is already done with CMOS cameras by binning after the sensor is read out. However, with this after-the-fact approach, sensitivity increase as the square root of the number of pixels binned, whereas with CCD binning before reaching the output amplifier the sensitivity increases linearly with the number of pixels binned. So for example, after-readout binning of a group of 4x4 pixels increases sensitivity by a factor of 4, while pre-readout binning of charge in a CCD increases sensitivity by a factor of 16. So how many of you would like to have a megapixel sensor during the day, but a lower resolution, WDR camera at night with 16x the low-light sensitivity at night time? This is what we are offering in a single sensor. If you are interested, please contact me at Pixon Imaging, rick.puetter@pixonimaging.com.

Why not assume EVERYONE is interested, and just tell us more about it here?

Hi Matt, our solution is very simple. People have been using binning on CCDs for years to increase sensitivity. In effect, you simply have a bigger light collecting area for the binned pixels that puts them over any read-noise floor that might afflict a single pixel in a low-light situation. This is the difference between "I can't see a thing", and "Oh, so that's what hiding in the shadows". Our trick is to change the readout multiplexers so that you can more creatively combine pixels to form multiple binned and unbinned images at the same time, giving you WDR capability and extreme low-light sensitivity in a single exposure. What you get is a high sensitivity, low-res image and a low-sensitivity, high-res picture. These can then be combined giving you a WDR image that has high-res where there is more light and low-res where the lighting is poor and where you wouldn't been able to see at all with a higher-res sensor.

Rick, that sounds like some pretty good imformation, but who are you offering to contact you too? Manufacturers? Integrators?

Hi Luis, We are interested mainly in getting in contact with sensor manufacturers. This is a sensor play and the magic needs to take place on the CCD before reaching the readout applifier. But I am also interested in hearing from the community at large to guage interest in this type of solution. I know WDR sensors are all the rage these days, and I know that people are lamenting the low sensitivity of HD sensors for night-time applications. So I'm betting that this should become a big win. Afterall, you get the best of both worlds, highest possible resolution in the day and a tunned, high sensitivity, WDR sensor at night, where the tunning trades off spatial resolution for sensitivity. Thanks for your comments. I hope this clarifies our goals.

Rick, you are certainly targeting real areas of customer pain - low light / WDR. The big question becomes how much better your image quality will be and at what price points you can deliver them. For example, if your technologies low light performance is about the same or only moderately better than large incumbents 'super low light' offerings, it's going to be a hard sell, unless you can offer it at a fraction of the price (which I doubt you can do with CCD and lower volumes).

However, if you can make a big jump over those offerings (e.g., a crisp detailed pictured when they are dark and noisy), you have a real shot. As our Integrated IR vs super low light shootout results show, super low light is still a long way from rivaling the range / darkness that added illumination can bring.

Hi John, I agree with all of your comments. There is nothing that helps picture quality more than more light. So turning on lights is one of the very best solutions. Indeed, the technology that we're talking about is also about getting more light. We do this by gathering the light from adjacent pixels together so that you get enough signal that you can see something. Now does this provide a "razor sharp" picture? Well you give up resolution for the greater sensitivity, and this is only helpful when you can't see anything with the higher resolution because the light is too spread out. So how much sensitivity can you gain? Well suppose we had a 1920x1080 sensor and could do 4x4 pixel binning. This would give you a 480x270 image. Now this would not be "razor sharp" compared to the 1920x1080 image, but the light sensitivity will have increased by 16 times, taking a 0.25 lux camera, say, into a 0.016 lux camera. And with our technology, you could arrange to have a higher resolution image (say 960x540 pixels) available for the regions of the FOV that are better illuminiated. Plus, you can always use it in normal mode for day time or for night time for that matter. In short, you give up no capabilities, you just gain capabilities by a flick of a switch. There are more details, of course, which I can't get into in such a brief space, but I think I am giving a reasonable, and defendable, account of the benefits of this technology. Will this cost more than a CMOS sensor? Yes it will. But the price will be just about the same as any other CCD sensor.

I think it's worthwhile goal to pursue.

As for this specific claim: "the light sensitivity will have increased by 16 times, taking a 0.25 lux camera, say, into a 0.016 lux camera."

What really ultimately matter is how much 'better' that image is. If at that same light level, top of the line cameras are really dark and grainy while your camera provides clear details (faces, license plates, etc.), and you can do it at roughly the same price, you have something very attractive.

The problem I have is trusting or comparing lux ratings, so it's hard for me to abstractly comment on the hypothetical advantages until I see real side by side pictures/videos.

Understood. Lux rating on cameras are inconsistently rated. The proof is in the picture. That is what IPVM is all about and what it does so nicely. Still, one can easily understand that adding the light of 16 pixels together before suffering the read noise will increase the sensitivity by a factor of 16. But this technology is not for reading license plates. It is for detecting that there was a car there in the first place, a car that you couldn't otherwise see. It's not a panacea for all problems. To read that license plate the best thing one could do is to turn on a flood light and use a hig-res sensor. Or one needs a lens as big as an astronomical telescope to collect enough light even under low-light conditions and then use that high-res sensor. In situations were this can't be done, however, it still might be useful to use our technology and to establish that there was a car there and perhaps its make and model, even if the license plate can't be read.

Thanks for your comments, John. As always, you are bringing out the critical issues. And we are in the process of trying to perform some demos so that the pictures can speak for themselves.