What's the deal with the uF of filter caps?

[Mbira]

There seems to be a huge range of the uF values in power supply caps. The older Fenders had 10-20uF. I understand that back then they didn't have anything higher at those high voltages. The peavey I'm working on now has 4200uF at 63V. I can't even find that value through mouser. So how important are those values? Do you just want the highest uF for the voltage you need or what? I have also heard people talk about benefits in using different value caps together, though I havn't seen that in any designs. Can I just use a 1000uF value cap instead of a 4200uF in the filter section?

Thanks as always to the teachers...

Joel

 

[Matt Syson]

Hi
High voltage caps used on valve gear only supply a low current so the 'ripple voltage' is relatively small and because the output transformer steps it down and push pull cancels, you can get away with a low capacitance. (all of this is a generalisation and you could get specific).
Low voltage (transistor) amps use high current and you need more capacitance to keep the ripple reasonable (and a load of other reasons).
4200uF at 63 volts is not a 'preferred value' although 4700 / 63 is so look for one of these although if your amp has 2, change the pair together.
Manufacturing tolerance is wild on these components typically -20 +30% so don't get hung up on details although 1000uF would be far too small.
Don't skimp on the voltage rating byt 63 is easily available.
Ripple current rating is important for big amplifiers so look at the caps you have and compare them with replacements.
Take care
Matt S

 

[Mbira]

OK, thanks. I finally found the proper caps thru an ebay seller...

 

[PRR]

To a first approximation, the filter cap is inversely proportional to the load resistance.

An old Fender may be 500V and 100mA. 500V/0.1A= 5,000 ohms.

A sand-state amp may be 50V and 1A. 50V/1A= 50 ohms.

Therefore it would be logical for the sand-amp to use a cap with 100 times more uFd than the Fender. 20uFd*100=2,000uFd. You (probably) really have two 4,200uFd caps in series, 2,100uFd. Nothing has changed, except we take lower lower voltage, higher current, and proportionally more uFd.

Matt is correct about ripple ratings, but in my experience audio amps rarely run into trouble. We "over-size" caps for ripple reduction, and that generally leads to a part with ample ripple current rating. Also we don't work at full power continuously. Do be suspicious of a first cap replacement that is much smaller than the original. Max ripple current increases with physical case-size but decreases with internal resistance. The newer caps have less internal resistance, and can take the same ripple current in a smaller case, but don't let this go to extremes.

> The peavey I'm working on now has 4200uF at 63V.

Just one??? That would be about a 32 watt 8 ohm amp, and it needs a second large cap for output coupling. It has generally (for the last 30 years) been cheaper to use two power caps and no output cap.

> Can I just use a 1000uF value cap instead of a 4200uF in the filter section?

Amp will run, hum will be 4 times higher, and when clipping the buzz-modulation will be offensive. In sustained full-power use, you might actually run into ripple-current trouble.

3,900, 4,200, 4,700, 6,800 would all be appropriate sizes. In hi-fi, more is better yet and in repair, your labor costs dwarf the incremental cost of upsizing the caps; as long as the screws are out and the iron was hot, I'd go 10,000, even 22,000uFd if the rectifier was ample. Or get the ten-for discount: ten 2,200uFd as two parallel arrays of 5 gives two "11,000uFd caps". In gitar-amp work, oversizing may reduce buzz but can also "harden" the overdrive, which is not always preferable; I'd stay near original values unless you have an urge to play with a different sound.

DigiKey has power caps in about every size I've ever needed.

 

[Mbira]

the filter cap is inversely proportional to the load resistance.

What is the formula to figure out "proper uF"?

Thanks for the good explination!

The peavey does have two 4700uF caps, but they are giving a + and - voltage (+-50v)
There are also two 1000uF 35v caps in the power supply...

 

[Matt Syson]

Hi
There isn't a 'proper' value, it is what the orignal designer planned for that unit. You can work it out properly knowing the mains frequency, load current and capacitance but it is too late tonight for this!
Matt S

 

[CJ]

over filtering with too much C can make a git amp sound weird.
wats up with dat :?:

guess thats why jimi never used a macintosh.

