90 degree phase shifter

[crutschow]

The currents in the L branch and the C branch of a parallel circuit are 180° opposite each other no matter what the voltage frequency is.

What do you mean, they are 180° opposite each other? Do you mean the obvious -- that when the current is coming out of the inductor, it is going into the capacitor terminal connected to the same node?

How is that relevant to achieving a 90° phase shift? :?:

 

[Ratch]

rohitkhanna,

Here's a circuit outline to illustrate

Interesting circuit, but how does this new circuit with 2 opamps, 2 resistors, 1 coil, and 3 caps relate to a L & C in parallel?

Ratch

 

[rohitkhanna]

The extras are there to be able to realise a practical circuit... buffers/ isolation/ load etc ... if you wish.

The key L & C elements are in the center -- the 100nF & the 25uH in parallel.

Do try and pay attention Ratch. You're much smarter than this.

cheers!

 

[BradtheRad]

A 90deg shift is easily possible by using an L & C in parallel, and which are tuned to the exact carrier frequency.

I can confirm this concept. Simulations show resonating action around the tank loop. Current through the coil and capacitor are 90 deg. from the incoming signal, when it is anywhere near the center frequency.

To tap for a signal, it should be done across a resistor inserted in the loop, because you want to read current through the resistor. It will not work to tap for volt level across either the capacitor or coil, because their volt reading is close to being in sync with the incoming signal.

 

[Ratch]

crutschow,

What do you mean, they are 180° opposite each other? Do you mean the obvious -- that when the current is coming out of the inductor, it is going into the capacitor terminal connected to the same node?

I mean that according to the impedance triangle for currents in a parallel circuit, the currents in the capacitor and inductor are 180° from each other.

How is that relevant to achieving a 90° phase shift?

The relevancy is that it does not achieve a 90° phase shift. That is what I am trying to point out. Now if one adds a resistor in parallel to the L & C, then a 90° phase shift can be achieved with respect to the R and L or R and C.

Ratch

 

[rohitkhanna]

I can confirm this concept. Simulations show resonating action around the tank loop. Current through the coil and capacitor are 90 deg. from the incoming signal, when it is anywhere near the center frequency.

To tap for a signal, it should be done across a resistor inserted in the loop, because you want to read current through the resistor. It will not work to tap for volt level across either the capacitor or coil, because their volt reading is close to being in sync with the incoming signal.

Oh good !! Then can you simulate the circuit I've attached above, and see whether we really DO need the resistor you've mentioned ? I believe the voltage tap is fine...

 

[Ratch]

rohitkhanna,

The key L & C elements are in the center -- the 100nF & the 25uH in parallel.

Now, there are more than just L & C elements. There is a resistor also, so it is not the same circuit you first described.

Do try and pay attention Ratch. You're much smarter than this.

I do and I have. That is why I called it to your attention.

Ratch

 

[rohitkhanna]

Yes yes you are right Ratch.... it's a TOTALLY different circuit. How remiss of me to have (gasp) changed my circuit from an LC-R to an R-LC one !!!
And to have also (horror of horrors) added buffer stages & a load resistor tooooo.....
Ouch ! Thank you THANK you for pointing it out & bringing to my attention. I am now a better man.

 

[Ratch]

rohitkhanna,

Thank you THANK you for pointing it out & bringing to my attention. I am now a better man. .

You are welcome.

Ratch

 

[BradtheRad]

I admit I was skeptical... however I realized the concept might work, having experimented on simulators for countless hours with capacitors and coils combined in roles of LC tank loops, bandpass, bandstop, oscillators, phase shifting, waveform modification, etc.

Below is the layout that I found works. All resistors are necessary in order to give the tank loop some degree of isolation. The 1 ohm resistor in the loop is needed so that current through it will create a voltage across it, and thus obtain an output.

The volt levels at top and bottom of tank loop show a smaller degree of lead or lag, hence they cannot be used for outputs. Moreover, current through the bottommost (10 ohm) resistor is in sync with the incoming signal when it is at resonant frequency.

--------------------------------------------------------------------------

By the way I have a Youtube video showing a tank circuit in animated action:

www.youtube.com/watch?v=cX1-CpzGUHc

 

[BradtheRad]

And the corresponding phase plot ..

I set this up in a simulation, and I applied a frequency sweep.

It works, and it yields 90 deg phase shift... although you get 90 deg. only when the input signal is exactly at the resonant frequency. At other than that frequency, the shift quickly diverges to either side of 90 deg. (as seen in the phase plot).

I also observe that the resonating oscillations are easily triggered even if the incoming signal is far off the resonant frequency. The loop has no damping (which is associated with a high Q).

Using my posted schematic, and taking the volt reading across a resistor in the loop, I find that the phase shift is 90 deg (or not far off) for a larger portion of the frequency sweep. Furthermore inserting a resistor has the effect of damping oscillations (even though it reduces Q). However this could be preferable if we do not want the resonant loop to be oscillating at odd times.

It depends on what behavior we want when the input signal is at a different frequency than the resonant frequency.

 

[crutschow]

Seems that the All-pass circuit I posted which uses 3 resistors, 1 capacitor, and 1 op amp is a lot simpler. ;-)

 

[rohitkhanna]

Seems that the All-pass circuit I posted which uses 3 resistors, 1 capacitor, and 1 op amp is a lot simpler. ;-)

Yes it is :smile: , and in fact I'm going to use it in one of my projects.

