Understanding power supply filters in your amp

A sub forum for our most useful repair, maintenence & upgrade threads.
Post Reply
User avatar
øøøøøøø
PAT. # 2.972.923
PAT. # 2.972.923
Posts: 11912
Joined: Sun Jan 06, 2008 8:26 pm
Location: Brooklyn, NY
Contact:

Understanding power supply filters in your amp

Post by øøøøøøø » Wed Mar 09, 2011 11:42 pm

This is some stuff I've been gaining a better understanding of lately, and I thought I'd share. If anyone is smarter than me, feel free to correct anything I say here. I'm just a hack, and have no formal training. Anyway, here goes.

So we often talk/hear about "filter caps" or "smoothing caps" in an amp's power supply. We know that their job is to take pulsating DC voltage from the rectifier and turn it into smooth, battery-like DC voltage. They do this by charging as voltage rises, and discharging as voltage drops.

But there's another way to look at it. This is that the "pulsating DC" is DC but with an "AC component." That means that basically, it's alternating current and direct current mixed together. Realizing that this is so, we can remove the AC component and we'll have DC left over. To see why this is so, let's first take a look at a graphic of the output of a full-wave rectifier.

Image

This is a very instructive graphic. You see the alternating current on the left. It's a sine wave, swinging positive (above the zero 'line'), then negative (below the 'line'). When it's rectified, the part that was below the line gets flipped above the line, and what was a sine wave is now a series of 'humps' that never goes below zero. Make sense? We've talked about all this before. If you hooked this signal up to a speaker, you'd get 120Hz hum in the US, and 100Hz hum in some other parts of the world.

Here's where it gets interesting. Now look at the 'green bits' of our rectified signal. Imagine we drew a line right through the middle and made this the "new zero." It would look something like this:

Image

Well, it looks almost like a sinusoid, but the 'green' is in the wrong places, so let's fix that.

Image

That's better! Now this is interesting. It looks sorta like our sine wave from before, except kinda pointy on the bottom. It almost looks like a triangle-shaped wave that has had its tops rounded off, or soft-clipped... but only on the top. THIS is what it would look like if we subtracted ALL DC from the rectified signal and were left only with the AC component. Put another way, our rectified, pulsing DC is just THIS waveform, but offset in the positive direction by a pretty large amount. Now stay with me, I'll get to this in a second.

When dealing with AC signals, there's a name for this phenomenon--appropriately, it's known as "DC offset." This is when you add some DC to the AC, and it shifts the whole wave toward the positive direction. In audio, this happens occasionally, and is a "nasty..." it happens on accident, and it can wreak havoc on headroom, and sometimes gear.

Our pulsating, full-wave rectified DC is really just the AC waveshape shown above with a LARGE amount of DC offset. Does that make sense? Basically, one way of looking at it is that it's a bunch of DC and a bunch of AC added together, and it looks like our rectified signal on the right of the first picture above.

What we want is smooth battery-like DC, so we want to remove all the AC component and JUST leave the DC. How do we do this? same way we remove frequencies in audio... with EQ filters!

That's right, your power supply filters are really just specialized EQ filters... in this case low-pass filters.

Let's look at the schematic for a low-pass filter. This is an EQ filter that "allows low frequencies to pass" but blocks high frequencies. It looks like this:

Image

It's just a resistor and a capacitor. Simple. Here's what it does to alternating current (remember that audio signal is alternating current):

Image

The higher the values of the components, the lower the "corner frequency," or "cutoff frequency." This is the frequency at which signal is attenuated 3dB. Above this frequency, AC signal rolls off at 6dB per octave (AKA 20dB/decade), and below that frequency, it remains untouched. Your guitar's tone control is a variable RC low-pass filter just like this. Its corner frequency somewhere pretty high up in the audio range, where it's useful in that application.

