# How are pixel arrays read

How is a pixel array read, such as in a LED TV, digital camera, microbolometer, etc. All of these things have an array of pixels that either transmit or receive. Each pixel must be read somehow, is this done with a multiplexer and do you need more and more multiplexers the more pixels you have.

How many analog outputs does a multiplexer have? If you have a lot of pixels would you have to have multiplexers that are reading other multiplexers or just one big mulitiplexer would make more sense since your scanning an array.

So then the analog outputs of the multiplexer would then connect to a micro controller and presumably the controller could tell an LED screen which scan goes in which LED pixel. So say the multiplexer scans sensor 11 and sends it back in a micro second or less the micro controller then displays that signal in the LED screen 11. So the controller would have to know the scan order to know where to display.

Is this correct thinking?

Are there solid state lab companies that can take a design and make a chip for you as a proto type?

For large LCD/LED arrays - are they addressed individually?, I agree a MUX would be difficult.

Baluncore
So would the CCD multiplexer have to be etched on to the same chip as the array or is it possible to hard wire them as 2 separate devices. Though this would likely result in a lot of little tiny wires going from each pixel into the CCD which may not be practical.

Baluncore
The CCD shift register and the holding register must be manufactured as a single integrated unit with the pixels.
The CCD is not a multiplexer. It is a long shift register.

A SIPO shift register requires a single data input and a clock connection. Those two connections are at one end of each shift register. A display has many edge connections so many shift registers can be operating at the same time in parallel with different data.

So what is a multiplexer used for?

It almost seems like a multiplexer could be used in conjunction with a micro controller instead of a CCD shift register. Is this correct thinking?

Baluncore
No.
A multiplexer is a tree structure. For 1024 pixels it would require an address depth of 10 levels. It takes less time to load a line of pixels through a shift register because binary pixel addresses do not have to be distributed throughout the display.

A shift register needs only 2 layers. One to shift the data and one to hold it for display.

Would either of these 2 texts cover it:

https://www.amazon.com/dp/0121197905/?tag=pfamazon01-20

It seems like these courses are more modeling and less dealing with device design, would CCD and multiplexers fall into VLSI design?

I am actually looking into doing a masters in EE but am finding it difficult to nail down where device design is actually covered (also such as the detailed internals of a op amp).

The math is fun and all and i'm reasonably good at it but I really would like to design and contract out proto type devices to be built. I will also take the other classes such as fourier optics and I desperately need to take a matlab class, I was hoping to cobble it along as I went but ended up with a C in signals processing because my matlab skills are not really up to par, they are not horrid as I can program pretty well in visual basic but all the ploting in matlab is kind of quirky since it does not automaticly 0 your plots you have to do weird manipulations.

I think I will try to find an MIT open course ware you tube videos which is what I used for discreate signal processing, I absolutely hated using properties and tried doing the integral methods which was ok for homeworks but in tests you don't really have time and it hurt my grade but all properties are is lame short cuts for the calculus you should be doing.

Last edited by a moderator:
Doing a little more digging, what is the difference between a microbolometer and a CCD. Does a microbolometer need a CCD or is a microbolometer a special type of CCD with one analog output to connect to a micro controller?

Baluncore
Modelling is sufficiently accurate to be used for testing the concept before making a chip. The problem is identifying the best concept to implement. That will take years of experience to get right. To work with hardware you need to study the concepts of “Digital Design”.

“Field Programmable Gate Arrays” make it possible to implement hardware without high costs.
http://en.wikipedia.org/wiki/FPGA

New chip design is now so specialised that you cannot expect to work in that field. Your books are about signal and image processing. That is numerical programming to implement mathematical algorithms. It is different to digital hardware design.

Try not to be distracted by a fascination with electronics, the field changes very rapidly. Electronics is now embedded as a modular tool used by engineers to manage the bigger systems that change more slowly. The most reliable employment is in Electrical Engineering.

Baluncore
A microbolometer measures IR radiation. It senses element temperature and does not generate a charge directly so it requires an interface different to a CCD.

A microbolometer measures IR radiation. It senses element temperature and does not generate a charge directly so it requires an interface different to a CCD.

That is what I was reading as well, it uses a sensor material that acts as a variable resistor when exposed to IR with a reflective metal underneath to give better sensitivity. However, there is still a large array of these resistor pixels that have to be analysed. I was reading that a multiplexer is used to scan this array of resistors to give a single analog output, as I understand the multiplexer will scan the array and give a single time varying analog output which can be read by a micro controller to process and display.

