Hum in active speaker caused by floating flyback SMPS

Hi there,

for an active speaker I build a 150W flyback converter delivering 30V to an power amp and +-18V to linear regulators for some analogue OPamp circuits.

The power supply is floating, having no protective earth connection, to save space and this is where the trouble starts. The primary and scondary grounds are connected with a Y-cap 2n2.

The input circuit following a neutrik XLR connector is a symmetrical opamp buffer stage. An highpass before that with 10kOhm to GND keeps the impedance low. When the whole PCB is connected to the ground of my measurement system or with a cable to protective earth the speaker is silent. But when disconnecting the GND from the board, a big hum comes out of the speaker. Measuring that, you can see 50Hz/60Hz with harmonics till some kHz. Even without any audio cable at the input, the hum is huge. When putting in more Y-cap this gets worse. Without the Y-cap it gets a little better, but still huge. When I get closer with my hand to the XLR or the PCB the hum gets louder.

Does anybody know that behavior? And what is the connection between having no earth connection and this big hum?

This is the first time that I design a SMPS without protective earth. With an earth connection, there are normally Y-caps to filter for CM-noise. How can I do that when I have no earth?

Best wishes and a great weekend
 

29285

Disabled Account
2010-01-17 9:20 pm
Yes, this is a well known effect and the reason why wall warts generally introduce much noise into audio systems. The Y-cap conducts 100Hz-noise generated by the primary rectifier into the secondary circuit. All in all this is common mode noise current growing with Y-capacitance carrying many harmonics caused by the primary rectifier.
As you mentioned secondary grounding helps.

I would recommend to connect the Y-cap with primary PE instead of secondary gnd. This should work even with secondary floating.
 
Last edited:
Does anybody know that behavior? And what is the connection between having no earth connection and this big hum?

Its common-mode noise, as you mention in your next paragraph. And introducing the earth provides a lower impedance path for those CM noise currents to flow, so they don't go via your input circuitry back to the source component (or measuring device).

This is the first time that I design a SMPS without protective earth. With an earth connection, there are normally Y-caps to filter for CM-noise. How can I do that when I have no earth?

You'll need a common-mode choke. Try one from Murata. But beware it probably won't eliminate the hum, just reduce it. You'll need to re-think the input circuitry most likely.
 
Hi there,

for an active speaker I build a 150W flyback converter delivering 30V to an power amp and +-18V to linear regulators for some analogue OPamp circuits.

The power supply is floating, having no protective earth connection, to save space and this is where the trouble starts. The primary and scondary grounds are connected with a Y-cap 2n2.

The input circuit following a neutrik XLR connector is a symmetrical opamp buffer stage.

Surely symmetrical? Instrumental amplifier?

An highpass before that with 10kOhm to GND keeps the impedance low. When the whole PCB is connected to the ground of my measurement system or with a cable to protective earth the speaker is silent. But when disconnecting the GND from the board, a big hum comes out of the speaker. Measuring that, you can see 50Hz/60Hz with harmonics till some kHz. Even without any audio cable at the input, the hum is huge. When putting in more Y-cap this gets worse. Without the Y-cap it gets a little better, but still huge. When I get closer with my hand to the XLR or the PCB the hum gets louder.

Does anybody know that behavior? And what is the connection between having no earth connection and this big hum?

Y capacitor sets your reference potential ("signal GND") to a power line potential, and the stray capacitance to the environment injects noise currents into inputs. Probably input impedances are not equal, therefore noise currents don't cancel each-other and your shielding system is not complete, therefore stray capacintances "connect" to a very different potential. High harmonic content is because of the (stray) capacitive coupling of noise.

This is the first time that I design a SMPS without protective earth. With an earth connection, there are normally Y-caps to filter for CM-noise.

Y caps are for switching frequency and above. (Common mode choke also, so it will not help with this problem.)

How can I do that when I have no earth?

Tricky question, no simple answer.

But as starting: shielding is a must, and symmetrical inputs must have exactly the same impedances.
 

basreflex

Member
2007-10-09 7:30 pm
the converter's flyback transformer needs a tight coupling , hence a lot of common mode noise injected into the audio gnds. if you have a chance use a flyback transformer with dual shield windings between prim and sec. or redesign to a LLC resonant design with intended weaker coupling between prim and sec. this way you can really minimize the noise injection..
 
the converter's flyback transformer needs a tight coupling , hence a lot of common mode noise injected into the audio gnds. if you have a chance use a flyback transformer with dual shield windings between prim and sec. or redesign to a LLC resonant design with intended weaker coupling between prim and sec. this way you can really minimize the noise injection..

