About ac analysis by LTSPICE of AD8606

Remember that SINE of value 1 in transient analysis is not the same as AC1 in ac small signal analysis, see  What does a dB reference in voltage in LTSpice 

Check your input signal amplitude in the .tran analysis by plotting V(acv) as well and look at the rms values of each signal.

I changed the amplitude value of V1 to 1.4142V and did the tran analysis result, but it was 33mVrms/1Vrms ≈ 30dB at frequency = 20MHz.
The AC analysis result was -3.87dB around 20MHz, so there is still a big difference.
Please let me know what causes this difference.
thank you.

Could you upload your LTspice schematic .asc file, please?

Upload the LTspice schematic .asc file.

thank you.

自動平衡ブリッジ_AD8606(周波数特性の確認)2.asc 

Thank you.

If you zoom into the transient waveform you will see the problem.  Rather than a nice sinewave, it is quite triangular - a sure indication that you are running into the slew rate limit of the amplifier.  Checking the datasheet, it is 5V/us (typ).  For a sinewave, the max slew rate is A*2*pi*f.  Rearranging this for signal amplitude, at 20 MHz, max A = 5000000/125663706 = 40mV.  If you drop the source below this level, you start to see attenuation closer to the expected result from the AC small signal analysis (~-4dB at 20MHz).  The ac analysis is 'small signal analysis' so will not take into account slew rate, only the bandwidth of the amplifier.  However, you will see 'large signal' effects like slew rate limitation in the transient simulation.  This is actually quite a good example of the difference between amplifier bandwidth and slew rate.

I hope that helps.

Run the following simulation and check the error log for the apparent attenuation due to slew rate for different values of input signal amplitudes:

SlewRateLimit.asc

I understood the content.
Thank you very much.

I understood the content.
Thank you very much.