Now look at ANY schematic of a guitar amp, and you'll see a bunch of these low-pass filters in a row in the power supply. Usually the resistors are moderately high values, and the capacitors are several microfarads. This makes for very LOW corner frequencies... something between about 2 and 10 Hz, usually.

Let's take an example. Look at this schematic for the 5E3 deluxe. The two low-pass filter resistors in the power supply are circled:

Image

the caps are 16µf. The two low-pass filters are preceded by another 16µf cap to form a capacitor input filter, or pi filter. But that's beyond the scope of this.

OK so our resistors are 5000 (5k) and 22k ohms, and our caps are 16µf.

Here's a calculator that you can play around with to find the corner frequency of the power supply filters in your amp. Actually, it's designed to calculate EQ filters for stompboxes, but it works just the same for our application.

http://www.muzique.com/schem/filter.htm" onclick="window.open(this.href);return false;

Plug in 5000 and 16µf, and you get a corner frequency of 2 Hz. Pretty good. Plug in 22000 (22k) and 16µf and you get a corner frequency of 0.5 Hz. That's REALLY good, except you lose a LOT of power through that 22k resistor. In this case, that's intentional-- it feeds the preamps. In most cases, this is not so good. In this case it works great, because the preamp has to be quiet and considerably lower voltage.

Back in the days of the 5E3, capacitors were expensive in large values. Now, not as much. If we were to use our 5000 ohm resistor (5k) and double the capacitor value to 33µf, we'd get a corner frequency of 1 Hz. Now you may say to yourself "1 Hz doesn't sound like a huge improvement over 2 Hz," but looks can be deceiving. Since these filters roll of at 6dB/octave, and an octave is a doubling of frequency-- moving the corner frequency from 2 Hz to 1 Hz gives you an extra 6dB of ripple attenuation.* Remember that the difference between 10,000 and 20,000 Hz is 1 octave, and the difference between 1 and 2 Hz is also an octave. It's NOT linear!

Now if we're in the USA, our ripple frequency is 120 Hz. If the corner frequency is 4Hz, then we're down 3dB at that point, and we lose 6 more dB with each additional octave. By the time we get up to 120Hz, we're attenuating that frequency many tens of dB. Arrange a few of these in a row and you can get some pretty serious attenuation of the ripple frequency.

Now there's another type of low-pass filter... the LC filter. This looks much the same, but instead of a resistor, is made of an inductor and a capacitor. The choke in your amp is the inductor in the LC filter. LC filters have the advantage of having steeper filter orders... like 12dB/octave. They also have the advantage of greater filtering with less power loss (like the power loss in that 22k resistor above). This can give you SERIOUS attenuation, if the corner frequency is low enough. They look like this:

Image

The disadvantage is that they can be very expensive. Inductors are used in many console EQs, and in fact are sought after for their 'sound.' The varitone in a Gibson ES-345 has an inductor in it, and your Vox or Crybaby wah-wah has an inductor in it, though it's a different type of filter (band-pass with a steep "Q"). Lots of speaker crossovers also use inductors in them, to make precisely the types of high-pass and low-pass filters we're describing here. In speaker crossovers, the steeper the slope, the more precise you can get with tuning the crossovers, in many cases. It's actually not too terribly different from what we're doing here.

In our application, basically, it's like we're turning a big "tone control" all the way down to filter out the unwanted signal, in this case 120Hz ripple voltage (and its harmonics). We have two types of tone controls... the resistor-capacitor "guitar tone control" type, and the inductor-capacitor type.

Done properly, this effectively removes the AC component from the pulsating DC voltage and leaves us with ONLY the DC component. We use this clean power to supply the critical parts of our amp. It really goes to show you that there are a few 'building blocks' that are used all over the place. Yes, there are low-pass filters in your tone stack. But they are also in your power supply!



*In guitar amps, it's not necessarily the case that "more is better," though. Sometimes too much filtering can make an amp feel "stiff," because in the "real world," capacitors are not ideal, and part of what we've grown to like about these amps is that they don't always 'hold together' that well when fed lots of volume and bass. So the power supply turning to mush is part of 'the sound' sometimes.
Last edited by øøøøøøø on Thu Mar 10, 2011 11:23 am, edited 4 times in total.

