Totem pole filter cap Q

[Chuck H]

I'm designing an amp for a friend. It will have five power supply nodes, a filament elevation and an unloaded voltage of 495V. Other than Rubys, Atoms and a couple of other "tube amp parts" type caps I can't find caps rated over 500V. I can't get specs on most of these caps and they're typically overpriced. I'd just as soon use 350V Nichicon or Panasonic caps in series at each node. I trust these caps and it'll actuall cost less. Are there any considerations about using totem arrangements all the way into the preamp? My concern is that all those resistor circuits paralleled in the power supply could add up to a load that must be taken into account.

TIA

[Enzo]

So? CAlculate the current demands of those 220k or whatever loads on the B+. 10ma maybe? A couple 220k in series is 440k, and that draws about a milliamp from 500v.

SO use tandem caps for the first node. Those 500v caps are 500 working volts ratings, they have higher surge limits to handle power on spikes. Myself, I wouldn;t sweat the later nodes with 500v caps. A bazillion Fenders and Peaveys, and who knows what all use 500v caps with 500v B+ or close to it. And none of them are cap eaters.

If you want to tandem them all anyway, I see no reason you couldn;t other than space.

[pdf64]

The shipping cost from Germany may be prohibitive for you, but Ask Jan First has axiel 550V / 600V surge F&Ts Frag Jan zuerst- Ask Jan first: Roehren und mehr
If you go with the series connected caps, then each node will use about 1 watt, assuming 2x100k balancing resistors; probably won't make much difference, and the last one could be used for the heater elevation anyway.
Pete.

[R.G.]

When you put caps in series, you need to use resistors to equalize the DC voltages. This is done by putting a resistor in parallel with each capacitor in the series stack. The resistors should all be equal valued (assuming, that is, that the caps are equal valued; if they're not close to the same value, transients can overvolt the lowest-value capacitors anyway because of capacitor imbalance).

The resistance value needs to be set to "leak" at least ten times the end-of-life leakage current of the caps. If this is true, one leaky cap can't put all the voltage across the least leaky cap, popping that one, and then popping the leaky one in a chain failure. This isn't too hard. As Enzo notes, the values for the equalizing network are generally in the 100K to 330K range. This doesn't wind up being all that wasteful of B+ power.

And it's sure better than an early death.

[Chuck H]

Thanks gents. I roughed it out in my head to around 60k total load. So about 8mA. No big. I posted right after drawing it up. Even though I did rough calculations it just looked funny. So my lack of design experience is equal to my lack of confidence.

Thanks for the cap link Pete. I do have another Q... How come I always read about builders elevating the filaments from the last rail node. Shouldn't the idea be to keep any heater circuits away from the early preamp stages? In less ambitious builds I have installed a "bleeder" circuit for safety and just made that a voltage divider for the elevation with the grounded resistor bypassed by a cap.

[J M Fahey]

I do not trust 500V caps. Period.
Nor do many modern designers.
The consensus seems to be (I'm talking commercial, $$$ driven designers, not "Boutique", Mojo driven ones) just use caps in series.
2 x 350V caps in series stand 700V *rated* voltage, and higher surges, of course.
And 220K in parallel with each, as Enzo suggested, is a good bleeder value.
In Argentina I must foreign special order 500V caps at high cost, high International postage, and then struggle at Customs explaining them they are not Nuclear Bomb detonators ( lots of paranoia around lately, and I'm not kidding); while at any Mom and Pop TV parts shop I can find *excellent* high value 350V or 400V caps intended for TVs and SMPS , abbundant and for peanuts.
So it's a no brainer for me.
And I know that in, say, 10 years time they will still be around .... not so sure about 500V ones.
.

EDIT: and you need 500V resistant caps only at the plates, screens, and *maybe* the PI ; after that, once you are below 400V, there's tons of caps.
I use excellent EPCOS (ex Siemens) ones.

[Chuck H]

I've resolved to use the 350V Nichicon PW's. Rated for something like 9000 hours and very low impedance, so much the better if they're going to be in series. I can get them for $2 to $3 US. All the poor spec'd (or NO spec'd) 500V caps cost more for one than I will pay for two Nichicons. The specs for the FT caps that Pete linked look pretty good too. I was worried about the voltage spike at switching out of standby. Not so much now. But I'll probably just totem the whole thing anyway because: 1)I can order all the parts from the same place. 2) It'll cost less. 3) The specs on the Nichicon PW caps is superior. And 4) Absolutely no wories about rated voltage if the supply runs unloaded.

This amp will also use a Pi filter type arrangement as the main filter AND a three cap totem on the third node to derrive a specific voltage as well as the other four nodes and the cap bypassing the filament elevation divider. For a total of fourteen caps in the power supply. Good thing they're smaller than Atoms or vintage caps!!!

[Gingertube]

Chuck,
Good choice. You will get far better lifetime (like x4 minimum) with this arrangement.
This will tell you why (if interested) and give you the info to calculate voltage share resistors, capacitor lifetimes etc).

For reading
http://ken-gilbert.com/images/pdf/el...c_appguide.pdf

In brief:
People say that caps share voltage according to their leakage currents, that is true for absolutely equal values of caps. More though, they share voltage according to their capacitance value and remember that typical electolytics have a tolerance of +50% -20%, that is 2 caps of the same "nominal" value may be very different in "real" value, so it is even more imporatnt to have voltage share resistors.