These are very good advices again sw freq noise!

However this doesn't solve the problem of floating and not properly shielded system.
 
no, the solution is in posting #3, obviously ignored by the experts here.

1: Y capacitor is needed to short capacitive coupled high freq noise. You will get 100 kHz (+ modulation + harmonics) noise on output GND instead of 50 Hz. Not a great deal.
2: Even if there were no discrete capacitor from mains voltage, stray capacitances would still be there. As basreflex pointed well: "flyback transformer needs a tight coupling". This would still couple 50 Hz as well.
 
i'm working on a dcdc usb isolator using the llc principle. only 2pf coupling between prim and sec.. (toroid sectional winding)

Do you mean effective transfer capacitance for switching freq. or capacitance can be measured directly by a meter?

I know effective transfer capacitance can almost be eliminated by matching starting phases of near windings in case of certain turns ratio, but simply primary to secondary cap being 2 pF is a little hard to believe for me.
 

29285

Disabled Account
2010-01-17 9:20 pm
1: Y capacitor is needed to short capacitive coupled high freq noise. You will get 100 kHz (+ modulation + harmonics) noise on output GND instead of 50 Hz. Not a great deal.
2: Even if there were no discrete capacitor from mains voltage, stray capacitances would still be there. As basreflex pointed well: "flyback transformer needs a tight coupling". This would still couple 50 Hz as well.

I know this all, working in emi/CE field for some time. The most audible and rf noise is injected into the secondary through the y-cap - typically 2.2nF. Winding capacitance is a fraction of this. Disconnecting the Y-cap from secondary circuitry and grounding it to PE avoids that.

Obviously wall warts lack any PE-connection which makes them unsuitable for powering audio devices. But laptop bricks do - that makes the difference.
 
Last edited:

29285

Disabled Account
2010-01-17 9:20 pm
Do you mean effective transfer capacitance for switching freq. or capacitance can be measured directly by a meter?

I know effective transfer capacitance can almost be eliminated by matching starting phases of near windings in case of certain turns ratio, but simply primary to secondary cap being 2 pF is a little hard to believe for me.

Agreed. But this is not the point. The point is the Y-cap.
The best I ever found was a wall-wart made by friwo with an effective coupling capacitance of 0.7pF. No Y-cap required!
 
Last edited:

29285

Disabled Account
2010-01-17 9:20 pm
i'm working on a dcdc usb isolator using the llc principle. only 2pf coupling between prim and sec.. (toroid sectional winding)

Did this, and it works fine. Using a LLC-bobbin with 2 separated chambers gives similar coupling capacitance. But coupling capacity is not a problem with the Y-cap grounded at PE.
On the other hand, I used laptops bricks in many audio applications. This works fine, too, although these are always flyback converters, sometimes with a PFC-booster.
 
I know this all, working in emi/CE field for some time.

If you know, why deny it?

The most audible and rf noise is injected into the secondary through the y-cap - typically 2.2nF.

Don't think they are the same! There is a source and a transfer impedance for each type of noise. Source of RF noise is the switching node, and transfer impedance is trafo capacitance. The current of this is shunted by Y cap, therefore no-load voltage is divided. Without Y cap in a fly-back SMPS in bad case you get about Iswnoise=50 kHz*200V*100pF*4=4mA. The other disturbance (50 Hz and harmonics) comes from power line, and coupled by both y cap and trafo capacitance. In a system not grounded at all, and have high primary to secondary capacitance audio "GND" potential will be close to the half of the power line voltage whether you connect Y cap or not, since there is no shunt impedance. This is what the topic starter already experienced and stated. Typically Ihum=50Hz*6,28*230V*2.2nF=0.14 mA.

Winding capacitance is a fraction of this.

Yes, it should be, in order to maintain low RF voltage.

Disconnecting the Y-cap from secondary circuitry and grounding it to PE avoids that.

As we saw it doesn't. (Please don't use the word "it" in a sentence that contains more than 1 subject, because it is ambiguous. "That" is also problematic, since you wrote about many different things. If you try to speak and write unambigously, you have the chance to understand things better.)

Obviously wall warts lack any PE-connection which makes them unsuitable for powering audio devices. But laptop bricks do - that makes the difference.

This is a false generalisation. For example I have some HP SMPS without PE I used as a replacement for laptop to make them usable as audio source, after original PSUs (with PE connection) generated too much noise.
 
Last edited:
Hum issue

Hi there,

thanks for all the replies, this really helps!

I did some further investigations on the hum. The FFT of the hum shows, that it is 50Hz, 150Hz, 250Hz, etc up to 5kHz. This signal stays the same, when changing polarity of the AC-plug (swap L-N) and I can measure this signal consistently on different PCBs.

