0
\$\begingroup\$

(Please excuse my English)

I have two types of DC-DC step-down(buck) converter modules, and the switching frequencies of the converter modules are 1.4MHz and 150kHz each. The two converter modules supply 3.3V power to two sensitive sensor boards. I should remove the switching noises of the converter modules.

However I cannot modify the circuits of the converter modules. The converter module circuits are fixed on PCB. So I want to append some circuits to V-out(3.3V) of the converter modules.

What are the simplest circuits for my purpose?

\$\endgroup\$
15
  • 3
    \$\begingroup\$ Welcome to EE.SE! Probably by fixing your layout and circuit first and filters secondary. Please show schematic and layout. Your English is fine by the way. \$\endgroup\$
    – winny
    Commented Jan 9, 2019 at 14:19
  • 2
    \$\begingroup\$ We need more data. e.g. if you 'sensor module' does not require too much current you could replace the converter with an LDO. \$\endgroup\$
    – Oldfart
    Commented Jan 9, 2019 at 14:30
  • 3
    \$\begingroup\$ @pdh0710 - Welcome to the site :-) In the comment above, you said: "My situation is more complicated than expressed here." You need to give us all the relevant constraints. Otherwise you risk wasting our time (and yours) writing answers, which end up being unsuitable due to missing information. You also risk getting downvotes or even the question being closed, if readers believe you are withholding relevant information. Please read the tour and help center for more information about how to use the site. Can you provide some photos of the existing buck converters, and list all constraints? \$\endgroup\$
    – SamGibson
    Commented Jan 9, 2019 at 14:53
  • 1
    \$\begingroup\$ It's not difficult to understand. You still need to show schematic and layout if you want any specific help. What you have going for you is that 1.4 MHz is faily low so differential filters do stand a chance here. \$\endgroup\$
    – winny
    Commented Jan 9, 2019 at 15:18
  • 2
    \$\begingroup\$ Without quantitative details, the only answer that can be given is to "add a filter." Maybe an RC filter or maybe an LC filter. If you can tell us how much ripple you have on the 3.3V rail and how much ripple you can tolerate, we can give you some ideas about what kind of filter may be required. Also, I am assuming that your sensitive sensors consume a steady current. But if that is not the case, if there is ripple in the load current, that needs to be specified also. An extreme filter solution would be to boost up to 4V or 4.5V and then use an LDO down to 3.3V. \$\endgroup\$
    – user57037
    Commented Jan 9, 2019 at 16:42

2 Answers 2

4
\$\begingroup\$

1) All DCDC converters will produce some ripple (switching noise) at their output. It is unavoidable and inherent to the way switched DCDC converters work.

2) Before you continue, read and understand 1) and if you understand 1) correctly then your conclusion should be that you will not be able to completely remove the switching noise. So you will have to determine how much switching noise is acceptable. So how much lower does the ripple need to be, for example 10 dB lower at 150 kHz or 20 dB or 40 dB? Note that if you say: "But of course I want 40 dB!" then I'll say, OK, that will cost more (more complex filter, more expensive components) etc. but also: Would you like 80 dB suppression? And if you then say "yes please" we will go on and on and your design will be infinitely expensive and have way too much suppression. So figure out what is really needed. Yes, that might not be easy.

3) Once you know how much switching noise is acceptable you could choose to filter the output voltage of the DCDC converter. A filter like this could do the job:

enter image description here

Another option would be to use an LDO (linear regulator) for supply voltage regulation. An example: a DCDC converter is used to convert 12 V to 5 V, that 5 V with the switching noise is then regulated by an LDO down to a far less noisy 3.3 V.

\$\endgroup\$
9
  • \$\begingroup\$ O.K. I understand 1) and 2). I'll try the L-C filters as you recommended in 3). (As mentioned in the comments above, I cannot change the converters) \$\endgroup\$
    – pdh0710
    Commented Jan 9, 2019 at 14:44
  • 1
    \$\begingroup\$ @Bimp Unless you know the X(f)/R ratio. this will amplify the noise \$\endgroup\$ Commented Jan 9, 2019 at 15:47
  • \$\begingroup\$ if 3 people approve of this answer , they could be making a big mistake. \$\endgroup\$ Commented Jan 10, 2019 at 2:44
  • \$\begingroup\$ Do an error-budget, for the signal-noise-ratio you need. If you cannot generate a useful error-budget, then ask an experienced person. \$\endgroup\$ Commented Jan 10, 2019 at 4:26
  • \$\begingroup\$ @SunnyskyguyEE75 I think the switching noise may be amplified or attenuated depending on the values of L and C. Before I posted the above question, I tried a simple L-C filter without any calculation. Fortunately, it slightly attenuated the 1.4MHz switching noise (10mV to 8~9mV). Of course, this result was not satisfactory. So I posted the above question. \$\endgroup\$
    – pdh0710
    Commented Jan 10, 2019 at 14:45
2
\$\begingroup\$

Another way is to specify the load current Min and max then the low frequency ripple peak min,max Rs=ΔV/ΔI and choose the series drop Rs ( Load regulation )

  • then select LC values for the notch filter. where f error contributes to variation in attenuation with Q and external impedances matter!!

enter image description here enter image description here If you search my answers for RLC impedance nomograph you can chose these values on a graph. enter image description here

\$\endgroup\$
2
  • \$\begingroup\$ Hmm... This is not simple circuit I expected. However, I think this circuit will work better for the sensitive sensor boards. I will try this circuit. Thank you. \$\endgroup\$
    – pdh0710
    Commented Jan 9, 2019 at 15:54
  • \$\begingroup\$ Oh, thank you very much for your various filter circuits. They are very helpful to me. \$\endgroup\$
    – pdh0710
    Commented Jan 9, 2019 at 23:49

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.