# Shift Phase of 100 kHz Signal 90 Degrees

1. Jul 2, 2014

### Fischer777

For a school project I need to be able to shift the phase of a 100 kHz sine wave by 90 degrees (either positive or negative). I have tried several passive integrators and differentiators, both with poor results. I have also tried using an op-amp integrator and differentiator, but in both cases I got a heavily distorted waveform that didn't seem to be 90 degrees shifted. I'm running out of ideas, so I'm wondering if anyone here has some pointers or tips for building a relatively simple circuit for a phase shift.

Thanks!

2. Jul 2, 2014

### jim hardy

Pointer:

It's hard to do in one stage.
An R-C integrator with a long time constant will do it but it reduces the amplitude drastically.
You might try three 30 degree shifts in cascade
but be sure each stage is ~10X higher impedance than the preceding one else the loading messes up your
transfer function.

An active integrator can do it nicely but they're difficult to keep zeroed.
Active differentiator is a high pass so will amplify noise, but integrator is low pass so attenuates noise.

Let us know of your progress.

3. Jul 2, 2014

### Fischer777

Thanks for the advice. I tried using an integrator with a large time constant a resistor network that creates an identical attenuation with no phase shift so I can get both the shifted and the not-shifted wave to have the same amplitude. Even after wiring it to a transistor amplifier the output is still low (.01 volts peak-to-peak) but it should be enough to work with.

4. Jul 2, 2014

### Fischer777

On another note: I'm currently using laboratory equipment to generate the 100kHz signal, but at some point I have to create a working oscillator that will be able supply the signal, and for that I was going to use a crystal oscillator. I'm curious if there are any low-powered oscillator configurations that will do well at such frequencies. I've tried numerous clapp/collpitts configurations and none of them work so I'm guessing they do better at higher frequencies.

5. Jul 3, 2014

### Staff: Mentor

Quite possibly the frequency is too high for the common low frequency OP-AMPs. Did you buy a higher frequency OP-AMP for your tests? Otherwise, the results will depart markedly from your idealised design.

Besides the passive circuit that Jim described, there are two techniques for producing a 90° shifted sinewave. The simplest is to use a phase-locked loop IC. The venerable LM565 is an analog PLL and contains its own VCO which generates the phase-shifted sinewave. All you have to add to the 565 are a few capacitors and resistors, for example, see the figure at the foot of this page. http://mysite.du.edu/~etuttle/electron/elect12.htm

The other method is to use an all-pass filter. It has a flat gain, but the phase changes across its frequency range. It will be worth buying a wider bandwidth OP-AMP for this, so that the results are close to what you design for. http://www.ecircuitcenter.com/Circuits/op_allpass1/op_allpass1.htm

Good luck.

6. Jul 3, 2014

### Staff: Mentor

Does you application demand the accuracy & stability of a crystal oscillator?

On that note, what is your application that it needs quadrature signals?

7. Jul 3, 2014

### Fischer777

The op-amp I was using for the application specified it had a 3.5 MHz bandwidth and the data sheet states it has a slew rate of 13 volts/usec, so it should work for a 100 kHz application.

I can try the all-pass filter design, based on the article if it works it will do exactly what I need it to do.

The end product will use a QAM signal to send two audio channels, and the device will need to be very small and low-powered so an LC oscillator would be too bulky. Other oscillators based on ceramic resonators could work, but the principal would be the same as a crystal oscillator I think, which is proving to be very difficult to design.

Last edited by a moderator: May 6, 2017
8. Jul 3, 2014

### sophiecentaur

One easy way to get two signals with a well defined phase (for example, quadrature) between them is to start with a higher frequency square wave and two divider (counter) circuits to produce the wanted lower frequency (dividing by three four five six or whatever you want). Use a 400kHz square wave oscillator and two divide by 4 circuits, with some logic so that one counter waits one cycle before its counting. The two output square waves will be in quadrature, whatever the input clock frequency. If you want two good sinusoidal outputs, two band / low pass filters can easily be made with matched phased characteristics but for a QAM signal, the final (combined) output can be filtered after the modulation of the two carriers has been done.

9. Jul 4, 2014

### jim hardy

Last edited by a moderator: May 6, 2017
10. Jul 4, 2014

### Staff: Mentor

[strike]An easy way to get a sinusoidal oscillator is to use the VCO section of a PLL IC. The NE565 family has such an oscillator.[/strike] Check the data sheets to see whether its current falls within your idea of "low power". If not, there are other PLLs to look at. So you'd need the PLL chip plus a couple of C's & R's, and a potentiometer if you need to fine tune the VCO frequency. Its frequency stability will be set by the stability of the timing C & R. Will you need exceptional stability?

EDIT: Actually, I'm not sure that the 565 oscillator is sinusoidal, I think it may be a squarewave. I know its phase detector is designed for sinusoids, that is why it is marketed as an analogue PLL. But the VCO could still be a squarewave. So would not be a source for a sinusoidal VCO.

