# Why do tube amplifiers have such (relatively) small reservoir capacitors?

In learning about guitar amps. I've been studying the schematics for solid-state and tube amplifiers, and I found a pattern that I've never seen an explanation for.

If you consider any of the main styles of tube amplifier, you'll usually find a reservoir capacitor with a value of <100 μF @ ~500 V - regardless of brand.

Even on a fairly meager solid-state amp however, I regularly see values in the 5000-10000 μF @ ~50 V range.

Obviously, these amps don't run at the high voltage that we see in tube amps, but I'm not sure how that affects this.

The only theory I've come up with so far is that because MOSFETs are so fast, the power supplies need more reservoir capacity compared to tubes.

• Also, 100uF @ 500V is 12.5J of energy. 10mF @ 50V is also 12.5J of energy. So both capacitors can deliver the same amount of energy Oct 3, 2018 at 22:56
• Tube amplifiers use large transformers to convert the tube loadline to the speaker impedance. That transformer allows smaller B+ capacitors. Oct 4, 2018 at 3:31
• Also, in the specific case of a guitar amp, as opposed to something supposed to sound clean, it is often felt to be desirable that the rail sag when played hard. The tube rectifiers have sufficient impedance to cause the cap to take a meaningful time to recharge (Way more then the single half cycle that a modern diode would take) and the audio compression resulting from the rail drooping on an amp that is pretty much in saturation is apparently musically useful. You pretty much have to view guitar amps as being part of the instrument effects chain, not as a classical amplifier. Oct 5, 2018 at 11:28
• Have another look at the schematics and see if there is an inductor / choke in the DC supply output. That will further filter the DC. Oct 20, 2022 at 16:49

The energy stored in a cap is proportional to the square of the voltage -- so a 47uF cap on a 500V rail is storing as much energy as a 4700uF cap on a 50V rail. That's going to be part of it.

Anything else is guesswork -- but I suspect that tube amps are less sensitive to power supply variation, and that people are just used to tube amps having a bit of hum. This is audio, so someone will refute me.

• Yup, Tubes are high impedance devices and don't need as much current. That, and its often "desirable" to use less than ideal power supplies (they still use tube diodes, or throw in a current limiting resistor to emulate tube diodes) to encourage amplifier sag. Oct 3, 2018 at 23:01
• Thanks for the info. Did not know that cap energy storage was prop to the sq of the voltage. That explains the missing order of magnitude of power that I was really confused about. Oct 3, 2018 at 23:38
• Are you saying that it's not to do with filtration of sound, but rather it's simply that the value is selected to achieve some total reservoir energy? e.g. solve for capacitance given voltage and reservoir energy desired. Oct 3, 2018 at 23:53
• If I understand your question, it has to do with the filtration of the power supply! The impedances involved all scale as the square of the voltage, so selecting for the same total reservoir energy will -- more or less -- result in roughly the same percentage-wise hum reduction on the power supply leads. Oct 4, 2018 at 1:22
• P=The output power of a tube amp is transferred through an output transformer so it can deliver speakers the current needed. Tubes are voltage driven devices with a high internal impedance. Depending on the topology used (single ended, push pull, circlotron etc.) they may be very sensitive to power supply hum and nobody like constant 50 or 60Hz hum in a loudspeaker.
– greg
Oct 4, 2018 at 7:29

It has to do with the current consumed by the amplifier. A general value to use was 2,000 uF per amp of current used, to keep ac ripple to less than 100mV. Watts is equal to volts times amps, so tubes had high voltage but low current and transistors had lower voltage at much higher amps, given amplifiers of the same wattage consumed. A tube amplifier might consume 100 mA or 200 mA as a push-pull output, but at 500 volts it was still 100 watts of power. A transistor amplifier would use +/- 120 volts at 10 amps to get 2,500 watts of power. Yes, tube amplifiers waste a lot of power as they are a class 'A' design.

The old tube amplifiers used 500 volts at a much lower current to drive large 6L6 tubes which drove a massive impedance matching transformer. It is basically a step-down transformer with good audio specs to match the high impedance of the tubes to the low impedance of the speakers. At 500 volts the current was low so little filtering was needed, 200 uF instead of 20,000 uF, but with transistors and 50 to 120 volt supplies the current went way up as the amp directly drove the speakers, which could be as low as 2 to 4 ohms.

Now you see the contrast in high voltage vs. high current amplifiers, and tubes vs. transistors. A 5,000 watt amplifier with +/- 120 volt power rails may have 50,000 uF to 100,000 uF capacitors or those values from arrays of capacitors. Another target for low ripple was to keep it to 0.1% of the DC voltage.

In both cases a high ac ripple might get into the pre-amp stages and cause an annoying hum in the speakers.

You mentioned MOSFET amplifiers. Well, they tend to not be used at 100 volts or more as it is difficult to make matched P-channel and N-channel MOSFET's for audio, so they take advantage of their low ON resistance and build amplifiers that can drive 1 ohm loads, which bjt amplifiers normally cannot do without overheating.

Todays bjt and MOSFET amplifiers take advantage of 24 bit and 32 bit audio which can have pounding deep bass (ala Dr.Dre), something tubes and a output transformers or old er bjt amplifiers with output capacitors could not come close to replicating. Can you say 5 HZ? You can bet they have 100,000 uF capacitors in them, maybe several of them.

• Are you saying that because solid state amps drive the speaker directly, they don't get the additional filtration of the output transformer? (Does AC noise travel differently on high vs. low voltage?) Oct 3, 2018 at 23:54
• @idbentley. Any audio grade transformer has a lower limit of 30 to 40 HZ at best, plus an upper limit way below that of a high performance transistors amplifier. Solid state amplifiers would go down to DC but most woofers cannot handle that. The voice coil would burn up. The 5 HZ cutoff is to protect the woofers from DC currents, but that is still very deep bass. (kick-drum)
– user105652
Oct 4, 2018 at 0:06
• @idbentley. AC noise (ripple-hum) is treated as a percentage of the power supply voltage, and if not filtered enough both tube and transistor amps could have hum and noise in the speakers.
– user105652
Oct 4, 2018 at 0:10
• @user105652 "Any audio grade transformer has a lower limit of 30 to 40 HZ at best, plus an upper limit way below that of a high performance transistors amplifier." The claim there is no output below 30~40 hz of a transformer in an audio amp, I have some McIntosh amps from the early 50's(A116) and early 60's (MC250) both go below 30 HZ. Oct 20, 2022 at 16:11

There are two main engineering reasons you see small reservoir caps in tube amps:

1. In the Before Time(tm), i.e. when tubes weren't obsolete, a 47uF reservoir cap was an expensive and massive beast. It may be hard to believe these days, but massive "choke-input" arrangements were a viable and not-uncommon alternative at the time - the choke being about half the size of the power transformer itself.

2. Rectifier tubes will self-destruct due to surge current if the reservoir cap is too large.

P = ½×C×U²

In a tube amplifier the energy of the high impedance output stage (huge voltage variation, milliAmp current variation) power goes through an output transformer to reach the low impedance speaker (few dozens volts variation, Amp current variation). Transistor stages do not need this transformer so the power supply must be able to deliver lot of current on demand.

In guitar amps, the topology used is often a single ended stage because of the nice distortion it generates (lot of even harmonics) and this topology is really sensitive to power supply noise so it requires extra µF in the power supply. Nobody likes constant 50 or 60Hz this is why a self is commonly used to filter the high voltage (adding extra Kg to the beast).