I'm currently driving my two speakers with seperate TPA3118 PBTL amplifier boards, sharing the same 19V laptop power supply. The problem is, there is a constant background noise similiar to cicada noise.
I analysed the noise spectrum and saw two appearent noise bands around 5khz with harmonics(attached under the post).
The reason i came to the conclusion that the noise is due to the power supply, is that when i plug off the psu from 220v mains, it continues to supply the amplifier for around 2-3 seconds more, during this time, that cicada-like noise is absent.
I'm currently planning to put a low pass filter between the amplier and the power supply, i'll be using this circuit:
This circuit is intended to remove 50-60Hz ripple, so i decided that i'll not need a 10H inductor. Through some calculations and considering the ferrite cores i have, i've decided to use a 200mH inductor and a 3300uF capacitor after that. The circuit with those values have around -30dB filtering at 50Hz, -60dB at 100Hz and so on.
On the calculations i've considered the amplifier as a 8ohm load, which corresponds to ~2 amps at 19 volts.
Am i doing this right? Would it work? It is the first time i'll be doing such a circuit so there might be something so basic i'm missing.
If I'm correct, i'll be buying some copper inductor wires and capacitors tomorrow and build the circuit.
I analysed the noise spectrum and saw two appearent noise bands around 5khz with harmonics(attached under the post).
The reason i came to the conclusion that the noise is due to the power supply, is that when i plug off the psu from 220v mains, it continues to supply the amplifier for around 2-3 seconds more, during this time, that cicada-like noise is absent.
I'm currently planning to put a low pass filter between the amplier and the power supply, i'll be using this circuit:
This circuit is intended to remove 50-60Hz ripple, so i decided that i'll not need a 10H inductor. Through some calculations and considering the ferrite cores i have, i've decided to use a 200mH inductor and a 3300uF capacitor after that. The circuit with those values have around -30dB filtering at 50Hz, -60dB at 100Hz and so on.
On the calculations i've considered the amplifier as a 8ohm load, which corresponds to ~2 amps at 19 volts.
Am i doing this right? Would it work? It is the first time i'll be doing such a circuit so there might be something so basic i'm missing.
If I'm correct, i'll be buying some copper inductor wires and capacitors tomorrow and build the circuit.
Attachments
At what current is your 200mH inductor going to saturate?
I often use use a 19V "brick" for my amplifiers. I use 10000uF at the output (before the amplifier) because the 19V adapter is intended to look into a battery and not designed for audio use. Haven't you got an old 2200uF-10000uF (min. 25V) capacitor you can use for a quick test?
I often use use a 19V "brick" for my amplifiers. I use 10000uF at the output (before the amplifier) because the 19V adapter is intended to look into a battery and not designed for audio use. Haven't you got an old 2200uF-10000uF (min. 25V) capacitor you can use for a quick test?
uhh, i don't know at what current will it saturate because i just looked it up to learn what inductor saturation is and i havent wound the inductor yet, thanks! I don't have any capacitors above 1000uf ready at hand, what will i test with it?
The easiest method to measure the saturation current i can find up until now seems to be putting the inductor under DC pulses, in series with a shunt resistor, measuring the current by an oscilloscope and looking at the screen. The problem is i don't have an oscilloscope.
Does it become a problem when the DC current going through the inductor exceeds the saturation current, or should i just care that the ripple current is staying below the saturation current rating? This video got me confused about that: YouTube
I also wonder why i experience this problem because people have reported they are happily using their tpa3118 amplifiers with laptop power supplies and my laptop power brick is a Delta brand one. As far as i know these are among the highest quality products between all laptop bricks.
@voltwide , what will i achieve by increasing the DC loading of my power supply?
This online calculator gives me a crude Isat value of 0.19A (for 200 turns, 0.8mm wire diameter, 16mm inner diameter, 25.4mm coil length, and 120 relative permeability, 148mH): Magnetic core saturation
I also found out that, for the E ferrite core i'll wind inductor on, if i reduce relative permeability to 5 by leaving an air gap between halves, i can increase Isat value to 4.5A, while reducing the inductance to 6.2mH. I later put that 6.2mH in my low pass calculation and i still get a gain of -70dB at 4500Hz. Since the noise in question starts around that frequency and i have no low frequency noise problems, it seems building the coil this way should still solve my problem.
