DC-DC conv. Ripple affecting Speech Recognition - Capacitance multiplier

[Vatech]

TL;DR

A dc-dc converter causes high ripple that affects the Speech Recognition function . Suggested solutions?

Into a project I am working is use 12Vdc SMPS to power an audio amplifier and some more components. We have no “noise” on that stage.

Then I use a step down converter to convert the 12V to 5V and power a mic preamplifier (Respeaker – usb Mic array) and a development board. There is a lot of background noise, probably caused by the high frequency ripple of the step-down converter.

If I power the preamp directly from a cheap 5v power supply, or a usb port the noise disappears. The problem is that I only have one power source, the 12v, so I need a solution.

Unfortunately, I don’t have oscilloscope. I tried to add 105(1μF) capacitor with 1000μF electrolytic capacitor with no improvement at all.

Should I implement a Capacitance Multiplier

Thank you

 

[Baluncore]

Use a 7805 or similar linear regulator to produce quiet 5V power for the preamp only.

 

[Vatech]

I am sorry, i forgot to mention that i tryed different modules and also voltage regulators:
1) XH-M404, 2) XL4016, 3) XL4015 , 4) voltage regulator LM1084IT-5.0
5) voltage regulator lt1083cp with no improvment.
i watched the following video , and he explains that voltage regulators have almost no effect in high frequency ripple caused by step down converters

 

[DaveE]

Should I implement a Capacitance Multiplier

Maybe. But for high frequency noise, normal capacitors will usually do well enough.

voltage regulators have almost no effect in high frequency ripple caused by step down converters

Yes, it's true. But they are good at the low frequency stuff. This leaves HF noise that is easier to filter. Ferrite beads or small powdered iron toroidal inductors can be really helpful in making effective LPFs. Also, if the current is low, shunt zener diodes can work well, they are very fast (low dynamic impedance).

First make sure that you have adequately separated your signal traces (including grounds and reference voltages) from the other stuff (power, digital, etc.). Your audio signal path should have no signal path to the power supply or any switching circuit. Using the PS for offsets (instead of references) is a common rookie mistake. Maybe I'm telling you something you've already done. But your experiment of switching PSs may also have changed the signal routing.

 

[Tom.G]

Use a 7805 or similar linear regulator to produce quiet 5V power for the preamp only.

As usual, both @Baluncore and @DaveE have good suggestions.

Or if you don't want to build another regulator, you could try using small-value monolithic, multilayer, ceramic capacitors, instead of or even in addition to the 1000uF you have.

Also, the switching supply may be radiating the noise in addition to conducting it on its output. You can test this by powering your amplifier with a quiet supply and then turning on the switching supply while it is close by.

The noise from switching regulators is rather high frequency and electrolytic capacitors do not work well at high frequencies. Multilayer ceramic capacitors are optimized for high frequency operation.

Since these capacitors are low cost (USD $0.26 each), I suggest connecting 3 different values all in parallel across the 5V supply near the amplifier. I say near the amplifier so that the tiny inductance in the wiring from the switching power supply can help filter the high frequency noise.

Capacitor values of 0.1uF, 0.01uF, 0.001uF (value codes of 104, 103, 102 respectively) would be a good selection. Considering their low cost and shipping cost, you may want to order several to keep in stock for future projects.

Here is a link to one supplier:
https://www.digikey.com/en/products/detail/vishay-beyschlag-draloric-bc-components/K103K15X7RF5UH5/8557355

And a link to the capacitor datasheet:
https://www.vishay.com/docs/45171/kseries.pdf

There are several suppliers and what is available where you are will vary.

Hope this helps!

If you still have problems, posting a few photos of the project here will help us help you. Fighting noise such as you have can sometimes seem to be magic, and there are several people here that seem to have "Magic Wands."

Cheers,
Tom

p.s. Please let use know about your progress/results.

 

[Rive]

Then I use a step down converter to convert the 12V to 5V and power a mic preamplifier (Respeaker – usb Mic array) and a development board. There is a lot of background noise

- Could you please give us the type of the DCDC PSU? Is it some isolated type?
- Do you have a spectrum for that noise?
- is your grounding is solid, through the whole setup? Honestly, for me your problems feels more like some grounding problem than a DCDC noise.

 

[Vatech]

- if you mean about the dc-dc converted i used the modules based on 1) XH-M404, 2) XL4016, 3) XL4015 , 4) voltage regulator LM1084IT-5.0

-No , unfortunately i don’t, i could upload a record from the noise of the amplifier
-all the connections have common ground, continuity tested to verify each connection

 

[Baluncore]

Only the LM1084IT-5.0 is a linear regulator.
Did you put any capacitors from ground to the input and output terminals?

 

[Vatech]

i tryed 105(0.1μF capacitor) combined with 1000μF electrolytic capacitors, on both input and output. No improvment

 

[Vatech]

 

[Vatech]

As usual, both @Baluncore and @DaveE have good suggestions.

Or if you don't want to build another regulator, you could try using small-value monolithic, multilayer, ceramic capacitors, instead of or even in addition to the 1000uF you have.

Also, the switching supply may be radiating the noise in addition to conducting it on its output. You can test this by powering your amplifier with a quiet supply and then turning on the switching supply while it is close by.

The noise from switching regulators is rather high frequency and electrolytic capacitors do not work well at high frequencies. Multilayer ceramic capacitors are optimized for high frequency operation.