 

[PRR]

> The peavey does have two 4700uF caps, but they are giving a + and - voltage (+-50v)
There are also two 1000uF 35v caps in the power supply...

For analysis, the main power supply is one (center-tapped) 100V supply with 2,350uFd total capacitance. This is not very different from a Fender tube amp, allowing for the different voltage/current ratio.

The couple lower-uFd lower-volt caps are probably a second level for the preamp stages. A Fender tube amp has a 20uFd off the rectifier to feed the power tubes, a resistor and a 10uFd to feed the driver, another resistor and small cap for the preamp. You usually have several layers of filtering: somewhat dirty for high-level stages, cleaner power for low-level stages. That's a lot cheaper than making the main supply clean enough for the preamps.

> What is the formula to figure out "proper uF"?

> There isn't a 'proper' value,

Right.

There is not a perfect answer, so no formula. Even if you specify an "acceptable answer", the formulas are ugly. I'm not sure there is an exact formula; we "solve" by random guessing and approximate checking.

> it is what the orignal designer planned for that unit.

On existing gear, that's always your best starting point. Someone else already did the guesswork already; assume he was not a total fool and use the original values.

But round up. The designer was hoping to live on the profit from thousands of units, so there was a good reason to use the lowest-cost (least uFd) cap that didn't sound awful in the showroom. If the unit is loved enough to be worth repairing, the economics turn upside down. A high-value technician hand-replacing low-value parts: I'd charge you $30 to replace $5 of caps. If the amp could benefit from smoother more stable power, put in $7 or $10 of caps: it does not really change the labor cost which is most of the job-cost.

For new design, Ohms * uFd = 100,000 will give you DC that does not suck too bad for a push-pull power amp. Not too bad, but not very good: consider that an absolute minimum, and try to go 2 to 5 times higher if it does not break the budget.

> over filtering with too much C can make a git amp sound weird.

Plucked strings should have a loud attack and a softer sustain. When you push a gitar amp into overdrive, attack and sustain want to come out the same, the amp's maximum power. That's not natural, and boring. But historically (and perhaps by design), many push-pull amps have a "soft" power supply that will deliver high power for a few milliSeconds, falling to a lower sustained power. This is more like what we want to hear.

> jimi never used a macintosh.

I believe the Dead did, in one version of their wall of speakers; something modest like a 50W into a couple dozen 12-inch guitar speakers. But then, they didn't do Heavy Metal.

 

[alk509]

C = (I x T)/Vripple(p-p), where:

C = capacitance in Farads,
I = load current,
T = rectified AC wave's period in seconds (1/120 or .008333 in the U.S. , 1/100, or .01 in EU), and
Vripple(p-p) = desired ripple voltage (peak to peak).

Suppose you live in the U.S., your circuit pulls ~20mA from the PSU, and you can live with a 100mV (peak to peak) ripple:

C = (.02A x .0083sec)/.1V
C = 1666uF

So you's stick a 2200uF cap in there and go to bed. Actually, you usually end up using a 4700uF cap anyways! :grin:

Peace,
Al.

 

[NewYorkDave]

I've heard that a stack o' Macs (MC3500) was used for sound reinforcement at Woodstock. I can't say for sure... I wasn't there. I was "sittin' in Queens, eatin' refried beans." :razz:

http://www.mat-hifi.co.jp/amp/amp2%20007.jpg

http://www.mat-hifi.co.jp/amp/amp2%20011.jpg

http://www.ys-audio.com/wce/mclntoshmc3500.JPG

 

[Mbira]

Thanks guys, I really appreciate it...

 

[The Kid]

C= I(DC)÷ f × V(rip)

So, if you had 200mA load current, and no more than 3V (p.p) ripple, then: C= 200mA÷ 120 (full wave rec.) Hz × 3V=555 µF.

(Same as Al's, just inverted!)