@FauxFrench , i can't understand how can you filter any noise with a capacitor(10000uF in your case) in parallel alone after the power supply output, as far as is know a capacitor in parallel isn't supposed to filter anything unless there is a resistor or inductor in series before it.
I also found out that, for the E ferrite core i'll wind inductor on, if i reduce relative permeability to 5 by leaving an air gap between halves, i can increase Isat value to 4.5A, while reducing the inductance to 6.2mH. I later put that 6.2mH in my low pass calculation and i still get a gain of -70dB at 4500Hz. Since the noise in question starts around that frequency and i have no low frequency noise problems, it seems building the coil this way should still solve my problem.
@FauxFrench , i can't understand how can you filter any noise with a capacitor(10000uF in your case) in parallel alone after the power supply output, as far as is know a capacitor in parallel isn't supposed to filter anything unless there is a resistor or inductor in series before it.
Last edited:
If a choke is used above its saturation current, it is more or less just a short-circuit. It looses its function as an inductance.
C-L-C filters that has to carry some current are often of no use for lower frequencies because the choke has to be quite big for carrying the current without saturation.
I have two 19V "bricks" and both work well. As I said, they are designed to charge a laptop battery. For that, they need not have much capacitance at the output. Amplifiers like having good capacitance across the supply terminals to stabilize the voltage. It is possible that too little capacitance may cause distortion or oscillation a bit like what you experience.
If you have one or two 1000uF capacitors (25V or higher), put them across the power supply input terminals of the amplifier and see if the noise disappears.
C-L-C filters that has to carry some current are often of no use for lower frequencies because the choke has to be quite big for carrying the current without saturation.
I have two 19V "bricks" and both work well. As I said, they are designed to charge a laptop battery. For that, they need not have much capacitance at the output. Amplifiers like having good capacitance across the supply terminals to stabilize the voltage. It is possible that too little capacitance may cause distortion or oscillation a bit like what you experience.
If you have one or two 1000uF capacitors (25V or higher), put them across the power supply input terminals of the amplifier and see if the noise disappears.
Last edited:
Yes i understand that 19V bricks arent supposed to perform good as a amplifier power supply but people have praised tpa311x amplifiers in terms of PSRR. Maybe my power supply is just too noisy.
i've tried two 1000uF capacitors in parallel to the power line and i can't hear any effect.
The 6.2mH choke i told in my last reply is going to be indeed big, like 5cm*5cm assembled with 58 grams of copper wire in it. Or were you talking about something even bigger?
i've tried two 1000uF capacitors in parallel to the power line and i can't hear any effect.
The 6.2mH choke i told in my last reply is going to be indeed big, like 5cm*5cm assembled with 58 grams of copper wire in it. Or were you talking about something even bigger?
6.2mH and 1000uF should leave a resonance frequency of 64Hz. Thus, not useful for 50/60Hz. Then you either increase the capacitors to very large values or you increase the inductance value in which case the saturation current is lowered for the same inductance core.
I prefer good regulators instead. I continue a bit later.......
I prefer good regulators instead. I continue a bit later.......
U might want to try this. You can increase it in say 1A steps untill you hear a shift in frequency of the noise. If it shifts you must increåse the load untill you no longer hear any noise. Your powersupply might simply be too powerful for the application and enters skip/burst mode. Very difficult to filter that, easier to eliminate it by proper loading of the supply.
One reason some folks dont experience noise issues is b/c their smps is not over dimensioned and thus switches continously without skipping.
1000uF with 6.2mH have 64Hz resonant frequency but:
-It's pretty dampened, and the resonant gain isnt more than a few dB's
-I don't have problems with noise at 50/60Hz
-On my new design i plan to use 10000uF cap with 6.2mH inductor whose resonant frequency is 20Hz with only 1 decibel of gain, and has gives -20dB more rejection at the remaining frequencies.
Both 6.2mH/1000uF and 6.2/10000uF bode plots included in the attachments, I guess i'll go with 6.2mH/10000uF
I'm afraid of going into building a good regulator, passive filters with simple calculations are much easier for me now
Attachments
My speaker system is rated at 70 watts RMS and the power supply i use is 90 watts. I'm already at the limit. Right now, the only way i can see to increase the DC load is to turn up the volume or use a seperate weaker power supply for lower volumes. If thats the situation, i should filter it anyway.
