1
\$\begingroup\$

I am confused about 2 different online active filter calculators giving different GBW requirements for the same frequency range. One calculator is giving GBW requirements in the kHz range while the other calculator is saying I need a very high GBW in the MHz range - up to 40 MHz.

I am trying to calculate an 8 pole bandpass filter in the range of 5400 Hz - 7500 Hz and this is what each says about GBW requirement for each section:

first website is https://tools.analog.com/en/filterwizard/:

GBW for each section with the above calculator: 768 kHz, 654 kHz, 741 kHz, 547 kHz

2nd website is https://webench.ti.com/filter-design-tool/filter-type:

GBW for each section with the above calculator: 3.4 MHz, 4 MHz, 27.4 MHz, 39.6 MHz Analog.com calculator values

Ti.com calculator values

\$\endgroup\$
1
  • \$\begingroup\$ Comments are not for extended discussion; this conversation has been moved to chat. \$\endgroup\$
    – Voltage Spike
    Commented Dec 21, 2020 at 6:34

2 Answers 2

2
\$\begingroup\$

Without knowing details about your filters (topology, gain, approximation), the "required" GBW depends primarily on the required accuracy of the filter function. Of course, an "ideal" opamp would be best - however, it does not exist.

Therefore, we require an opamp "as good as possible" under technical, economic and application-specific considerations. In this context, it makes no sense to ask for an accuracy that is better that the unavoidable tolerances of the passive parts (R,C), which may be different for the variuos applications. More than that, some filter topologies are more sensitive to GBW limitations (multi-feedback) and some other are less sensitive (Sallen-Key, GIC-structures).

Therefore, it is common practice to consider a certain "safety factor" as far as the recommended (or required) GBW is concerned. And it is no surprise that each filter design software will contain its own "safety factor".

Comment: Instead of a cascade of 4 second-order stages, a direct realization might be better - starting from a fully designed passive RLC bandpass of eighth order. There are different strategies to convert such a passive filter into an active realization.

\$\endgroup\$
3
  • \$\begingroup\$ "a direct realization might be better". This is an interesting suggestion. In what way, managing GBW? and why "might"? \$\endgroup\$
    – P2000
    Commented Dec 21, 2020 at 16:18
  • 1
    \$\begingroup\$ In electronics, everything is a trade-off between conflicting requirements. Therefore, to answer your question it would be necessary to define the term "better". Oner thing is sure: Such "direct realizations" have the best properties concerning "passive sensitivities to parts tolerances". However, these are not the only considerations which have to be taken into acount. And because I am not informed about specific application-oriented requirements I have chosen the word "might". \$\endgroup\$
    – LvW
    Commented Dec 21, 2020 at 16:47
  • \$\begingroup\$ These are very good points and I am glad you bring this up here. Meanwhile, I did check the tool and indeed by considering the power/noise factor the GBW is greatly altered, for the same topology. \$\endgroup\$
    – P2000
    Commented Dec 21, 2020 at 19:21
0
\$\begingroup\$

The recommended op-amp AD4096 is selected based on the low power requirement.

As you pointed out in the comments, after defining the filter specs, the component picker reveals the GBW:

  1. "Pick for me"
  2. Optimization "Low Power"
  3. "I want to choose" shows the ADA4096 for a GBW of 786KHz, and they calculate the required GBW at 178KHz, which seems quite reasonable

However, if you choose Optimization "Low Noise", then the "I want to choose" option takes you to the LT6233, with GBW 60MHz.

I didn't check the TI settings, but it's possible that their tool defaults to low noise rather than low power.

Set optimization to low noise:

enter image description here

I want to choose:

enter image description here

See the GBW:

enter image description here

This was an excellent question. User beware.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.