About the -3dB bandwidth of TIA

Jack schrieb:

What is the optical modulation format? If it is NRZ (non-return to zero) than the most energy in the power density spectrum of your signal is contained up to 0.6 or 0.7 times the bit symbol rate. And -3dB BW does not mean that any frequencies above 700 MHz are sharply cut-off, either. So 700 MHz is just OK for 1.25Gbps NRZ without significant penalty on eye-opening.

More bandwidth would reduce the rounding of the 0-1 edges somewhat but would mean more noise power which is proportional to the TIA bandwidth. So 700 MHz maybe indeed a good choice. Be sure that your photo diode has only the specified 0.5 pF capacitance or you will see a penalty for reduced bandwidth and reduced eye-opening.

If your modulation format is RZ than 700 MHz would be to small for 1.25 Gbps.

Regards Rainer

Hi Rainer,

Thanks for your kind reply!

The TIA is designed for GPON/GEPON ONU and 1.25G SFP applications,would you please explain the concern of TIA bandwidth for the applications,respectively?

Thanks and Best Regards,

Jack

R.E. wrote:

Hi, Jack:

I'm not the expert for GPON, but as far as I know its a optical NRZ format. But 700 MHz TIA bandwidth really seems to be too small. A common choise is TIA-BW = 0.75 * bitrate. A smaller bandwidth will reduce signal power and edge-steepness of the pulses, a much higher bandwidth will generate more noise. Both will increase bit error rate. See also the application note from Maxim:

For example, if you look on the Maxim chipsets, the MAX 3266 designed for 1.25 Gbps has 920 MHz TIA BW.

I'm not related with Maxim in any way. But I know their chip sets and as an academic teacher I'm quite happy with the detailed specifications and application notes they provide. Good for use when training students.

Did you ask your manufacturer of the 700 MHz TIA if it is really good for 1.25 Gpbs? What did they say?

Regards Rainer

Jack schrieb:

Hi, Jack:

I'm not the expert for GPON, but as far as I know its a optical NRZ format. But 700 MHz TIA bandwidth really seems to be too small. A common choise is TIA-BW = 0.75 * bitrate. A smaller bandwidth will reduce signal power and edge-steepness of the pulses, a much higher bandwidth will generate more noise. Both will increase bit error rate. See also the application note from Maxim:

For example, if you look on the Maxim chipsets, the MAX 3266 designed for 1.25 Gbps has 920 MHz TIA BW.

I'm not related with Maxim in any way. But I know their chip sets and as an academic teacher I'm quite happy with the detailed specifications and application notes they provide. Good for use when training students.

Did you ask your manufacturer of the 700 MHz TIA if it is really good for 1.25 Gpbs? What did they say?

Regards Rainer

Jack schrieb:

R.E. wrote on 22-Jun-06 06:30 :

That's obvious.

[...]

If "Jack" is still reading this...

"0.75 * Baud" is related to transmitted signal power of an NRZ data stream. If you were to integrate the power spectrum of NRZ PRBS you would find that 94% of total power is contained in a bandwidth that is roughly 75% of the line rate.

Go ogle this:

One reason to worry about that effect is SNR. With wider bandwidth you can preserve more of a transmitted pulse shape, but you also get more noise. So the goal is to restrict bandwidth enough to reduce noise below a tolerable level -- thus, preserve sensitivity at a TIA -- and avoid unacceptable pulse distortion. Hence, choose a cutoff frequency of ~0.75 x Baud.

If more than one element in a serially-connected, mostly-linear link has bandwidth of 0.75 x Baud, then overall bandwidth will be = 0.75 Baud. My favorite choice for those other elements is at least

4x higher bandwidth.