Anyone know of a design for a -simple-tunable- low distortion sine wave oscillator (audio frequency range)

Thanks. I'm looking into it.

You would have more choice if you could live with 0.01%.

Another thought is if you cant get well behaved tracking pots put them in series with a pair of switched precision resistors to halve the relative tracking error. Not very convenient but it might get it back in the right ball park at the expense of narrower frequency range.

The output of a digital pot moves in steps. If you do this to a sine wave amplitude then it creates interesting intermodulation distortion. Engineers call it zipper effect because it sounds a bit like one. A big problem on an audio mixer desk gain, but not on most test gear. I suspect but have not tested it that if you do this to the sine wave frequency or phase the "zipper" effect will be all but inaudible.

Regards, Martin Brown

Several of the hobby magazines have published articles on it - kits are available. Velleman do a kit for a PC controlled audio one for about £130 which really only requires you to be able to solder (

The clock to the filter is a multiple of the fundamental, maybe 64 times, and quite low level to start with so I would think that wouldn't be too difficult. Even a DAC is going to have quantization noise to get rid of.

Mark.

ns

o

ing

on

- 10

cally

the

e

y?

The general idea is that clock frequency will be a lot higher than the sine wavefrequency. My proposition suggested 32 times higher. If you put the 3dB point of your analog filter at three times the frequency of the sine wave you want, a five pole filter would attenuate a full- scale clock feed through by a factor of 100,000.

Since the clock frequency feeds through as the steps on the staircase approximation - essentially a sawtooth error waveform - a 32 times faster clock would give you a manximum step of sine 2.pi/32, or about

10% of full scale, so a four pole filter would do

John Larkin's 18-bit DAC plus look-up table would let you use a higher clock frequency. To take full advantage of the 18-bits you'd clock a system gnerating a 60Hz sine wave at about 50MHz, which isn't practical. Commercially available 18-bit DACs are aimed at the audio market, where the data is sampled at 44.1kHz. In practice, everybody uses over-sampling to make the analog filters simpler and cheaper. This Analog Devices part can generate new samples at up to 850kHz, though the test circuit uses 350kHz.

The maximum step size at 350kHz seems to come out a 0.05% of full scale, which wouldn't meed much filtering. You'd probably see more of the clock frequency leaking in through the power rails.

-- Bill Sloman, Nijmegen

--
What frequency and amplitude range and output impedance are you
looking for?

tions

y to

x 12

ugly

he

rces

38 .

LUT

ho

s l d

ted text -

Hmm, when you change frequency the distortion changes? I never saw anything like that. Mostly I found that the distortion changes with the drive level. If I ran things too clsoe to the rails there would be more distortion. How are you controlling the gain? (I use a light bulb... or three) Is this a 'one of' or something for production?

George H.

Some variation of this...

...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

text -

If the pots don't track well then after you tweak it all up changing frequency alters distortion. Also, distortion increases with frequency as there is less surplus gain for negative feedback.

I've been using a JFet with just a simple diode rectifier. In simulations it actually gave me the spec's I needed which was a surprise. I have a handful of CDS cells coupled to LEDs and also LT1228's but I don't think they are going to be needed.

This is a one off (for my use) but I hope the result is something others (hobbyists) will find useful.

Oh, great.

A big

Might be a problem for a frequency sweep.

Amplitude control is probably going to be a Jfet - the digital pots are for shifting frequency. Frequency control to 1% will be good enough. It could be .1% if I get excited and put 100k and 10k digital pots in series.

There was a recent on in Silicon Chip Magazine - 12-bits I think. It was almost there, and a big step up from most of the stuff recently published (in Oz)

about 5Hz - 200kHz, the distortion figures are only important to about

50kHz. Output impedance will be 600 ohm. I'm not sure if it is best to calibrate the output in dBm or Volts p-p.

About 3 volts out +12dBm

Nice output, especially considering the diode stabilisation. How do you tune it?

With the C's, or a ganged pot for R4, R5. My example has unequal R4, R5 values to set frequency with standard values. Just use a 2-gang pot (rheostat mode).

I don't know how ugly it'll be at your wide frequency range.

You might want to switch in capacitors for "range", and use the pots for tuning within the range. I would guess you _can_ make a decade of adjustment.

(*) The diodes just clip the amplitude down so the core of OpAmps (acting as an active filter) can clean it up the resulting square wave.

I have, on a aeronautics-related project, replaced the diodes with comparator active clamps that were adjustable... allowing setting amplitude as well as frequency. Think LM339 (in place of R6) squaring "OUT", but varying the LM339 pull-up voltage to set amplitude.

IIRC, your high-end need was around 200kHz? Make sure you use high GBW OpAmps. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Two possibilities: XR2206 is a mangled-triangle generator with the tunability you'd like. And, National Semi's LM13700 data sheet has a two-chip VCO design that does rather well (see figure 17).

rations

ity to

2 x 12

ly ugly

the

ources

8038

ow..

, LUT

who

ess

tal

and

ed

n

quoted text -

ght

I played around with a Jfet as control for a bit. (I had this occasional motor boating problem. I discovered that it all went away when I used good low dissipation caps in the bridge.) And so the fet wasn't needed. I never got the fet to give as low a distortion as from the light bulb. Is your distortion problem mostly at the higher frequencies? Have you tried it with a light bulb? It's really so simple the bulb does everything for you (rectifing, averaging, and nonlinear resistance.) As I recall I had a hard time getting the fet to be perfectly symetrical.

George H.

Fred Flintstone simply puts a toucan in a big rock box, and shoves its head out the other side, and reaches in and grabs it by the ass. Instant siren.

When you run it up and down (ie: one end to the other)...