Oscillator Design

Oscillator Design (Phaser project)

Hi, I am designing an oscillator. The output voltage will control the resistance of 4 JFETs.

This is the first thing I've ever designed really from scratch, so if its totally a horrible design you should say so.

**Anyway I don't have the values all picked out yet.** The values listed for all of the components are simply a product of copy/pasing. I pretty much would like to see if this idea is feasible before I start really getting into it. I have however, done some basic testing in PSpice and the results are promising.

I used a filter on the output of the phase shift oscillator because I need the signal to be smooth, but I also need a fast set time. I can achieve a fast set time with a big Rf in the phase shift oscillator circuit, but the output signal is very distorted. I used the filter to attenuate the harmonics. It works nicely, according to Pspice.

The JFETS I'm controlling will have gate voltages between -0.8 and -3 volts. I used the pair of inverting/noninverting amplifiers like I did because I want 2 of the JFETS to be at -0.8 Volts while the other two are at -3 volts. I hope this makes sense.

Right off the bat I'm pretty sure I can knock out one of the op amps by using a Chebyshev filter. The butterworth filter cannot provide a gain of less than 1, and the output from the PSO is always of a greater amplitude than needed (maybe I can power the PSO's op amp with less than +/-15 volts). THe trouble is is that I'm not all that good at designing Chebyshevs.

Also, I hvae the following problem: The magnitude of the output of the filter changes as the PSO changes with frequency (as in 0.2 and 1.75 Hz). I can't think of a good way to regulate the peak voltage, so I think I'm just going to make some comprimises.

Thanks for reading this!
and thanks for your help!

What's the purpose of this circuit?

It's helpful to show us the bigger picture as you might be doing this the hard way.

How are you going to change the frequency of your oscillator? It needs 3 capacitors or 3 resistors to be changed at the same time.

You can't use a Chebychev filter because its output changes in its passband.

Of course the output from your Butterworth filter changes when the frequency changes, it is attenuating frequencies above about 0.2Hz.

Your phase-shift oscillator uses differentiating filters. Why not swap the Rs and Cs to make them integrating filters? Then they are a 3rd-order lowpass filter that reduces the clipping distortion from the first opamp. Use a second opamp as an amplifier and buffer.

Hi, thanks for your help.

audioguru--I will try this oscillator circuit. I've never really learned anything about oscillators, and the one I used was copied straight from a book. Thanks for the tip.

It is preferable to adjust all 3 resistors or capacitors in tandem to change the frequency, but I'm quite sure you can do it by adjusting just one.

Hero--The whole project is a phaser. https://en.wikipedia.org/wiki/Phaser_(effect) I will be using this LFO to vary the poles of the all pass filters.

I'll be using this all pass filter design, except I'll replace the R with a JFET so that I can control the pole.
https://en.wikipedia.org/wiki/Phaser_(effect)

Thanks again
Sean

Oops, here is the link for the all pass filter:

https://en.wikipedia.org/wiki/Image:Allpass.svg

SeanHatch said:

Hi, thanks for your help.

audioguru--I will try this oscillator circuit. I've never really learned anything about oscillators, and the one I used was copied straight from a book. Thanks for the tip.

It is preferable to adjust all 3 resistors or capacitors in tandem to change the frequency, but I'm quite sure you can do it by adjusting just one.

Hero--The whole project is a phaser. http://en.wikipedia.org/wiki/Phaser_(effect) I will be using this LFO to vary the poles of the all pass filters.

I'll be using this all pass filter design, except I'll replace the R with a JFET so that I can control the pole.
http://en.wikipedia.org/wiki/Phaser_(effect)

Thanks again
Sean

Click to expand...

Hi Sean,

If you want an easier-to-use variable resistor, look at the H11F1M at http://www.fairchildsemi.com . They are MUCH nicer to use than JFETs.

The H11F1M is current-controlled, and can vary the resistance of two of its pins from less than 100 Ohms to several hundred MegOhms with a control current from about 30mA to 0 mA, with 99.9% linearity, and has 15us to 25us (or so) on and off times. The resistance pins can handle 60V p-p and +/-100mA.

You could implement proper voltage/current control for both your oscillator and your filters, with them.

Also, there's a great automatic filter design program available as a free download from Texas Instruments, at http://www.ti.com . Just search for Filterpro.

- Tom Gootee

**broken link removed**

I tried some H11F1 opto-FETs in a Wien-bridge oscillator for AGC because they say "distortion-free". It has the same fairly high distortion as ordinary JFETs that have 50% signal feedback for low distortion.

audioguru said:

I tried some H11F1 opto-FETs in a Wien-bridge oscillator for AGC because they say "distortion-free". It has the same fairly high distortion as ordinary JFETs that have 50% signal feedback for low distortion.

Click to expand...

Bummer. What sort of operating conditions was it at?

I guess that if one wanted a "pure" resistance that was similarly current-controlled, one could use a Vactrol (e.g. VTL5C2), or just an LED encapsulated with a photocell. But then it's not linear. And it's relatively slow.

