0
\$\begingroup\$

First off, the terminology I'm using could be complete rubbish - I'm still learning

I have a mixed signal board for measuring bio-signals. The 4 layer board will have a split ground plane, analog ground and main ground (for a microcontroller, SD card, extra sensors, etc), star connected at the power pins through a ferrite bead. Previous versions (which I did not design) have had significant problems with noise.

Should the analog circuitry (mostly op-amps, etc.) have additional decoupling capacitors near the physical chips, in addition to the caps at the outputs of the DC/DC and charge pump that are powering them? similar to the picture below.

Differential amplifier

Assuming that it should have the decoupling caps, what size and why? The DC-DC and charge pump are in the range of 100s of KHz to a few MHz, while the analog signals we are interested in are less than 100Hz, centred around 10Hz

Finally, all componants are Imperial 0805 packages (for convenience, still prototyping). The op-amps are 0PA2277 (additional question: is it better to use OPA4277, with 4 op-amps per package instead of 2, in terms of noise performance?).

\$\endgroup\$
7
  • \$\begingroup\$ If you have a noise problem, you need to figure out where it's coming from before you can try to fix it. \$\endgroup\$
    – Matt Young
    Jul 10, 2015 at 14:52
  • \$\begingroup\$ @MattYoung The primary noise source is 50Hz - Picked up by the origianl board which has a flooded ground as opposed to a dedicated ground plane, and the sensor leads (even though they are shielded). <ost of this is removed by a notch filter. The remaining noise source is the supply rails - being supplied by a switching DC-DC and charge pump, we have observed large quantities of noise (almost up to a volt). \$\endgroup\$
    – Mauvai
    Jul 10, 2015 at 14:56
  • \$\begingroup\$ You should post more information if possible. \$\endgroup\$
    – scld
    Jul 10, 2015 at 15:00
  • \$\begingroup\$ @scld As I said, I'm still learning - what information would be useful? \$\endgroup\$
    – Mauvai
    Jul 10, 2015 at 15:03
  • 1
    \$\begingroup\$ You say bio signals, and my mind goes to the 50Hz noise probably (certainly) coming from any off board leads connected to high impedance nodes. Decoupling is important, but it's not going to help you here, nor is splitting the ground plane. The standard way to get rid of the 50Hz crap is in software with a FIR notch filter. \$\endgroup\$
    – Matt Young
    Jul 10, 2015 at 15:26

1 Answer 1

Reset to default
4
\$\begingroup\$

All active components should have decoupling capacitors. The PCB traces between your part and your power source act like parasitic resistors and inductors, and if you don't decouple your ICs, then when their power requirements change quickly - for instance, when they try to change their output signal in response to something - the changing power requirements will cause voltage drop and overshoot due to the long PCB traces. A nearby decoupling capacitor eliminates that high frequency noise by providing for those short-term spikes locally.

0.1uF capacitors are a reasonable default for decoupling; if your device may have particularly large power draw requirements, you should add a 1uF or larger capacitor in parallel.

Regarding grounding, you will find a lot of conflicting advice on this on the Internet. Split ground planes are often less simple and more problematic than you might suppose, because return currents prefer to flow on the reference plane underneath the signal trace. If any of your traces cross the split in the ground plane, you will force that current to deviate around the split in the ground plane, causing a lot more noise than you might have eliminated by splitting the ground plane in the first place.

This article provides an excellent description of why you should consider using a single ground plane combined with careful routing.

\$\endgroup\$
3
  • \$\begingroup\$ Is it relevant That both power and ground have planes? The Power plane is split even more (four different voltage requirements), but even so, the "traces" for the power are very wide. \$\endgroup\$
    – Mauvai
    Jul 10, 2015 at 14:58
  • \$\begingroup\$ The answer to this very much depends on your full design. Proper power traces with enough bypass capacitors are fine in many cases but it really depends. \$\endgroup\$
    – scld
    Jul 10, 2015 at 14:59
  • 4
    \$\begingroup\$ @Mauvai Doing so helps reduce the impedance - but nothing eliminates the need to place decoupling caps immediately adjacent to all active devices. \$\endgroup\$
    – Nick Johnson
    Jul 10, 2015 at 14:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.