Paralleling class 1 & class 2 MLCCs for audio?

[Zoios]

Hi everyone,

I've been looking for any information on using X7R and C0G capacitors in parallel for passing audio signals. I've seen some examples of this technique used for DC blocking in mixer schematics - would using a 0.1uF C0G cap in parallel with a 22uF X7R, for example, be better than just using the 22uF cap?

I've already ruled out using electrolytics for their size and tantalums for their bad reputation, so this leaves me with ceramics and the aforementioned question. Thanks in advance for any input!

 

[Newmarket]

Film caps too large ?

 

[Zoios]

There are very limited options when it comes to getting film caps larger than 1uF where I live, and I'd like to stick to SMD if possible.

 

[JohnRoberts]

This sounds like they are resurrecting old audio myths, about paralleling small high quality capacitors to fix large low quality capacitors.

Back in the 70s I bench tested this theory and found that the small high quality capacitor needed to be at least 10% of the value of the larger so-so capacitor to make a measurable improvement. To make your 22uF X7R behave you would need a 2.2uF COG. Good luck with that.

JR

 

[john12ax7]

Is X7R better or worse than a bi-polar electrolytic in the audio path?

 

[JohnRoberts]

Is X7R better or worse than a bi-polar electrolytic in the audio path?

Different horses for different courses... Back in the 70s my bench testing was looking to improve the ESL of a 22uF electrolytic capacitor in the gain leg of simple RIAA phono preamp. I was measuring (not hearing) tens of degrees of phase shift at 20kHz. In the course of that investigation I also discovered that some 22uF tantalums I had on hand had much lower ESL and reduced the phase shift significantly.

I would expect a ceramic X7R capacitor to have lower ESL than an electrolytic but would have to look at data sheets of both to compare voltage coefficient that I would expect to dominate distortion. I'm not sure I have ever seen voltage coefficient specs for electrolytic capacitors (but maybe they should).

JR

 

[merlin]

would using a 0.1uF C0G cap in parallel with a 22uF X7R, for example, be better than just using the 22uF cap?

No, because the 0.1uF has 220 times higher impedance than the 22uF.* So only 0.5% of the current is gonna flow though the 0.1u cap.

*Ok this ratio will change a bit at high frequencies, but probably above the audio range.

 

[JohnRoberts]

No, because the 0.1uF has 220 times higher impedance than the 22uF.* So only 0.5% of the current is gonna flow though the 0.1u cap.

*Ok this ratio will change a bit at high frequencies, but probably above the audio range.

Nonlinearities above the audio band can generate IMD products that are down in the audio passband. Of course it depends on the specific application.

Note: modern electrolytic capacitors are a much improved from what I had on my bench a few decades ago.

JR

 

[ccaudle]

would using a 0.1uF C0G cap in parallel with a 22uF X7R, for example, be better than just using the 22uF cap?

Possibly marginally, but the difference would likely be academic, i.e. you may be able to measure a difference, but unlikely to be a noticeable improvement.
Merlin argued from a view of current division between the two capacitors, but another way to think about it is that the voltage coefficient of the X7R capacitor is going to dominate the distortion performance. Look at the voltage change across the capacitor (DC bias - voltage at negative signal peak, vs. DC bias + voltage at positive signal peak) and see how much the capacitance of the X7R changes with that voltage change.
Then see how much the capacitance of the parallel combination of X7R and C0G changes with that voltage change. The C0G cap has no voltage coefficient to speak of, so it will be a small constant capacitance in parallel to a larger variable capacitor. The larger the C0G cap, the less that the combination capacitor varies with voltage, but you would likely need to get the C0G up to a significant fraction of the X7R capacity (of course depending on what performance you are willing to accept).

Is X7R better or worse than a bi-polar electrolytic in the audio path?

Worse. Not only does X7R have a noticeable voltage coefficient, unless you have a super small size limit you should be able to get an electrolytic in higher capacitance, which reduces the AC voltage across the capacitor, further reducing any effect on signal quality.

If you really have such a tight size limit that you cannot fit electrolytic, then consider increasing the load impedance and including the capacitor in the feedback loop if you have to use X7R, or if you can increase the load impedance enough perhaps you can use parallel C0G capacitors to get the value you need.

 

[Newmarket]

I would expect a ceramic X7R capacitor to have lower ESL than an electrolytic but would have to look at data sheets of both to compare voltage coefficient that I would expect to dominate distortion. I'm not sure I have ever seen voltage coefficient specs for electrolytic capacitors (but maybe they should).

Me neither. But for a 'lytic I'm usually well exceeding the voltage it will see and be required to work at - to greatly increase the expected lifetime. But - yeah - might be interesting from an audio pov.

 

[JohnRoberts]

Me neither. But for a 'lytic I'm usually well exceeding the voltage it will see and be required to work at - to greatly increase the expected lifetime. But - yeah - might be interesting from an audio pov.

