90 deg phase shifter

[curious_mind]

I am looking for phase shifting 50hz sine wave by 90 deg. I found two solutions, one using a differentiator and other with all pass filter. Which one to use?

 

[KlausST]

Hi,

Indeed I never used an all pass filter.

Thus I'd use a differentiator (not knowing if an all pass filter is better in this case)
An active differentiator (using Opamp) is pretty accurate according 90° shift.
The downside is, that it over ephasizes overtones (higher frequencies).
Thus any higher frequency (amplitude) will become amplified with respect to 50Hz.

If in your application a microcontroller is involved, I'd go the digital way.

Klaus

 

[curious_mind]

In differentiator, whenever there is a change in input signal slope, the output is noisy at the change over point.

 

[KlausST]

Hi,

There are two major sources of noise at a differentiator output:
* the OPAMP noise
* the input signal noise. Usually is the bigger one, especially, because higher frequencies get the higher gain.

So you say "change in input signal". Is it the "expected change" or is it an erronous change?

Klaus

 

[BradtheRad]

Sinewaves in a capacitor or inductor universally exhibit 90 degrees shift between voltage across it and current through it. Measure current by installing low-ohm resistor inline.

 

[LvW]

What about an inverting integrator (Miller integrator)? A differentiator has always inherent stability problems.

 

[danadakk]

https://www.analog.com/media/en/training-seminars/tutorials/MT-202.pdf

Regards, Dana.

 

[KlausST]

Hi,

Ah, this is an all-pass solution.

I just saw the graph with phase vs frequency. At 50Hz the graph is most steep. This means at this pont the phase shift is rather sensitive to frequency change and part value (tolerance and drift). Especially C1 and R3.

So if you are after very accurate 90° phase shift, then there should be better solutions.
But if you are after constant amplitude, then it has a very impressive performance.

Klaus

 

[danadakk]

Not too difficult to replace R for phase control with fdbk loop controlling a R type element,
and use auto zero techniques. Errors reduced to something on the order of 1 / Gloop

Or use hi res DDS phase control.....even that has error, but a whole lot less.

Here is a simple solution, good to 8 bits, one chip, 98% of chip resources unused available
for other tasks.

Regards, Dana.

 

[FvM]

All these solutions work, with specific advantages and drawbacks. Without an application specification, it's hard to decide which one to use.

 

[LvW]

Here comes a very exact but - more or less - exotic solution: PLL with a multiplicative phase sensitive block.

 

[FvM]

I don't think it's not particularly exotic, it's one of the solutions to process the voltage reference of grid tied Inverters. Usually by DSP techniques as "all digital pll".

 

[crutschow]

I am looking for phase shifting 50hz sine wave by 90 deg. I found two solutions, one using a differentiator and other with all pass filter. Which one to use?

An op amp differentiator or integrator could be used, but their amplitude varies with frequency.
An all-pass filter will have a constant amplitude, but the phase-shift varies with frequency.

An integrator circuit may be preferable, since it suppresses any high frequency noise with a 1-pole rolloff, and the phase-shift is not significantly affected by component tolerances, although it will affect amplitude.

LTS simulation of example op amp integrator circuit below:
The phase from the output to input appears leading because the op amp inverts the output signal.
If desired you can invert the output with a second op amp to give a lagging phase-shift.

 

[danadakk]

If 50 Hz is line based this discusses freq accuracy in page :

Utility frequency - Wikipedia

en.wikipedia.org

Regards, Dana.

--- Updated Jan 26, 2023 ---

Integrator response (excellent phase stability as crustchow states) but
amplitude an issue :

Regards, Dana.

 

[crutschow]

Integrator response (excellent phase stability as crustchow states) but
amplitude an issue :

Not likely if the source is the 50Hz mains.

 

[KlausST]

Hi,

The whole thread misses informations from the OP.

We need to do a lot if guessing.

I had to design similar circuits.
The problems:
* While mains frequency and amplitude is fairly stable here, we had problems when the user switched to generator supply (during mains power fail)
* The integrator solution has a (drifting) DC shift problem, caused by integrating DC offsets (Opamp, and leakage currents on PCB and capacitor) and amplitude vs frequency.
* The differentiator has the noise problem, and amplitude vs frequency
* The all pass filter has a phase accuracy / precision problem (vs frequency)
* PLL solution is good, it may even back up short time power loss, but has increased part count
* then there are several digital solutions ... using additional hardware and software

We successfully used
* the differentiator solution with aditional filters and
* the PLL solution (with microcontroller in feedback)
(microcontroller to avoid hardware modifications. So we just had to update just the software if any problems arise. Also because we had to switch between 50Hz and 60Hz mains frequency)

Klaus

 

[crutschow]

The integrator solution has a (drifting) DC shift problem, caused by integrating DC offsets (Opamp, and leakage currents on PCB and capacitor) and amplitude vs frequency.

True, I didn't have a bleed resistor across the capacitor in my circuit to minimize any DC shift from input bias or leakage currents.
Below is the circuit with that corrected:

And I wouldn't expect the amplitude variation to be a problem unless there is a significant change in frequency (assuming he is likely looking at the zero-crossing phase-shift).

 

[danadakk]

Some guidance on distortion verus sample size :

64 samples

8 samples

Regards, Dana.