I never said they were a mighty helper for all problems, but we are talking filtering a SMPS module... cant understand your reply or the generic comment, I am just pointing out what would be a better and recommended solution.
Its what is recommended or a polymer, SMD package low parasitic inductance low ESR an electrolytic is the wrong type... Murata recommend their GRM42 series.
Its what is recommended or a polymer, SMD package low parasitic inductance low ESR an electrolytic is the wrong type... Murata recommend their GRM42 series.
Can't find a statement in the DS why exactly 10uF. You mean a 10uF Type or a 10uF at desired voltage? (Dc bias) Case size, ESL/ESR? Without knowing what to dampen, it's just general like "put 100nF" right near an IC.
DS says "High efficiency with no external components"..
So without exactly knowing what noise is coupling into the system it's just try n error.
My 2c.
DS says "High efficiency with no external components"..
So without exactly knowing what noise is coupling into the system it's just try n error.
My 2c.
Last edited:
No it isn't try n error its a SMPS output.... the noise will be high frequency due to the nature of the beast, I do plent of layouts with SMPSs such as this and worked with many TI?National?linear etc application engineers on these layouts due to the equipment they are going in. And if they say they need a low ESR MLCC or polymer it is because they have done extensive testing of the circuit and know what is best, so follow the recommendation.
If you read the data sheet you will find what you need to know...
http://power.murata.com/data/power/oki-78sr.pdf
This in particular caught my attention...
If you read the data sheet you will find what you need to know...
http://power.murata.com/data/power/oki-78sr.pdf
This in particular caught my attention...
The reason to not use electrolytics for SMPS filtering is the usable bandwidth is only 10s of kHz, while MLCCs or polymers can be into the 10s of MHz. Smaller caps will have less ESR and ESL, so paralleling them can increase capacitance at higher frequencies. The reason for recommending SMD components is leads add inductance and resistance reducing the capacitor's effective bandwidth.
Noise != ripple.
I give it up. With no measurements it can't be fixed the "analytical" way. But if it helps..
I'd first remove the AGND connection, limit the bandwith on the analog connection and further put ferrites within the lines.
Sorry to revive this thread, but I'm facing the exact same problem: I have a tpa3116 powered by a 18V, a raspberry powered by a step-down SMPS, and a noisy ground loop through the audio channel in the form of the classic digital switching noise.
The noise disappears with the use of a transformer in the audio signal path, but I wanted to resolve it in the "correct" way, eliminating the ground loop. The trafo I've used to test is small and cheap and a good audio trafo is pretty expensive.
What solution did the OP find? Did you try different things? Which one worked out better?
Others mentioned the use of a small resistor in the groud path: is that supposed to be put in the ground between smps and the arduino/raspberry boards? (did I understand correctly?)
I've also tried to put one of removable ferrite bead lugs on a couple of spot along the lines to apparently no solution. The suggestion here is on the output of the SMPS, right?
And lastly: if I were to try and design my own amp board, would I get rid of the problem by keeping the AGND of the amp separated from the power ground and going back only to the raspberry/arduino side?
A lot of questions, I know, but thanks for the infos.
The noise disappears with the use of a transformer in the audio signal path, but I wanted to resolve it in the "correct" way, eliminating the ground loop. The trafo I've used to test is small and cheap and a good audio trafo is pretty expensive.
What solution did the OP find? Did you try different things? Which one worked out better?
Others mentioned the use of a small resistor in the groud path: is that supposed to be put in the ground between smps and the arduino/raspberry boards? (did I understand correctly?)
I've also tried to put one of removable ferrite bead lugs on a couple of spot along the lines to apparently no solution. The suggestion here is on the output of the SMPS, right?
And lastly: if I were to try and design my own amp board, would I get rid of the problem by keeping the AGND of the amp separated from the power ground and going back only to the raspberry/arduino side?
A lot of questions, I know, but thanks for the infos.
Think differently
I have encountered this problem before. Sorting it out makes one's head spin sometime. Now I would consider this problem in current domain instead of voltage domain.
Thinking about signal current coming out of ADC output through amplifier input impedance and coupling capacitors, then loop it back, this current is really tiny, if you have input voltage of 1V and amplifier input impedance of 100K, the current is 10uA at its peak. And tracing this current you can find the path sometime sharing power supply currents of various blocks like Bluetooth module, supply current is very dirty compare with signal and magnitudes time larger, this create problem if there is any impedance on the path they are sharing. To eliminate this current sharing, a perfect star connection ground is the best, in many cases, people made mistakes or compromises. Finding the current sharing can certainly help to sort things out. Single ended signal output can create big problem if different blocks sharing ground and supply, because there is no way to separate supply current and signal current on supply wire/traces. This is true for separate blocks even in the very same amplifier.
I have encountered this problem before. Sorting it out makes one's head spin sometime. Now I would consider this problem in current domain instead of voltage domain.
Thinking about signal current coming out of ADC output through amplifier input impedance and coupling capacitors, then loop it back, this current is really tiny, if you have input voltage of 1V and amplifier input impedance of 100K, the current is 10uA at its peak. And tracing this current you can find the path sometime sharing power supply currents of various blocks like Bluetooth module, supply current is very dirty compare with signal and magnitudes time larger, this create problem if there is any impedance on the path they are sharing. To eliminate this current sharing, a perfect star connection ground is the best, in many cases, people made mistakes or compromises. Finding the current sharing can certainly help to sort things out. Single ended signal output can create big problem if different blocks sharing ground and supply, because there is no way to separate supply current and signal current on supply wire/traces. This is true for separate blocks even in the very same amplifier.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.