Manufacturers, Shoot This Idea Down If You Can!

Make a High MP camera with multiple sensors but just one lens.

Here's a picture of a sensor. Others are different, but generally you have an array of photodiodes (pixels) with sensor output leads, coming out below and to the side of the chip itself. Because of this encumbrance, you can't get them adjacent to each other to make a seamless array.

/

So, would it be a worthwhile idea to make sensors so they can be placed right next to each other with no gap? Like make the wires come out the back of the sensor more instead of the side. So you could do this:

Then you could take 4 4K sensors and make a 48 MP camera with one lens instead of the 4 you need with normal multi-view cameras. Send 4 independent streams to the VMS and do a simple, borderless quad, (writing a simple driver if necessary). Or send 4 streams to 4 monitors and put the monitors in a bezel-less display wall quad.

And unlike any typical multi-imager/lens quad that might suffer from pin cushion lens distortion in the corners, the sensor array with shared lens would be seamless, naturally.

So maybe it would be a good idea if they did make sensors that can just be arranged domino style to make a larger area sensor of varying dimensions. But I doubt it actually is, since they don't seem to make them, but I'm not sure why.

Maybe:

  1. It can't be technically done, not at a reasonable cost at least because ?
  2. Even if you could make sensors like that for the same cost, the "innovation", (1 lens, 4 sensors) does not provide any value because ?
  3. They actually do make sensors like that them but ?
  4. Other

What do you think, shoot the idea, or steal it?


I don't understand the benefit of this. Is there really that much demand for a really high MP camera (Avigilon has had the 29 and now 30MP offerings, and they're not top sellers overall).

Your biggest problem would likely be finding a suitable lens. Even the Canon EF lenses that the Avigilon cameras use are only designed to cover a 35mm sensor area. You'd have to keep your multi-sensor array smaller than the "sweet spot" area of commonly available lenses, which would generally be the Canon or Nikon mounts.

While not technically impossible, there would be a little bit of precision needed to read data from the sensors in a syncronized manner that kept the 4 individual streams in perfect sync so they truly looked like 1 stream.

In the end, I think you'd end up with a $12,000+ MSRP camera with a very limited market.

Thanks for the well thought out objections!

Let me give a more concrete example, so we can see the exact problems. Say we take an Arecont 40MP multi-view with 4 1/2.3" sensors costing around $2000. Electrically we change nothing about how the camera operates. Physically we arrange the sensors in a contiguous quad. Note that this is not possible today, but we are allowing it for the time being.

Each sensor has a diagonal of < 8mm and the diagonal of a full frame 35mm sensor is > 40mm so the image circle of an EF is more than big enough to contain 4 Arecont sensors.

Assuming that we can buy one high quality lens for the price of the 4 that come with the Arecont, we could put together a 40MP camera that would act like a single sensor camera for around $2000. If the Arecont's frames are synced in their camera, then so are ours.

Not saying it would be all around as good as a 29MP Avigilon, but it would have 10 more MP for 10 grand less, so that would be the value proposition.

But of course the real sensors don't physically allow the needed placement, so the question is still open as to whether they could be made that way economically.

What do u think?

One general issue is the relatively small volumes in the surveillance market and the corresponding inability to amortize significant development costs over them.

For example, how many of these cameras do you expect to sell over 3 years? 10,000? 100,000? It's certainly going to be a niche within a niche.

How much do expect it to cost to develop this type of novel camera?

That is a typical fundamental objection to novel / technically complex new developments and why almost every 'manufacturer' is doing the same off the shelf combinations of sensor / chip / housing / etc.

One general issue is the relatively small volumes in the surveillance market and the corresponding inability to amortize significant development costs over them.

Agreed. Especially any development dealing with fabs or package size...

On the other hand, I would expect that being able to flexibly create sensors of varying sizes and shapes without wafer redesign/retooling would be of value to the entire imaging industry, not just security imaging.

Though I'm old enough to realize there is probably a damn good reason they just don't make the leads come out of the back.

Good thoughts.

However, there are already 40MP sensors available, we don't have to stitch 4 8K sensors together to make a high-res camera. And I think even if you wanted a 64MP camera it would be easier to fab a 64MP sensor to current methods than to do a respin of an existing sesnor to make it an "edge to edge" format with a ballgrid array or something.

There simply isn't much market demand for a super hi-res camera, especially if it involves using non-COTS components (as John mentioned above).

This was my initial thinking as well - what's the benefit of multiple sensors butted together when you could just have one big sensor? No matter how close you trim them, there's still a seam there.

Your biggest problem would likely be finding a suitable lens. Even the Canon EF lenses that the Avigilon cameras use are only designed to cover a 35mm sensor area. You'd have to keep your multi-sensor array smaller than the "sweet spot" area of commonly available lenses, which would generally be the Canon or Nikon mounts.

