Best low-voltage blocking caps (PC mount)?

[BackwardsBoB]

I'm designing a high-performance analog output off an audio DAC. I thought I could do the whole thing differential with no blocking caps, but no such luck. My balanced output circuit converts to unbalanced nicely, but doesn't have enough common mode rejection. I need to add some blocking caps. I only need a 16V rating. Most of the rest of the board is surface mount, but radial leads would be OK. I need about 10uF. What to use?

THD + N is somewhere better than 100 dB without the caps, so ceramics won't cut it. Polarized is OK. What's the good stuff?

[Le Basseur]

I'd give a try (listen) to the following:
from Elna-Cerafine or Silmic
from Black Gate-no specific type,they're all great "by default"
HTH!
Regards,

[Steve Conner]

Douglas Self has actually measured capacitor THD. His conclusion was that it's unmeasurable, provided you size the capacitor big enough that the AC signal voltage across it is negligible, which it should be for a DC block capacitor anyway.

So my answer is "anything will do as long as it's big enough".

I never believed that anyone could hear the difference between different capacitor brands, and wouldn't waste my time auditioning them.

[tedmich]

Douglas Self has actually measured capacitor THD. His conclusion was that it's unmeasurable, provided you size the capacitor big enough that the AC signal voltage across it is negligible, which it should be for a DC block capacitor anyway.

So my answer is "anything will do as long as it's big enough".

I never believed that anyone could hear the difference between different capacitor brands, and wouldn't waste my time auditioning them.

I concur, and would skip the "silk and hemp" dielectrics and use a good 16v 10uf Tantalum, BUT this doc suggests a decent polyester cap would have a lower noise floor if you are chasing perfection:
pdfserv.maxim-ic.com/en/an/AN3171.pdf

[BackwardsBoB]

Thanks. The good news is that I dumped the quasi-floating balanced output circuit I found in the trash on the web into LTSpice (had a wonderful first experience - highly recommended), and found that it's architecture does support good common-mode rejection with proper resistor matching. Changing four resistors, the DC offset is gone, and I now have a DAC with -124 dB THD + N on differential current outputs running differential all the way to the output, no caps, and just a pair of 10K ground ties at the output to keep it from pegging a rail with no load, using two op amp stages based on op amps with noise levels of 1.7nV/rootHz. I think I can claim any performance level I want, since the distortion is unmeasureable (the stimulus bill has already created or saved 1.3 million jobs!).

Of course, Boss wants me to make crummy audio because he thinks it will save time. He just doesn't understand motivation.

I saw an article on Audio Design Line that claimed performance will be cut to -55dB with a ceramic blocking cap, but I can no longer find it. The main problem was capacitance change with respect to voltage.

Tantalum is alright I guess, but I (almost) never use caps that blow up, catch on fire, and emit noxious smoke.

Anyway, thanks. I knew I could count on a few valuable opinions here.

[dai h.]

here's an interesting page on caps, FWIW:

Capacitor Sounds, Speaker Cables and Crossover Inductors.

[sebi]

I would suggest Panasonic FC or Oscon dry electroyt and parallel them with a decent PolyesterCap 100nF. Many "audiophile" do so and I did for example in several channels of an old yamaha mixer and it made it sound more open and clear. Very cheap improvement.

[Steve Conner]

The "paralleling with a plastic film cap" is an old myth that just won't die. But it deserves to die. Look at the ESR of an OS-Con or similar high-quality electrolytic at 20kHz, compare it with the reactance of 100nF at 20kHz (which is 79.6 ohms if you can't be bothered working it out) and then tell me if you think the polyester cap is doing anything.

[Bruce / Mission Amps]

The "paralleling with a plastic film cap" is an old myth that just won't die. But it deserves to die. Look at the ESR of an OS-Con or similar high-quality electrolytic at 20kHz, compare it with the reactance of 100nF at 20kHz (which is 79.6 ohms if you can't be bothered working it out) and then tell me if you think the polyester cap is doing anything.

I get this plastic cap in parallel with an E-cap demand on builds all the time.... and I'm pretty lazy Steve... can you show the math on all that?

[Steve Conner]

I've already presented the argument in as much detail as I thought it needed. But I'll try to spell it out more carefully.

Here is a page with a graph showing impedance vs. frequency of various electrolytic caps, including Os-Cons.

DIY Paradise - Sanyo OS-CON

Notice that the impedance is below 1 ohm for the whole range above 3kHz. (Below 3kHz, it's just the capacitive reactance of 47uF.)

