controlling widths of optical pulses

[sdc18]

Hey all,

I am trying to set up a system that reads in an optical pulse of light (~nanosecond time scales), calculates the input width x of the optical pulse, and then outputs a pulse of optical light at a different width y through any function y=f(x) with little latency. Any suggestions?

 

[jiripolivka]

Hey all,

I am trying to set up a system that reads in an optical pulse of light (~nanosecond time scales), calculates the input width x of the optical pulse, and then outputs a pulse of optical light at a different width y through any function y=f(x) with little latency. Any suggestions?

The straight known method is to convert the light pulse to an electrical pulse, then use it to activate a fast LED or laser to get an output.

Other physical methods can be used, like phosphorescence, gas ionization and similar, but to my knowledge it is difficult to control the y=f(x) function, the delay and pulse width.

 

[sdc18]

The straight known method is to convert the light pulse to an electrical pulse, then use it to activate a fast LED or laser to get an output.

Do you know of a reference in the literature?

Once it is an electrical pulse, how can you implement the y=f(x) pulse width change?

 

[jiripolivka]

Do you know of a reference in the literature?

Once it is an electrical pulse, how can you implement the y=f(x) pulse width change?

Electrical pulses are easy to control. Pulse width can be set in any monostable multivibrator.
As you mentioned only nanosecond pulses, then he task is a bit more difficult. Check Hittite for very fast devices you can use.
The first device to be used to convert an optical pulse to electrical, is a photodiode. For nanosecond pulses, again special PIN photo devices should be used.

I do not remember any particular reference. Use google, also there are magazines showing new development, like IEEE Optical Transactions and others.

 

[sdc18]

Electrical pulses are easy to control. Pulse width can be set in any monostable multivibrator.

Doesn't a monostable multivibrator give only a single output pulse width? That would mean that y = f(x) = c, where c is a constant. What about y = f(x) = 2x?

 

[mtwieg]

For such short pulses I'd recommend trying something with tapped logic delay lines. One on the receiving end which measures how many gate delays the input pulse lasts, then uses a combinational logic lookup table to select the right number of gates to use on the output delay line to produce the output pulse.

 

[sdc18]

What you mean with a "tapped delay line of logic gates" for measuring? If the pulse is traveling down the waveguide of gates, how do you measure its width?

Moreover, how, without using a processor, can you tell another tapped delay line to implement f(x)?

 

[FvM]

All this is surely possible with programmable logic, e.g. a FPGA.

 

[sdc18]

All this is surely possible with programmable logic, e.g. a FPGA.

I do not see how, without using a processor, you can tell it to output f(x).

 

[FvM]

f(x) can be e.g. implemented as lookup table or polynomial.

 

[sdc18]

f(x) can be e.g. implemented as lookup table or polynomial.

A function for pulse widths as a look up table?

I can imagine having a clock that counts the pulse-width in cycles in stores it in a register and then performs the polynomial operation or uses a look up table, BUT this imlies a slow system at the clock speed.

I am looking to do it with an asynchronous mode with no latency with a resolution well below the clock cycle, and have a continuous input.

 

[FvM]

"No" latency sounds irreal. Depending on the function f(x), the output pulse can't be started before the input pulse has finished and it's duration has be determined.

I believe that you have an idea about intended latency range, but it has be neither stated nor the purpose of low latency explained. There may important reasons in your application, or the specification is purely arbitrary, we don't know.

You can probably implement certain conversion functions asynchronously with low latency. Without restricting the class of functions, low latency is a problem.

 

[sdc18]

Sure, "No" latency is impossible - but my question to you is simply:

How can one implement a conversion function that maps an input width x to an output width y such that y = f(x) = 2x using asynchronous logic gates? You have mentioned a tapped delay line, but I have no idea how that would work. Can you elaborate at all?

 

[FvM]

This can be made with a combination of delay lines and logic.

 

[sdc18]

Can you give me a hint on what you mean or refer to something online? I have tried looking and came up with nothing.