The advantage that the LC has is that you can make it work for much higher frequencies, where the R & C values of your circuit would be a challenge.

from what I see, your RC should = 1/ w for 90 deg shift point.

---------- Post added at 09:53 ---------- Previous post was at 09:49 ----------

I set this up in a simulation, and I applied a frequency sweep.

It works, and it yields 90 deg phase shift... .

thanks Brad
cheers!

 

[BradtheRad]

Seems that the All-pass circuit I posted which uses 3 resistors, 1 capacitor, and 1 op amp is a lot simpler. ;-)

Yes. I have been running a simulation of this type as well. The diagram is a screenshot of my layout using Falstad's simulator. For convenience I am using a center frequency under 1000 Hz.

It provides a 90 degree phase advance. Looks as though it does so for a range of frequencies around 50% to 150% of the center frequency.

Up until today I am used to my simulations showing that the RC-to-ground yields reduced amounts of phase shift. So I suppose the op amp is key to the success in this case. Because usually we think of the series capacitor providing greater phase shift, and never more than 90 degrees.

(The all-pass filter can provide more than 90 degrees advance.)

Now I recall, an earlier run of the all-pass had the capacitor and resistor positions reversed. This yields a phase delay instead of a phase advance. I'll try it again and see if it provides a 90 degree delay.

 

[crutschow]

Yes. I have been running a simulation of this type as well. The diagram is a screenshot of my layout using Falstad's simulator. For convenience I am using a center frequency under 1000 Hz.

It provides a 90 degree phase advance. Looks as though it does so for a range of frequencies around 50% to 150% of the center frequency.

Up until today I am used to my simulations showing that the RC-to-ground yields reduced amounts of phase shift. So I suppose the op amp is key to the success in this case. Because usually we think of the series capacitor providing greater phase shift, and never more than 90 degrees.

(The all-pass filter can provide more than 90 degrees advance.)
......................

The phase-shift is 90 degrees at the -3dB frequency of the RC network. With a 1000kHz -3dB frequency (159Ω, not 300Ω, and 1µF) I simulated 127° shift @ 500Hz, 90° @ 1kHz, and 67° @ 1.5kHz, Don't know why you said it provides 90° over that range. :???:

Yes, that All-pass configuration with an op amp generates a phase-shift between 0° and 180° with frequency as compared to the 0° to 90° shift of a single RC filter. It also keeps the output constant with frequency, rather than rolling off past the -3dB frequency at the usual -6dB octave.

 

[BradtheRad]

The phase-shift is 90 degrees at the -3dB frequency of the RC network. With a 1000kHz -3dB frequency (159Ω, not 300Ω, and 1µF) I simulated 127° shift @ 500Hz, 90° @ 1kHz, and 67° @ 1.5kHz, Don't know why you said it provides 90° over that range. :???:

Yes, that All-pass configuration with an op amp generates a phase-shift between 0° and 180° with frequency as compared to the 0° to 90° shift of a single RC filter. It also keeps the output constant with frequency, rather than rolling off past the -3dB frequency at the usual -6dB octave.

I have every reason to believe your results are correct.

My statement was a little overblown, after having seen how much it could do with just the one capacitor. And its behavior is stable, making it easier to predict how changes in component values affect the outcome. Yet as you point out, it is versatile enough to generate a wide choice of phase shifts.

Anyway this after I'd been watching tiny sine waves traveling across the screen. The input was the upper trace, the output was the lower trace. I could barely discern much difference between 90 and 120 deg., or between 90 and 60 deg.

 

[crutschow]

If you connect the two signals to generate a XY (Lissajous) display you can readily see the affect of phase-shift between the two. Or superimpose the signals on top of each other.

 

[BradtheRad]

If you connect the two signals to generate a XY (Lissajous) display you can readily see the affect of phase-shift between the two.

Yes, I tried this. However I concluded that I could not rely on what I was seeing, because there were times when:

* the output (Y axis) changed amplitude at important moments in the sweep, causing the lissajous figure to change shape unexpectedly

* the figure became a rounded oval at times, making it hard for me to tell when its axis was rotating past the horizontal / vertical

* the X and Y axes were uncalibrated, making it hard for me to gauge orientation in degrees off-axis

* I had trouble keeping an eye simultaneously on the frequency and the lissajous figure

Or superimpose the signals on top of each other.

I'm not sure this can be done in Falstad's simulator.

Also, the great range of phase shifts I observed (in the all-pass type) would make it hard for me to figure out whether it was delaying or advancing the input signal.

 

[crutschow]

You might try doing it in a Spice simulator such as LTspice which is a free download from Linear Technology. That has more options for observing and measuring the signals.

 

[BradtheRad]

You might try doing it in a Spice simulator such as LTspice which is a free download from Linear Technology. That has more options for observing and measuring the signals.

I tried LTspice after seeing several here recommend it. I see it has a lot going for it. Versatile. Many features to offer (once I have learned what they all are and how to use them).

Given time I could learn a lot with it, except I have been working to develop my own simulator for years, which portrays circuit dynamics in motion, and which lets me change component values easily on the fly. I believe there's a call for a simulator of this kind. In this age of specialization, it's something I've found to focus on.