To buffer, or not to buffer?

[DrMag]

I'm always trying to learn more and understand why things are done one way and not another. That sometimes gets me into trouble, because I have a hard time settling for "It works; good enough."

A project I've had in mind requires generating audio-frequency sine waves, and I've explored a number of ways to generate those. One that's caught my interest lately I spotted in a perusal of The Art of Electronics, where they suggest that a good way to generate a low-distortion sine wave from a 555 timer is to use a monolithic active elliptic filter, such as the MAX7426. (I believe AoE suggests a different one; an 8-pole filter that costs more, so I'm playing around with this 5-pole filter for now.)

Obviously, the square wave into the filter needs to be 50% duty cycle to get the low-distortion on the sine wave output. I've seen essentially three different suggestions on how to do that with a 555:

• Bypass the discharge resistor with a diode and compensate for the voltage drop by adjusting the resistor values.
• Use the simpler, one-resistor circuit that automatically has a 50% duty cycle.
• It's impossible to get a true 50% duty cycle with a 555; instead, have it oscillate at twice the frequency and connect the output to a flip-flop. This has the added benefit of buffering the 555 output.

It's that last statement that caught my attention. I'm not sure I know the answer to when you need to buffer a signal and when you don't. Impedance matching is still vague voodoo to me, and plus, as I understand it, an ideal buffer has infinite input impedance and 0 output impedance, which doesn't indicate to me that there's any impedance matching at all.

So here's my question: what is different between connecting the 555 output directly to the filter IC and putting a flip-flop in between them? (aside from the obvious cost, part-count, yada yada...)

And before someone tries to dismiss the question, yes I realize that a 1 khZ wave is unlikely to be significantly affected by any of this, but let's pretend just for a while that we live in a universe where those effects are important. I'm not asking which is the better approach, I want to understand what is different between these approaches and why you would choose to buffer the output with a flip-flop instead of directly driving the filter.

Thanks, everyone!

[ataradov]

The obvious answer - to isolate source from the load. The frequency of 555 oscillator will depend on the load, and you will have to factor that into your calculations.

If you have constant load, then it is easy. If not - buffer makes it constant, and in turn handles the variable load itself.

[helius]

They mean, use a flipflop as a ÷2 divider, by connecting its /Q to its D, and the 555 output to the clock input. Then each time the 555 goes from low to high, the Q output of the FF will change state (alternate). This will always be 50%DF no matter what the 555's DF is.

Whether it is effectively "buffering" the 555 output depends on the current supply capability (output impedance) of the FF in relation to the 555. This has nothing to do with impedance "matching", which is only relevant to transmission lines. The total parallel input impedance of the downstream components should be 1000 times greater than the output impedance of each stage for stability reasons. If you need to fan out a signal to many consumers, it may need to be buffered, depending on the technology being used and how much fanout it supports. A 74-series TTL output has a maximum fanout of 3, for example. Buffering also adds delay to the signal propagation.

[David Hess]

they suggest that a good way to generate a low-distortion sine wave from a 555 timer is to use a monolithic active elliptic filter, such as the MAX7426.

A switched capacitor filter is a great way to generate variable frequency low distortion sine waves.

...

3. It's impossible to get a true 50% duty cycle with a 555; instead, have it oscillate at twice the frequency and connect the output to a flip-flop. This has the added benefit of buffering the 555 output.

It's that last statement that caught my attention. I'm not sure I know the answer to when you need to buffer a signal and when you don't. Impedance matching is still vague voodoo to me, and plus, as I understand it, an ideal buffer has infinite input impedance and 0 output impedance, which doesn't indicate to me that there's any impedance matching at all.

So here's my question: what is different between connecting the 555 output directly to the filter IC and putting a flip-flop in between them? (aside from the obvious cost, part-count, yada yada...)

It is not buffering so much in this case although that may be a consideration but that if you use a CMOS flip-flop, then the output voltages will be very well controlled so the amplitude of the filtered sine wave will be more stable.

[Audioguru]

Instead of the 555 you should use a CD4047 which is a simple RC oscillator plus a divide by 2 divider that produces exactly 50% duty cycle.
My sinewave generator also uses a switched capacitor lowpass filter but its distortion is extremely low because it uses a CD4018 digital counter and a few resistors to make a stepped "sinewave" and the switched capacitor filter simply smoothes the high frequency steps.

[edavid]

It is not buffering so much in this case although that may be a consideration but that if you use a CMOS flip-flop, then the output voltages will be very well controlled so the amplitude of the filtered sine wave will be more stable.

I hope OP was planning on using a CMOS 555.

However, a switched capacitor filter emits a certain amount of clock noise through its input, so a buffer is probably a good idea anyway.

Instead of the 555 you should use a CD4047 which is a simple RC oscillator plus a divide by 2 divider that produces exactly 50% duty cycle.

A 555 has better frequency stability.