Please . . . advice on filter caps

[ByronInLawrence]

I'm designing my first power supply for a class A/B amp. Searching for filter caps, I found they are often described by the manufacturer as "audio grade", "computer grade", "general purpose" or other descriptions. For a given capacitance and voltage value, the electrolytics described as "audio grade" were among the least expensive, and often the ESR and leak were not listed in the data sheet. Capacitors with the lowest ESR and leak current (and highest price) are usually described as "general purpose", "computer grade", or as being for SMPS. Another thing I found is that all electrolytics described as "audio grade" were only available with radial or snap-in terminals. However, I know that many DIY builders and high-end companies use screw terminal in their PS filtering caps. I would prefer screw terminals, as I'll be point-to-point wiring this PS.

So, to finally get to my questions:

• What, exactly, does a manufacturer mean when they describe their caps as "audio grade"? Nichion's data sheet describes this as "Realization of a harmonious balance of sound quality, made possible by the development of new electrolyte." And why would they bother with the hype for their least expensive caps?
• How important are the values of ESR, leak current, and ripple current in PS filtering caps?
• What are all these high-end, high power amps with screw terminal PS filtering caps using?
• Just where is the point of diminishing returns for filter cap quality?

I've posted a screen-shot of my planned PS below. The transformer will be a torroid, 40V + 40V, 300VA. A third set of 10,000uF, 100V electrolytics (not in diagram) will be on the board. This will be a dual-mono amp - two power supplies, one for each channel. I plan to drive 4 ohm speakers. Don't worry about the amp. It's designed for slightly higher rail voltages, and more current than my transformer can supply.

Any advice from more experienced builders is greatly appreciated.

[metalsculptor]

The design looks like overkill so the capacitors will be not be stressed. Even 80V pcb mount types should do. SMPS rated devices usually have low ESL and ESR. 105C rated devices last longer.

Use the equivalent circuit of the components in the simulation to simulate what effect of ESL leakage and ESR are. Google and wikipedia will help explain both transformer and capacitor equivalent circuits.

Audio is not a demanding application so that is why audio capacitors should be cheaper. Forget any marketing diatribe, that is what happens when anything enters the consumer market

[indianajo]

There are radical difference in the expected life of caps. Depends on whether they use the rubber sealant, or the expensive stuff. I buy the one with the 8000 hour life rating at 105 deg C if I can get it, 2000 hr if I can't. A lot of 1000 hour caps are still stocked for repair people whose customers don't care. Usually ESR doesn't matter in these life ratings, but 10000 uf comes in a much cheaper "tall' version with *****y ESR. Replacing some 10000 uf @ 100V recently, I bought twice as many 4700 uf caps at 5/4 the ESR and 1/3 the price each. Then I doubled them up on a little CB I made with lexan, 18 ga wire, and steel angle brackets. Glued the caps down with silicon seal, but wall board adhesive will work as well. Weren't a lot of choices in 10000 uf caps, looking for in stock items only. These were radial, soldered the wire to the leads through the new CB. The stock spacing (10 mm) and the old "4 peg" spacing, did not match.
The only thing I've noticed about "audio" caps is that mcmelectronics charges about 4 times as much for them as the series I find in AB motor drives- the ones that run 8 years in an industrial oven motor compartment.

[Mooly]

The LED won't last long

By far the biggest contributer to noise/distortion/hum etc etc etc is your physical layout of all the tracks on the PCB for the main amplifier and what connects where on those tracks.

PSU's are important but just throwing lots of capacitance isn't the answer IMO. A good amp rejects rubbish on the rails anyway.

Also the more capacitance on the rails and the shorter the time the bridge conducts for... but the same energy as lower capacitance rails still has to be put back for a given load. That means the peak currents are correspondingly higher, copper losses become significant and the transformer may well technically be pushed beyond its ratings.

[ByronInLawrence]

Thanks everyone! I was worried no one would respond to me post. Then all of a sudden . . .

Originally Posted by metalsculptor

The design looks like overkill Even 80V pcb mount types should do.

Yes, when unsure of what I'm doing, I prefer to overbuild than underbuild. I may, at some point in the future increase transformer voltage. I'm using these 40V transformers because I already have them. The board is designed for 64V rails. I'm really insisting on screw terminals because I'm not going to make a board for this.

