Shift Phase of 100 kHz Signal 90 Degrees

[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!

 

[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.

 

[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.

 

[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.

 

[NascentOxygen]

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.

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
Obtain manufacturer's Application Notes for extensive design information.

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.

 

[NascentOxygen]

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.

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

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

 

[Fischer777]

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.

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.

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
Obtain manufacturer's Application Notes for extensive design information.

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/Circui...p_allpass1.htm

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.

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

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

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.

 

[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.

 

[jim hardy]

A big "Thanks!" to pf contributor dlgoff for finding this treasure:
http://bitsavers.informatik.uni-stu...dataBooks/1972_Signetics_PLL_Applications.pdf

I think it's the best ever introduction to pll's

print yourself a hardcopy. Or buy an original on Ebay.

Op-amp sinewave generators are a little tricky to keep "balanced".
http://www.ti.com/lit/an/snoa665c/snoa665c.pdf
http://www.ti.com/sc/docs/apps/msp/journal/aug2000/aug_07.pdf
http://www.ti.com/lit/ml/sloa087/sloa087.pdf

 

[NascentOxygen]

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.

[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.

 

[jim hardy]

565 family won't give a sinewave, just triangle and square

but they sure are fun

 

[NascentOxygen]

565 family won't give a sinewave, just triangle and square

but they sure are fun

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

 

[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.

 

[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.

 

[sophiecentaur]

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

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.

 

[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.

 

[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

 

[NascentOxygen]

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

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?

 

[Fischer777]

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

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

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

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.

 

[NascentOxygen]

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?

 

[Fischer777]

Sorry to resurrect this thread, but I recently ran across a phase-shift design used in phase discriminators for FM detectors that uses a capacitor in series with a parallel LC circuit. The original design can be found here: http://www.ajay.keckist.edu.np/ec2/fm_detectors.pdf After running a few tests on the circuit I found it works extremely well for shifting the phase of a signal, however, depending on the capacitance in series it can be shifted anywhere from 0 degrees to 180 degrees, and I was wondering if there was a way to determine which value of capacitance will yield the theoretical closest shift to 90 degrees.

Thanks

 

[NascentOxygen]

Sorry to resurrect this thread

I didn't know it was finished. I've been hoping to hear a report of success, or pending success.

After running a few tests on the circuit I found it works extremely well for shifting the phase of a signal, however, depending on the capacitance in series it can be shifted anywhere from 0 degrees to 180 degrees, and I was wondering if there was a way to determine which value of capacitance will yield the theoretical closest shift to 90 degrees.

A neat arrangement. I'm sure calculations could be done, but you'll still need to fine tune it in situ to set that 90° shift accurately. By the time you add a buffer the component count will be about the same as the APF that I pointed out.

 

[Baluncore]

@ Fischer777. I do not believe it is the solution to your phase shift problem.

The LC circuit will need to be resonant at close to the FM centre frequency. It will then be tuned to just above or just below the FM carrier frequency. The bandwidth of the LC circuit must be wide enough to pass the FM modulation on the linear flank of the frequency response.

To reduce sensitivity to component values and frequency, as a phase shifter, you would need to lower the Q with a parallel resistor.

 

[Fischer777]

I didn't know it was finished. I've been hoping to hear a report of success, or pending success.

Sorry about that. I've succeeded in making an oscillator, and I have also succeeded in using an integrator as a phase-shift circuit, however, due to the level of attenuation it creates I'm looking for a better option, which was why I asked about the circuit configuration used in FM receivers. I haven't yet been able to get the APF to work even though the op-amp I'm using is rated to work at several megahertz, although I might try again using an op-amp with an even greater slew rate.

The LC circuit will need to be resonant at close to the FM centre frequency. It will then be tuned to just above or just below the FM carrier frequency. The bandwidth of the LC circuit must be wide enough to pass the FM modulation on the linear flank of the frequency response.

I'm not demodulating an FM wave with the circuit, I'm only using the concept of a 90 degree phase shifter to shift the phase of a constant amplitude, constant frequency 100 kHz wave. But in retrospect, I think given the size of the components to create a resonant circuit at 100 kHz I'll stick to a solution involving an all-pass filter or integrator.

 

[NascentOxygen]

Sorry about that. I've succeeded in making an oscillator, and I have also succeeded in using an integrator as a phase-shift circuit, however, due to the level of attenuation it creates I'm looking for a better option, which was why I asked about the circuit configuration used in FM receivers. I haven't yet been able to get the APF to work even though the op-amp I'm using is rated to work at several megahertz, although I might try again using an op-amp with an even greater slew rate.

First, construct an amplifier with unity gain, no capacitors at all. Get that working, before trying for anything more ambitious. Any OP-AMP will work, though a gain-bandwidth of at least 4 MHz should give near-ideal filter performance.

 

[berkeman]

This has been a fun and interesting thread, but I have to confess that I haven't read all the posts in detail, so apologies if this solution has already been covered.

