Filter capacitors???

[Chuck H]

Or maybe more properly "power supply capacitors".

I read (don't remember where) that it actually takes only a few MFD to "filter" hum from the circuit. And that additional MFD are there as a sort of resevior to help smooth out the DC. I ran some tests with Duncan PSU and now I'm wondering...

How many MFD does it take (to screw in a lightbulb...no) to filter out the 60Hz hum? Let's assume for the purposes of this Q that we aren't trying to lessen ripple.

And, if the above is true, why are the power supply caps on the preamp end of the supply typically so large for the task. As in 20u to 50u. A preamp tube doesn't draw much current. So if it takes a few MFD to dump everything AC down to 20hz, wouldn't that already be large enough to smooth out the power on that end of the supply?

With few exceptions it just seems like most of us just build our power supplies like they've always been done without questioning why. So I'm interested in anyones take on it.

Chuck

[tubeswell]

Or maybe more properly "power supply capacitors".

I read (don't remember where) that it actually takes only a few MFD to "filter" hum from the circuit. And that additional MFD are there as a sort of resevior to help smooth out the DC. I ran some tests with Duncan PSU and now I'm wondering...

How many MFD does it take (to screw in a lightbulb...no) to filter out the 60Hz hum? Let's assume for the purposes of this Q that we aren't trying to lessen ripple.

And, if the above is true, why are the power supply caps on the preamp end of the supply typically so large for the task. As in 20u to 50u. A preamp tube doesn't draw much current. So if it takes a few MFD to dump everything AC down to 20hz, wouldn't that already be large enough to smooth out the power on that end of the supply?

With few exceptions it just seems like most of us just build our power supplies like they've always been done without questioning why. So I'm interested in anyones take on it.

Chuck

Interesting question about the pre-amp supply caps. My guess would be that the pre-amp is more sensitive and any ripple that gets into it will be amplified through more of the signal path . However having said that, those early champ schematics had 8uF caps in there, (and were they hummy?) I found almost no difference between using 10uF and 22uF for the pre-amp filter supply in a 5F2A I built.

[Enzo]

Look up power supply design in an electronics text. Asking how many microfarads it takes to filters like asking how big a nail is needed to hold up a shelf. Is the shelf going to hold a couple litle figurines, or a half dozen bowling balls?

If you rectify some AC and put no load on it at all, a very tiny cap will make nice smooth DC. The ripple will come from the little losses in the cap due to its own leakiness and ESR. After all, if a cap charges up to the peak value, unless it discharges a little, there is no place for ripple to form.

The more current you draw from a cap, the lower its stored voltage drops. When it is constantly being replenished by rectified DC, then 120 times a second it gets recharged. SO the classic ripple sawtooth waveform happens. The start of each cycle is the cap voltage charging up to the peak of the AC. Then the following slope down is hte voltage dropping as current flows from the cap into the load.

Make up a small rectified AC supply. Put a small cap on it - 10uf for example. And make up a variable load - a pot or something. Scope the filter cap. With no load, it should be pretty smooth DC. Now add the pot as a load and slowly turn the resistance lower and lower. As the load increases, so will the ripple. Use a low current transformer like a little wall wart to make it more obvious.

SO power supply design has an aspect wherein you must decide on the current it will source. Don't recall the numbers but seems to me there is some sort of farads per amp formula to follow.

In the typical tube amp, the first filter cap is the main one. It does a half-assed job and makes relatively smooth DC. SInce the ripple cancels in a push pull output stage, it doesn;t have to be perfect. Then the screens take off from node 2. At this point the DC should be pretty well clean. How large are these caps? How much current do they need to supply? A 400v supply in a 100 watt amp is not the same demand as a 400v supply in a preamp only unit.

If you use too small a cap, it is not like the DC won't be filtered, but what will happen is peak demands won't be met. Play real loud and the ripple will color your sound, peaks will get mushy, etc. Crank the bejesus out of some small amp and scope the first B+ node. Watch how when the guitar is muted, the B+ climbs to its top value and looks clean. Now bang out a loud chord and watch as B+ drops then recovers and also how ripple increases.

How can we assume we are not trying to eliminate ripple? Unfiltered DC is nothing BUT ripple in a practical sense. The circuits draw their current from the caps, not the transformer, the transformer and rectifiers' job is to replenish the caps. You could make smooth DC with a lot smaller caps, assuming you were not going to actually play the amp. There you go, let's assume we built the amp not to play, but to just sit there powered up. We could use smaller caps.

You got a minute? Take the HV winding of some PT and rectify it - no filter. SCope it. You'll see the endlessly repeated stream of half sine waveforms. Now add a small cap across it. I mean like a .047 or a .1uf film cap. See how much cleaner the DC got. It would be useless for audio, but that tiny cap does filter the supply. The impedance of your scope and the leakage of the cap are the load and will be the cause of the ripple you see. Oh there is probably also dielectric absorbtion and other things going on too.

