Decoupling capacitors on headphones Amps.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi!

I'm planning on making an headphone amplifier based on the Objective 2. I'm also thinking on adding a PCM2912 as a DAC. The objective of this project is not really the improvement of a already well renowned amplifier, but a learning intention instead.

t will be a desktop version, so i'm not constrained on the PCB by those two batteries. That gives me more space, and searching around the web I found some designs exemplified here that have 1000 uF caps, along side the typical 0.1uF decoupling ceramic caps, parallel to the power traces.

The reasoning behind is that they will act as a "reserve power source" when the amps are hit with bass heavy signal, and this will improve their response.


My doubt is on how accurate this is, if it will indeed improve the design, or if actually even make things worse!?

Thanks in advance!:):hbeat:
 
Member
Joined 2009
Paid Member
I found some designs exemplified here that have 1000 uF caps, along side the typical 0.1uF decoupling ceramic caps, parallel to the power traces.

The reasoning behind is that they will act as a "reserve power source" when the amps are hit with bass heavy signal, and this will improve their response.

So the problem with power supplies is the wiring or PCB traces between them and their loads. This wiring adds inductance. Too much inductance can turn your amplifier into an oscillator, which is a problem.

Typically a medium-to-large bulk capacitor is added to a circuit right at the power connector. These capacitors tend to be at least 10 uF for line level circuits and often exceed 1000 uF in power amplifiers. The purpose of these capacitors is to lower the power supply impedance "seen" by the amplifier - or in other words "cancel" the effects of the inductance between the power supply and the amplifier.

The bulk capacitance is the energy storage. Those capacitors supply the current needed to produce a sharp transient.

The problem with large electrolytic capacitors is that they tend to have relatively low self-resonant frequencies (SRF) - typically in the 0.5-5 MHz range. This means the capacitor will in effect turn inductive beyond the SRF. If the amplifier has gain at frequencies beyond the bulk capacitor's SRF, it might oscillate at these frequencies. To combat this, a smaller capacitor (0.1 uF) is added as local decoupling right at the amplifier's supply connection.

There's much folklore and rules of thumb regarding the selection of bypass capacitors and most of them are wrong. If I was to recommend something that will probably work for you, I'd go with 1000 uF where the supply enters the board and 22 uF in parallel with 0.1-1 uF right at the amplifier power pin. To fully optimize the decoupling network, I would simulate the PCB layout in a circuit simulator.

For a more technical explanation, I suggest reading through my Taming the LM3886 - Supply Decoupling article. While it is specific to the LM3886, the underlying concepts are applicable to any circuit.

Best of luck with your build.

Tom
 
Thank you for your amazing response!!

Since the power supply section will have multiple electrolytic caps, for the AC-DC conversion, those already act as a reserve then, right?! That avoids me using more caps than what are needed. Good

I never placed another cap besides de 0.1uF ceramic one, to decouple the IC, but I'll certainly consider adding the 22uF now.

Thank you for you help!!
 
Founder of XSA-Labs
Joined 2012
Paid Member
Another consideration is that if you have two separate power amplifier IC’s (or in case of discrete amp, output devices such as MOSFET) that may share same single power supply, adding a larger bulk reservoir cap next to IC or MOSFET with 0.1uF decoupling cap can improve stereo separation and imaging. I have measured improvements in stereo crosstalk going from -50dB to -84dB with added 2200uF cap at each power output device.
 
Another consideration is that if you have two separate power amplifier IC’s (or in case of discrete amp, output devices such as MOSFET) that may share same single power supply, adding a larger bulk reservoir cap next to IC or MOSFET with 0.1uF decoupling cap can improve stereo separation and imaging. I have measured improvements in stereo crosstalk going from -50dB to -84dB with added 2200uF cap at each power output device.

Damn. That's not negligible at all.

Since my project is based in the Objective 2, it will probably have 4 4556s as drivers, which can indeed pull quite a bit of current. I'll definitely consider your advice!
Thank you!
 
Damn. That's not negligible at all.

Since my project is based in the Objective 2, it will probably have 4 4556s as drivers, which can indeed pull quite a bit of current. I'll definitely consider your advice!
Thank you!

Bias current is static, the dynamic output current is the one needed to be worried more. The output current of a headphone amp is usually less than 10mA, you probably can get away with much smaller caps. Anyway, the big caps do not hurt.
 
I'll probably place somewhat big caps in there (not too big, taking in account what tomchr said), even if for most of the time it will be driving light loads. That is if I have the space. I'm limited to 100x100 mm since beyond that the fab house will charge a loooot more. Let's see how this goes
 
The figures I gave were for SE Class A amp running about 65mA bias current. Simple MOSFET follower SE Class A amps have no PSRR to speak of, so large caps or independent PSU’s are needed for good stereo separation.

I won't be using discreet components for the driving section. That job will be done by some NJM4556, which fortunately have 90 dB of SVR according to the datasheet. This might make the "big caps" redundant. I have to further research this
 

PRR

Member
Joined 2003
Paid Member
Indeed! Did I do something wrong?...(

It's complicated.

Yes, you get more answers.

More answers is not always nirvana.

To the point: if a point is covered in another forum, why type our fingers to the bone repeating the point? Or if a DIYA poster says "blue", and the Reddit poster says "red", we may wish to know of the disagreement and a possible reason to support the "blue" assertion.

If all in one thread, we see all disagreements and can defend or walk-back as appropriate. If we "have to" run around multiple forums (I have seen one question in four places) then our feet get tired.

So it is different from our usual in-one-place discussions. And inevitable in the expanding InterWeb. Not "wrong".
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.