Charge pump question

[Mark Hammer]

Some of the charge pump chips, like the 1044, have the capacity to double the clock frequency so as to minimize the audible ripple on the supply lines.  Bridge two pins and there you go.

Having bought a bunch of plain vanilla, 7660 chips, though, I worry about the audibility of such ripple when the chip does NOT have that capability.  So, four questions:

1) Is the ripple audible as a whine in the background that a higher gain circuit might make VERY audible?

2) Is it possible to decrease the ripple by simply upping the capacitance at each stage in the charge accumulation?

3) Is any potential ripple really just the sort of thing one would normally address with a 47-220uf cap on the power lines for the typical V+?

4) When using the same chip as an invertor to make +/-9v from +9, same sort of approach to ripple?

[R.G.]

1) Is the ripple audible as a whine in the background that a higher gain circuit might make VERY audible?

It depends on the circuit. Most opamps have a high rejection of power supply noise on their power pins. If the noise leaks in on the signal/ground connections, they amplify it as though it was signal, which it is. Single ended circuits like simple transistor and JFET amplifiers have essentially zero power supply rejection, so the power supply has to be made about as clean as you want the quiet parts of your signal to be.

Yes, it's audible as a whine in the background.

2) Is it possible to decrease the ripple by simply upping the capacitance at each stage in the charge accumulation?

Not really. The bucket caps only need to fill nearly full by the end of the pump's clock cycle. More is heavily decreasing return on capacitance. The output filter can be increased a lot, though.

For a capacitor, Q  = C*V, Q being the internal charge. Each bucket of charge dQ raises the capacitor voltage by dV=dQ/C, so bigger C means littler dV (i.e. ripple voltage) per bucket of charge dumped in. Raising the size of the bucket caps increases dQ and makes dV bigger per step.

3) Is any potential ripple really just the sort of thing one would normally address with a 47-220uf cap on the power lines for the typical V+?

Mostly. A small resistor from the bucket cap to a second, bigger filter cap would help, as would keeping the whole charge pump/pump filter setup from contacting ground except at the negative terminal of the second filter.

4) When using the same chip as an invertor to make +/-9v from +9, same sort of approach to ripple?

Yes, except that this now forces you to connect the same ground wires to the charge pump that you did to the output voltages, and the chances of polluting your ground with pump noise goes up.

[armdnrdy]

Yes
No
No
Yes

A charge pump is also known as a switched capacitor DC to DC convertor. The "whine" that we've been toiling with is the frequency of the switching...but it is in the power supply proper.

I have had good "luck" with the LT1054. It has the capability to change the switching frequency with a small ceramic cap.

One thing that I've noticed in the data sheets is the use of 10µf filter caps. That is not very realistic for the best ripple filters used in our "audio" applications.

I use as big a cap as space will allow, ranging up to 220µf and 470µf.

Edit: It looks like R.G. and I both pulled the trigger at the same time.

[puretube]

the "whine" most of the time is heterodyning of the chargepump`s frequency against "an other" frequency, somewhere in the circuit/pedal surroundings...

[Mark Hammer]

In which case, would it be wiser (although a little more complicated) to sidestep the 7660 entirely and go with a 4049-based charge pump, where the clock frequency can be dictated by the user, such as in this circuit? http://paia.com/prodimages/siabsch.pdf

[armdnrdy]

What type of power are you looking for?

Bipolar? Boost?

[Mark Hammer]

Good question. I guess a bit from column A and column B.

[R.G.]

the "whine" most of the time is heterodyning of the chargepump`s frequency against "an other" frequency, somewhere in the circuit/pedal surroundings...

... if there is such an "other" frequency, and if the fundamental of the charge pump itself is above the audio range...

If there's not such a signal in the circuit, as in a purely analog amplifier, there's no heterodyning to be had. Well, I guess the clock frequency heterodyning against DC counts, maybe, but that's kind of sophistry.

I suppose I would say that there is another source of such background whining - one is the base frequency of the charge pump, another is the sum/difference frequency generated if there is another regular signal to beat against. That's one way to get a difference frequency within the audio range if neither of the other two is.

Connecting up some second circuit that pulls its own regular(ish) pulses from the charge pump power supply, like a digital circuit approximating analog, can  cause a heterodyne. But bad wiring and layout can do it all by itself.

[lowbrow]

Stick to the 7660S chips, the S denoting you can jump pins one and two and double the frequency out of the audible range. You can buy bulk lots of 50 7660S chips from China on eBay for about $14. I have found them to be mostly good...maybe 10% of them won't produce the rated voltage. But at that price, I'm perfectly ok with that, I just socket them and if its not up to snuff I replace it.

[PRR]

> Most opamps have a high rejection of power supply noise on their power pins.

Many have poor *supersonic* rejection on one rail. The compensation capacitor throws "all" the rail crap right into the main gain stage.

http://www.ti.com/lit/ds/symlink/lm308-n.pdf  page 5, figure 8. At an extreme, 500KHz on the rail is *amplified* to the output. Not much above the audio band, PSRR drops below 40dB (100:1).

Opamp rails clean-up moderately easy. 100 ohms and 10uFd adds many dB rejection.

As in all HF hassles, layout and trash-current-paths are critical, and often non-intuitive.

[merlinb]

You can buy bulk lots of 50 7660S chips from China on eBay for about $14.

You can get the TC1044 for free from Microchip samples!