So, we are "bypassing" a 1 ohm impedance with a 79 ohm impedance shunted across it. Since the 1 ohm is an ESR, and the 79 ohms is a capacitive reactance, we have to use complex arithmetic to calculate the resulting impedance, which is the square root of 1/(1+1/79^2) : 0.999 ohms.

This is what's technically known as "F**k all difference", especially since the ESR should have been negligible in a DC blocking cap in a well-designed audio circuit, so we're talking about an 0.1% improvement in a parameter that didn't even matter to start with. I think this proves that it's not worth the extra expense and board space.

The above is also the worst case, because the impedance of the bypass cap (being capacitive reactance) gets larger as the frequency drops below 20kHz, but the impedance of the electrolytic is still dominated by ESR.

Similar arguments hold for power supply filter capacitors in audio amps. The sad fact of engineering is that the customer has no clue about what he/she wants, but is always right nevertheless.

[BackwardsBoB]

Steve, I think you may be being too emperical. Perhaps the Spirit of Music fears OS-CONs and digs polyester. Can you prove that the Spirit of Music has no preference?

The papers from C. Bateman are real eye-openers. I've also seen another analysis that looks at speaker cable impedance's effect on damping factor. Even modelling a speaker cable as a resistance points to a serious erosion of a damping factor of 100 driving a 2 Ohm stack of speakers with a cable resistance over 0.01 Ohms. Bateman's analysis of the effect of capacitor dissipation factor at audio frequencies does appear to show that capacitor construction matters alot.

Parts I'm using have THDs measurable in ppm. Component deviations from the ideal apparently become quite significant.

Of course, it's all silly. Why in the world would you need a D/A converter with -124 dB THD+N when you're going to use it to generate the signal to drive a speaker? At least Mr. Bateman was able to use his years of analysis to figure out why his amp kept blowing up.

Still, I need -100dB+ margins on THD and noise to "keep up with the Joneses". At least I don't have to drive five bits into a three foot cable before the first bit pops out at the other end.

The universe is too strange to have happened by chance.

[Steve Conner]

It's also too strange NOT to have happened by chance.

I'll meet you halfway though, I'll prove that the Spirit Of Music has no preference (for instance by swapping the caps in Eric Johnson's pedals when he's not looking, which I bet he wouldn't notice) but first I want you to prove that it exists in the first place.

[JoeM]

Forget the film cap in parallel, what you really need are wood encased electrolytics:

|| DHTRob - Projects: Rebuilding a capacitor (1).. ||

[BackwardsBoB]

It's also too strange NOT to have happened by chance.

I'll meet you halfway though, I'll prove that the Spirit Of Music has no preference (for instance by swapping the caps in Eric Johnson's pedals when he's not looking, which I bet he wouldn't notice) but first I want you to prove that it exists in the first place.

You're basically asking me to prove that it's not true that there's no Spirit of Music. If you can't prove a negative, how in the world can I prove a double-negative?

[Steve Conner]

I'm actually begging a really deep philosophical question here. I'm asking you to ponder whether the "Spirit Of Music", aka "Mojo", is an actual thing that can be objectively measured.

If it is something measurable, then according to what we know about human hearing, you need to change acoustic power at some frequency by a dB or two to affect it. And I proved that choice of DC block capacitors won't do that.

If it can't be measured, then we're into a whole different ball game. It's basically up to the advertising guy to write bumf that makes the customer feel good, then he'll believe that the mojo is present, and so it will be for him. And it's up to you, the designer, to think up some crazy new circuit and parts choice that will let this advertising guy go to town with his copywriting. This is actually easier because now technical performance doesn't matter at all. (Have you ever checked out the performance of these single ended DHT amps on paper?)

This is how we get to startling "innovations" like the wooden electrolytic, which will probably die prematurely because the original case was broken open and replaced by a non-airtight one, and you could buy 50 electrolytics for the price of the wood and your labour to make it, but of course you believe it sounds better so it's worth it!

But is it worth it? What if you decided to stop believing in the hype and grant yourself the mandate to enjoy making and listening to music on any kind of equipment at all? Yes, even laptop onboard audio.

This would make your life so much easier as a designer too, just get an AKM codec or something, build it up with the datasheet reference circuit and vanilla parts, put it in a nice shiny box and leave the Spirit Of Music to the marketing department. Hell, rebox a M-Audio Transit with gold-plated RCAs and resell it at a 300% markup.