Re: Continuously variable gain amplifier for a low distortion 1kHz Wein bridge sine wave generator.

On 25/01/2025 4:08 am, john larkin wrote:

On Fri, 24 Jan 2025 08:06:20 -0500, legg <legg@nospam.magma.ca> wrote:
 

On Thu, 23 Jan 2025 23:34:54 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:
>

We've been messing about using a FET as a variable resistor to try to
control the amplitude of a 1kHz Wein bridge sine wave oscillator for
months now.
>
It works, but it does introduce some harmonic content into the sine wave.
>
A good four quadrant analog multiplier can do a better job, but the
AD734 isn't cheap. An asymmetric current mirror can do the job more
cheaply but with even more components, and seems to introduce even more
distortion - not all that much, but enough so that it isn't a good choice.
>
All we need is a controllable gain element that can adjust the gain
around the Wein bridge to sustain oscillation at a constant amplitude
despite component value drift with time and temperature.
>
Linear Technology and Burr-Brown both used to sell amplifiers where you
could vary the gain continuously with a control voltage - I used both
together in one project - the expensive Burr-Brown part managed the
signal gain part, and the cheaper and slower Linear Technology part
managed the DC offset feedback path.
>
The AD8330/1/2/6 parts all seem to do much the same job, as does the
AD603. None of them are cheap, and the are all a lot faster than the job
requires. Anybody know of anything more suitable?

>
Fiddled with RC4200 at one time. Not sure where you could
buy them nowadays. Digikey hands you off to Rochester, which
seems to handle 'off-market' or old-stock type sources.
>
RL

 A mosfet run ohmic, with millivolts of swing, should be very linear.

Why do you think that?

Two fets antiparallel would be better.

But you have to provide two separate bias voltages, and since both FETs are doing  the same job there no obvious way of working out the two separate bias voltages. Edward Rawle was posting examples a few months ago, but he didn't have any kind of scheme for geting two different bias voltages out of a single output amplitude.

The trick for low distortion is to give the gain control element very
little influence.

You can buy 0.1% resistors easily enough, but getting better than 1% tolerance capacitors involves buying a lot of them and sorting them into well-matched pairs. Trimming potentiometers can take you a long way, but it is strictly a small volume production solution, and it isn't cheap.

A dual-integrator oscillator can add a free -12 dB/octave harmonic
attenuation.

So post a simulation.

The opamps should be the distortion limit, and I suspect that Spice
models that badly.

Different Spice op amp models model it more or less precisely. Some of them are computationally expensive, and the manufacturers of the op amps (who supply most of the models) don't have a lot of motivation to make the models all that accurate. Jim Thompson used to bitch that while he could produce pretty accurate behavioural models of op amps, hardly anybody wanted to pay him to do it for them.
Your willingness to suspect that all Spice models are inaccurate comes from your wish to be seen as some kind of expert without having to bother to put in the work it takes to look even moderately expert.
Looking expert here mainly depends on not carrying on about stuff you don't know much about.
--
Bill Sloman, Sydney