Since these capacitors are low cost (USD $0.26 each), I suggest connecting 3 different values all in parallel across the 5V supply near the amplifier. I say near the amplifier so that the tiny inductance in the wiring from the switching power supply can help filter the high frequency noise.

Capacitor values of 0.1uF, 0.01uF, 0.001uF (value codes of 104, 103, 102 respectively) would be a good selection. Considering their low cost and shipping cost, you may want to order several to keep in stock for future projects.

Here is a link to one supplier:
https://www.digikey.com/en/products/detail/vishay-beyschlag-draloric-bc-components/K103K15X7RF5UH5/8557355

And a link to the capacitor datasheet:
https://www.vishay.com/docs/45171/kseries.pdf

There are several suppliers and what is available where you are will vary.

Hope this helps!

If you still have problems, posting a few photos of the project here will help us help you. Fighting noise such as you have can sometimes seem to be magic, and there are several people here that seem to have "Magic Wands."

Cheers,
Tom

p.s. Please let use know about your progress/results.

i tryed to use 105 capacitors combined with electrolytic 1000μF with no improvment. lower value , 103,104 could help but i do not have on stock and i have to deliver the project soon, so waiting to receive is not an opption. I have inductors and mosfets, zener , resistors and alla common used parts, what else could i try? capacitance multiplier?

 

[Vatech]

Maybe. But for high frequency noise, normal capacitors will usually do well enough.Yes, it's true. But they are good at the low frequency stuff. This leaves HF noise that is easier to filter. Ferrite beads or small powdered iron toroidal inductors can be really helpful in making effective LPFs. Also, if the current is low, shunt zener diodes can work well, they are very fast (low dynamic impedance).

First make sure that you have adequately separated your signal traces (including grounds and reference voltages) from the other stuff (power, digital, etc.). Your audio signal path should have no signal path to the power supply or any switching circuit. Using the PS for offsets (instead of references) is a common rookie mistake. Maybe I'm telling you something you've already done. But your experiment of switching PSs may also have changed the signal routing.

Seems you were correct, although i don't know how to move on.

I tried to power externally the 5v line and I had no noise.

BUT when I bridged the common line (ground) between the 5v power supply and the common line of the 12v Source NOISE STARTED.

What should I do? I never faced that before

 

[Rive]

- if you mean about the dc-dc

I meant the 12V => 5V dcdc.

This 'respeaker' thing is some digital equipment? Do you have digital or analog signal there? Is it the uC which makes it analog?

 

[Vatech]

I meant the 12V => 5V dcdc.

This 'respeaker' thing is some digital equipment? Do you have digital or analog signal there? Is it the uC which makes it analog?

https://wiki.seeedstudio.com/ReSpeaker_Mic_Array_v2.0/
I tried to power externally the 5v line and I had no noise.

BUT when I bridged the common line (ground) between the 5v power supply and the common line of the 12v Source NOISE STARTED.

 

[Rive]

ReSpeaker_Mic_Array_v2.0

I assume you are using the analog output, then?
If so, then my bet is still on the grounding issue: a ground loop. You have one ground connection through the jack there, and a second through the bridge you made.

 

[Vatech]

I assume you are using the analog output, then?
If so, then my bet is still on the grounding issue: a ground loop. You have one ground connection through the jack there, and a second through the bridge you made.

Yes, seems you are correct, how do i fix this? on amplifier audio common and power common are bridged (as they should), what should i do?

 

[tech99]

You want to avoid any of the units sharing a single ground wire at any point. Use star system of grounding.

 

[Rive]

What I would do, is to make sure that the power ground is lot thicker than the signal ground. Then I would just lead the signal wire right near the power ground line.
If that is still not enough, then there are some more ideas in the 'solutions' part of the linked wiki article.

 

[tech99]

I think that if you share the (noisy) current in the ground wire from the various units, the resistance of the wire will create a volt drop which couples noise from one unit to the other.

 

[Vatech]

Thank you all for sharing your experience with me . Following your suggestion problem solved.

Seems that adding a second cable to ground the Respeaker solved the issue. Although I didn’t understand good what was the issue since the problem solved by following the same route. The module was connected via a 20cm usb cable direct to Tinkerboard, but the Ground wire was directly connected to DC converter. Would the issue be that the audio ground cable had lower resistance? What I did was just to add an additional cable and connect it again to dc converter and that solved the issue.

 

[Baluncore]

What I did was just to add an additional cable and connect it again to dc converter and that solved the issue.

https://en.wikipedia.org/wiki/Ground_loop_(electricity)

 

[tech99]

Yes, sounds like a lower resistance ground connection, so that there is less resistance in the common path.

 

[DaveE]

Seems you were correct, although i don't know how to move on.

I tried to power externally the 5v line and I had no noise.

BUT when I bridged the common line (ground) between the 5v power supply and the common line of the 12v Source NOISE STARTED.

What should I do? I never faced that before

You've gotten some great replies so far. I'll not repeat them.
In a general sense, you want to treat your signal traces/circuits separately from everything else to protect them from noise. Let's say your signal has a 5V range. Don't think of that as a 0-5V signal with respect to ground, think of it as a differential signal (signal+ and signal -) with no ground reference (or a single point ground connection, if necessary). This may extend to the pre-amp power and ground too if they haven't followed this advice and are coupling noise into the signal path internally.