want to build a circuit having 300 Led and 300 photodiodes. The leds will be turned on one by one and the values of all the 300 photodiodes will be recorded and send to computer.

The problem is i want to do this processing at a very high fequency. i should be able to take all the 300 reading 100 times a second.

The solution should be scalable to larger number of leds and photo diodes.

Is it necssary to use a micro controller/processor? Since i don't want to run the circuit independent of the computer.Can i use serial or parallel ports? i suspect the data rates will be too low. how about using USB?

What the maximum frequency at which these diodes can work? I assume it must be in Ghz range since they are used in fibre optics. cost is not the issue, i want to know the upper limits of this scenario.

 

What is the precision of the readings? 8 bits, 12 bits, 16 bits? At 30,00 samples/sec, you may have a challenge finding an A to D converter plus switching that can keep up.

 

as high as possible 32 bit will be good.

how can i do it using mimimum electronics ( asking because i am a software engineer). budget is not a constraint.

what are the data rates supported by standard anaglog to digital convertes?

there are high speed port avaliable e.g. FireWire 400. so, i think it should be feasible

i am basically making a custom camera

 

Actually it is only 30 kHz. or 33.33 microseconds per LED. The problem is that LEDs and photodiodes don't necessarily respond as fast as you would like. It was a matter of great frustration to me when I found out how slow opto-isolators are. I think your project is doomed based on what I think your expectations are.

There are 32 bit DACs out there, but you will have trouble finding and using a 32 bit A2D converter.

BTW aren't cameras made from CCDs (Charge Coupled Devices)?

 

my final version will be ccd based.
i want to start experimenting with easier components as a proof of concept.

so, you mean it will be difficult to get led which can be turned off and on 100 times a second? then how do fiber optics work?

note that in my scheme although high data rates are required but individual leds don't flip flop more than 100 times a second.

 

The partial answer to your question about the LED sources used in fiber optic transmitters is that they are expensive. Although they may operate at high frequencies they have significant delays unless you pay extra for something like the HCPL 0710 (an opto-coupler) which has a 40 nsec. propagation delay.

Point is, you will get nothing like that performance for garden variety LEDs you can purchase from Digi-Key or other sources for this experiment.

Try it your self and measure the delay from switching the LED on to when the photo detector turns on. Then notice what happens when you turn the LED off and it has, OMG - persistence!

You will be quickly disappointed unless you have some tricks in your bag we don't know about. I'm not saying you can't find a way to accomplish your goal; I'm just saying it won't be easy.

 

Wave shaping, and detection can be enhanced by software techniques, but the basic response times will still complicate coding into something I would call 'a nightmare'.

A scheme in which you setup time delay sampling to 'bypass' long response times and detect a slope in the rise or fall times would be how I would approach this.

You will not easily be able to process 300 LED ckts at a time, but 16 or 32, or even 64 parrallel inputs could be do-able. Repeating this process across multiple PC platforms and networking them together, would be the approach most attainable.


adaq- analog-data aquisition units are available with large numbers of daughter card input modules and advanced software programming options that could do all of this very easily, but are VERY pricey. Is the price something you are trying to avoid or find a work around for?
Are you researching peak output, or response times?

Fleshing out your needs some more would help narrow suggestions and help into the specific forms you could use.

Kermit