One important information is that the hum just occures, when no input source is connected and no earth is given to the system. Even if no earth is connected but the inputs are connected to a floating source, e.g. a mobile phone, there is no hum and everythink is fine. I found that a magnetic field is coupling into the input stage, which is a highpass followed by an opamp configured as a buffer each for the positive and negative XLR input. A reduction of the hum could be realized with reducing the input impedance to a minimum of required value of 10kOhm. But the main reduction can be done with a shielding plate (connected to secondary GND) physically between the XLR connector + input opamp and the SMPS, which is on the left side. It is the AC input with EMI filtering, NTC, Rectifier which seems to radiate. When the Y-cap of 2.2nF between primary and secondary GND is removed the hum is still the same. If the Y-cap is increased to 4.4nF the hum increases as well.

Anyway this is just fighting the error signal and not reducing the error source, so I still hope to find another solution and avoid the shielding.

Did I understand you correctly that the stray capacitance between the transformer primary/secondary winding is the main reason for the hum? When reducing this capacitance the radiated magnetic field gets weaker?


Regards
 
now we are getting somewhere. here we are talking of capacitive coupling between primary side and input, capacitance is going through the air not through the transformer. capacitance behaves as a current source so that's why the hum reduces when the input impedance is decreased. while having a balanced input, I would expect that the 2 inputs cancel, but when the source is unbalanced the hum could be there. as you have a 50 hz component and not pure 100hz and harmonics, this is the effect of the primary side capacitively radiating.
there is hardly any magnetic component of 50hz as the converter operates at high frequency and the core does not radiate 50hz. if some sort of PFC is implemented you would see 100hz magnetic noise. if you connect the tip of your scopeprobe to the ground alligator forming a small loop and cranking up the scope sensitivity, you have a magnetic sensor to search for noise sources.

back to the capacitive pickup, if the physical location of the board and XLR, such that
one node picks up more than the other you have the root cause. one possibility is to make a small radiator using a coin connected to the generator emitting at 10V 1khz, generator gnd to secondarygnd, and find out where the sensitive nodes are while positioning the coin over the sensitive areas..
 
Hi there,

thanks for all the replies, this really helps!

I did some further investigations on the hum. The FFT of the hum shows, that it is 50Hz, 150Hz, 250Hz, etc up to 5kHz. This signal stays the same, when changing polarity of the AC-plug (swap L-N) and I can measure this signal consistently on different PCBs.
...

Regards

Hi,

Can you post the schmematic? Maybe it's easier to analyse and refer to components... A picture of the circuit also can tell many things...

One important information is that the hum just occures, when no input source is connected and no earth is given to the system. Even if no earth is connected but the inputs are connected to a floating source, e.g. a mobile phone, there is no hum and everythink is fine.

Of course. Capacitively coupled noise current to input is shunted by low impedance source.

I found that a magnetic field is coupling into the input stage,

I wonder why do you think this. Everything you wrote points to the conclusion of parasitic capacitive coupling.

input stage, which is a highpass

It's usually called AC coupling. Probably capacitively. The term "highpass" refers to a filter where cutoff frequency is within audible range (20Hz...20kHz).

which is on the left side.

Quite meaningless without a picture, don't you think?

It is the AC input with EMI filtering, NTC, Rectifier which seems to radiate.

Or not. Maybe (very probably) the amplifier is a radiating antenna, and everithing that is not on the same potential, draws current from it, and some part of this current comes from the input. As you said your hand also makes hum stronger when you approach it. Isn't this ring any bell?

When the Y-cap of 2.2nF between primary and secondary GND is removed the hum is still the same. If the Y-cap is increased to 4.4nF the hum increases as well.

This is quite controversial, and still we don't know where exactly did you place Y-cap. What you call primary GND? Schematic please!

Did I understand you correctly that the stray capacitance between the transformer primary/secondary winding is the main reason for the hum? When reducing this capacitance the radiated magnetic field gets weaker?

Well, not really. The reason is (floating voltage caused by (Y capacitor + primary-to-secondary transfer capacitance)) and (input stray capacitance to different potential) and (input resistance multiplied by gain) and (asymmetry). Any factor you eliminate, the effect will disappear accordingly.

But: Y capacitance is needed mainly because of primary-to-secondary capacitance. If you can reduce latter, you can decrease the former, and get better result. But without earth connection you can never expect realy clean GND potential.

BTW: You can't tell seriously, that you expect magnetic effect from a capacitor. This is plain simple capacitive field. No black magic.