I would expect voltage coefficient to only be important for in-band audio frequency filters that generate significant AC terminal voltage across the capacitor. Most electrolytic capacitors are used in DC blocking applications with extremely low audio voltage across the two terminals.

Of course passive loudspeaker crossovers using electrolytic capacitors are a concern. It can be a significant cost increase to use film in place of electrolytic in passive crossover filters.

JR

 

[Newmarket]

I would expect voltage coefficient to only be important for in-band audio frequency filters that generate significant AC terminal voltage across the capacitor. Most electrolytic capacitors are used in DC blocking applications with extremely low audio voltage across the two terminals.

Of course passive loudspeaker crossovers using electrolytic capacitors are a concern. It can be a significant cost increase to use film in place of electrolytic in passive crossover filters.

JR

Yes. I'm not familiar with LS crossover design but the cost difference film vs 'lytic can be large both in cost and size. And tbh I'm not sure how the polarity aspect works out in LS crossovers... Bipolar ? "Back to Back" ?

 

[thor.zmt]

I've been looking for any information on using X7R and C0G capacitors in parallel for passing audio signals. I've seen some examples of this technique used for DC blocking in mixer schematics - would using a 0.1uF C0G cap in parallel with a 22uF X7R, for example, be better than just using the 22uF cap?

Funny, I just had occasion on something related...


In this picture are C0G (light Grey), X7R (lighter coloured Shit) and Y5U or similar (darker coloured Shit). Real devices.

For real C0G I have seen "Cream" colour to white to light grey.

X7R are a lighter brown shade, Y5U can go nearly black.

C0G is also sometimes called NP0 and they are also available as Through Hole.

Here WHY only C00G/NP0 belongs into audio circuits (bottom left hand corner):


So, no. Bypassing 22uF X7R with 0.1uF C0G doesn't really do much.

C0G in 1206 SMD come in up to 0.47uF and you can parallel more than one.

Try refactoring your circuit so that this is enough.

If you use OPA1678/79 you have a Fet input Op-Amp that allows very high value DC feedback/bias resistors (say 3.3M which with 0.1uF form an 0.5Hz turnover) and can allow circuits to be re-arranged to make do without large value audio-path capacitors

I've already ruled out using electrolytics for their size

Why. Many modern SMD types are very small.



The 100uF/6.3V bipolar electrolytics in the pic are tiny. The microphone on the right is a 12mm capsule.

If you go 22uF this same size will probably go up to 50V.

tantalums for their bad reputation

They are bad. I mean really terrible. But they are still 10 times less bad or so (even less with DC bias) than X7R!

so this leaves me with ceramics and the aforementioned question.

I think the solution is to:

1) Abandon stupid prejudices and go fact and evidence based and also avoid sh!t coloured capacitors in audio, or you sound matches the colour
2) Design so you can use low value, Class 1 dielectric Ceramic Capacitors or Film wherever possible
3) Bite the bullet and use Bipolar SMD electrolytic capacitors where larger values are unavoidable (low impedance, low noise)

I think these 3 points can get you far.

Thor

 

[JohnRoberts]

Funny, I just had occasion on something related...


In this picture are C0G (light Grey), X7R (lighter coloured Shit) and Y5U or similar (darker coloured Shit). Real devices.

For real C0G I have seen "Cream" colour to white to light grey.

X7R are a lighter brown shade, Y5U can go nearly black.

C0G is also sometimes called NP0 and they are also available as Through Hole.

yup, they've been around longer than SMD but were only available in tiny values.

Here WHY only C00G/NP0 belongs into audio circuits (bottom left hand corner):

yes the data sheets reveal near ideal performance for cog/npo...

So, no. Bypassing 22uF X7R with 0.1uF C0G doesn't really do much.

yup

C0G in 1206 SMD come in up to 0.47uF and you can parallel more than one.

Large value COG/NPO (like 0.47uF) are a relatively recent development. I am optimistic that we may see even larger npo/cog values in the future to meet high performance designer's desires.

Try refactoring your circuit so that this is enough.

If you use OPA1678/79 you have a Fet input Op-Amp that allows very high value DC feedback/bias resistors (say 3.3M which with 0.1uF form an 0.5Hz turnover) and can allow circuits to be re-arranged to make do without large value audio-path capacitors

Back in the day I used many 0.33uF or 0.47uF film (polyester/Mylar) capacitors with impedances scaled up appropriately. High impedance bifet op amps have been widely available since the late 70s. This worked well for line level audio paths but large resistor values can compromise noise floor due to Johnson noise (thermal noise). Mylar is not as linear as npo/cog but cheap and manufacturing process friendly while better than aluminum electrolytic.

Note: DC servos are just an active variant on scaling impedances to use higher linearity film capacitors for DC blocking.

Why. Many modern SMD types are very small.



The 100uF/6.3V bipolar electrolytics in the pic are tiny. The microphone on the right is a 12mm capsule.

If you go 22uF this same size will probably go up to 50V.