T = rectified AC wave's period in seconds (1/120 or .008333 in the U.S. , 1/100, or .01 in EU),

That's for a full wave rec. (double the freq.) Adjust accordingly for half wave.

 

[Larrchild]

Mother of Sweating Jesus! 8 6LQ6 Horizontal Output tubes with individual trimpots in that Mc Intosh 3500 amp, Dave.

I bet the Dead were Greatful.

 

[ciminosound]

over filtering with too much C can make a git amp sound weird.

Plucked strings should have a loud attack and a softer sustain. When you push a gitar amp into overdrive, attack and sustain want to come out the same, the amp's maximum power. That's not natural, and boring. But historically (and perhaps by design), many push-pull amps have a "soft" power supply that will deliver high power for a few milliSeconds, falling to a lower sustained power. This is more like what we want to hear.

From what I learned from the "amp gurus" this is the "compression/sustain/distortion" effect we like in the older, (particularly) small tube amps. The larger amps like a Twin or Marshall 100 watters do it but at a ridiculous SPL. Of course the tube rectifiers also can be "slower" and "less stiff". Reminds me a little of an optical type compresser. Peaks get through then get squashed. The amp kind of runs of of juice.[/quote]

 

[Brian Roth]

When using a regulator after the rectifier/filter cap, it's important that the "lower end" of the sawtooth doesn't fall below the drop-out voltage of the regulator. Page 4 of this National app note touches upon the subject:

http://www.brianroth.com/library/national-ps-design.pdf

In the example using a 2000 uF cap with a 1A load, the ripple is 3V. Thus, if the "raw" DC without a load was 15 VDC, and you were attempting to derive a 12 VDC rail via a common "3 legged" regulator, you would run into problems since the raw DC rail would sag down below the drop out rating of the regulator 120 times per second.

Just another 2 cents...

Bri

 

[Pbassred]

Do you just want the highest uF for the voltage you need or what?

I don't think that this actually got answered (simply). As far as a power supply is concerned, if he found some 8200uF instead of 4200uF, would that be a bad move?

 

[PRR]

> 8200uF instead of 4200uF, would that be a bad move?

In commercial work: you are cutting your throat. You only make a buck profit per box, the big caps cost $2 more, now you lose a buck a box.

In handiwork: yeah, reasonable upsizing is usually not bad. I had a 1938 amplifier with 16uFd main cap, I put in two 470uFd caps. No hum, and it kept playing almost a minute after unplugging.

There can be gotchas. Vacuum rectifiers can bust a gut trying to charge-up a big cap at switch-on. Often 20uFd-40uFd is the suggested max cap, unless you have a very sharp pencil and days of time (or a PC) to calculate peak currents. My amp had the wrong rectifier bottle, and I detest hollow rectifiers anyway, so I just did a couple 1N4007. Even so, I would not expect 4007 to live through millions of hot-starts with 470uFd at 400V; but I was just fooling around.

That was just over-kill. I didn't need anywhere near 470uFd. As proved by the reason I had the spare 470uFds: I was attacking a hum in a signal generator. I threw in some big power caps. Hum did not change. Original caps WERE big enough. The hum was picked-up on the tuning circuits, not coming in the power supply. So I abandoned that plan and had the 470s laying around when the other amp fell into my hands. The original wax cardboard electrolytics (the oldest I have ever seen) were clearly past their "best by" date, looked too old even to blow-up, were just dead meat. I needed something more than 15uFd, all I had was the 470s, and thanks to Progess, they fit fine.

 

[Svart]

PRR brings up a good point.. charging a large cap.

But first, a public service rant...

As I have seen (not repeating as truth mind you) large caps often mean something is terribly wrong in the PSU world. Needing tons of charge like that usually means your trafo is a little on the strained side. I would upsize the trafo in VA and downsize the caps in uF.

Back to the discussion...

I've seen trafos really strain to charge a large ass cap, as well as overpower insufficient rectifiers that were already close to the minimum specs to even work without burtsting into flames, which is strange to me since buying a much more sufficient diode is likely a few cents more at the most.