I have been playing with Wein bridge and other oscillator configurations without any joy. I want low distortion foremost but also the ability to easily tune it over an octave or better tunable over a decade.
Wein bridge works fine, but there is no low cost supply of close tracking pots (1%) available to me and that really sucks. Using a 2 x 12 position multipole switch also doesn't work - at the moment I am thinking of 8 - 10 DPDT stitches on an R-2R network which is really ugly and possibly inconvenient to use.
My question is "is there a low distortion oscillator design where the frequency is tunable by changing a single resistor or a source of electronic pots that track within 1 or 2 percent?
There is a technique involving switched capacitor digital filters. Basically you clock the filter at, say, 64 times your intended fundamental (or whatever the multiple is for that particular filter), and feed it with the fundamental to filter. What you get out is a very clean sine wave at the fundamental at a known amplitude.
Have a search for digital filters and sine wave generation.
There was a thread on generating low distortion sine waves. I forget who asked, but their requirement for distortion was was about an order less than mine - I would like .001% - less would be better, but I could (happily) live with .005.
My problem is tuning the sucker. I'm going to go back and try a digital pot (MCP42010). The dual pots don't track as accurately as the data sheets imply because the terminals and wipers are at different potentials and the SPI basically means some form of digital control and noise.
Asking for 10ppm precision sine waves makes things difficult. You are asking for a 17-bit or better D/A converter if you want to do it digitally.
One option is to make a finite impulse response filter for a square wave by feeding your square wave through a long - say 32-bit - shift register, use your 32 outputs to control 32 single pole double throw analog switches, and connect 32 high precison resistors between a summing junction and a positive reference voltage or an - equal - negative reference voltage.
The resistor values follow the sinc function - (sine x)/x - windowed by a raised cosine Hamming window over at least 90 degrees worth of x. I think I stretched it further when I did something similar back in
1978, but I can't remember why.
The lowest value resistor needs to be a 0.01% part, and even the E196 values are too coarse to let you get away with single resistors any place after the first, so you'd have to figure on measuring the actual resistance (ideally including the on resistance of the analog switch) and padding or trimming the nearest E196 value up to the resistance you actually need.
Truncating the length of the finite impulse response filter gives you Gibbs oscillations on the output (whatever they are) and the Hamming window minimises this problem. I found out about it the hard way and had to rebuild my resistor array to get the desired effect.
I still think the switched digital filter approach is best (see other post), i.e. clock a switched digital filter (multi-pole) with a multiple of the fundamental, say 64 times depending on filter used, and feed the square wave fundamental into the filter input. You'll get a clean sine wave out, and should be independent of frequency.
You're basically trying to build your own circa 18-bit DAC from discretes. It's a whale easier to just buy a 20 bit DAC, and drive it from a sine lookup table.
That's not bad. Use a single pot that tunes an oscillator. The oscillator clocks the switched cap filters, and a div-by-64 or some such divider to feed the filter input.
Some cautions:
The filters will alias the square wave input and make small errors. A little analog pre-filtering will help.
Switch-cap filters make output spikes, so you need an analog output filter as well.
Swcap filters are noisy, have rotten PSRR, and have a bit of distortion of their own. They are, in general, kinda junky.
Probably the best way to get low distortion, tunable sines is by feeding data into a good DAC. A DDS chip does all that for you, for a few dollars.
What kind of distortion level are you trying to achieve. I think I get something < -70dB down with a Wein bridge. (It's kinda hard to measure except when I borrow an SRS spectrum analyzer.) We use these dual ganged pots from Bourns. I can look up a part number if that would help. (The Newark number is 62K3333.. the picture on Newarks website is crap.) I don't know how well they match, but I'm not sure it makes that much differernce in the distortion.
There is also a phase shift oscillator that Phil A. turned me on to. This is nice cause you can get both quadratures out, but I could not get the distortion down as low as the Wein bridge.. and it still needs two pots.
Back in the old days one simply used a ganged variable capacitor for the variable element in a Wien bridge. Values of 30-450 pf with accurate tracking, certainly better than 1%, were readily available. Every superhet receiver ( ie virtually every radio ) had one.
Now you need a DDS to do this simple job.
--
Regards,
Adrian Jansen adrianjansen at internode dot on dot net
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.