"Taming" charge pumps

[Mark Hammer]

Prompted by member SNK's battle with noise from a charge-pump bipolar supply, I thought I'd ask folks here: in your experience, what are the most effective approaches to keeping clock noise (because all charge pumps employ a HF clock to do what they do) to an audible minimum?

[R.G.]

I use a local "bucket" capacitor of at least 10x the pump capacitors, a 10nF to 100nF ceramic in parallel; a 10R to 100R series resistor depending on the current (and hence the voltage drop in the resistors); scrupulous ground routing so the charge pump ground connects to the rest of the ground only at the negative terminal of the board decoupling bucket cap, and ground plane under the charge pump chip and its local decoupling caps. The various parts attached to the pump chip I put as close to the chip itself as possible, on its local ground plane.

[taudio]

If you can tolerate using a higher voltage supply, along with bias networks and input/output decoupling capacitors (instead of an inverting charge pump), you could try a boost converter.

I've had good luck using a switched-mode boost converter operating at a much higher frequency, instead of using a charge pump (which will typically operate at 45 kHz or so).

Specifically, I've used an AP3012 followed by a linear regulator. The combination of a much higher switching frequency (1.5 MHz), combined with a low noise linear regulator provided low output noise (assuming good layout practices) and operation over a wide range of input voltages.

Here's the datasheet for the AP3012: https://www.diodes.com/assets/Datasheets/AP3012.pdf

As an added bonus, the AP3012 has a maximum input voltage of 20 volts -- much higher then a TC1044S (12 volts maximum input).

If you haven't designed a boost converter before, Phil's Lab has a great tutorial video:

[Rob Strand]

Here's a kind of god view layout.   (All RG's comments valid and can be appended.)


An interesting variant came up in this thread.  The same principle applies.
The take home message is throwing ground planes at the problem doesn't
help.  Keeping current pulses local and off the grounds/ground planes does.
https://www.diystompboxes.com/smfforum/index.php?topic=130597.msg1266753#msg1266753

There was another recent thread but I can't remember exactly when it was.  IIRC
the posted followed the recipe and had no noise.

[R.G.]

Yeah, I'm a great believer in ground planes not necessarily fixing things. I did throw a local plane under the chip and associated caps as a king of Hail Mary move, trying to keep the current resolution paths local, but running the ground separately to the power source on the board without any overlying planes probably did the actual work. Any trace inductance helped to keep the noise currents local to the decoupling cap for the chip itself.

Maybe. 8-)

[Phend]

Sounds to me that these chips need TLC.
Would this layout on a solderable BB work or would probably be noisy ? (PNP Fuzz Face)
J = Jumper 
0.0022uF is film

Thank you for any comments..

[snk]

Oh!
I just noticed this thread!
Thank you Mark for starting it !
A lot of interesting stuff in there, I need to take the time to read it carefully.

From my simple user experience, I have built several circuits using chargepumps, and they all performed well so far.... until recently when I used a couple pedals on another system (the instrument being an electric piano going to an amp, the pedals being powered by an Asmuse AP01. I must also add that this (cheap) power device also worked flawlessly for 2 years with all my gear (NOT using any chargepumps).

So, it seems that sometimes it can be a bit hit & miss

[drdn0]

I've used quite a few charge pumps and have never had an issue with noise (touch wood), and there's nothing TOO crazy that I've done that I think is above and beyond 'best practice'.

• always use a LT1054 instead of MAX1044/7660
• big old cap (100-470uf) right at the DC input, small series resistor and decoupled with MLCC
• charge pump stays as close to the DC input as possible, with as much copper above/below it as I can get connected just at the DC input
• pump caps as close to the IC as possible with fat tracks
• decouple every IC on both positive and negative rails

My place is GENERALLY fairly noisy as I live at the base of a radio tower, so every single pedal has a 10k/100pf filter right at the input.

So far, so good!

[mzy12]

What about inductors? Couldn't you make an CLC filter like a lot of switching power supplies do to create a cut-off filter a decade below the switching frequency of the charge pump?

[PRR]

Couldn't you

Of course. But the usual market for charge-pumps does not want to deal with chokes. It's a cultural thing, or how far your education got. Also chokes "can" radiate the crap they were asked to block. And they have funny numbers (what is a nanohenry? how many zeros?). And the current rating is sometimes pretty fussy. And the bandwidth is not infinite (all chokes become capacitors at some high frequency).

But your question IS quite appropriate for Mark's point.

[mzy12]

What about something like the AP3015? https://www.diodes.com/assets/Datasheets/AP3015_A.pdf

It won't do negative voltage like a charge pump, but I don't see much need for negative voltage in most pedal builds. Stuff like PNP fuzzes can be built with a power supply 'flipped upside down', and if you need a 18Vpp headroom on an opamp or transistor stage, the only benefit I can think of for a dual rail power supply is the lack of the current needed for a voltage divider and the voltage noise that comes with it, the former being generally negligible and the latter easily made redundant by the noisy charge pump.