Bandpass filters, Pi vs T topology?

[TERRA Operative]

I'm looking to whip up a 10MHz bandpass filter to smooth a square wave to more of a sine wave and I can't seem to find much of a straight answer online about the pro's and con's of a Pi-filter vs a T-filter for this application.
Can anyone fill me in on why I'd choose one rather than the other?

Also, besides physical size and parts count, what are the other drawbacks of higher order filters (assume a butterworth filter).

Thanks!

[PKTKS]

Well pretty much that depends on other factors like you said..

Size cost part number..

But TODAY? I would not even consider a discrete filter against a clever OPAMP one.

Just not cost effective.

I would put all time effort on a opamp one.
**IFFF**  the net result is more important than sizexcost

Paul

[TERRA Operative]

I'm looking to build a passive filter to just be used inline, BNC coax in, BNC coax out. Don't want to have to mess with power supplies and stuff for this one.

[T3sl4co1l]

From what source?

If it's directly on a voltage source like a CMOS output pin, it's not going to appreciate the low impedance (~short circuit) at high frequencies of the pi filter.

And since inductors are generally more bothersome than capacitors, you might as well go for an extra pole (4th order) LCLC.

If the load is say 50 ohms all the time, all you need is a filter designed for source and load respectively (the source will be fairly resistive if CMOS logic, but if it's substantially lower than 50 ohms (say from a gate drive IC?), consider either ballasting it up with a series resistor, or using a one-port-shorted type filter -- such types are found in more complete tables [1][2].  Or if it's current-sourced, a one-port-open (shunt C input) type will be needed.

[1] Electronic Filter Design Handbook, 4th Ed., Williams and Taylor
[2] Handbook of Filter Synthesis, Zverev

If the load is not a constant resistance in general (e.g. sometimes poorly terminated cable?), consider using an internal filter, then a buffer amp.  This would be the way to go for a signal generator, for example.

Tim

[Kleinstein]

With not so good termination a attenuator can improve thing, but off cause it has it's downsides.

[ogden]

Check https://rf-tools.com/lc-filter/. 5th order lowpass Chebyshev, series first, E12 components, 1dB passband ripple is good place to start playing with

[KE5FX]

I'm looking to whip up a 10MHz bandpass filter to smooth a square wave to more of a sine wave and I can't seem to find much of a straight answer online about the pro's and con's of a Pi-filter vs a T-filter for this application.
Can anyone fill me in on why I'd choose one rather than the other?

Also, besides physical size and parts count, what are the other drawbacks of higher order filters (assume a butterworth filter).

Thanks!

Be sure to check out Charles Wenzel's notes on this question if you haven't already.  Some of the points he brings up aren't immediately obvious.

[TERRA Operative]

I'll be sticking it on the output of a GPSDO (Leo Bodnar GPS thing) to remove harmonics and somewhat smooth the square wave to a sine wave because reasons. It'll be fed into a 50ohm impedance load.

Thanks for the references, I'll check them out.

[David Hess]

In a timing application which is sensitive to phase, it is better to notch out the low harmonics and maybe use low and high pass filters than to use a bandpass filter because a bandpass filter will create much more fixed and variable phase shift at the fundamental frequency.

[viking]

A lot later: If you want to smooth out a square wave, what you want is a lowpass filter. A Π filter will reject high frequency components better, but give more phase shift but only require 1 inductor.
Chebyshev will give faster rolloff but more phase shift than Butterworth
Use for example https://rf-tools.com/lc-filter/ to build a 3rd order Chebyshev lopass Π filter:
Inline inductor is 910nH, and 2 shunt capacitors are 330pF.
-3dB at 14MHz, and -12dB per 2x rolloff after that. You can take a Fourier transform of the filter response, and see your rounded square waves before you build. Use for example Excel. The square wave components are of course (2n+1)*10MHz= 10MHz, 30MHz, 50MHz, so you dont need to scale more than the first 3 to clearly see the result. Set the rest to 0 since 70MHz is already  <-46dB
I wrote a fourier expansion of a square wave in desmos: https://www.desmos.com/calculator/tzkpxnaq5u  .You can turn on or off different harmonics, and multiply with the filter parameters if you will.

This is for a 50Ω design, but is easy enough to adjust.

[jonpaul]

Bonjour à tous

We use Leo Bodnar GPS at 10.0000 MHz without any filters, the semi square wave is fine for the RÉF freq in on HP, TEK, ÉPI.

These are very forgiving in the levels and waveform.

We made an 8 way splitter out of Mini Circuits 2..4...6 splitter transformers in series for house clock.

Mini Circuits also have LP,BP, HP passive filters, inline with BNC, SMA and other connectors.

Bon courage

Jon