Need help. Is the diff Amp BW double as single amp.

[rats]

I design a limiting amplifier, it is 5 identical cherryhooper amp chain.The simulation result of -3dB BW of one end is about 7GHz, but the rise-fall time is less than 30ps, it looks like the -3dB BW is larger than 10GHz, I can't explain this. Can one fellow help me? One guy told me in one book has a chapter explain this, but I couldn't find this book, I urgently need yours help. my email. tanghav@yahoo.com

 

[krashkealoha]

Since your information posted is limited, I'm going to be assuming a lot. I'm assuming that you feel the bandwidth is larger than 10G is because the trise < 30 ps and it follows Bandwidth = 0.35/(rise time).

Your 7GHz bandwidth is the result of a small-signal simulation. I don't know what the total gain of your cascaded stages. If your input voltage for the transient simulation is larger than the minimum voltage required to have the amplifier limit, then you would be able to drive it past the small-signal 3dB point.

On the other hand, make sure your output voltage you are measuring is not attenuating due to bandwidth limitations. This will give a lower trise than expected because of the lower swing.

Feel free to post more info on your circuit to help debug your problem.

 

[rats]

The Gain is 49dB, the circuit is 5 stages LA, every stage is identical. the rise time is 26ps.

 

[krashkealoha]

With 49dB of gain, your input sensitivity must be less than 10 mVpp-differential input voltage. In another words, your amplifier will be operating in small-signal with an input less than ~10 mVpp-differential or you will not be limiting. If you are applying an input voltage greater than 10 mVpp-differential, then your large-signal bandwidth will look higher than your small signal-bandwidth.

Another thing you can do is to plot gain (or S21) vs. Frequency after each amplifier stage. From these plots, you can determine which stage is limiting your bandwidth of the entire amplifier.

If you are targeting for 10 Gb/s applications, the 7 GHz bandwidth and the 26 ps edges should be good enough, but check your eye diagram simulations.