Re: Final final 1kHz oscillator

On Thu, 14 Nov 2024 11:14:28 -0500, "Edward Rawde"
<invalid@invalid.invalid> wrote:
>

"JM" <sunaecoNoChoppedPork@gmail.com> wrote in message news:t5fajjdteskfftvkf84iqsp2vc4b9ta5kj@4ax.com...

On Fri, 8 Nov 2024 15:43:41 -0500, "Edward Rawde"
<invalid@invalid.invalid> wrote:
>

This is the simplest circuit I can come up with.

>

>
Using an NSL-32SR3 doesn't really gain you anything over a FET.  If
you keep |vds| much less than |Vt-vgs| in the circuit and compensate
for the vds/2 variation on the gate drive the distortion levels
produced will be lower than almost all opamps out there.  Here's a
link to a sim using an LDR.
>
https://1drv.ms/f/c/1af24d72a509cd48/EkjNCaVyTfIggBrrAQAAAAABHXNKs52BoVHi7zaxKmdMMQ?e=oVD0JS
>
I've added some 3rd harmonic distortion to a linear model to give
distortion figures roughly in the ballpark of Fig 6 in the attached
application note (for small signal swings across the device).  In the
oscillator o/p the 3rd is about 150dB down.  If you use the linear LDR
model, or a FET, it will be much the same. (The opamps used in the sim
actually have distortions specs of about 130dB in reallity.)

>
Thanks for that. I left it overnight for 100 seconds of simulation. It looks like it has better harmonic performance than my
circuit.
Why use two op amps in the oscillator instead of one like my circuit?

>
Because of common mode (non)linearity.  In most op-amps (especially
older designs) the distortion generated by the common mode signal is
typically one to two orders of magnitude larger than the transfer
distortion.  If it is not clear from the data sheet, it is best to
measure the CM performance or just assume that it's 40dB worse than
the specified figure.
>
Note that it is seldom modeled in a (free) spice model.
>
This is a link to an extract discussing it's measurement.
>
https://1drv.ms/u/c/1af24d72a509cd48/ESErQT4Elb5FsgcgsOp1J3wBWZSYHvooHblQJPA3DmydbA?e=wwqtSy

>

What happens without the initial conditions? Does it take a long time to settle?
I'll be away until next week.
>

>
No idea.  LTSpice is useless for handling large simulation data files
so I don't bother trying to use it to check for that sort of thing.