No phase shift filter

All real filters only have information from the past to work with. As a result, they all make phase shift. Do you want a linear phase shift ie: a constant delay?

People often think they want a linear phase filter when in fact that is nothing like what they want. A sharp cut off or a sharp notch and linear phase is requires a lot of electronics.

--
--
kensmith@rahul.net   forging knowledge

DSP

He'd need a DSP and a time machine to do what he's asking for. Since what he probably needs is a linear phase filter, a digital implementation of a finite impulse response filter might do his job, but might also be more expensive than necessary.

We need a lot more detail about the application, which we probably won't get - someone who thinks that a Sallen-Key implementation of a particuar sort of filter (say Butterworth or Bessel) has a different phase shift from say a passive implementation of the same sort of filter is going to need quite a bit of education. But I can remember back to when I was just as ignorant ...

--
Bill Sloman, Nijmegen (but in Sydney at the moment)

For linear filters, no poles in the right half plane, there is a direct one to one relationship between phase shift and attenuation. Hence the rule is for zero phase shft you are stuck with zero attenuation.

Howard

Henry W> Can anyone provide information on active filter designs that

From what I understand, Henry wants no change in phase shift in the frequency range of interest when input frequency gets closer to the cutoff point. I may have made the wrong assumption though.

Johnny

I'm not absolutely sure (my filter bible is back in Nijmegen), but I think that you can get close to that by adding a purely phase-adjusting all-pass filter after your attenuating filter, but completely eliminating the need for phase compensation would seem to be physically impossble (at least if you haven't got a time machine).

--
Bill Sloman, Nijmegen (but in Sydney at tne moment).

Hey, da Fonz is cool!

For any given frequency response, there is a corresponding minimum phase shift vs frequency. The minimum phase is achieved by ordinary RLC ladder filters, regardless of their detailed design. (Terman proves this somewhere.)

The actual phase shift is sometimes very important, especially when the filter forms part of a feedback loop, but usually what we care about is the whoop-de-doos in the phase vs frequency characteristic, because they cause bad behaviour of various kinds. Specifically, the desired phase vs frequency characteristic is a constant delay. If you plot phase vs frequency, this gives you a straight line passing through the origin (zero phase at zero hertz--you can't phase shift DC).

Being able to tolerate delay is a good thing, because building filters with less than the minimum delay requires a time machine.

Cheers,

Phil Hobbs

(who quit watching TV about the time 'Happy Days' came on the air)

Nope, an Allpass will add additional phase shift without affecting the frequency response. This will delay low frequencies even more(exessive phase) and is just the opposite of what he wanted. Anyway, phase shift is difficult to hear, exept phase jumps in the range up to 1.5kHz. Notch filters might do this if they have infinite attenuation. The phase shift becomes very important when adding several sound sources with crossover filters for a speaker box. They tend to create a lot of excessive phase, often the bass sounds are 'limping' behind. Any decent analog filter will have a minimum phase characteristic, it doesn't depend on the topology. The Q-factor of these filters can be deliberately chosen to fit your needs. You should be familiar with the transfer function in Laplace form. This helps putting the appropriate coefficients into simulation programs. You can then play with the values, look at the impulse response, overshoot etc. and get a feeling of how filters work yourself.

--
ciao Ban
Apricale, Italy

I've always wondered what became of Robert Lerner's linear-phase filters. He came up with the idea of using non-minimum phase techniques to build filters with linear phase _and_ sharp cutoffs (the usual linear-phase filter, Bessell, has very broad cutoffs). But I've only ever found two articles of his, and an occasional reference from others who tried the technique.

There was also something called the Papagoulis (?) filter, used in high-end audio FM receivers for a while. Are these still around? I've never had to deal with filters, myself, so have never had more than an interested layman's acquaintance with filter design.

You'd think that something like that would have taken over the world in short order, so there were obviously some disadvantages. I've never heard of anyone spelling out the disadvantages, though. Any comments from filter experts?

John Perry

I've never heard of Robert Lerner. Do you have any referring URL's?

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Filters are the implementation of differential equations. They cannot do whatever is wished. The common behaviour is that in the passband the phaseshift is marginal and increase towards the stopband. The phase shift is n times

90 degrees per order or something similar. Since the signal practically vanishes in the stopband anyway the phaseshift there is not really interesting. No ?

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net

For the low pass, do the obvious, i.e. Bessel or Gaussian. If you need a sharper cut off, then you need to add all pass networks. I have a few LCR filters that are Bessel filters with zeros inserted, making sloppy but decent linear phase low pass filters. I haven't done the LCR to leapfrog conversion on them.

The notch is a bigger problem.

Henry, usually the phase shift is inoffensive in audio application. Maybe you could elaborate what are you doing ?

best,

There are thousands of google hits on "linear phase lerner", but almost all have to do with biology. There were also a few hits on others who used Lerner's methods, but all I could find require IEEE or APS membership, or payment to DTIC for the full papers.

John Perry

Is it worth the expense to subscribe to ieeexplore ??

My membership only includes SSCC stuff.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Not to me. To you, I couldn't say.

All I remember is reading Lerner's first paper in someone else's IEEE magazine (Proceedings, maybe?), in which he discusses the usual technique of designing minimum-phase networks (as Phil Hobbs mentioned in the post I was commenting on), and that with minimum-phase you could have either sharp cutoffs and distorted phase, or broad cutoffs and linear phase.

He then describes a technique of allowing a longer phase delay, which allows linear phase _and_ sharp cutoffs. My impression is that the more delay you allow, the sharper cutoff you can get without phase distortion

-- but, since I've never had to dig into linear theory to any extent, I didn't really understand what he was saying, and couldn't pursue it myself.

It just seems to me that so many systems would benefit from linear phase

-- anything that needs waveform fidelity, for instance -- and sharp cutoffs -- telecom channel filters, for instance -- that these things would be in common use. But I never see them mentioned.

John Perry

The following is from the Wikipeda article on MinimumPhase.

A minimum-phase system, whether discrete-time or continuous-time, has an additional useful property that the natural logarithm of the magnitude of the frequency response (the "gain" measured in nepers which is proportional to dB) is related to the phase angle of the frequency response (measured in radians) by the Hilbert transform.

Howard

hrh1818 wrote:

So, Fonzie grew up to be an EE?

-a