Higher current voltage doubler

[Mick Bailey]

Well, my boost converter didn't actually work out when installed in a pedal. I installed it in a TC Electronics stereo chorus/flanger as a testbed and it powered up just fine, but I though I had a fault with the compander or BBD device. The pedal was super-silent with nothing being played, but when I played anything there was a load of hash, clock noise and distortion that accompanied the notes. After lengthy experiments I disconnected the converter and used a regular 18v DC supply with no problem. When I reconnected the converter and scoped the pedal there was 500-odd khz all over the place, so this will be interfering with the BBD clock signal and causing all kinds of intermodulation. Even the leads from the converter have 12mV of RF on them.

Damn.

Anyhow, I'll see what some ferrite beads and a screening can will achieve.

[darkfenriz]

Anyhow, I'll see what some ferrite beads and a screening can will achieve.

Wrong wrong wrong. Solve the problem where it happens.

1. The boost inductor should be EE, EI or toroid, unshielded drum is not recomended, but usually not the biggest source of problems.
2. The boost diode must be hyperfast or schottky. Equally important is the small package and minimized leadout area. SMD (such as DO-214AA or DO-214AC) is best, upright mounted with huge lleadout loop is no-no.
3. Input and output caps must not be small aluminium electrolytics. Tantalums, X7R, Y5V are acceptable, bigger low-ESR aluminium electrolytics might be acceptable.
4. Short and simple ground path is required, preferable a plane.

Screening cans and ferrite beads are helpful at dozens of MHz, your problems are at 100's of kHz.

Disclaimer: not being dismissive, rather helpful

[Mick Bailey]

The caps (low-ESR Panansonic intended for SMPS) and diode are as per application notes. Inductor is a standard drum. You can see my layout in post #9. By comparison I have a rather crude DIY boost converter built from discrete components that's in a strung-out layout and no noise.

I re-read the documentation and it says "A core geometry like a rod or barrel is prone to generating high magnetic field radiation" then goes on to say "To reduce the noise generated by the inductor, insert a 1.0 uF ceramic capacitor between VCC and ground as close as possible to the chip. I don't have that capacitor but will see if that makes any difference.

[Mick Bailey]

Here's a follow-up;

The original layout turned out to be fine. The suggested cap didn't make any difference, but I needed a 200pf cap between pin 1 and ground and this took away most of the noise. I then subbed in a toroidal inductor and this fixed the rest. I would redesign the layout anyhow now I know more about how these things operate. There are two grounds - one for the switching side and another 'clean' ground which should be used to tie the divider resistors. The application notes shows these tied together in the schematic, but the narrative suggests separating them.

[darkfenriz]

Great job, these little animals are so unforgiving re components and layout, you never know what trick makes them behave.

[MarkusBass]

The original layout turned out to be fine. The suggested cap didn't make any difference, but I needed a 200pf cap between pin 1 and ground and this took away most of the noise. I then subbed in a toroidal inductor and this fixed the rest. I would redesign the layout anyhow now I know more about how these things operate. There are two grounds - one for the switching side and another 'clean' ground which should be used to tie the divider resistors. The application notes shows these tied together in the schematic, but the narrative suggests separating them.

The grounds should be separated but finally at some point they have to be connected.
The datasheet of CS5173 explains the analog ground:
"Analog ground. This pin provides a clean ground for the controller circuitry and should not be in the path of large currents. The output voltage sensing resistors should be connected to this ground pin. This pin is connected to the IC substrate."
and provides guidelines for the layout:
"Separate the low current signal grounds from the power grounds. Use single point grounding or ground plane construction for the best results"
It means that that the power ground (high current) should be separated from the analog ground (low current) but finally they should be connected. I hope that you did it in this way. For the power ground it is suggested to use a ground plane, which suggests that better layout could be achieved with SMD version of the circuit.

Mark

[Mick Bailey]

Thanks for the clarification - I haven't yet done an SMD layout. The grounding reminds me of automotive ignition circuits where there's a heavy ground current when switching the coil primary that can interfere with the microprocessor due to 'ground bounce' if not properly designed.

I've seen the artwork for the development board so this should provide a reference layout.