While Not Quite a Camera Obscura, This Obscure Camera Has Captured Our Interest!

Pim de Groot (@mifune) has a thing for cameras — that much is evident from the fact he owns a Hasselblad — one of the de facto high-end camera makes, the name of which is recognized far and wide around the globe.

Lesser recognized, though we'll do something about that if we can(!), might be this interesting take from @mifune — a study in developing his own take on a digital camera, using his own design of scanned photodiode array.

The initial impression you might get from the image above, is that this project is merely some CCD on a USB interface board, with a jury-rigged E-Mount adapter, to get some more professional optics mounted in place.

But as with all the bets photography, there is far more to this than meets the eye. If you were to remove the lens from its mount, you'd be greeted with an image sensor array, as you would be any camera. The thing is, there's a bit more of a story to this design!

The above image shows what you would typically see if you were to remove a lens of a high-end camera from its mount — a large sensor array that is comprised of many millions of individual detector elements, or pixels, as we more commonly refer to them in this context.

There have been two methods of sensor technology primarily employed within digital imaging — the first, CCD, or charge coupled device, has seen reduced popularity in recent years, as the performance of active-pixel sensors has increased, removing the previous advantages that CCD used to offer over APS.

An active pixel sensor is comprised of a photodetector — a device which is able to represent the amount of light it is subject to, as an electrical signal. Typically, the low-level device employed to do this is a PIN photodiode, and in the arrays seen within modern day digital cameras, many many millions of such photodiodes are crammed into the footprint of the image sensor located within the heart of the camera body.

With the appropriate filters, these photodiodes can be grouped to respond to light of varying wavelength, allowing us to capture full colour images by way of capturing the response for red, green and blue light that falls upon the sensor. Throw in some high-speed readout and processing, and you have a digital camera. Simple, right?

In theory, yes. In practice, you'll need some multi-million dollar specialist wafer fabrication and assembly processes, clean rooms, and all that fun stuff that goes with lithographic wafer production — in order to meet the ever expanding demands of the market, and their never ending thirst for more pixels, and higher resolutions.

@mifune has decided to take things in the other direction, and while his camera sensor might not meet the demands of today's professional photographers, it's been completely manufactured by himself, without any need for all of that "messy" clean room work.

The trade-off is that the resolution is a somewhat... limited resolution of 24 pixels, mounted in a 6 x 4 array. Instead of that hefty image sensor seen in the Sony A7 above, let's take a look at what's under the hood of @mifune's camera body.

While you probably aren't going to be using this camera to take any of your upcoming headshots, that's not the point of this build.

Perhaps wanting to expand upon his experience gained while building his mechanically scanned, single-pixel camera — an impressive build in its own right! — or perhaps wanting to shave a few minutes(!) off the exposure time needed to construct an image, I can only suggest these as two possible reasons behind @mifune's motives. Regardless of reasoning, this is a fascinating exploration of what is needed to construct a camera from scratch.

With the photodiode array laid out in a grid, @mifune needed a way to read in the voltages being output by the photodiode junctions as they respond to the varying amount of light falling incident upon them.

WIth a Microchip SAM D11 MCU being put to good use as the camera controller, @mifune has a few analog-capable pins at his disposal, internally mapped to the ADC channels of the D11.

One slight issue, is that the D11 doesn't even have 24 pins, and of the pins it does have, well, even fewer can actually be mapped to ADC channels, and well, it's only got a maximum of 10 channels that be read by the 12-bit ADC.

The trick to making this all work, is the venerable ol' 74HC4051 analog multiplexer chip — capable of turning 1 analog I/O into eight — as long as you don't mind a bit of timesharing!

The 3 signal select pins of the device (S0:S2) allow for selection of any one of the 8 analog channels, mapping that specific I/O to the common pin (COM).

If any of this seems like it's ringing a bell, that might be because we've recently seen a similar trick from Greg Davill - who choose to use a very similar part - the 74HC4067 - on his OrangeCrab (🍊🦀!) board, allowing him to implement multi-channel analog capability, with only a single ADC present in the ECP architecture!

With the array of 6 x 4 photodiodes giving a total of 24 analog signals to monitor, @mifune has elected to use 3 of the 74HC4051 parts, allowing him to abstract the 24 signals down to 3 banks of 8 signals, stepped through in sequence, to be passed into 3 of the SAMD11 ADC channels.

We can take a peek at the screenshot of the schematic provided by @mifune below for some further details!

So! What's the picture quality like? Surely this labor of love must have a certain characteristic to its image that makes this all worth while?

It certainly has character, if not much detail... But then again, that's not the point to such a build. As many of us know, we don't often build something more readily available on the market because we can do it better. No, we design and build things that interest us, because we want to better understand the world around us. Even if our devices have a slightly fuzzier view of things then we do!

Make sure to follow @mifune on Twitter to keep up to date on further developments with this exploration into digital exposures!