Harros said:
Hi, I have built the BP filter with Q = 1M/70k = 15 approximately, and ~50 ohm for input/output impedance. Any comments on this filter? However the rolloff at upper frequencies seems very slow, will it influence the performance of this filter? Besides, there is a great change of phase response around 1MHz region, will it cause the phase delta measurement to be inaccurate?
The way that you couple the 50 ohm source directly to the input and the 50 ohm load directly to the output is severely reducing the Q of the first tank and the third tank, so the overall filter Q is not good enough. This is evidenced by your measurement of Q which is only 10. This is quite poor and I expect at least 50 and perhaps as high as 200 in the simulation. It will be lower in practice, depending on the Q that you have assigned to the coil in the sim.
Imagine that there is no source attached to the input. In that case, the loaded Q of the first tank circuit might be as high as 50 to 100 or more. However, when you attach the 50 ohm source directly to the top of L1 like that, it is identical to putting a 50 ohm resistor in parallel with L1 which dramatically reduces the loaded Q.
The fact that the rolloff is slow, and levels out, at frequencies above the center frequency of the filter is due to the same problem. As the frequency goes above 1 MHz, the impedance of the top coupling capacitors C2 and C4 goes lower plus the effect of L1, L2 and L3 disappear and so there is nothing with an increasing impedance vs frequency in the way of the energy, so the output levels out at about -50 dB.
The next step to fix these problems is to change the connection point of R1 and of R2. For example, with R1, you should disconnect it from the top of L1 and create a "tap" on L1 about 1/10 from the bottom (that is, 10% of the total number of turns of wire) and then connect R1 to that tap. Similarly, disconnect R2 from C5/L3, create a tap on L3 at the same height as you did for L1, and reconnect R2 there.
This method of tapping the coil will improve the impedance match between the 50 ohm source/load and the tank circuit, it will dramatically increase the loaded Q of the first and third tank, and it will cause the ultimate rejection above resonance to keep going down without levelling out. You may adjust the height of the tap for best overall performance, but it should be kept fairly low on the coil.
The change in phase at the resonant frequency doesn't matter. Ultimately, you will be operating at only one frequency and your circuits will cause a total phase shift of some unknown amount. It will be necessary to calibrate out this phase shift. The important thing is that the phase shift isn't varying with time, and that won't happen in this circuit.