Group delay is d(phi)/d(omega), and is always positive, unfortunately.
One approach would be to put a bit of delay in the reference arm, if possible.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
On 07/08/13 18.48, John Larkin wrote: ...
Hi John
If f and f*1.001 are sent through a transmission line that delays by
0.000,000,1 second (30 meter or less; depends on the dielectrics), then try to calculate the phase change of the two frequencies individually, because of the delay./Glenn
The thing I care about is the phase shift introduced by the filter. The reference phase for the filter output is the filter's input phase.
No, that would be linear-phase. I want zero phase change over some modest frequency range.
I think I can remember that.
(What is with this "Remember..." thing that people keep doing lately?)
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
Maybe your newsgroup viewer is dropping lines. I said "filter with a reasonable Q, 25 or 50"
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
Posted that already.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
Hey, count your blessings--it's much more polite than the way most "friendly corrections" are delivered round here. ;)
The missing piece is that group delay is d(phi)/d(omega), so since it is always positive, you can't really have zero phase slope with frequency. It shouldn't be that hard to fake, though, assuming that you also control the phase reference.
Cheers
Phil
(*) In the EE sign convention, it's -d(phi)/d(omega)
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
I'd like one too. Unfortunately it's called a "time machine." ;)
One day I beat my brains out trying to do that with an RL equivalent of the usual op amp RC all-pass, till I discovered my algebra mistake(*) and realized the more fundamental nature of the problem. Sure would have improved my PLL if I could have done it!
Cheers
Phil Hobbs
(*) The RL one has the opposite sign of phase shift, but the same sign of phase slope with frequency.
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
What's wrong with a phase equalizer? I don't need zero delay, just minimal phase shift as the frequency changes. Lasse posted a patent, and I posted a Spice file, for filters that have substantially flat phase shift near the center of the passband.
Is it fundamentally impossible to make the bandpass equivalent of a baseband allpass? I don't even need gain flatness, just a phase/frequency curve in the opposite direction from a simple LC bandpass filter.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
On 07/08/13 20.22, John Larkin wrote: ...
Have a look on these slides and article:
Lecture 12 Relationship between Magnitude and Phase:
Determination of the phase of a transfer function: theory versus practice:
/Glenn
Yes, it's impossible. The group delay is the derivative of phase with respect to frequency, so a box whose slope goes the other way is precisely a time machine.
You can make filters whose group delay is quite small away from the passband edges, but you can't make it absolutely flat. From my quick look at the patent, it was a HP/LP cascade that worked more or less that way.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Missed the point about the baseband allpass. You can't make one of those with the opposite slope either, for the same reason. That's what I spent a day trying to do, some years back.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
I could use a trombone delay line and a stepper motor to change the delay (and phase) of a bpf output as a function of frequency, any amount in any direction, at least until the line crunches at zero length. That's not a time machine, it's just a variable delay.
I could also make two bandpass filters that peak a bit apart, and sum their outputs. Then add delay lines to both, and tweak the delay lines so that either filter has any phase shift that I like. As the frequency changes near the center and the output sum shifts from one filter peak to the other, I can change the phase any way I like.
-- John Larkin Highland Technology, Inc jlarkin at highlandtechnology dot com http://www.highlandtechnology.com Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom laser drivers and controllers Photonics and fiberoptic TTL data links VME thermocouple, LVDT, synchro acquisition and simulation
As I said in an earlier post, you can probably fake it by playing with the reference channel. You can make phase offsets (in one direction only) with delay lines, and you can make the phase at a given frequency be anything you like, agreed.
But what you can't do is build any causal network whatsoever whose d(phi)/d(omega) has the opposite sign from that of a piece of coax.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
On 07/08/13 20.22, John Larkin wrote: ...
Some years ago I found this odd phase changer with constant magnitude:
/Glenn
Right, that's the usual all-pass filter that many of us know and love.
Figure 1 of that article,
H(w) = (jwRC-1)/(jwRC+1) .
If you try replacing the capacitor with an inductor, as I once did, the gain is now +1 at low frequency and -1 at high frequency, but the transfer function is exactly the same except with an overall minus sign, which doesn't affect the phase slope at all.
Cheers
Phil "unsuccessful would-be time machine builder" Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Did you also try to NIC a capacitor or inductor - and place in the all-pass filter ;-)
The time machine might not be so far out ;-)
/Glenn
You can make acausal networks. They're called "oscillators". ;)
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
--- Are you concerned at all about delay, or do you just want a region in the passband with low phase shift VS frequency?
I don't think you can get _no_ phase shift, so what can you tolerate?
-- JF
So a notch at the same frequency will give the other sign of phase shift. (Grin)
I think what you want is a BP filter with a bit in the middle that's *really* flat. I can imagine it 'in theory', but 'in practice' may be harder.
What 'cat' are you trying to skin? Maybe there is some other way? Can you digitize the signal?
George H.
-- Well, you do seem to be struggling... Why not tell us more about the center frequency, the deviation, and the phase change you can tolerate?
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.