The way I've done this in the past is to use 2 ROMs with sine wave data in them that is offset by 90 degrees (for I and Q demod, but the technique works for encode as well). They are clocked together, and their outputs are driving R-2R ladder DACs with filter buffers. This gives you 2 very nice sine waves that are 90 degrees out of phase.

You can then use those waveforms for your 4QAM modulation circuit.

 

[Baluncore]

Analog Devices make a single chip CMOS 300 MSPS Quadrature DDS. AD9854
http://www.analog.com/en/digital-to...al-synthesis-dds/ad9854/products/product.html

 

[sophiecentaur]

It really seems a bd idea to me to produce any phase shift with analogue components when it is not strictly necessary. It's just another 'adjustment' to make which can drift off. If you are working at UHF or microwave frequencies then it might make sense to think in terms of a directional coupler, for instance, which will give you 90 degrees over a wide bandwidth but 100kHz is just crying out for a digital solution.

Waveform synthesis (a-la Berkeman) would be a nice 'pure' solution. I read somewhere in the thread that QAM was the aim (??) so the whole thing could be done directly with DSP, requiring the easiest analogue post-filtering to tidy things up. Not a single line-up adjustment needed. Processors take very little power these days, when operating at a few tens of MHz, which would probably be quite adequate.

 

[Fischer777]

It really seems a bd idea to me to produce any phase shift with analogue components when it is not strictly necessary. It's just another 'adjustment' to make which can drift off. If you are working at UHF or microwave frequencies then it might make sense to think in terms of a directional coupler, for instance, which will give you 90 degrees over a wide bandwidth but 100kHz is just crying out for a digital solution.

Waveform synthesis (a-la Berkeman) would be a nice 'pure' solution. I read somewhere in the thread that QAM was the aim (??) so the whole thing could be done directly with DSP, requiring the easiest analogue post-filtering to tidy things up. Not a single line-up adjustment needed. Processors take very little power these days, when operating at a few tens of MHz, which would probably be quite adequate.

A digital process hasn't been ruled out yet, especially given that the output of the 100kHz oscillator is a square wave. I am still testing various methods of getting this phase-shift signal.

 

[Baluncore]

The digital generation of 4-QAM is very easy. See the attached circuit.

Edit: Fixed terminal name in diagram.

 

[Baluncore]

It might be a good time to consider that, no matter how the 100kHz 4-QAM carrier is generated, the serial data stream will need to be delivered to the modulator at a rate of 200k bits per second.

So the clock that moves the serial data to the modulator will somehow need to be synchronised with the modulator clock. That may be difficult with a 100kHz crystal oscillator. It may need a PLL.

Will that be two separate channels of data, clocked at 100kHz, or one channel clocked at 200kHz, that needs to be split into two staggered bit streams.

 

[sophiecentaur]

It might be a good time to consider that, no matter how the 100kHz 4-QAM carrier is generated, the serial data stream will need to be delivered to the modulator at a rate of 200k bits per second.

So the clock that moves the serial data to the modulator will somehow need to be synchronised with the modulator clock. That may be difficult with a 100kHz crystal oscillator. It may need a PLL.

Will that be two separate channels of data, clocked at 100kHz, or one channel clocked at 200kHz, that needs to be split into two staggered bit streams.

If the modulating signal is going to be binary and all filtering done post modulation then there is no need to have a fast clock for the data. Yes there may be a small 'beat' where the data transitions strobe through the 100kHz but that will be ignored by a demodulator. (I reckon?)
If the data is to be digitally pre-filtered (much healthier) then it will need to be sampled at a high rate. It could be better to choose a clock frequency for the data samples that will strobe well out of the transmitted band. But if the system is all in one box, why not use the same clock source everywhere?
The choice depends on how the system is planned to work. It is good to have an overall plan, involving such choices early on.
If this is a first time project then it may be better to keep everything as simple as possible and then evolve it into something better later.

 

[Fischer777]

With regards to the 4-QAM; I'm not sure if it's an unrelated discussion but the idea behind this circuit was to try and send two audio channels, a left and right, by QAM to a receiver. A 4-QAM device would have to be used at a much higher frequency to send any sort of "good" sound if used digitally, so for now I'd just use the resultant quadrature output as the means by which to obtain a reliable 90 degree shift between the two waves, and do the rest using analog signal processing

Thanks for the ideas/suggestions/advice though, they've been a lot of help.

 

[tfr000]

Have you looked at polyphase networks?

 

[Baluncore]

Polyphase networks, or polyphase filters, are a stack of cross-coupled low-pass and high-pass filters, in other words, integrators and differentiators. Polyphase filters, tend to be insensitive to component variation, but they are quite bulky as they employ multiples of 8 components. They are usually cascaded and applied to broadband signals such as in SSB modulators that employ the audio phase shift method. Used where space is not a problem, they require differential drive and produce differential quadrature output, which makes two inputs and four outputs.


QAM is classed as a spread spectrum technique. I believe it is being used here as an example of how to get a data bit rate that exceeds the carrier frequency by a factor of from two to maybe six times on a quiet channel. The bandwidth required remains twice the carrier frequency.