Not that difficult to find a lens for such a thing - there are several medium- and large-format digital cameras out there from the likes of Hasselblad, Mamiya, Pentax, Leica, Rollei...

But as you say, such a camera would be hugely expensive on its own, as would the lenses (nobody makes CHEAP medium-format lenses - that would be sacreligious).

And you run into the other issue you see with the very-high-resolution Avigilon cameras: low framerates. I don't know if it's improved, but I recall the 29MP models wouldn't do over 1fps. Network traffic becomes a concern unless you're compressing massively, which starts to negate some of the benefit of all those pixels.

Not that difficult to find a lens for such a thing - there are several medium- and large-format digital cameras out there from the likes of Hasselblad, Mamiya, Pentax, Leica, Rollei...

Though for the Arecont repurpose example above, the 4 8mm diagonal sensors easily fit into the image circle of the standard 35mm format.

They're not cheap lenses either, but you are replacing 4 smaller ones. In any case, if you could align the imagers, I believe you could make the equivalent of the 29MP Avigilon with more MP for a lot less.

35mm lenses don't HAVE to be expensive - the Canon "nifty-50" EF-50mm f/1.8 can be had used for <$100, or brand new for around $120 - it's a simple design, which makes it small, light, and amazingly sharp for a <$100 lens. Most Avigilon Pro demos I've seen are using Canon L-series lenses, which are their cream-of-the-crop lineup, and priced accordingly.

Update: so they got these. Just snap'em together. Yeah, sure... But still cool looking.

The four PanSTARRS cameras will each be the largest digital cameras ever built. Each camera will have about 1.4 billion pixels spread over an area about 40 centimeters square. For comparison, a typical domestic digital camera contains about 5 million pixels on a chip a few millimeters across.

The focal plane of each camera contains an almost complete 64 x 64 array of CCD devices, each containing approximately 600 x 600 pixels, for a total of about 1.4 gigapixels. The CCDs themselves employ the innovative technology called "orthogonal transfer", which is described below. The individual CCD cells are grouped in 8 x 8 arrays on a single silicon chip called an orthogonal transfer array (OTA) , which measures about 5 cm square. There are a total of 60 OTAs in the focal plane of each telescope (The four corner OTAs are omitted because they are too far from the opic axis of the telescope to collect useful data)

Undisclosed A, you can test the validity of your hypothesis in the privacy of your own home without a chip fab. How? Arrange mirrors to redirect the image to off-the-shelf arrays. Please let us know how it works out.

Diagram please... :)

Multi FPA with One Lens

Nice work!

I'll get right on it. I believe I have one of those old Michelson-Morley interferometers lying around in the garage. It should work.

And to think I almost thru it out last week on 'old appliance collection day'.

Don't forget to account for the fact each sensor will only be receiving half the light.

That decreased sensitivity would matter in production, but it should be adequate for proof of concept.

I swear I had no knowledge of this new smartphone camera technology from Light:

Light is planning to use an array of small cameras — very similar to those you would find in a smartphone today — to mimic a larger camera. By precisely aligning and calibrating the individual cameras, the images from them can be combined to produce a single image of very high quality. By having some of the cameras feature a wide-angle lens, and some a telephoto lens, an optical zoom capability is also possible...

Light’s CTO, Rajiv Laroia, explains that by treating the ten or more small sensors as pieces of a single, large sensor, Light’s camera module can add together all the photons that reach a particular location in each of the sensors to mimic larger photosites

Anyway it's different because it looks like they have many lenses (at different focal lengths) AND many imagers, kind of like a micro-sized Dalliamer Panorama...

And Horace, it looks like they got some of your mirrors in there too... :)

Related sensor stitching work from Albert Theuwissan.

I'm still not clear what the benefit would be. From your earlier posts it seems like your idea is that you can turn a ~$2,000 Arecont camera into a ~$2,000 40MP camera to compete with a much higher priced super high resolution cameras.

It still seems that the cost of making sensors that can be set side-by-side, without creating a notable gap in the image, needs to be addressed. Once you do that, I'm not convinced the R&D effort pays off relative to existing options.

It still seems that the cost of making sensors that can be set side-by-side, without creating a notable gap in the image, needs to be addressed...

Yes, but keep in mind the surveillance industry does not drive the massive imaging industry. Security scavenges whatever comes, whenever the price comes down to a reasonable level.

So, this buttable wafer technology (with only 1 pixel gap between sensors) is initially being marketed to the medical and scientific (telescopes) industries. The hope would be that the economies of scale trickle down to us, in a few years, to provide the technology at prices that are not "astronomical".