bandpass filter

you really mean no phase shift? not linear phase shift

-Lasse

Yes. A plain old 2nd-order bandpass filter has no phase shift at the center of its passband.

Define "around". For that matter, define "no phase shift".

Or maybe state what you really want, with engineering units and tolerances and all that important stuff.

I want a filter that does not change the phase of a sine wave for small changes of frequency about the center frequency. In other words, the slope of the phase shift is zero at CF. Say, a 10 MHz bandpass filter with a reasonable Q, 25 or 50, and very little phase shift over

There would of course be some nominal phase shift and delay, but we'd want the phase change to be minimal when the frequency changes or because the temperature of the filter changes.

The answer is "a phase equalized bandpass filter", but I was wondering if there is some filter form like this that has a name.

Most good filters do. But they are called Minimium phase filters. And may have some phase shift. Watch the group Delay, thats a good indicator. Some Elliptic Filters are considered minimun phase. But they have passband ripple. A buttworth may be better.

Cheers

Fig 6 of the patent is about what I had in mind. And it's expired!

Thanks

On Tue, 6 Aug 2013 14:56:49 -0700 (PDT), Lasse Langwadt Christensen wrote:

OK, that patent inspired the thing below. It's just a top-coupled bandpass with two peaks, and each amplitude peak is also a delay peak, so there's a delay saddle between them. Not quite flat on phase, but close.

Version 4 SHEET 1 880 680 WIRE -64 112 -144 112 WIRE 64 112 16 112 WIRE 128 112 64 112 WIRE 192 112 128 112 WIRE 304 112 256 112 WIRE 352 112 304 112 WIRE 368 112 352 112 WIRE -144 160 -144 112 WIRE 64 160 64 112 WIRE 128 160 128 112 WIRE 304 160 304 112 WIRE 368 160 368 112 WIRE -144 272 -144 240 WIRE 64 272 64 224 WIRE 128 272 128 240 WIRE 304 272 304 224 WIRE 368 272 368 240 FLAG 128 272 0 FLAG 64 272 0 FLAG 304 272 0 FLAG 368 272 0 FLAG -144 272 0 FLAG 352 112 OUT SYMBOL ind 352 144 R0 WINDOW 0 63 33 Left 2 WINDOW 3 68 70 Left 2 SYMATTR InstName L1 SYMATTR Value 1 SYMBOL ind 112 144 R0 WINDOW 0 50 44 Left 2 WINDOW 3 55 78 Left 2 SYMATTR InstName L2 SYMATTR Value 1 SYMBOL cap 288 160 R0 WINDOW 0 -53 22 Left 2 WINDOW 3 -55 58 Left 2 SYMATTR InstName C1 SYMATTR Value 1.2 SYMBOL cap 48 160 R0 WINDOW 0 -50 17 Left 2 WINDOW 3 -53 52 Left 2 SYMATTR InstName C2 SYMATTR Value 0.9 SYMBOL cap 256 96 R90 WINDOW 0 -54 29 VBottom 2 WINDOW 3 -40 27 VTop 2 SYMATTR InstName C3 SYMATTR Value 0.05 SYMBOL voltage -144 144 R0 WINDOW 0 -118 22 Left 2 WINDOW 3 -134 62 Left 2 WINDOW 123 -138 102 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value SINE() SYMATTR Value2 AC 1 0 SYMBOL res 32 96 R90 WINDOW 0 -52 57 VBottom 2 WINDOW 3 -40 60 VTop 2 SYMATTR InstName R2 SYMATTR Value 100 TEXT 0 344 Left 2 !.ac oct 200 0.05 1 TEXT -8 -104 Left 2 ;BANDPASS FILTER WITH TEXT -24 -64 Left 2 ;ALMOST FLAT PHASE SHIFT TEXT 32 -24 Left 2 ;JL Aug 6, 2013

(Designed entirely by fiddling.)

A Q of 50? Really. You are going to manufacture this yourself? If you are going to pursue this project, I'd suggest synchronously tuned LC filters. Most likely trimmed. Good luck.

Have you considered transversal SAW filters?

According to mini-circuits then yes:

Band Pass Filters with Linear Phase Response: http://217.34.103.131/app/AN75-004.pdf

September 03, 2012, How linear phase filters can still cause phase distortion:

The Filter Wizard Blog:

September 17, 2012, Turn linear phase into truly linear phase:

/Glenn

Color me confused, I thought you always got a phase change with a change in gain. (Kramers-Kronig and all that.) So maybe a wide bandpass with a flat part in middle.

Or two BP's near each other, and a flat bit in the middle?

George H.

Maybe the OP has confused phase delay and group delay. Two signals at different frequencies are, by definition, such that a constant phase difference does not exist.

For constant delay, it is possible to construct linear-phase FIR filters or nearly-constant delay Bessel filters.

That's linear phase; I want zero phase shift over some frequency range.

He has not.

Two

I only plan to have one signal. I want a bandpass filter that bandlimits it but has constant phase shift over some modest range of frequencies about the center frequency.

Linear-phase is what I don't want. I want zero phase change vs frequency.

Hi John

What you want is presumably that the group-delay is constant.

PS: You can not get a zero group-delay. So you can not get a zero phase change for all pass-band frequencies.

If you get constant group-delay, the pass-band signals frequencies do not change phase with respect to the other frequencies, hence the signal shape is preserved. Is it that what you want?

/Glenn

I guess what I want is the narrowband equivalent of an all-pass network. That could be cascaded with an ordinary bandpass filter, but if it has the opposite phase/frequency slope it would cancel the (possibly linear phase) slope of the main filter.

What bandpass characteristic are you seeking... single complex-pole pair or multi-pole? How tight the phase flatness? ...Jim Thompson

Phase as such does not exist without some reference phase. If your signal changes frequency, its phase starts to creep away from the phase of the original signal, at a rate determined by the frequency change.

I still think that you're after a constant-delay filter. Remember that there is no realizable zero-delay bandpass filter.

What I want is phase shift to be constant.

Naturally not.

I only want zero phase change over a narrow frequency range near the center of the filter's passband. There's nothing impossible about that.

The only signal will be a single sine wave. I don't want its phase to wander around if its frequency changes a bit, or if the filter temperature changes a bit.

You need to define "bit" and how much phase change is acceptable. ...Jim Thompson

Minimum-phase doesn't mean zero phase slope, it means that the phase shift matches the minimum value set by the causality condition.

All ladder filters are minimum-phase, for instance, but lattice filters generally aren't. (That's why an LCR all-pass filter is always a lattice.)

Cheers

Phil Hobbs