I'm afraid of going into building a good regulator, passive filters with simple calculations are much easier for me now
Simplest is best, but do take into consideration the prior comments with respect to increasing the load on the supply. I have a couple of SMPS supplies -- most of them look to see a constant current draw. The buck regulator's inductor value is inversely related to the current drawn.
Not to gild the lilly too much -- it's good practice to filter the a.c. input to the supply, wrapping the cord a few times through a torroid designed for the purpose,..
further lilly gilding -- Wenzel Associates had this neat trick of sampling the power supply output and inverting the noise as an "error," then summing the inverted error and output.
Putting an inductor in series is another way of making the current draw more constant, right? An inductor tries to resist changes in the current flowing across it afaik
Sorry for the quick exit......I was just about to miss an appointment.
Thanks for trying the 1000uF as power line decoupling.
There is the possibility the power supply runs in burst mode and you hear the burst being radiated to the amplifier.
You say it sounds like "cicada"-noise. I understand this as the noise made by an insect looking a bit like a grasshopper.
I frequently have this noise that origin from my smartphone polling the transmitters in the neighborhood. The smartphone has to be away from the amplifier not to disturb the amplifier with radiated noise.
Do you have the same problem when the amplifier is not connected to any source and the inputs are shorted to ground?
Thank you, it's okay. You're helping me voluntarily.
i didn't try to connect the input to the ground and will not be able to in the next 6 hours but as i said in the beginning, when i pull the mains plug of the power supply, i don't hear that noise in the next 2-3 seconds, then the psu gets completely depleted. of course the amplifier continues to work in that period. So my best guess is, what i experience is not a line level noise.
By the way, i just got the coil wires and the 10mF capacitor, will try them tonight.
i didn't try to connect the input to the ground and will not be able to in the next 6 hours but as i said in the beginning, when i pull the mains plug of the power supply, i don't hear that noise in the next 2-3 seconds, then the psu gets completely depleted. of course the amplifier continues to work in that period. So my best guess is, what i experience is not a line level noise.
By the way, i just got the coil wires and the 10mF capacitor, will try them tonight.
Try the filter. Here we do not have important HF noise on the power distribution grid. I know in certain countries you use the same power grid for sending information being modulated on the wires as higher frequencies. Such HF noise may pass the power supply and a filter may help.
NB: For the next three days our Internet connection will be gone!!
NB: For the next three days our Internet connection will be gone!!
No, though you got a good answer the wrong way.
A single totem-pole amp driving 8 Ohms looks like ~~48 Ohms to the power supply.
Your amps are bridged, so you have two t-p outputs each seeing 4 Ohms. So two 24 Ohm loads makes 12 Ohms.
Two such channels makes 6 Ohms. So in-sight of your "8 Ohm" number, but you can't just take the speaker impedance as the amp impedance without working through the current paths and tossing some pi at it.
BUT: this is at FULL output which you will never reach. And playing FULL you will not hear your cicadas. The problem is apparently at low/zero volume. When DC current is very low. For class D, maybe even 1%. So the amp draws 0.03 Amps and acts-like maybe 600 Ohms.
Large-uFd caps will not have low-low impedance in the chirp frequencies because they are physically large and thus have some inductance. Modern caps are much better than the old ones (because switchers are now the primary market). Still I would not favor a physically large cap.
Using simple numbers: 1uFd against 600r is a 256Hz corner, well below your 5KHz chirp. 0.4H is also 600r @ 250Hz. At 5KHz these are 30 Ohms loading 12,000 Ohms, apparently 52dB attenuation, which is fictional because parasitics will leak more than that. But a huge cut.
We'll probably get similar real-world results with lesser parts. As the coil is a costly mystery object, and caps are cheap and good, I would cut the coil to like 100mH (which should be available pre-made) and use 10uFd-100uFd capacitance.
A hard-core coil will saturate. 100mH is a bit large for air-core with low DCR. However when the current is high the music is loud and the chirp is masked. The main thing is to have high I at your lower currents, perhaps up to 0.3 Amps.
> noise made by an insect looking a bit like a grasshopper.
A common name is "17 year locust". Like the Biblical plague. Not at all the common locust, not a lot like a common grasshopper, and has a fascinating life cycle. But for our purposes we only consider the "chirp". In 1959 I was in a cicada event in Missouri and in mass they are LOUD. Fortunately gezzer has only one/few 'cicadas' not millions.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.