DSPRelated.com
Forums

I need help

Started by Shatinouex 8 years ago6 replieslatest reply 8 years ago151 views

Hi everyone,

I will appreciate if you can help me with the following:

I am working in visible light communication and want to show the effect of multipath indoor propagation on my system. The multipath channel impulse response is generated using barry et al model;

http://barry.ece.gatech.edu/pubs/journal/jsac93.pd...

My problem is that as I increase my bit rate, my bit error rate increases then start to decrease which is very strange to me. It means that I actually get a better performance at very high data rate when compared to lower data rate (see attached figure).

When I had a look at the channel samples, I saw a lot of zeros between the LOS and the first reflection which might be the reason and I also notice that the change in the BER occurs when my samples/symbol becomes smaller than the length of samples between the LOS and the first reflection.

It seems that at high data rate , only the LOS has effect which does not make much sense.

Can someone give a better explanation? 

A sample of the impulse response of the multipath channel I am considering mPath impulse responseis also attached in the txt file.

Thank you very much.



ber_58660.png





[ - ]
Reply by Tim WescottMay 17, 2017

This makes sense to me.  If a symbol is destructively interfering with itself, then it can reduce the signal strength to an unlimited degree.  If, on the other hand, the symbols are not much correlated over time (which makes sense if you've got a good FEC and are sending at least semi-random data), then a bunch of paths at different delays will just ("just") tend to raise the noise floor.

Dunno if this helps...

[ - ]
Reply by SlartibartfastMay 17, 2017

If you're using only a single example of possible channel impulse responses, then, yes, that can happen.   Is this a free-space application, or...?   Regardless, characterizing the behavior over multiple possible channels may reveal how the channel delay profile affects the performance.   Not all channels are created equal.


[ - ]
Reply by stephanebMay 17, 2017

Hi! Please suggest a new title for this thread and I will edit it.  And next time please use a title that is relevant to the discussion and a little less generic, thanks!

[ - ]
Reply by ShatinouexMay 17, 2017

Effect of multipath channel


Thanks

[ - ]
Reply by artmezMay 17, 2017

I like to not only look at the math, but the physics too. Some of this could be analog frequency related, the photon saturation characteristics of the photo sensor, or the frequency response of the receiver circuit. Are you using a laser or noncoherent light source? If you are using a monochromatic light source (LED or laser), then maybe a filter that only passes that light's wavelength will help a lot, especially for saturation.

Other factors could be the time of day, location, or even bad luck. One test does not a correlation make. Carefully define the problem and then define test conditions as both are critical for any analysis that will produce consistent results. Then once the results are consistent, eliminate sources of error. Simulate the error source(s) and see if that is consistent with test results.

It could be something as simple as time of day creating the higher error rate due to sensor saturation when the sun is overhead or a reflection is impinging on the sensor. Even wind could contribute if it causes some reflections to oscillate at the carrier frequency. Don't assume that with a simple test you have a definitive answer. Sometimes it's easy, but many times, it's not.

[ - ]
Reply by fkimpelrockwellMay 17, 2017

I have a few questions about the communications link:

1. Assuming an AC coupled receiver, what is the high pass break frequency of the receiver?

2. Not sure about the number of contiguous '0's in the data but do you maintain CDR lock at lower frequency. Not sure of the modulation format. M-ary PPM or a more standard format that might use 8B/10B encoding to insure CDR lock?