They are bad. I mean really terrible. But they are still 10 times less bad or so (even less with DC bias) than X7R!



I think the solution is to:

1) Abandon stupid prejudices and go fact and evidence based and also avoid sh!t coloured capacitors in audio, or you sound matches the colour

maybe avoid scatalogical advice... the electrons do not know the color of their packaging. Instead trust the specification sheets.

2) Design so you can use low value, Class 1 dielectric Ceramic Capacitors or Film wherever possible

+1 npo/cog = good.

3) Bite the bullet and use Bipolar SMD electrolytic capacitors where larger values are unavoidable (low impedance, low noise)

Tuning DC blocking poles lower can keep mischief below and out of the audible passband.

I think these 3 points can get you far.

Thor

We pretty much agree...

JR

 

[thor.zmt]

maybe avoid scatalogical advice... the electrons do not know the color of their packaging.

But they sure seem to know the colour of ceramic dielectrics...

Thor

 

[JohnRoberts]

But they sure seem to know the colour of ceramic dielectrics...

Thor

I don't think electrons have eyes...

DIY hobbyists do, so if building projects with random SMT capacitors maybe note the colors.

I prefer to use new components labelled with what dielectric they use.

I don't melt much solder these days so what would I know.

JR

 

[Newmarket]

I don't think electrons have eyes...

DIY hobbyists do, so if building projects with random SMT capacitors maybe note the colors.

I prefer to use new components labelled with what dielectric they use.

I don't melt much solder these days so what would I know.

JR

ime the height of a chip capacitor is a more reliable indicator.

 

[Zoios]

If you use OPA1678/79 you have a Fet input Op-Amp that allows very high value DC feedback/bias resistors (say 3.3M which with 0.1uF form an 0.5Hz turnover) and can allow circuits to be re-arranged to make do without large value audio-path capacitors

Could you provide examples for such an approach? I don't think I've ever come across this solution. I use JFETs in current mode for audio switching in my circuit, with 22uF caps directly before them to remove the DC that causes switching artifacts - would there be a way to implement higher resistances with, say, a 220nF C0G in such case?

 

[thor.zmt]

Could you provide examples for such an approach? I don't think I've ever come across this solution. I use JFETs in current mode for audio switching in my circuit, with 22uF caps directly before them to remove the DC that causes switching artifacts - would there be a way to implement higher resistances with, say, a 220nF C0G in such case?

I routinely use J-Fets in audio switching in DC coupled circuits. Why do you need coupling capacitors?

Unbiased the J-Fet is just a low value resistor. Biased off, a very high value resistor in parallel with a puff (pF).

DC doesn't come into it.

Correctly applied modern active parts will have effectively no DC offset, so many coupling capacitors become optional.

The main area where the common SMD Film and C0G Capacitors (0.1uF....1uF) cannot be used are DC blocking in phantom microphone inputs.

Of course, we might instead use a DC coupled flying frontend, so the problem goes away.

I would need to see what circuit you are trying to realise, to be specific.

But take a (J)FET Op-Amp (say 1/2 OPA1678 place 3.3M from +in to gnd, another from -in to output, an 0.33uF C0G in line with each input and you have an AC Op-Amp with no further coupling capacitors needed.

The noise contribution from the resistors and capacitors is usually low enough to not raise circuit noise (~ 3.3nV|/Hz at 20Hz, falling with a first order function (so 0.33nV|/Hz at 200Hz). For 0.1uF coupling capacitors the noise at all given frequencies is up ~10dB which depending on application may still be fine.

For the OPA1678 1kHz noise is 4.5nV|/Hz and 10nV|/Hz @ 100Hz. As noise adds Rootsumsquare 3.3nV|/Hz@ 100Hz will not significantly add to the noise of the Op-Amp, suggesting 68nF capacitors suffice.

The Ein for that circuit (OPA1678 + 2 X 3M3 + 2 X 68nF) is around -118dBu unweighted, DC offset guaranteed below 2mV, 20kHz THD of -114dB, total BOM cost < 0.80 USD for two channels in 1kU.

Plus, subjectively (for those who care and not violently disbelieve) sounds good.

Very hard to beat all said.

Thor

 

[AngryAudio]

Could you provide examples for such an approach? I don't think I've ever come across this solution. I use JFETs in current mode for audio switching in my circuit, with 22uF caps directly before them to remove the DC that causes switching artifacts - would there be a way to implement higher resistances with, say, a 220nF C0G in such case?

Are you sure you need the coupling caps in this application. When using JFETs in current mode to switch, you must slow down the gate drive to prevent the gate change from producing an audible transient in the signal chain. This may be what you're hearing when you switch, not a DC offset component.

I always try to reduce the number of coupling caps in the signal chain. Maybe the strategic placement of a few caps or DC servos (which can use much smaller value caps), some attention to DC offset contributors in the signal path, etc., might allow you to make a cleaner device and reduce the need for (as many) large caps?