SO it doesn;t take a lot to filter a pulsating DC voltage, but for it to be useful the filters have to be larger.

Farther downstreasm in the preamp, the filter caps are there more for decoupling than for ripple filtration. They are still serving to filter, but instead of ripple, now they are filtering traces of signal that can get impressed on the B+ rail. And if they get there, they can sneak into other stages.

What you don't want is the strong signal at the PI for example, getting into the B+ that feeds the first stage - the sensitive input stage. If you use 10uf instead of 20uf for the first stage node, it won't sound much different, but you might notice a difference in stability or in odd artifacts in the sound, or even parasitics.

[Chuck H]

Wow. Thanks Enzo. Great laymans explaination (perfect for me). I will run some scope tests, when I get my scope, and learn to use it. Still waiting for it to arrive.

You were very helpful on my "Oscope advice" thread too. If it's not too much trouble, could you swing back by that thread and critique my load box. I think it's pretty good but it's stuck in the "more replies below current depth" range. So I havent gotten any critique.

Thanks

Chuck

[loudthud]

To expand on what Enzo posted, here's some other things you'll notice if you stare at your scope long enough. Consider the classic 5F6a Fender Bassman and derivitives like many Marshalls. When the amp is idle, there is not much ripple on the main B+. The peak might be 470V and the trough (bottom of the ripple) might be 450V. Now look at the screen supply. It should be about 460V, half way between the peak and the trough minus the loss caused by the DC resistance of the choke. And there will be very little ripple left at the screen node. If anything you might see a small distorted sine wave. Now hit a big power cord. The main B+ will sag a little (depending on transformer and rectifier losses), and the ripple on the main B+ will increase. The peak might be 460V and the trough might be 420V. The screen will drop down to 440V fairly quickly but still not have much visible ripple. That's the magic that a choke performs. It turns that nasty sawtooth into little sine waves but the screen voltage has saged 20V.

Now the downstream filtering for the preamp has to deal with the 20V sag. If you look at the top of the unused volume control (of the channel you're not using) you will see the voltage will go negative very slowly when you hit the power cord and climb back slowly and go positive slightly when you mute the cord. You can slow the horizontal sweep speed of the scope way down to like 500mS per div. This sub sonic signal gets to the input of the tone driver stage and subtly shifts its bias. The size of the preamp filter and the size of the coupling and cathode caps control how much the bias will shift and how fast. I think this is why original 5F6's had 250uF on the common cathode of the first stage.

Another thing you can look at. If your scope has a "line sync" function use that to lock the sweep to the ripple. Now connect the input of the scope to the speaker. When you hit that big power cord you can see the little divits that the ripple makes in the tops and bottoms of the square waves of the clipped waveform. I've made several amps with Pi filtered B+. There is a big choke and a second filter cap on the main B+ and very little ripple makes it the the OT. Something like a 2V pk to pk sine wave. You won't see the little divits on the output square wave. But if you look at the main B+ you start to see a rectified version of the guitar signal's lowest frequencies, especially with a bass guitar. The value of the second filter capacitor affects how much signal there is on the B+ and how firm the bass sounds through the speaker.

For a guitar amp, those little 120Hz divits are actually audable and you will notice if they are gone. But if they are too big they make lead notes sound out of tune or like there is an out of tune instrument playing behind you. So getting the right amount of filtering is one of those things that falls into the personal preference catagory. Experiment with different values of capacitance until you find what sounds good to you.

[mooreamps]

. However having said that, those early champ schematics had 8uF caps in there, (and were they hummy?) I found almost no difference between using 10uF and 22uF for the pre-amp filter supply in a 5F2A I built.

Ya, they were hummy due to the Fender style chassis grounding schem, and not using a dedicated dc / star bus ground. Also, plates on the preamp and power amp can tolerate a few volts of ripple, since the delta change in plate current is very low with respect to the delta change in plate volts. Probably why you saw very little difference between 8uF, 10uF, and 22 uF.

Now, having said that. Ripple on power tube screens, different story. Must be very clean and well filtered. Fortunity, not hard to do since the power tube screen pull very little current. In fact, if you have a very very well made power tube, where the wires on the screens are "exactly" in the mechanical "line of sight" alignment with the wires on the control grid, then screen current is very low.

-g

[Enzo]

And a Champ, being single ended, has no power stage B+ ripple cancellation, so it tends to be hummier for that reason too.

[dai h.]

some of the Champ layout grounding schemes look wrong to me also. Maybe I'm overthinking, but I've wondered if Fender did it on purpose to encourage people to upgrade from the "student model". "...and that's why you need to upgrade to a Twin" or whatever.

