2
\$\begingroup\$

Following on from this question. Can I place the capacitor anywhere in the circuit before the IR Distance Sensor? In my project I have 2 PCBs: A power circuit and a Processing Circuit. In the Diagram below you see the two PCBs Power Board and Processing as well as the Sharp distance sensor. Following the diagram, can I place the capacitor on the power board or will it be ineffective by the time it reaches the sensor? I don't want to damage my expensive sensors. 

---------------         12V             ------------------
|             |-------------------------|                |
| Power board |          5V             |  Processing    |
|             |-------------------------|     Board      |
|             |         GND             |                |
|_____________|-------------------------|________________|
                                          |       |     |
                                          |       |     |
                                      GND |   Out |  5V |
                                          |       |     |
                                        ------------------
                                        |    Sensor      |
                                        ------------------
\$\endgroup\$
0

2 Answers 2

Reset to default
4
\$\begingroup\$

See my answer to your previous question here, specifically this image:
this image.

You want the caps to be as close as possible to the components which are drawing current. The voltage varies due to to trace/wire inductance, so it doesn't help to locate the cap before this inductance. The caps absolutely must be on the sensor. You should already have output caps on your power board and decoupling caps near every IC on your 'processing board'.

\$\endgroup\$
3
\$\begingroup\$

All of the above.

There should be bypass caps on the power board at the regulator output, perhaps more than 1 depending on the type of regulator and ripple required.

There should also be bypass caps at the input to every IC, again more than 1 may be required and every power pin should be bypassed. A 0.1uF and a 10nF on each pin is a good general practice. You may also need a 10uF or more on each rail on each board for bulk bypass.

The sensor board likely has bypass caps on the sensor board for the +5V line. If not, add some, 0.1uF is a decent start but you may need some more storage there, 10uF or something to supply the IR pulse current.

Your best bet is to post your entire schematic, especially the interface between the sensor and the "processing board". There are likely things that can be done to improve accuracy in your preprocessing of the sensors output signal before sampling it. For instance, what does your ADC anti-aliasing filter look like?

\$\endgroup\$
8
  • \$\begingroup\$ 10nF? That's new for me - I use .1uF/100nF 0603s everywhere, and parallel two where that's not enough. What kind of clock speeds are you designing for? \$\endgroup\$
    – Kevin Vermeer
    Commented Jan 11, 2011 at 23:49
  • \$\begingroup\$ Often system clocks of 24-25Mhz (digital audio) with USB/Ethernet interfaces (higher clocks). 0.1uF ceramic caps have fairly high impedance over 100Mhz or so and to limit emissions to pass FCC you'd need cover to 250-300Mhz or so. I've had to use 10nF caps even with PIC microcontollers with Ethernet on board though. \$\endgroup\$
    – Mark
    Commented Jan 12, 2011 at 0:05
  • \$\begingroup\$ @Mark, from what I've seen, package inductance pretty much dominates high frequency response, making the capacitor value almost meaningless. I've also seen studies showing the LC resonance from using multiple different cap values leads to overall worse bypassing over the entire frequency range, which is why some EMC guys recommend just sticking to one value, 0.1 or 0.01 all over the board (other than bulk) and that mixing caps (such as 0.01 with 0.1 should actually be avoided) - see Henry Ott's book. But, have you failed EMC testing before, and then passed after changing a cap to a 10nF? \$\endgroup\$
    – bt2
    Commented Jan 12, 2011 at 1:10
  • \$\begingroup\$ @bt2 I've failed FCC testing using just 0.1uF 0603 capacitors and then passed it adding 0402 10nF caps to all IC power pins. This was a PIC18F97J60 which has a current mode ethernet transceiver, same operation at enc28j60 which is known to have EMI issues at ~200Mhz. This particular board also had a DSP and 2 stereo class D amplifiers that were rated at 18W a channel (72W total) among other analog audio ins and outs. You are correct that package is critically important but capacitance is not irrelevant. There would have to be able evidence to discount years of use of decade capacitors. \$\endgroup\$
    – Mark
    Commented Jan 12, 2011 at 1:38
  • \$\begingroup\$ @Mark, interesting, thanks for letting me know. I'm actually working with that exact PIC at the moment, though most likely won't be going through FCC testing. Years past, decade capacitance worked for low frequency design, but in high speed digital designs using it can be like playing Russian roulette. The resonances between the different value caps produce impedance spikes. If some clock harmonics should fall on or near the frequency of these spikes, the power to ground noise can actually increase. Please check out this study, or at least the summary \$\endgroup\$
    – bt2
    Commented Jan 12, 2011 at 3:12

Your Answer

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

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