...and I think I'll need to have an offset since the gate of the jfets will oscillate between 0 and -vto or something.

is this thread long because im dumb? i hope you guys are having fun. i am getting somewhere, on the oscillator, though.

 

Five 1st-order filters in series make a very droopy filter that begins attenuating frequencies far below what is wanted.
A Butterworth filter uses positive feedback to boost the response so it is flat right up to the cutoff frequency.

But the filter needs to be tuned so that it still greatly attenuates the 3rd harmonic of 0.2Hz (0.6Hz) and allows a flat response at 1.75Hz.

A switched capacitor Butterworth lowpass filter IC has its 8th-order cutoff frequency changed with a single resistor on its built-in high frequency oscillator.

__________________
Uncle $crooge

 

Hi audioguru,

I need to quit posting when I haven't had enough sleep. The five-stage filter I simulated was just to quickly see what it would take to get the THD below 1%. An active filter with a steeper roll-off could make it practical to use a filtered squarewave.

A switched-capacitor filter IC sounds like the thing to try, for that approach. It looks like Hero999's filtered squarewave idea, with your switched-capacitor filter IC, could result in a simple, two-IC solution, for the oscillator.

Do you happen to have a part number, for that 8th-order switched-cap IC (with internal clock oscillator) that's controllable with one resistor?

The dc-accurate LTC1062 (and appnote AN20) from Linear.com looks interesting, for low frequencies.

- Tom Gootee

 

Maxim-IC have many switched-capacitor filter ICs. Their MAX291 and MAX295 are 8th-order Butterworth lowpass filters with an adjustable cutoff from 0.1Hz to 25kHz and to 50kHz. I used to use ones from National Semi but they aren't made anymore.

The frequency is adjusted with its internal RC oscillator by changing the capacitor value, not the resistor value like National's ones. Or make your own RC oscillator.

__________________
Uncle $crooge

 

Hi, today I finished a sort of preliminary design for the phaser. I've included a picture of it in this thread. Here are some notes and things that I already know about:

1.) I haven't chosen all the values yet. Most of them are correct, but I'm still working out some loading issues.

2.) The lowpass/summer at the output is something I sort of made up. In simulation it seems to do what I want, but what do you think?

3.) The real product will have some switching circuitry that I didn't show on the schematic. I think it'll just be a simple switch. I'm not going to implement a flip flop switch or anything.

So yeah besides that, what else should I change. I know there's probably alot!

Thanks for everyone's help on the oscillator. I was able to get down to about 1% THD just with the circuit in the schematic. I did try some other things like the 555 to 8th order IC, but I don't think it was worth it in this case.

any suggestions are welcome...

thanks!

Attached Images

phaser.JPG (82.1 KB, 13 views)

 

Hi Sean,
Your oscillator uses differentiating RC circuits that pass harmonics. Didn't I show a phase-shift oscillator that integrates and reduces harmonics?
I think your oscillator will produce a square-wave and the simple filter won't do much.

The 741 opamp in the oscillator is driving a resistance far lower than its minimum allowed load of about 2k ohms. It will current-limit with square-waves most of the time.

The oscillator will stop when you change the frequency pot. All three resistors are usually changed at the same time.

The FETs need an input voltage from a negative voltage for a high resistance to 0V for a low resistance. Your oscillator will drive them positive which will forward-bias their input diode then they are like a short.

__________________
Uncle $crooge

 

Hmm I did not know anything about the minimum load of 2Kohms. I will have to fix things, then.

I tried that phase shift oscillator schematic you showed me a while ago, and I couldn't get it to work. It'd be nice if I could, though.

I've simulated the oscillator I have now in SPICE, using the uA741 subckt model I copied from my textbook. The results are pretty good (post filter), ill post some images tomorrrow. But, could the model be missing something that would mislead me?

Right now the output of the oscillator filter swings between -0.5 V and -2 V, which shouldn't short the FET, but is a pretty big sweep in regards to distortion; i think you told me this earlier.

do they make triple pot kind of things? i've never seen them.

 

I made a phase-shift oscillator with a triple pot I removed from something. I buffered each RC stage with an opamp so the loss wasn't high. It worked well until the triple pot got dirty and now the frequency jumps all over the place when I turn the pot.

