A question about charge pumps and pcb design

[rockhorst]

I've got a few pcb designs, like a modded blues breaker style circuit, that have a charge pump on the PCB to supply 18V (either 0 - 18 or split rail -9V to 9V). I've got a very large stash of ICL 7660CPAZ chips that I'd prefer to use. I've noticed that sometimes I get a lot of 'noise' especially when the guitar is not being played. Sort of mid/high pitched pulsating sound. It depends on location and the power supply being used. A 1044 chip sometimes reduces the problem to useable levels. I've noticed that I can usually circumvent issues by building the charge pump on a separate pcb stuck to the inside of the enclosure on a 90 degree angle to the main board (no clue if rotating the board makes a difference).

Anyway, I'm not describing anything new, these are known problems. I'd like to take the 'voodoo' out of it. What would be best/good design practices to have the charge pump integrated on the main board without induced noise? Should I feed a status led from the 9V or 18V tab, use a ground plain or remove the ground plain under the IC so noise can't be induced on it? Stuff like that. All suggestions much appreciated.

[Derringer]

that charge pump may not be the best for audio application

from the data sheet
"The oscillator, when unloaded, oscillates at a nominal
frequency of 10kHz for an input supply voltage of 5.0V. This
frequency can be lowered by the addition of an external
capacitor to the "OSC" terminal, or the oscillator may be
overdriven by an external clock"

10kHz is well within the audible range.
Chips like the 1044 (when using the boost function) and Lt1054 oscillate at a frequency above our hearing range.
I don't see, in the data sheets I just perused, a boost function for the ICL7660

[rockhorst]

My bad, I should've stated I use ICL 7660 SCPAZ chips, which do have the boost function.

[thomasha]

I used the ICL7660s with boost function in the murder one amp designed by frequency central and the noise was never an issue.

Check your layout, and keep the traces as short as possible, and far from the audio circuit. Anything that could work out as an antenna is a bad idea. Using a separate ground also helps.

[Kipper4]

I might be barking up the wrong tree here but you know me dive in with both feet.....

Is'nt there a pin that you hang a cap off to reduce oscillation? in the 7660

[amz-fx]

You don't have to use a ground plane. Taking the grounds from the charge pump directly to the power jack, separately from audio grounds, will probably work better. Grounding is odd and not every solution works with specific pc boards, but making a star ground with the audio and charge pump grounds has the best chance for low noise.

Increase the value of the capacitors used on the charge pump. I see 1uF sometimes used and that value is too low. You can also add a 0.01 to 0.1uF ceramic paralleling the electrolytics on the voltage output pins to improve filtering at high frequencies.

The ICL7660CPAZ does not have the boost function on pin1 that allows the internal oscillator to operate at a frequency above the audio range. A chip that can run the oscillator higher will likely work better for you.

Best regards, Jack

[rockhorst]

Hi Jack, thanks for the input. The 7660 SCPAZ (note the S) does have the boost function, that is not the problem here. I've designed boards with no issues, and boards that 'whine' or 'pulse' and I'm having a hard time figuring out what it did right on one design but wrong on another. What I called 'voodoo' in the first post.

[McCamish]

I have built many charge pumps using 7660s, 1044s, and 1054s. From that experience,  I now use only LM1054 because of the inaudible oscillation frequency without boosting and it's higher input voltage threshold. Many of the charge pumps I build are dual function and generate both -9v and +18v from a 9v source.  When using opamp I can use both + and - voltages with a 4.5v bias and have about a 26v swing. With the availability of +18,+9,+4.5 bias, and -9v the design options are greatly increased. For example, you could use +18 to feed a FET and -9 for a pnp germanium transistor in the same circuit, as well as a opamp with up to 26v swing. Tips from my experience are to parallel the electrolytic output caps with ceramics and tie all grounds to pin 3, then to the dc jack. I also tend to make the 9v input filter cap the largest. It is ok to be the same size as others, but not smaller. I also find that using schottky diodes for voltage doubling is much more efficient. I hope this makes sense and helps.

[sajy_ho]

I have built many charge pumps using 7660s, 1044s, and 1054s. From that experience,  I now use only LM1054 because of the inaudible oscillation frequency without boosting and it's higher input voltage threshold. Many of the charge pumps I build are dual function and generate both -9v and +18v from a 9v source.  When using opamp I can use both + and - voltages with a 4.5v bias and have about a 26v swing. With the availability of +18,+9,+4.5 bias, and -9v the design options are greatly increased. For example, you could use +18 to feed a FET and -9 for a pnp germanium transistor in the same circuit, as well as a opamp with up to 26v swing. Tips from my experience are to parallel the electrolytic output caps with ceramics and tie all grounds to pin 3, then to the dc jack. I also tend to make the 9v input filter cap the largest. It is ok to be the same size as others, but not smaller. I also find that using schottky diodes for voltage doubling is much more efficient. I hope this makes sense and helps.

+1 to 1054

[alanp]

General layout tidiness -- keep noisy clock stuff (this includes charge pumps, as well as LFOs and clocks for BBDs) away from analog parts of the circuit, especially any part that has gain on tap.

[KarenColumbo]

Sorry for intruding here! I hope it's okay to ask for a schematic (or a wink in the right direction) for making +/- 9V out of a +9V wall bug supply using a LT1054CN8. Can you help me here?

[rockhorst]

@KarenColumbo: Madbean's Road Rage will help you with that. Scroll down to 'other' here

[KarenColumbo]

@KarenColumbo: Madbean's Road Rage will help you with that. Scroll down to 'other' here

Exactly what i was looking for. Many thanks!

[lars-musik]

Whenever I had problems with charge pump noise it was due to digital pedals in the chain or pedals with excessive power consumption in a daisy chain. However, I also have a huge stock of 7660S but usually resort to LT1054 whenever even remotely necessary.

[rankot]

Hi Jack, thanks for the input. The 7660 SCPAZ (note the S) does have the boost function, that is not the problem here. I've designed boards with no issues, and boards that 'whine' or 'pulse' and I'm having a hard time figuring out what it did right on one design but wrong on another. What I called 'voodoo' in the first post.

The best you can do is to compare boards that you made and figure out what is common for those having or not having noise.

[rockhorst]

I've reworked one of the PCBs that had issues. I usually have a big +V plane on top and a ground plane at the bottom. In most circuits, the top plane doesn't do that much so I removed it in the rework. I'm thinking that's less chance of the charge pump inducing a signal on it. I'm in doubt about the ground plane: should it cover the entire board or would the odds favor notching out a recess in the ground plane under the charge pump (which is easily done). Here's a picture illustrating that: