There is a analog input that is totally isolated from a light source and a number of detectors. The detectors outputs are a digital representation of the analog input. The output code is developed during a pulse from the source.
That would be what ever you are wanting to measure. The sensivity of the input would depend on the material which is being "probed" by the light source. Maybe in the range of 0 to 10 VDC.
Like any other A/D integrated circuit. But intead of providing a convert input and waiting for the conversion time period, you have a pulse input which should give the digital output much quicker.
I'm confused. If this is an optical A/D, why are you measuring the input in volts?
I have to ask -- do you know how existing A/D converters work?
It uses optics to accomplish the analog to digital conversion. The input and output are electronic. So I guess it's really electro-optical.
Yes I do in general. Most, I think use fast successive approximation conversion. You'll find other inputs on the chips for voltage references, chip enable, etc.
This is true. For fast A/Ds, flash conversion is needed. These are the expensive ones. The idea I came up with should be much faster than these. You sample the outputs as fast as you can pulse the integrated light source.
Shoot. Your the right person to be talking to. I can try to explain it here. The design is quite simple.
I'm not familure with this way of clocking since it was several years ago that I was more up on existing A/Ds. Anyway it seems to me that speed would always be important for realtime processing.
Well this one should do the trick since we're only talking about a single source (LED) and detectors for the bits (photodiodes). The input has very high impedence so no power there. It can be put into an inline package. As to dynamic performance; I need one to test.
One LED and an array of photodetectors (I mentioned photodiodes but phototransistors is what I meant. However diodes could be used I think) but not producing any thermometer codes.
How many comparators do you need for say a 16 bit one. I'm sure the IC is very complex with many transistors (causing more heat with speed as one disadvantage). Here all you have are a few transistors, a diode and a three step film disposition (deposit aluminum, etch pattern, deposit active film, deposit aluminum and your done).
What's so complicated about it? They make ICs all the time. Do you mean that it would be hard to make an LED transmit light through the substrate? I don't see a problem here. Deposition the thin films. Still no problem.
There's lots of existing materials that can do this. Have you ever heard of the Kerr effect? Any optically isotropic substance can become doubly refracting. Even glass to some extent; however it take a high voltage to see the effect. There are polymers that have been designed that are very sensivity to an electric field. i.e. it has a large Kerr constant.
No. For a 24-bit AD you need 24 detectors.(arranged in a spiral fasion with the least significant bit being the furtherest from the center. i.e for the dielectric having an index greater than the substrate in this case.
Well I think if the openings (slits) are large enough there shouldn't be any problem. Since I'm not considering using a coherent source (laser) this will decrease inteference.
Yes this would be a problem. But since all detectors will be read at the same time during the source pulsing, there is really only one delay to consider. i.e. the total settling time. Don't CD players use electro-optical devices now?
Oh, and CD audio is 44 KILOhertz. Audio applications are "low speed" by just about any standard.
Then I don't grasp how do you produce binary code by 24 detectors that represents 16 million possible values. At which point does substraction and residue amplification happen? You can't produce binary code by proportioning alone. "Arranged in spiral" doesn't hit the light for me, but I'm curious.
How large is enough? There is a need to illuminate detectors in precise pattern, and have no "dead-spots" between readings, as well as no garbage-spots. Thus detectors must be closely spaced, and you have issue with either false concurrent detection or transient where no detection happens. Difraction makes this worse, and it happens quite easily.
Yes. The higher the light source frequency the better. Can InGaAs/InAlAs detectors be integrated in an InP substrate?
My first idea was to use a wedge between the mirrors. But I think that would be to difficult to fabricate.
Do you see any other applications for something like this?
Cut the angle of refraction in half and you get twice as many reflections and so on. The placement along the spiral should be such that the furthest detector will have twice as many reflections passing over it as the next closer one and so on. The inner most on will be off over the whole range of input voltages until all other bits are on. The placement is very critical.
Larger than many wavelengths. Very precise pattern. Not close around the spiral but the distance to the center circle may overlap a little. There may be a little light but still be in a zero/off state.
Couldn't it be done with a large enough scale where resonance doesn't come into play? i.e. just reflections and not "standing waves"? Am I missing something here?
Wish I could but I'll have to respectfully decline the offer.
