Can passive phase shifters be implemented without a variable delay element?

are most

lay

te

s a

ple

If you need a variable delay, you need a variable element. If you only want a fixed phase shift you don't.

Radar systems use the variable delay to scan the beam, so the need for vari ability comes from the requirement that you want them to scan.

I can conceive a of "radar" system with a fixed wide-angle illumination an d a lot of fixed detectors, each of which was most sensitive to reflections coming in an a different lines of sight, and that wouldn't need variable d elays, but you'd need a lot of fixed different delays.

are most

UM... why is that 'wideband'? A fixed delay is a different phase shift at each frequency.

ple

one can phase-shift based on the identity

sin(w*t + phase) = sin(w*t) cos (phase) + cos(w*t) sin(phase)

using mixer techniques (mixer :== analog multiplier) with a suitable way to generate (like, a lookup table) the phase-dependent parts.

Direct synthesis is another example.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

are most

lay

te

s a

ple

Phase shift IS delay. At least until you make the classic student mistake of basing your output o n future input. :)

NT

Not all phase shifts involve delay. An RC integrator has a pi/2 phase shift but no delay at all.

Group delay doesn't always equal true delay either. The confused terms cause all sorts of confused minds.

Cheers

Phil Hobbs

Sure. You make a Hilbert-transform filter (a linear phase bandpass filter whose phase extrapolates to pi/2 at the origin) and use that with an I/Q modulator.

Cheers

Phil Hobbs

On Sunday, 14 January 2018 13:27:45 UTC, snipped-for-privacy@gmail.com wrote: NT:

Last time I checked an integrator integrated the signal over time, namely the time between now and some point in the past. Ie its signal out is the sum of signals in over time elapsed.

NT

Full marks. ;)

You can undo an integrator with a differentiator, but to undo a delay you need a time machine.

Cheers

Phil Hobbs

On Sunday, 14 January 2018 16:07:33 UTC, snipped-for-privacy@gmail.com wrote: NT:

Heh. What the differentiator does is responds only to the immediate data in the integrated signal, ignoring the stored data, hence it recovers the original waveform.

On a simpler note, you can get a 90 degree phase lead filter, as long as you don't mind it really being 270 degree lag.

NT

On Jan 13, 2018, snipped-for-privacy@ieee.org wrote (in article):

All you say above is correct, but I?m asking a physics question, not a radar design question:

Is there any way to build a passive variable phase shifter (a physical device) that does not ultimately rest on some kind of variable delay device?

Joe Gwinn

On Jan 14, 2018, whit3rd wrote (in article):

Because many radar waveforms are wideband, one gigahertz being common in X-band radars. Typically, linear-phase response is required to avoid undue waveform distortion, which causes range blurring.

This approach is widely used, but it is active, not a passive physical device.

Joe Gwinn

On Jan 14, 2018, Tim Williams wrote (in article ):

This, while widely used, is an active approach.

Joe Gwinn

On Jan 14, 2018, snipped-for-privacy@gmail.com wrote (in article):

Doesn?t the RC integrator cause a delay? - the step response is a delayed ramp.

Hmm. In radar, we generally equate group delay with true delay of wideband waveforms. Can you think of any examples where group delay is not true delay?

Joe Gwinn

On Jan 14, 2018, snipped-for-privacy@gmail.com wrote (in article):

This one-liner is a keeper.

Joe Gwinn

Hmm I'm not sure this applies in this case. (I don't know what a Hilbert transform BP is.) But lots of times there's a difference in the transient and long term response of the system. I see this mostly in optics or acoustic 'wavelength' sized coatings.

George H.

On Jan 14, 2018, snipped-for-privacy@gmail.com wrote (in article):

Hmm. One can approximate Hilbert filters with analog filters (with some amplifiers), but as with the RC integrator, there will necessarily be a delay, and I?d hazard that the delay will vary as the implemented phase shift is varied.

I actually implemented a Hilbert Filter in the signal processing path for an IRIG-B AM signal demodulator, to demodulate the 1KHz AM signal into the baseband digital modulation. It did work, but was *very* noisy in practice with low-SNR input, so I used some other approach that implicitly integrated (versus differentiating).

Joe Gwinn

Yes, it is possible, but usually not feasible.

Making two copies of the input signal, with a phase difference and a linear combination of the signals, you can adjust the phase of the combined output with the relative amplitudes of the components.

The phase adjustment is most effctive when the signals are in quadrature (phase difference of pi / 2), but other combinations will also do.

For a passive network, some phase shifts will demand heavy attenuation of the output (but you did not ask about it).

phase

Nope and nope.

First, ideal integrators have exactly zero delay. Otherwise you couldn't un do the effect with an ideal differentiator, which you can.

Second, the filter is fixed-tuned--you vary the phase with an I/Q modulator , If you match the delay in the two paths, the delay doesn't vary with the phase setting.

Cheers

Phil Hobbs

We had a long thread some years ago in this very boutique all about that, started by JL iirc. I was strenuously defending the position you advance, and discovered that it was wrong.

Cheers

Phil Hobbs