__________________
Uncle $crooge

 

Hi Sean,

Spice models certainly CAN mislead. And if you don't know to watch out, then it's probably typical that they mislead. I usually try to read datasheets carefully, do "sanity check" simulations, and breadboard early. I've seen opamp spice models that would happily pump out many amps of current, into a low-impedance load. Most real opamp outputs can only deal with 20 mA to 40 mA demands, at best. And most opamps don't do very well into even a 600-Ohm load. [I'll try simulating your oscillator in LTspice, to see what happens, tomorrow.]

Regarding your need for a triple-ganged pot: This might be one place where something like the H11F1M, or some Vactrols, might be expedient. You could use a single pot to set the LED current for several Vactrols or H11F1Ms (with their LEDS all in series), setting the resistances of their other terminals. (I don't know how well the Vactrols might track each other; probably good-enough, though.)

Alternatively, very low-tech: maybe you could mount three pots next to each other, and use string or twine to gang their shafts together, with multiple turns around each shaft before going to the next one.

Another alternative might be to motorize the pots. For "manual" control, you could do it with three small stepper motors and basically just a pulse generator (i.e. no computer needed, at least).

And you could also fairly-easily make one of several possible types of mechanical couplings, that would allow you to use just one DC or stepper motor to control all three pots. e.g. one threaded rod from motor, w/one worm gear on each pot shaft.

OR, to make it easy, if you use three identical pots, with values such that you only need about 1/3-turn of travel, then you could just attach an "arm" to each pot shaft and connect a mechanical link across all three arms' ends. Make all shafts co-axial and it could even be decent-looking and fairly nice to use.

- Tom Gootee

http://www.fullnet.com/~tomg/index.html

 

THanks for your reply. I agree that I should breadboard early, I just don't have the equipment really. (scopes and whatnot).

I think I could sit here and keep tweaking and designing forever in simulation, forever, but I feel like I can't make too much more progress without sitting down and testing, so I think I'm going to go ahead and breadboard from here and see what I've got.

thanks
sean

EDIT: just one thing I'm REALLY unsure of...what do you think of the summer stage at the far right of the schematic? is this legit? Like I said, I sort of 'made this up', so I'm not sure if it'll work how I want. It's actually a first order butterworth with unity gain sort of combined into a summer. It feels like I could get into some kind of trouble with this

 

Hi Sean,

Your right-most opamp is just a unity-gain buffer. The two 10K resistors should work as a passive summer. And C3 will make a 1st-order lowpass at about 0.34 Hz.

Referring to the famous AN-31 and AN-20 op-amp application notes, from national.com, note that a non-inverting op amp summer is typically implemented slightly differently (i.e. with two more resistors, one from out to - input and one from - input to gnd).

Regarding your core oscillator circuit:

If you lower the caps to 22 uF, then 6.8k for each of the three resistors will make it go at 0.388 Hz, and 3x 680 Ohms will put it at 1.73 Hz. You will have to raise the R1 feedback resistance value (a lot), probably. [I guess the caps might need to be bipolar. For that, you could just put two 47uF in series, with either both - leads connected to each other or both + leads connected to each other, and have a bipolar cap of 47uF/2 = 23.5 uF nominal.]

Also, it should work better if you connect the bottom of R10 to ground and use the output of C11 as your feedback signal, with an additional series resisistor before the neg opamp input; probably something like 1K, if you have 330K or more for R1 (270k is probably the bare minimum).

If you use a decently-robust opamp, you can probably control the amplitude by just hanging a small resistor to ground, from the opamp output. I simulated it with, just as an example, 10 Ohms to gnd (with an LT1057A opamp), and the amplitude stabilized at 400 mV p-p. The resistor was passing +/-20 mA peaks and its average dissipation was only 2 mW, while the opamp's avg dissipation was only 54 mW. I was using +/-17.5v supplies.

The oscillator does seem to take quite a while to get its output amplitude up; about 50 seconds, in my simulation.

For your JFETS: You should use the linearization trick that was mentioned. i.e. Use two equal-value resistors, one in series with the gate and one from drain to gate. You can probably use any value from 10k to 100k [use as low as your circuit will handle, for less distortion]. That should enable you to use anything from about -6v to 0v on the gates, with good linearity and minimal distortion, for low-level signals.

- Tom Gootee

http://www.fullnet.com/~tomg/index.html