Originally Posted by indianajo

I buy the one with the 8000 hour life rating at 105 deg C if I can get it, 2000 hr if I can't.

I chose one with 8400 hour life rating, but at 85 C. Like Metalsculptor said, I won't be stressing these caps.

Originally Posted by Mooly

The LED won't last long

Also the more capacitance on the rails and the shorter the time the bridge conducts for... but the same energy as lower capacitance rails still has to be put back for a given load. That means the peak currents are correspondingly higher, copper losses become significant and the transformer may well technically be pushed beyond its ratings.

What is wrong with my LED? I admit, I didn't give it much thought. Can I just increase the value of the resistor in series with it, or should I get a beefy-er LED? What do you think of this LED? 4302F1-12V Chicago Miniature Standard LED - Through Hole

Transformer is rated at 300VA, and I'll be using two. It's a dual-mono design. I included the data sheet, but I don't know how to interpret all those numbers. Do you think they might be stressed? Of course, it's kind of late now.

Included is the data sheet for the filter cap I decided on. It's new at Mouser.

Thanks again for the advice!

[discrete]

Originally Posted by ByronInLawrence

What is wrong with my LED? I admit, I didn't give it much thought. Can I just increase the value of the resistor in series with it, or should I get a beefy-er LED?

...it is because your schematic in post#1 has a 'typo'. It shows the junction of the LED and its R tied to ground.
Use any LED, calculate R for 2-5 mA. It will last forever.

On a side note, the 500 ohm discharge Rs you have are somewhat counter productive (120 mA, 7W at 60V). I usually do not use them, but consider 10k minimum if you must.

[ByronInLawrence]

Originally Posted by discrete

...it is because your schematic in post#1 has a 'typo'. It shows the junction of the LED and its R tied to ground.
Use any LED, calculate R for 2-5 mA. It will last forever.

On a side note, the 500 ohm discharge Rs you have are somewhat counter productive (120 mA, 7W at 60V). I usually do not use them, but consider 10k minimum if you must.

I might have built it like that if you hadn't pointed it out to me.
Yeah, 500R is way too low. I'll up the bleeders to ~2kR.
Thanks for the help!

[twest820]

An intermediate value between 200nF and 10,000uFs will be helpful in suppressing LC resonances between the caps---I don't have the link handy but there's a great thread in this forum (or maybe solid state) with quite a bit of data on the issues. A good rule of thumb for avoiding the problem is two order of magnitude steps; it varies from design to design and board to board but I often end up with 100nF 0603 X7R || 4.7uF 1206 X7R || 220uF 5mm aluminum || 22,000uF 25mm aluminum as a starting point. At higher voltages you'll need a 'lytic for the low uF range and a larger package on the MLCCs.

Originally Posted by ByronInLawrence

Nichion's data sheet describes this as "Realization of a harmonious balance of sound quality, made possible by the development of new electrolyte."

If you look at the datasheets you'll see the KWs have slightly higher ripple current ratings than VRs, implying a little bit lower ESR. Otherwise they're identical, though it's possible the KWs might have somewhat higher self resonant frequencies---I've not seen measurements but I'd frankly be surprised if there's much difference beyond marketing induced placebo effects. If your amp's PSRR is low enough the ESR, ESL, and self resonant frequency of the caps is a concern you either need a better amp design, regulation, or caps which are speced in more detail (TDK's online app is great for MLCCs, Nichicon LF and HE are good staring points for higher performance cans). Generally it's easiest just to regulate the amplifier's control loop---with care one can get discrete PSRRs approaching 100dB at certain frequencies but an op amp for control running off some LM7xL15s is pretty turnkey and offers 120+dB PSRR over the entire audio band (some will say this is overkill but every time I run the math for my amps it's about what I end up with).

[HornTube]

Do yourself a favor and use Epcos LL caps.

[ByronInLawrence]

Thanks twest820. I understand your first suggestion, and I think I'll add a pair of 5000uF electrolytics.

I didn't follow the second part. I'm relatively new to this with no formal training in electronics. I don't know what you mean by "KWs" and "VRs". I also don't know what the amplifier's PSRR (Power Supply Ripple Resistance?) is, or how to figure out how good it is. I'm relying on the reputation of the designer. It's the DX Blame MKIII by Carlos Mergulhao.

HornTube, I assume these are the "long life" caps. How are these better than the myriad of other capacitors available?