[Chuck H]

Well, I don't think much of this has to do with what Mr. thud was talking about at this point. But I rebuilt a VibroChamp for a friend (not the same I know) and in the process changed it to a humbucking filament arrangement and elevated the false CT at the cathode R. And at that point changing the standard ground scheme did make it even quieter. So I suppose that the stock ground scheme in the older Champs was done for convinience figuring that it wouldn't make much difference to the hum level anyhow without also changing the filament wiring. More $$$.

Chuck

[Enzo]

There was no conspiracy, Femder made all his amps in pretty similar fashion. A single ended Champs is just more hum prone. The almighty $$$ was king there. Early Champs had a choke pi filter for the B+ before the output plate. Then they eliminated the choke for a resistor in its place - much cheaper. Then they moved to plate to the first cap, eliminating a cap. Cheaper yet. They were not trying to push you to a more expensive model, they were just going for cheap.

[jrfrond]

Champs also ground one side of the heater supply, but I've never really seen that cause a problem.

With all of the great explanations here, one thing I didn't see anyone mention (and I apologize if I just don't see it) is how filter capacitance affects bass response when an amp is push. If the supply is stiffly-filtered (100uF or better) and the amp is pushed into overdrive, the result is a percussive "thump" on transients, rather than the aforementioned spongy response and hum modulation that can ensue when smaller caps are drained during transients and ripple occurs. THIS is one of the points that sets a Marshall apart from a 5F6A Bassman, even though they are essentially the same circuit. In addition, the ubiquitous 4 x 12 stack cabinet emphasizes the thump, yielding that Marshall kick that we all know and love. Bumping-up the main B+ filtering in a 5F6A can be a real eye-opener.

[Steve Conner]

There was no conspiracy, Femder made all his amps in pretty similar fashion. A single ended Champs is just more hum prone. The almighty $$$ was king there. Early Champs had a choke pi filter for the B+ before the output plate. Then they eliminated the choke for a resistor in its place - much cheaper. Then they moved to plate to the first cap, eliminating a cap. Cheaper yet. They were not trying to push you to a more expensive model, they were just going for cheap.

I kind of agree. I've modded a Champ-style amp that had the pi filter with the resistor. It hummed quite badly, and swapping the resistor out for a choke made the hum just completely go away.

I also built a 1x12" combo amp with two 6L6s that uses the pi filter with 47uF before and after, and about 1lb worth of choke that passes the full plate current. It has plenty of bass, and when I crank it, it seems to squish like a compressor.

Oh, and that formula was 2200uF per amp of current draw. It's more suited to low voltage power supplies for transistor equipment. A tube amp can happily run with 20 or 50V of ripple on a 450V B+, but a transistor one wouldn't be so happy with 50V of ripple on a 24V rail!

So tube amps can use much less than the formula calls for. For example my combo amp draws nearly 300mA of B+ on full blast, which would need 700uF by the formula, yet I only have 100 in total.

If you use a choke, you can get away with smaller capacitors too, which lets you enjoy some sponginess without being bothered by hum. Except in single-ended amps, where sponginess isn't an issue. Being Class-A, the current draw doesn't change with the signal level, until you start really driving hell out of the power tube, and then the draw actually decreases.

[TheTinMan]

Another 5F2-A experience - increasing the values of the filter caps on the plates and screens significantly above vintage values added too much bass in my build without any noise-reduction benefit. Seems like there's a happy medium for each amp, depending partly on speaker & cab.

2 pico-cents

Chip

[Robert M. Martinelli]

Another 5F2-A experience - increasing the values of the filter caps on the plates and screens significantly above vintage values added too much bass in my build without any noise-reduction benefit. Seems like there's a happy medium for each amp, depending partly on speaker & cab.
2 pico-cents
Chip

Well, increasing the filter caps' value beyond certain points has a lot more to do with dynamics than with boosting frequencies, each PS should be designed with the load it will be connected to in mind, and the rectifier type will also make a huge difference, as the high differential resistance of a rectifier tube will cause the voltage to drop heavily under dynamic conditions, causing the amp to sag, while SS rectifiers have a low Rd so the HT will be more stable under dynamic conditions. The other problem with tube rectifiers is their low current capability, so it' s practically impossible to increase the filter capacitors' value without being rough on the rectifier tube; this is not an issue even with small diodes like 1N4007, which can handle a continuous current which is about four times the current a 5AR4 can take.

Talking ' bout Marshalls, they' re indeed the living proof of the above statements, being the kick they show the result of the combination of a SS rectifier AND over-dimensioned filter caps.

As already noted, bigger filter caps are an energy reservoir, so IMHO they tend to enhance the low end because they improve the amp' s ability to respond to transients ( the amp' s "dynamics" ). We perceive as a "bass boost" something that actually is the result of a "dynamics" improvement.

My .001 ( Euro ) cent

Bob