- Tom Gootee

**broken link removed**

I noticed that the H11F1 is being discontinued, and replaced by the H11F1M. I wonder if there's much difference between them?

The spec's for the H11F1 and the H11F1M are the same. The datasheet shows the "s-curve" of a peculiar type non-linearity that causes a weird type of distortion. The level must be extremely low for low distortion.

It sounds like one of you are talking about using JFETS in an oscillator circuit, and the other is talking about using them somewhere outside of the oscillator. Just to clarify, my JFETS are outside of the oscillator.

There shouldn't be any distortion in the all pass zero as long as I keep the JFETS in the ohmic region, right? I understand that if I did let the JFET slip into saturation, there could be issues, but as long as I don't let that happen...

JFETSs are used for automatic-gain-control in sine-wave oscillators and in audio equipment.

A JFET produces about 0.1% distortion when the signal across it is more than about 30mV RMS. If the signal is 300mV then the distortion is severe.
Adding a signal to the gate that is half the level of the signal at the drain reduces 2nd harmonic distortion a lot.
Years ago the "chorus pedals" used bucket-brigade echo/delay ICs that had 2.5% distortion.

I understand that you will use matched JFETs in all-pass filters as part of a phaser effect circuit. I think the old CA3080, LM13600 and LM13700 transconductance amp ICs were used in the 70's to do it.

Google is full of old flanger/phaser circuits.

Heh, yeah, I know, but I really wanted to design something from scratch.

Thanks for all the info about JFETS. That's the type of stuff they never tell you in school. This is the kind of thing I feel like I learn from by designing from scratch; good stuff.

btw, back to the original topic, I switched from the phase shift oscillator to a wien bridge oscillator. Much less distortion.

A Wien bridge oscillator has bounce, bounce, bounce, bounce, ... on its output level if it has the usual light bulb or temperature-sensitive-resistor as its level controller. When a JFET is used as its level controller then there could be distortion or "squegging".
It needs to have two capacitors changed or two resistors changed to change its frequency without changing its level.

I just had a wacky idea for your very-low-frequency oscillator:

Use a small DC motor. If you put a disc on the shaft, with a rod connected to it that could move like a piston rod on a crankshaft, and that rod was pinned (i.e. hinged/rotatable) to an arm that was firmly attached, at a right angle, to a potentiometer shaft, the motor's rotation would rock the pot setting back and forth. You could apply a voltage across the pot and get a sinewave from the wiper lug. You could also do it with a linear pot.

I guess if you made a fancy multi-rod crankshaft, you could run pots in the filters, directly.

The pots would wear out, of course, and might get noisy. But if you just use the one, in place of the oscillator, you could put a good low-pass filter on it, at 10 Hz or so.

Another easy way to get a high-quality sine wave would be to load a sine wave's datapoints into an EPROM, and clock them out through a DAC, at whatever speed you want. No cpu would be required. Then just low-pass filter the output, which is easy to do well if you have enough datapoints. The DAC (digital to analog converter) could be just an R-2R resistor ladder and an opamp, or a commercial IC. If you want to get fancy, you can store only 1/4 of a cycle's datapoints, and reverse and/or invert when needed.

Another way to get a sinewave would be to generate a triangle wave and then either filter the heck out of it or use a triangle-to-sine converter. I have a pretty-good design for a tri-to-sine converter, if you want to go that way. But the triangle's amplitude has to stay exactly right, for it to have less than about .3% THD. What I did, with that, was amplify the resulting sine, to say 20v p-p, and then run it through a filter with a servo (integrator) feedback loop that supplied the control current to Vactrols (current-controlled resistances) that controlled the cutoff frequencies in the multistage lowpass filter. It just keeps lowering the cutoff frequency until the desired amplitude (say, 5V p-p) is at the output, which also dramatically improves the THD (by an order of magnitude).

- Tom Gootee

**broken link removed**

Wow your motor, idea sounds expensive, power hungry and unreliable, I don't know how you got all three in one!

Filter any waveform with a 50% duty cycle enough and you'll get a sinewave. Build a 7555 timer circuit with 50% duty cycle, add a non-inverting buffer to the capacitor, stick a picky enough filter on the end and you'll have a near as damn it perfect sinewave.

A switched-capacitor Butterworth lowpass filter IC does wonders to remove harmonics. A square-wave can be filtered with an 8th-order IC and the results are low distortion sine-wave.

My very low distortion sine-wave generator uses a CD4018 to make a stepped sine-wave with 10 steps (10 times over-sampled) then is filtered by an 8th-order Butterworth filter IC. The distortion is 0.01%.

is this gonna be some strictly controlled output? you could keep it simple, LC tank. dang yall love things complicated more than they have to be.

crusty said:

is this gonna be some strictly controlled output? you could keep it simple, LC tank. dang yall love things complicated more than they have to be.

Click to expand...

No, we all read the requirements - how were you planning making a 0.2Hz to 1.75Hz variable LC oscillator?.

Nigel Goodwin said:

No, we all read the requirements - how were you planning making a 0.2Hz to 1.75Hz variable LC oscillator?.

Click to expand...

with lots of love and hugs.