I whipped up a few of these, using the layout that gaussmarkov had generously provided, and neither of them worked properly, producing an annoying pronounced flutter+tick when the frequency booster section was activated. After much trial and error, I realized that in gaussmarkov's layout and drawing, the 10uf electro on the ground leg of the input stage (labelled as C3) is shown as going to ground. It should go to Vref to work properly.
...
Really? That's very surprising to me. I have only ever experienced the opposite with that kind of stage. In my experience, connecting the ground leg to Vref can, depending on the impedances involved, create an LFO via the bias resistor on the non-inverting input of the opamp. You feed significant signal right into the bias network, after all. Lowering the values of the bias network resistors from 47k to 10k may help in such cases*. When the ground leg goes to "proper" ground (V- in most cases), the impedance on said ground is always so low that this effect is killed completely (or should be unless there is something wrong with the power supply). Additionally, the cap you are talking about is polar, so it should be happier to have 4.5V of DC forward bias on it than to have none (meaning it would swing into revers bias half the time).
I am very perplexed that you find the exact opposite to be true for the 49er. Maybe all those
inverters invertors make it do the opposite of what I would expect
Cheers,
Andy
*p.s. Fun idea: Why not use some those left-over inverters to make a very low impedance reference voltage? Just connect input and output via a 100k resistor and a 1u ceramic cap and use the output as a 1/2 V+ reference voltage, which can then be buffered by one of the remaining buffers. Should provide a very low impedance Vref, no?