User avatar
Soochdaddy
PAT. # 2.972.923
PAT. # 2.972.923
Posts: 2389
Joined: Wed May 30, 2007 9:47 pm
Location: SoCal

Re: Understanding power supply filters in your amp

Post by Soochdaddy » Thu Mar 10, 2011 12:20 am

øøøøøøø wrote: So basically, it's like we're turning a big "tone control" all the way down to filter out the unwanted signal, in this case 120Hz ripple voltage (and its harmonics).
Best. Analogy. Ever.
"I want to be able to purchase gin in bulk at Costco on a Sunday. So I totally voted." - chrisrnps

"Always socks on cocks-as to protect my low hanging guitar from getting 'cock rash' when playing in the nude, of course." - basique

User avatar
al30
PAT. # 2.972.923
PAT. # 2.972.923
Posts: 564
Joined: Fri Mar 12, 2010 6:55 am

Re: Understanding power supply filters in your amp

Post by al30 » Thu Mar 10, 2011 6:32 am

Nice post

User avatar
i love sharin foo
Mods
Mods
Posts: 4889
Joined: Tue Sep 12, 2006 8:26 am
Location: Pittsburgh, PA
Contact:

Re: Understanding power supply filters in your amp

Post by i love sharin foo » Thu Mar 10, 2011 7:33 am

I've always wondered what exactly the filter caps did too. THanks a lot for the post. I am accumulating all the parts to build my first amp right now, so it's nice to gain a better understanding of what everything does before I get started.
I've fallen it's true but I say unto you, hold your tongues until after I've spoken- Townes Van Zandt

User avatar
øøøøøøø
PAT. # 2.972.923
PAT. # 2.972.923
Posts: 11912
Joined: Sun Jan 06, 2008 8:26 pm
Location: Brooklyn, NY
Contact:

Re: Understanding power supply filters in your amp

Post by øøøøøøø » Thu Mar 10, 2011 7:47 am

I updated the post with some more graphics and more detailed explanations.

User avatar
OffYourFace
Mods
Mods
Posts: 12136
Joined: Thu May 03, 2007 2:59 pm
Location: Sherman Oaks, CA

Re: Understanding power supply filters in your amp

Post by OffYourFace » Thu Mar 10, 2011 11:17 am

dyn-o-mite!
synth tech at www.rosensound.com

User avatar
46346
PAT. # 2.972.923
PAT. # 2.972.923
Posts: 2559
Joined: Sun Feb 27, 2011 1:39 am
Location: South Los
Contact:

Re: Understanding power supply filters in your amp

Post by 46346 » Wed Mar 16, 2011 12:29 pm

dude! can we get you a grant or something so you can teach us amp electronics over OSG? i'm digging this!
well written - thanks!
Dream Apes, ACME, Malcolm Mooney, Cat Museum, Cooling Time

User avatar
jorri
PAT. # 2.972.923
PAT. # 2.972.923
Posts: 1848
Joined: Thu May 07, 2009 1:53 am
Location: bath, UK
Contact:

Re: Understanding power supply filters in your amp

Post by jorri » Sun Jan 01, 2017 7:45 am

can anyone summarise what happens when caps leak electrolyte and old? I'm told they can start blowing fuses and become noisy, so is this just an intermittent short to ground and loss of capacitance?

Just replacing some big 4700uF caps and hoping that would resolve a fuse blowing problem. They are after the transformer, which might explain the huge difference in capacitance due to impedance, as well as only being 50V rated (although i then bought 100v anyway) as opposed to the usual 10-220uF 500V caps.

Playing around with values in my own pedal designs is pretty important. I do notice a certain undesirable tightness as high as 220uF and its not really making it darker in the same way as a signal path or emitter cap would.

Post Reply