# Optical analog to digital converter

1. Jan 31, 2004

### dlgoff

I designed an optical A/D (on paper only) several years ago. I have no means to make one since small scale integration is required. Any ideas how to get someone interested?

If it works correctly, there would be zero conversion time and only settling time for the detectors outputs.

2. Jan 31, 2004

### chroot

Staff Emeritus
Can you explain a bit more about what an "optical" A/D would do?

- Warren

3. Jan 31, 2004

### dlgoff

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.

4. Jan 31, 2004

### Jimmy

5. Jan 31, 2004

### dlgoff

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.

6. Feb 1, 2004

### chroot

Staff Emeritus
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?

- Warrem

7. Feb 1, 2004

### dlgoff

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.

The thing with this idea is that you can get a fast 12,14,16,... bit A/D at a very low cost compaired to good electronic chips.

8. Feb 1, 2004

### chroot

Staff Emeritus
Very few commercial A/Ds are SARs. The reason? SARs are necessarily slow. There are no "fast SARs."

- Warren

9. Feb 1, 2004

### dlgoff

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.

10. Feb 1, 2004

### chroot

Staff Emeritus
Well, I still don't see how this is possible, but I'd like to see what you've come up with. I'm an applications engineer with National Semiconductor Corporation, Inc. I specialize in analog to digital converters.

I will have you know that virtually all "high-speed" (> 1 MSPS) ADCs are actually folded-flash architecture, or "multi-step flash." Only the low-speed industrial products are SARs.

I'll also tell you that the speed of an individual conversion is frankly not usually of much importance. Most modern ADCs are pipelined. For a chip with an 8-cycle pipeline latency, it means that a conversion takes 8 clock cycles. The data coming out of the digital pins was sampled eight clock cycles ago. Only in very rare instances does this pipeline delay cause any problems for a designer.

The biggest concerns for most designers are power consumption, footprint, and dynamic performance.

- Warren

11. Feb 1, 2004

### dlgoff

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.

12. Feb 1, 2004

### chroot

Staff Emeritus
If you're using LEDs and photodectors, I can only assume you're using an array of photodetectors, each with a different voltage offset, and producing a so-called "thermometer-code" output. In other words, for any given input, some of the low-offset detectors will be turned on, up to a critical point. Past it, all the detectors will be turned off.

You'd then have to do a "thermometer-code" to binary conversion.

- Warren

13. Feb 1, 2004

### dlgoff

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.

The LED source and photodetectors are integrated into the surface of a GaAs (or other electrooptical crystal materials) substrate. They are arranged such that the LED output is through the substrate out into the "analog input part of the circuit" then back through the substrate to the photodetectors.

Can you visualize this so far?

14. Feb 1, 2004

### chroot

Staff Emeritus
Through the substrate?

Out into the analog part of the circuit?

No, sorry, you're not making any sense.

- Warren

15. Feb 1, 2004

### dlgoff

Okay. Let me try it this way. This device is made in layers parallel to the substrate. Normally an LED would emit its light off the chip but in this case I want it to go through the substrate and encounter the polished interface which has a three layer film on it.

This film consist of a mirror(Al deposit) with an etched pattern that allows a circular ring of light to be transmitted through to the next layer where it will encounter another mirror. The material between the mirrors is electroactive. After multipul reflections, the light will exit through the etched pattern to the photodetectors. The location of these detectors is critical to the code (digital output).

Now apply a potential difference between the mirrors. This will change the index of refraction of the electoractive material between them causing the rings of reflection (perhaps 100s) on the etched mirror to move directly proportional to the applied potential.

Make any since now?

eddit:"...move directly proportional to the applied potential." Actually the index of refraction is proportional to the square of the field. But the angle of refraction is directly proportional to the index of refraction.

Last edited: Feb 1, 2004
16. Feb 1, 2004

### chroot

Staff Emeritus
Uh... and you think this is simpler than a comparator? LOL....

- Warren

17. Feb 1, 2004

### dlgoff

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).

How much does a 16 bit A/D cost now days?

18. Feb 1, 2004

### chroot

Staff Emeritus
I'm sorry, what you're proposing is one of the most complicated pieces of electro-optical machinery ever concieved. There is absolutely no chance of anyone attempting to build anything like this.

Besides, I think a lot of this idea rests on technology that doesn't yet exist -- "electroactive substances" with variable indices of refraction?

Not a chance, man. Sorry.

- Warren

19. Feb 1, 2004

### dlgoff

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.

added by edit: Anyone else care to comment?

Last edited: Feb 2, 2004
20. Feb 3, 2004

### dlgoff

a picture's worth 1000 words

Here's a simple section drawing of the A/D concept. From a top view, the emitters mirror slit is almost a complete circle with the detector mirror slits spiraling out from its center.

When an input potential is applied to the mirrors, an electric field is created across the dielectric causing the index of refraction (parallel to the field) to increase. This will make the angle of refraction of the light going from the substrate to the dielectric decrease.

The rings of light between the mirrors will "shrink" and move across the detector slits as the potential is increased.

Any thoughts?