In this case is a multiplexer appropriate or am I still way off base.

Last edited:
Baluncore
If the sensor generates a charge then a CCD is simplest.
Microbolometers are big, very slow and have relatively few pixels compared to optical sensors for cameras.
Multiplexers could be used efficiently for microbolometers.

A micro bolometer is also a powered device, the IR is not generating a charge simply a change in resistance so the array has to be powered and as I understand they chew up a lot of battery power to use, but that's just part of it. I thought about having only the array built and hundreds of tiny wires that are micro coated run from the array to a non-solid state multiplexer but that might be a giant head ach, however the less solid state stuff I have the less complicated, I don't know if you can buy a multiplexer with 600-800 inputs and a single output that is not huge though so I may have to have it built on the chip which would likely be expensive and introduce more noise.

Do you know of a solid state lab that will custom build proto type devices who do it as a business. IF I could design and have something similar to a bolometer built then I could interface it with a arguino micro controller and a small LED screen. It would be a fun little project for the winter.

Thank you so much for walking through this with me.

Last edited:
Baluncore
You will need more than $100k to invest in development of a new bolometer chip. You will need to negotiate with several fabrication plants that want to take your money. You will have little chance competing against military funded research programs. There is no advantage launching a product that is behind the current technology. I really don't think you are at a stage yet where you have a significant chance of success. Indeed, with respect, I believe you are so far behind that you think you are first. At this stage you should search the websites of the following list of manufacturers. http://en.wikipedia.org/wiki/Microbolometer#Manufacturers_of_microbolometer_arrays Look for a commercially available low cost microbolometer array at a reasonable price. I knew that I would not compete with military research, however, I was hoping there was a solid state outfit business that catered the hobbyist doing one off prototypes. I called FLIR and a proper flir unit to mount to a plane is over 10 grand and none of it is STC'ed anyways. As far as I know its not possible to buy just the bolometer thus forcing someone into either having one made or buying an entire unit. Plus I kinda like the idea of being able to pick which colors I assign to which temperature range. I have some experience in microelectronics fabrication as a chemical engineer and I am working on a masters in electrical engineering, unfortunately our university here does not have a solid state lab, I have been trying to petition for them to get one and they are going in the right direction but a solid state lab is a tall order. Also as much as they want for a flir you could be a small bit behind and easily become commercially viable as there is tons of wiggle room to make an more cost effective device that people could buy for say 4-6k instead of 10-15k Last edited: f95toli Science Advisor Gold Member I knew that I would not compete with military research, however, I was hoping there was a solid state outfit business that catered the hobbyist doing one off prototypes. I called FLIR and a proper flir unit to mount to a plane is over 10 grand and none of it is STC'ed anyways. There is no such thing as a "one off prototype" of something as complicated as a multiplexed bolometer. Even at the research stage it makes little sense to only make one device: the overhead is just way too high. Also, just to give you some idea of costs, even if you were to do all the actual work in the cleanroom yourself, they would stilll charge you about £1000/day just to allow you to enter the cleanroom and use the "basic" equipment. There are cleanroom that will do all the work for you, but expect to pay about £2000 for one wafer with a single metallization layer. Something like a microbolometer would cost orders of magnitude more, even if you do not include the time needed to develop the fabrication process. Note that these are the current prices in the UK, but I assume the price would be approximately the same in the US, The point is: this is very, very expensive stuff. There is a good reason for why only large companies can afford to create complettely custom circuits; everyone else just uses a few large foundries, but that effectivly limits you to standard processes such as CMOS (can't be used for a bolometer). There is no such thing as a "one off prototype" of something as complicated as a multiplexed bolometer. Even at the research stage it makes little sense to only make one device: the overhead is just way too high. Also, just to give you some idea of costs, even if you were to do all the actual work in the cleanroom yourself, they would stilll charge you about £1000/day just to allow you to enter the cleanroom and use the "basic" equipment. There are cleanroom that will do all the work for you, but expect to pay about £2000 for one wafer with a single metallization layer. Something like a microbolometer would cost orders of magnitude more, even if you do not include the time needed to develop the fabrication process. Note that these are the current prices in the UK, but I assume the price would be approximately the same in the US, The point is: this is very, very expensive stuff. There is a good reason for why only large companies can afford to create complettely custom circuits; everyone else just uses a few large foundries, but that effectivly limits you to standard processes such as CMOS (can't be used for a bolometer). Wow, thank you for the info. So if you will humor me, would it be possible to build a bolometer with the following equipment: vacuum deposition chamber (could you deposit more exotic metals like germanium with a simple heating coil?) plasma etcher (would this replace the need for HF) good printer for making masks the UV exposure lamp for the paint on chemical Oven (I know these are several thousand dollars) And here is where I'm really reaching, could hundreds of very tiny very thinly coated wires be micro soldered on to each pixel to feed the multiplexer so that the multiplexer did not have to be a solid state device? Could you buy used equipment for relatively cheap Baluncore Science Advisor How about a different approach. There were film based cameras long before digital cameras. So why not search for some weird IR film equivalent, hopefully reversible. If that film is temperature sensitive then you have a thermally sensitive image. You can then use a cheap USB camera to view the thin film. An oil film shows interference colours dependent on thickness. Thermal expansion or evaporation rate will be temperature dependent. Both heating and cooling change temperature, your film could operate above or below ambient temperature. If you selectively heat the image film, (based on the image), you can bias the temp sensitivity to where it is needed. A camera looking at the film can learn the mapping of the film to image, and the relationship to an indirect method of selective heating, such as a directed IR LED. Chemical reactions are dependent on temperature. If the products look different then consider it as a film. There must be some chemical temperature indicator solutions, you only have to find one that works for your application. The film could be an array of isolated separate cells, each cell with contents that can be exchanged or reversed. The viewing angle of LCD screens are often temperature sensitive. There are cheap battery testers that heat an LCD film. Way back in the 1980's, amateurs used dynamic memory chips as crude image sensors. I see no reason why a temperature sensitive film should not be used as an image converter. Baluncore Science Advisor The problem with the back layers of a microbolometer array is that some parameter must be read at each pixel location. The switches being used must not indirectly effect that measurement. For example, to measure the resistance of a pixel you could provide a calibrated current and measure a differential voltage. That requires three multiplexers, not just one. You will have problems delivering the current through the same multiplexer that measures the voltage, because of connection resistance. Consider an array of sensors. By including diodes in the array, each could be addressed / selected individually by row and column lines. But you will still need two multiplexers. You would be better making an array of PIR detectors. A 32 * 32 array would need 1024 sensors, at a bulk price of US$1 each that is a significant cost, but still cheaper than a microbolometer array.

