A variation on my current mirror low distortion sine wave oscillator - 10dB less distortion and much the same number of components

The 2N38906 has 10pF of input capacitance and 4.5pF of output capacitance. The resistor introduces about 1usec of lag, which degrades the high frequency performance. In a 1kHz oscillator this isn't going to worry anybody, and the LT1013 is slow enough that it won't matter - C9 kills any risk there - but the resistor clearly isn't doing anything useful, so one has to wonder why you bothered to add it.

Which is irrelevant for this circuit.

So why bother pointing it out?

In any real circuit I would generally not connect a low impedance output from an op amp directly to the base of a transistor, but this doesn't mean that there aren't cases where it's perfectly fine or desirable to do so. In this case it doesn't matter, so why bother pointing out that it doesn't matter?

You can also argue that R7 isn't needed, but in any real circuit I would include both resistors. I can always put 0 ohm in.

That is the question that matters.

You've been complaining that my circuits include too many components, even though each one of the serves a purpose. You should expect me to complain when one of your circuits includes a useless component which degrades it's performance even it it is only a very minor degradation.

I automatically put a resistor in series with the gate of a power MOSFET. Only some of them need it to kill high-frequency oscillation with the parts originally selected, but purchasing does like to swap in cheaper parts over the life of a product. This isn't a parallel situation.

But it doesn't matter to anyone else Bill.

So add another hundred ferrites and claim that each one serves a purpose if you want. I don't mind. You're still only going to get 60dB down in LTSpice 24.1.2

I do too. Then I can put any value resistor in place from 0 ohm to infinity ohm. Just like the resistor I put between U6 and Q1. In any case this resistor doesn't actually exist. As resistors go it's about as real as Ceci n'est pas une pipe. It only exists in the limited mathematical imagination of the computer in front of me. So getting worked up over its exact value isn't very productive, don't you agree?

What makes you think that? You may find it a comforting thought, but it strikes me a self-serving delusion.

But you still complain about it. The ferrites do serve a purpose, even if you can't see the point.

So LTSpice 17 and LTSpice 24 give different results - not a good reason fro trusting either of them. If you want to make a fuss about harmonic levels you have to measure them in a real circuit, which is expensive and time-consuming. John May has done it - neither of us have.

But the circuit won't work if the resistance is too high. And production want to buying and fit a single resistor value. Select on test resistors aren't popular - Cambridge Instruments used them from time to time, and production kept on proposing to use a fixed resistor. We were buying parts in six month chunks, and for that six months production always fitted the same resistor (and got bored). A new batch of parts would need a different resistor.

Far from it. You didn't know what it was doing, and didn't realise that you didn't need it.

And it's cheaper and quicker to model a circuit that it is to build and test it. Within the limits of the model it's quite useful, but the game to get to circuit that you can build, which is likely to work. No real circuit? No game.

If you are silly enough to think like that this isn't going to be a productive discussion, and your part of it hasn't been for quite a while.

It strikes me as an obvious fact. I'm not expecting anyone else to offer any comment but it's interesting that they haven't. And it's not like this group is only for the discussion of electronic matters, as the "Cracking Speech by JDV" thread shows. You seem to be enjoying yourself there.

I wouldn't call it complaining. Just pointing out the obvious.

Ok let me simulate your mode of response. This is just a simulation, it doesn't mean you actually said this. Here goes.

Why have you included useless ferrites in your circuit? Even a five year old should be able to see that a simulation with and without all nine ferrites produces exactly the same harmonic distortion result. (-57.5dB at 2kHz in LTSPice 24.1.2) You obviously don't know what those ferrites are doing, and didn't realise that you didn't need them.

Who cares Bill? I just put 1 Meg in there and ran a simulation. The result it exactly the same. I'm not suggesting I'd actually use 1 Meg in practice.

This circuit has nothing to do with production. It exists only in the mind of a few people and the memory of a few computers.

"From 1960 the company started to decline and struggled to turn a profit." Hmm

Now you're starting to border on telling blatant lies Bill. For example, in the circuit here (which turned up in a search engine search).

Why do you enjoy telling other people that they don't know what they're doing Bill?

I think the circuit JM posted is likely to work. Two BCM61B devices would probably be fine for the current mirror. A single LT1679 can be used for U1,5,6,8 with a cheaper device for the rest. And C6 can be made from two polarized capacitors. Q5, R25, R7 and C9 can be removed but keeping R7, C9 and a base series resistor for Q1 does no harm. The simulation result in LTSpice 24.1.2 is the same, about -142dB at 2kHz.

Do you have any productive comments yourself?

Are you done coming across as a grumpy old headmaster talking to a student who didn't do their homework? My approach to such a situation would be to offer the student help and encouragement. But it appears that yours would be along the lines of "You obviously don't know what you're doing you stupid kid" and "What you should have done is..."

Congratulations. You've said something sensible for once.

This is a less defensible observation.

U5,U6 and U8 are the phase shift oscillator. U1 should a cheap device that is part of the network that generates the amplitude feedback signal fed into the integrator wrapped around U7 to generate the gain control signal that modulate the level of the feedback sine wave that adjusts the output amplitude.

Sadly, you can't buy a pair of 940uF polarised capacitors. Two 1000uf polarised capacitors would be quite close enough

I(R17) runs at about 100uA and I(R21) runs at about 145uA -that is about

Putting 100uF polarised caps in parallel to R17 and R19 delivers harmonics about -135dB below the fundamental rather more cheaply, and won't upset people who get nervous about polarised caps.

As we know, LTSpice isn't all that credible when it predicts very low harmonic content, so 470uF is perhaps a a bit much.

That has been my approach, and it is what I mostly post. Some students don't notice that they are being offered help.

There's little point in being diplomatic with poor students. Quite a few of them are resistant to the idea that they've got stuff wrong, and being diplomatic lets them skate around the negative message.

I've have spelled this out quite explicitly from time to time over the twenty years I've been posting here.

Bill Sloman, Sydney

...

Why do you like to talk down to other people Bill? It doesn't make you look superior, far from it.

After more simulation I agree that all devices may as well be LT1679. Two quad op amp packages. This produces the lowest possible distortion in simulation.

There have been two polarised 1000uf capacitors on my schematic for the last few days.

Who said anything about poor students Bill? Talking to adults as thought they are children may cause issues for you.

I've been reading and posting in this and other groups for way more than 20 years.

Just rechecked this before deleting the simulation files. H2/H3 are

-157/-177 dBc. A hanning window is used to minimise spectral leakage. I usually add a pilot tone at -120dBVrms to the output signal to be analysed (at a non harmonic frequency) as a sanity check.