For an IC waveform generator that can output sinewaves, look for the XR-2206 or the ICL8038: http://www.circuitstoday.com/audio-oscillator-circuit-2
This won't be as pure as from a good LC oscillator.

Last edited: Jul 4, 2014
11. Jul 4, 2014

### jim hardy

565 family wont give a sinewave, just triangle and square

but they sure are fun

12. Jul 4, 2014

### Staff: Mentor

Thanks Jim. I just noticed that in a tutorial and came back to edit my post. But see you've already corrected me.

13. Jul 4, 2014

### sophiecentaur

The purity of the sine wave may not be particularly important if this is a signal for transmission. A suitable low pass filter will reduce harmonics to a practical level; they will be 'out of band and ignored bu any demodulation system - which is why you'd be using QAM. It's a different problem from that of audio oscillators which have tighter requirements.

14. Jul 4, 2014

### Baluncore

Here attached is a possible circuit for a 100kHz IQ modulator. The output is ('Left' * In phase) + ('Right' * Quad phase).

How it works: The two analogue input channels 'L' and 'R' are split into normal and inverted phase. An oscillator running at twice the output frequency, clocks the two D' flip-flops, (74HC74), twisted ring to generate digital quadrature signals. Those I and Q phase signals are used to select the inverted or non-inverted analogue inputs, which are then summed by the final amplifier. The final amplifier needs wide bandwidth. The 470 pF capacitor removes some of the higher odd harmonics of the I and Q signal switching. You can kill more harmonics by following it with an output coupler having a broad resonance at 100 kHz.

The digital parts, (oscillator and 74HC74), need Ground and +5V power. The 74HC4053 data selector needs Ground, +5V and -5V supplies, (The digital inputs using the Ground and +5V, while the analogue uses +5V and -5V). All the amps need +5V and -5V supplies.

Edit: The quadrature generator clock needs to be 4 times the output frequency, so it needs a 400 kHz oscillator.

#### Attached Files:

• ###### IQ modulator.jpg
File size:
35.1 KB
Views:
1,126
Last edited: Jul 4, 2014
15. Jul 7, 2014

### sophiecentaur

HA. I was trying to figure how it would work with 200kHz but gave up, assuming you were right. But with the same clock signal going to each half. . . . . These days I have to start from scratch with all those circuits. The configurations are just not as familiar as they used to be.

16. Jul 7, 2014

### Baluncore

Twisted ring counters, with n stages, ideally generate an output frequency = clock / (2 * n).
If n is above 2 they must be reset to prevent initial 0101 type states propagating.
The n = 2 ring shown in my circuit is well behaved without reset since all possible codes exist.

If the quadrature outputs from a twisted ring were low pass filtered, they would produce a closer approximation to a sine wave, along with some phase error. But in my circuit, digital signals are needed for the mixer switching.
By placing the LPF after the summing stage, the phase errors remain the same for both channels, quadrature is preserved, and only one LPF is needed.

The fundamental of the output spectrum is centred on 100 kHz. Without a LPF, there will be a 1/3 amplitude third harmonic at 300 kHz, 1/5 fifth harmonic at 500 kHz …

Since the I and Q channels will probably be recovered using a PLL there remains the problem of synchronisation.
Synchronisation would be lost if both channels had zero analogue signal input.
Where digital data is encoded with QAM there are always an even number of states.
Zero amplitude is therefore not present as a coordinate in the phase constellation.

Last edited: Jul 8, 2014
17. Jul 9, 2014

### Fischer777

Thanks for all the input guys. On the topic of crystal oscillators, is it the load capacitance or the shunt capacitance that is supposed to be placed in series with the crystal? So far I've been using the data sheet's specified load capacitance of 12.5pF, but the data sheet specifies a shunt capacitance of .7-1.5pF. If I'm supposed to be using the shunt capacitance that might explain why I'm having trouble.

Thanks

18. Jul 9, 2014

### Staff: Mentor

Your stray wiring capacitance is probably of that order, a few pF.

Can you give a link to a schematic of the circuit you are using? You really have a 100 kHz crystal?

19. Jul 9, 2014

### Fischer777

Wouldn't the stray capacitance be in parallel with the crystal, not series (and the crystal is a series-cut).

Here's a schematic of the oscillator I'm using (attachment). It's a butler oscillator designed for series-operating crystals. And the crystal I'm using is a 100 kHz crystal, but it's a very tiny one in a cylindrical package.

#### Attached Files:

• ###### BUTLER-OSC.jpg
File size:
18.6 KB
Views:
354
20. Jul 9, 2014

### Staff: Mentor

I think that trimmer capacitor provides the series capacitance. Do you have any test gear to determine whether the circuit is oscillating in any fashion at all? Are you constructing this on veroboard or something?