I have drawn up a sketch of what I would like to design, I will clean it up and scan it in tomorrow. The chemical film would not work as I would like an active sensor, something mountable on a plane I could use to see though clouds if I got in a bad situation on accident.

The IR will impart different resistances across each pixel, these pixels could all be ran through a multiplexer but your right the multiplexer itself will have some resistance, perhaps the resistnace vs current could be measured and a curve generated for the multiplexer and that could be corrected in the micro controller. Then since the multiplexer would have only one output you could attach a current measuring device that feeds a micro processor (with corrections for the multiplexer resistance).

I will have to google PIR detectors and see how large the array would be, it cant be too massive for my applications. I have done a lot of research before starting this thread and I don't think I can get away without using at least a solid state array.

f95toli
Gold Member
Wow, thank you for the info. So if you will humor me, would it be possible to build a bolometer with the following equipment:

vacuum deposition chamber (could you deposit more exotic metals like germanium with a simple heating coil?)
plasma etcher (would this replace the need for HF)
the UV exposure lamp for the paint on chemical
Oven (I know these are several thousand dollars)

And here is where I'm really reaching, could hundreds of very tiny very thinly coated wires be micro soldered on to each pixel to feed the multiplexer so that the multiplexer did not have to be a solid state device?

Could you buy used equipment for relatively cheap

That equipment would be start, assuming you have a cleanroom to put it in. You would also need a few other pieces of kit (ashers, mask aligners etc etc). Also, I am not sure what you mean by "good printer". Assuming you can get away with using photolithography you would need to order the photo masks from an outside company (unless you have a laser lithographer or -even better- an e- beam lithographer). You can't use a printer for doing photolithograhy.

My guess -based on recent experiences upgrading our own cleanroom and talking to a collegue who just built a new cleanroom from scratch- is that if you stick to buying used equipment etc, you might be able to get everything you need for less just under $1M, if you are lucky (a basic RIE etcher is about$100K)
Then you only have to spend a few years learning how to use the equipment.