Dual sine wave generator with variable frequency and 90 degree phase difference

On a sunny day (Wed, 03 Sep 2008 09:42:57 -0400) it happened Steve wrote in :

2 x EPROM sine and cosine lookup table, 4046 VCO variable clock generator, binary counter on EPROM address lines, 2 x 8 bits wide DA converter, 2 x lowpass.

For a 256 values per sine wave form, your clock should be max 2560 Hz.

Steve a écrit :

If you don't mind distortion, a Johnson ring counter plus an oscillator will do what you want.

--
Thanks,
Fred.

I've used a software DDS running on an MCU with a DAC for that sort of thing. It's much easer to design and build than any analogue circuit.

Leon

This all depends on the frequency resolution, i.e. VCO step size. A coordic would be a better approach.

I'm always amazed at how often the coordic is not suggested as a solution. I suspect it isn't taught in college courses anymore. I haven't found a good write up on the net regarding the algorithm either. It's decades old technology.

an

A small pic is the way to go, or any other small micro, almost made for this job. You would have to learn to program but thats a requirement these days anyway.

df

Next to last item in the PDF. About 8% distortion It'd need feeding with a 0.6Hz to 120Hz clock. Could well use

1/4 of a CD4093 as the clock osc' and another 1/4 in place of that 4001 inverter.

an

I could make software for such, whats it worth to you? SndScopePro 1.5 Dual Trace Sound Card Scope Professional Dual Trace Oscilloscope AC DC Triggered Input Single / Dual channel Volts / DB Digital Readout Time / Frequency Digital Readout Auto Gain Function Plus Much More By using the sound card line inputs (left/right channels) a low cost Dual Trace Scope is possible, scope has many adjustments and can be used as is without additional hardware, with a small box with Calibrated 10 turn pots and freq dividers, you can use the scope for any amptitude or frequency.

With only a little trickiness, you can get two variable gains from one pot. You ground the wiper and use the two ends as variable resistances.

Th classic 3 op-amp "state variable" oscillator naturally makes two outputs at 90 degrees. The tricky bit is making their amplitudes equal. For this you need two equal variable gains. I think I have seen a circuit that does this but I can't remember it right now.

You are thinking of the state variable filter where you get LP, BP and HP at the same time. With a 3 op amp filter, you can scale the amplitude at each op amp output. This is known as dynamic range adjustment.

A quick google search didn't find a write up on filter dynamic range adjustment. I guess it's another stone age thing not taught any more. However, you can visualize this adjustment if you understand signal flow graphs. Say the output of an op amp is 3 db too high. Then what you do is reduce the gain of all paths into the op amp by 3db AND amplify the paths of signals leaving that op amp by 3db. This leaves the filter denominator untouched.

Generally what you do is sweep the filter over frequency and note the peaking of each node (op amp output). Using a spreadsheet, you can tweak the component values such that all outputs peak at the same level (though not necessarily the same frequency if the filter is a leap frog design).

Dual channel DDS on a cheap 8-bit uC using pwm sine / cosine approximation. About the quality of talking moose on a fat Mac. What in tarnation are going to do with that?

To find it use CORDIC (COrdinate Rotation by Digital Integrating Computation ). For sine / cosine DDA is simpler. At 8-bits lookup tables are easier still. It will leave you time for really fancy pwm methods to do error spectrum shifting / noise shaping like better CD and semipro (and above) audio does. Besides, a uC DDS solution allows variable phase relationships that most other solutions have trouble with.

It's not completely 'analog', but an MF10 filter can do the job. See National Semiconductor AN-307, figure 7...

NOT analog. Use a ramp oscillator for constant amplitude; one stage generates a square wave for integrating to the ramp. Run a comparitor off the ramp (triangle); that will be 90 degrees WRT the square wave. The 2 square waves can be filtered with a simple 3-stage phase retard filter.

Digital, NOT analog.

Digital, NOT analog.

No I'm thinking of the oscillator of the same basic topology and offering it as a solution to the OP's problem.

Making two gains that depend on one pot would be the tricky part.

If the OP is willing to use a ganged pot, the state variable oscillator may be exactly what he wants. It would be nicer to use a

10 turn pot because of the 200:1 range that the OP wants.

).

an

e e

ve

I was thinking of the state variable as a phase shift network (narrow band Hilbert transformer). Wrong application obviously. I suppose the same network could be used as an oscillator with positive feedback. Or you could make a sine wave oscillator, then feed it to the state variable filter to make quadrature signal. This could have the additional feature of filtering the sine wave oscillator.

Now that I think of it, there IS an analog solution: the old CA3080 transconductance amp application notes had the key item, I think it was ICAN-6668. National Semiconductor LM13080 data sheet, fig. 17, is similar.

Basicallly, it's a phase-shift oscillator using four amplifiers with four matched capacitive loads on the outputs. One tunes the frequency by modulating the output impedance of the transconductance amplifiers (all four amps tuned with a matched variable current). Your adjustment pot feeds four limit resistors into four PNP emitters (matched transistors), with four bases grounded; the four collectors are matched current sources for the transconductance amps' program pins.

Dual OTA amps are still available (LM13700), though not the '3080' (so sad, RIP). The oscillator with three sections rings at 60 degree phase shift, you want to use four sections, 45 degrees shift on each, to get sine and cosine outputs. Remember, an odd number of sections have to be inverting amplifiers!

^^^^ ITYM COordinate. ;-)

You only need to store one quadrant of sin(x) for both sin and cos eg. 0