Capacitor lifetimes depend upon the self heating which occurs internal to the capacitor. That is a ripple current squared x Equivalent series resistance thing (power dissipated internally).

Because the 2 series connected caps are x2 the value of the required "composite" cap you will find that they have about x2 of the ripple current rating and lower ESR giving much longer lifetimes. They will also have lower leakage currents than that a single higher voltage part run near its limit. The benefit is that you will not need to recap the amp for 30+ years instead of every 10 to 15 years.

The voltage share resistors also act as the "essential" bleed resistors so you don't get nasty surprises.

Cheers,
Ian

[Chuck H]

Sounds great. I'm about to read the link, but this popped out at me:

Because the 2 series connected caps are x2 the value of the required "composite" cap you will find that they have about x2 of the ripple current rating and lower ESR

How can two caps in series have half the ESR?!? In fact I chose low impedance caps partly because I'll be putting them in series. This quandry assuming that to replace, say, a 22uf cap you would put two 47uf caps in series for an approximately equal UF value. O.K. In case I'm jumping ahead, I'm going to read the link now.

[Gingertube]

Sorry I was'nt clear.
I was actually trying to say that (for example) a 47uF cap will have lower ESR (generally) than a 22uF. Yes you have to add them together when you series connect them.

The point is that not only do you get lifetime benefits from dividing the self heating between 2 devices but the total self heating will actually be lower as well. So you get a double dip on lifetime improvements.
Cheers,
Ian

[Chuck H]

Thanks Ian!

About 80% of the article goes over my head. Fortunately much of it is explained in simpler terms. So it looks like the relief of two caps sharing the load (with balancing resistors) regulates self heating better than a single cap of equal rating. Bonus. Since most specs are heat sensitive then you actually will get a small reduction in ESR. Perhaps not less for two 47uf's in series than a single 22uf. But probably close enough to be neglected as a consideration. I just learned more about how to treat caps than any normal human should have to

[pdf64]

Ian, thanks to you (and Ken) for that evox-rifa document.
I wondered if it was still available on the manufacturer's site; the brand owner is now Kemet, they have a lot of technical papers available Technical Papers but not that one.
Pete.

[Gingertube]

pdf64,
I have it on my work home drive from "yonks ago" so I knew what search terms to use.

ASIDE: I had this Application Note as I was looking for a 10,000uF cap for a laser diode array driver. These laser diode arrays are used to optically pump the slab in our main high power laser (day job). I have to deliver 50us wide 100A current pulses at 4kHz repetition rate. Voltage of the supply is 50V. When I went looking for a cap and did all the calcs it turned out that I needed to use a 10,000uF 250V rated cap even though there were a smaller 10,000uF 63V and 100V rated caps available. That was to handle the 28 Amps RMS Ripple Current.

So why bring that up - I build a few cathode biased guitar amps, 6V6 and EL84 in particular, bias voltages are typically around 25V and 12V respectively. For the 12V bias EL84 I find a 63V rated bypass cap works best and for the 25V bias 6V6 I use 100V rated bypass caps. The higher voltage part has higher ripple current rating (ripple current here is the actual audio signal) and lower ESR so I can expect those caps to last much longer. A power supply cap which fails will just stress the rectifier and power tranny which is easily fuse protected, an output tube cathode bias bypass cap failure stresses the output tubes and the output transformer as well so there is more likelyhood of doing damage to expensive components and it is best to design for longevity of those caps. An extra dollar at build time can save an extra 100 dollars at repair time.

Sane arguement applies to power supply caps as stated in the thread above. If you can manage it (by series connecting caps etc.) electrolytics work best and last longest when operating at no more than 2/3 of their voltage rating. This mainly applies to the first power supply cap where the stresses are largest. For further down the Power Supply chain (in the preamp) I'm happy to simply use 450V rated caps. In fact I keep a bag of 22uF/450V Panasonic ED for use in the preamps.

In the same "vein" where the voltage of the main power amp section supply is suitable (low enough) I use those same 22uF/450V, I just parallel 2, 3, 4 or 5 of them as required. The same cap lifetime benefits will result as you are still spreading the self heating between more components.

Of-course I build one offs so I don't have to argue with accountants. I am haveing an arguement with a few friends at the moment. I work 4 days a week and am thinking about reducing that to spend more time on guitar and HiFi tuibe amps. The ones I have built so far I have "Tongue in Cheek" called DUD Amps for "Down Under Designs". When I said I was thinking of registering that as a buisiness name the response was very negative and most suggested I just used "Miller Ampflication" instead - how boring is that?

Cheers,
Ian

[defaced]

Thanks for the cap link Pete. I do have another Q... How come I always read about builders elevating the filaments from the last rail node. Shouldn't the idea be to keep any heater circuits away from the early preamp stages? In less ambitious builds I have installed a "bleeder" circuit for safety and just made that a voltage divider for the elevation with the grounded resistor bypassed by a cap.

I generally elevate off the screen node. This was adopted from current high gain amp building practices.

[Gingertube]

Me too - for both Guitar and HiFi Tube Amps.
Cheers,
Ian