As for power supply filters, it's very much like the situation found with loudspeaker crossovers. Higher-order filters suppress more in the stop band. Also, whatever slope is used (first-order, second-order, etc.), frequencies further into the stop band receive more attenuation. So the idea is to make a filter that's low-pass, since all we want is DC, or 0Hz. Pretty much the larger the values, the better. A full-wave bridge presents a 2X (100Hz or 120Hz) pulse train instead of 50Hz or 60Hz, so the doubled frequency is further into the stop band, so more attenuation. And by using larger values and more filter poles, you increase attenuation of the primary ripple frequency as well.

This is all pretty much just basic filter stuff. You make a basic low-pass filter for power supply, and possibly a series choke or resonator between the supply and load to reduce ripple further. You can visualize the circuit having perfect inductors and capacitors for this part. But one thing to consider is that the inductors have a series resistive component, and capacitors act like they have a resistance in series and in shunt. Large electrolytic capacitors, in particular, lose their effectiveness at high frequencies, so it sometimes makes sense to use a small value capacitor directly across large ones, or in strategic places in the circuit to counteract for some of the effects of real-world imperfections in parts.

That's why I mentioned having some small capacitors across diodes. They'll shunt the switching spike. And small capacitors might also be employed across large electrolytics, to reduce the relatively high frequency components that pass through the power supply. Most of the noise coming out of a rectifier circuit is low frequency stuff. But there is some high frequency energy there too. It's attenuated, but if the filter components are ineffective at midrange frequencies, some noise might get through. Using smaller value bypass caps helps to reduce this.

There are two things at issue here, one that is probably more important in small and medium power supply circuits like are used in home hi-fi systems. There is a time when a rectifier goes open, which introduces a small switching spike. This can be pretty significant on high-voltage, high-power systems and you'll see small caps across solid-state switching components in circuits like those. The spike from going open in a high-current circuit can be pretty large under certain load conditions. Even though the crossover voltage is low, at very high currents, the switching spike can be alarmingly large.

All line-cycle rectifiers generate a series of pulses having sine curve half-cycles broken by sharp edges where rectification switching has occured. These transitions correspond to high frequency artifacts. So while the power supply filter is generally thought of as a very low-pass circuit, maybe with line-frequency or 2X notch filters, it still should have some ability to attenuate frequencies several times the line frequency. If the power supply filter becomes ineffective at high frequencies, there will be some noise at high frequency multiples of the line frequency. So that's why quality power supply capacitors are important, and if large electrolytics are used, smaller value bypass caps might be used as well.

As for LED-based CCS circuits, the switching hash is directly present at the tube.

Also keep in mind that the cause of hum in an amp may be totally unrelated to B+ ripple. It can be caused by an improper grounding scheme causing ground loops, inductively produced eddy currents in a steel chassis used as the circuit ground, very low level hum "piggybacked" on top of an ultrasonic oscillation and amplified, running signal lines too close and/or parallel to AC heater wires, and filament-to-cathode induced hum.

I've built C-R-C filter amps with a modest amount of capacitive filtering that were dead silent, by using proper wire routing and a star point grounding system with the RCA jacks, volume pot, and negative speaker binding post all isolated from the chassis.

Take a Dynaco FM3 tuner for example. FM3s always have a little hum. Their RCA outputs use the chassis as a ground. Running a jumper wire from the RCA jack to the ground buss on the multiplex board makes them dead silent. I know, I have one I'm repairing right now.

Another cause of hum is what I'd bet the farm yours is....the filaments of directly heated triodes. The direct AC voltage present on the cathode is one cause. Signal modulation caused by the AC hum is another. The other is the fact that the filament minutely cools when the AC sine wave drops to zero and starts back up. I know that sounds nuts, but it's actually been proven! The minute thermal expansion and contraction causes a resonance in the filament (that perfectly coincides with the 60Hz hum to make matters worse) that makes it create more hum much the same way as the springs in a guitar amp's spring reverb unit produce reverb.

Some megabuck DHT amps have proprietary filament hum reduction circuits. The AC filament voltage is tapped off, negative fullwave rectified (and purposely left unfiltered) and injected at the control grid, with an adjustment trimpot to adjust the phase so that the ripples in the DC will phase cancel the AC hum.

Some other high dollar DHT amps may use a series of LC bandpass filters on the heater leads set at octaves of 60Hz to solve the problem by "filling in the gaps."

Thermionic

I agree with you completely about ground loops. That's the most common cause, and (ironically) it seems to affect expensive separates more than inexpensive integrated receivers and amps. The reason is obvious, there are more separate grounds to tie to a common point without developing a loop. Seems like cable TV is a common culprit too, so if the cable is hooked in to the home theater, it almost always is a good idea to run the 75 ohm coax in through an RF transformer. That will knock off the hum in lots of cases.

I always thought that AC driven filaments would be a source of hum. I would have thought the cloud of electrons stays relatively constant but there certainly must be a small heatup/cooldown cycle from AC-driven heaters. I also would have thought the filament line would be a source of inductive coupling into the relatively high impedance signal lines right nearby. But I'm surely not an expert in tube circuits, and I'm really just kicking ideas around, sort of thinking out loud. I'm definitely soaking up the ideas around here, so thanks for your input.

My impression is that silence is not the only benefit of a good power supply: a clean DC will also lead to better sound? Or is dirt in the power supply ALWAYS translated as hum?

Finally, from a pragmatic standpoint, what i have learnt is: Use a tube or a full wave FRED rectifier, along with a CLC filter, with fast caps in parallel with the electrolytic caps, or use blackgate/cerafine. Coorect me if i am mistaken!
This knowledge itself is very useful for us newbie amp hobbyists.
thanx
-akhilesh

Akhilesh, trash in the DC does not manifest itself as hum, only excessive ripple from the residual AC component. Trash manifests as a lack of focus or coherency,a graining or veiling, or other anomalies. Indeed, IMO, tube rectifiers or FREDs with a choke input (L-C) is the ultimate. Even if you have no choke, adding film/foil parallel caps to the aluminum electrolytics can make a difference. Fast Caps, Cerafines, or Black Gates, anything to ditch the aluminum electrolytics, will yield a blacker background, smoother presentation, and often better microdetail.

Wayne, you probably have very little residual ripple. A bit of hum is just gonna happen with DHTs. I know, I'm a fellow 2A3 fanatic too!

Thermionic