It will depend on the reasons why the eye is poor.

What is the symbol rate?  What is the fibre dispersion at the operating wavelength?

Are you using single mode fibre or multimode?  DFB, VCSEL or LED source?

Is the poor eye opening due to distortion or because of inadequate signal to noise ratio?  Is it degraded by amplitude noise or timing jitter?

Do you use a directly modulated laser or an external modulator. Direct modulation of semiconductor lasers can be acceptable at lower bit rates, but requires appropriate bias control to achieve acceptable eye opening without operating too close to threshold.  Too high a drive level will increase laser chirp, and increase the susceptibility to fibre chromatic dispersion.

External modulation (e.g using an Electro-Absorption or Mach-Zehnder modulator) generally gives a cleaner optical waveform with less spectral broadening, and will be more tolerant of chromatic dispersion in single mode systems.  A Mach-Zehnder modulator with opposite chirp to that introduced by the fibre can improve tolerance to chromatic dispersion.   Modulation formats such as phase duobinary or carrier-suppressed RZ may also show advantages.

Are you limited by the electrical bandwidth at either transmitter or receiver?  Low bandwidth will give poor eye opening, even when operating over a short fibre link.  Too wide a bandwidth at the receiver will increase noise and degrade performance over longer links.

Are the transmitter clocks synchronised?   Can the receiver adapt to changes in clock phase between different transmitters?

Thank you very much ALan.

Let me try adjusting some of the parameters that you raised and I will let you know the progress.

Well this is an Externally modulated TDM PON using Mach-Zehnder modulators, a CW laser at 1550nm, Pseudo Random Bit sequence (around 7Gb/s),NRZ, and time delays (at 1/sample rate*0.5 seconds), forming the Multiplexer. On the ONU i used a spliter, time delays at respective time slots on the transmitter side, optical receiver and then the BER analyzer.

Thanks once more Alan

Just to clarify are you using a "tree" distribution system, similar to EPON or GPON, or some other architecture with a single CW laser and multiple MZ modulators?

What is your sample rate?  Is it the same as your symbol rate (7 G Bd), or faster?

Is the poor eye at the ONU (subscriber) terminal, at the OLT (head end or central office) or both.  Alternatively, is your architecture more nearly a peer-to-peer scheme?

It is an Unidirectional TDM PON, with a single CW laser and multiple MZ, has varied symbol rate from 2.5GBps to 7GBPs and its improving though not that significant. The eye diagram is on the ONU side (its a one way PON).

And by the way, is the 'tree' distribution (one CW laser and MZ; at the OLT side and splitter with time delays on the ONUs), a TDM PON?

Thank you

I have not followed the evolution of passive networks in detail, but here is my understanding of current practice.

GPON and EPON systems are widely deployed and typically use a passive tree or star topology to broadcast the same multiplexed signal to multiple ONU or ONT terminals from an OLT or central office server.  The signals to each ONU are time division multiplexed, optionally using encryption for security.  The OLT uses a single transmitter.  The system is bidirectional, with ONU transmitters using TDM at a different wavelength, each transmitting in a different time slot.

GPON and EPON are widely regarded as TDM PONs.  https://en.wikipedia.org/wiki/Passive_optical_network

I do not have Optisystem, so can't view your configuration from the file you forwarded to me.

I am unclear how the signal is distributed, where the splitters are, and what the path losses and net dispersions are. 

Does each ONU have a MZ modulator, which modulates a common carrier, and a receiver?  Is each receiver required to synchronise to multiple symbol rates between 2.5 and 7.5 GBd?  Is it a linear topology, a ring, or some other configuration?

Alternatively, do you have a single ONU connected to multiple transmitters?

http://electronicdesign.com/what-s-difference-between/what-s-difference-between-epon-and-gpon-optical-fiber-networks

https://sites.google.com/site/amitsciscozone/home/gpon/gpon-fundamentals

I also tend to think GPON, EPON, and even GPON in practices as you described heavily relies on splitting at the ONU side, thus forming the tree or, point to Multi-point/ multi-point to Multi-point connection via the same fiber.

In my analysis i used a single carrier signal modulating multiple message signals at the OLT, delaying them, and then feeding them to a single channel for transmission and receiving the same signals at the ONUs,thus used the modulators and Time delays at each ONU, to recover/demultiplex the respective signal. Thus it seems a single receiver is connected to a single transmitter from the OLT side.

**Progress

Thank you very much. I have tried several conditions you spoke about, specifically, symbol rate, dispersion and distortion and the eye diagrams have really improved.

If you come across the software its a very good one in the study of PONs. Downloadable from Optiwaves.

Thanks Alan

Basically Q factor and BER depends on various factors such as launched power, fiber reach, and also on transmitted bit rate. As you tried simulations for different input power and fiber length. you can check by varying the Data rate. i am sure that your BER wiil be improved.