Power Supply Design

The power supply of a power amp is very much the heart of the system.  Instead of blood, it pumps electrons ("current") through the vacuum tube heaters allowing them to do their thang.  The pressure at which the current is supplied is called "voltage".  In addition to the low voltage heaters, tubes require a high voltage (aka "high tension" or "HT"), on their anode.  It's this HT voltage that allows a tube to amplify the voltage of the incoming audio signal from less than a volt to hundreds of volts.

A good portion of the circuitry, especially the big and heavy parts, within a tube amp is actually the power supply circuitry.  There are numerous different common power supply topologies, each providing the general functions of power-line isolation, rectification, smoothing and stabilization.  Isolation is usually performed by a mains transformer.  It keeps the amp circuit electrically separate from the power line and usually steps up the voltage from 110V to a tube friendly voltage (200V to 500V).  Rectification turns the AC voltage from the mains transformer to DC used by the tubes.  Rectification can be done by a rectifier tube or by silicon diodes.  The voltage output of either type of rectifier is not pure DC, it's all positive relative to ground but still ripples (oscillates at 120Hz) around the nominal DC voltage.  To smooth out the voltage, we smooth out the ripple using honker electrolytic capacitors.  Those are usually the largest caps in an amp.  Capacitors take time to charge and discharge so they slow the rate of the voltage ripple almost to the point that it becomes almost perfect DC.

Based on the 6B4G output tube that I selected, I have the following power supply requirements:

• High Voltage (HT): 225v to 325v @ 120mA
• Heater (LT): 6.3v @ 2.3 Amp (1A x 2 for the 6B4G output tubes, plus 0.3A x 1 for the driver tube)

Choosing a power transformer is probably the most critical part choice in a power amp.  It's operating parameters and performance influence most every other part of the design.  R-Core transformers are fairly new technology.  They have the following claimed advantages:

• Lower profile and smaller size.
• Lower stray EM field from the round cross-sectional area and the balanced windings on either side.
• Lower core losses from no cuts the in core and minimized distance between the core and the windings. 
• Lower temperature rise and noise form the round cross-sectional area and tapered slitting.

The main disadvantage is that core saturation is more likely if not careful.  Since efficiency, size and low noise are 3 of my design goals, an R-Core seemed a perfect fit.  I found some nice ones built in China which included a nice set of windings (3 sets of 6.3v windings, 230v and 260v HT windings).

I obsessed for weeks to select a power supply topology.  I've been tweaking it for months since.  The basic topo can be seen in the schematic below.

Silicon diodes for rectification, then a capacitor-choke-capacitor smoothing circuit.  It's a common design for amps which spurge for the extra cost/weight/size of a choke.  I chose a choke which was large enough to do a decent job resisting current spikes, but still small enough that it wasn't hard to justify.

The capacitors I chose were a spurge by most designers standards but well justified by mine.  ESR (equivalent series resistance) of smoothing capacitors has a dramatic effect on their ability to smooth voltage ripple.  A perfect capacitor would have a ESR of Zero ohms.  Capacitors made for high quality switch mode power supplies have much lower ESR values than typical caps (even super expensive Audiophile caps).  I found some made by Mallory, which cost about $35 for the pair, but my simulations showed that using those, I could achieve a ripple voltage of about 50mV (.02%).  That's very quiet and DAMN good!

My next tweak was to add snubber capacitors to the choke to help suppress high frequency voltage spikes and to add snubbers across each rectifier diode to reduce switching noise.  Here's what it looks like:

Eventually, I'll probably replace the standard 1N4007 diodes (hidden under the brown snubber caps) with super fast switching hexfreds.

Side Project: a Switched Attenuator

I use a PC oscilloscope for many AC measurements and for measuring distortion.  My audio interface connected via USB provides the analog inputs (2 channels).  Unfortunately, they have a maximum input of 2Volts.  I need to measure up to 500 volts.  For a while, I used a couple high wattage resistors as a voltage divider to attenuate the voltage down to 2v or less.  Sometimes I need to measure low voltages, sometimes high, and switching the resistors became a pain in the ass.  I decided to build a switched attenuator.  It's basically just a ladder divider.  Since I'm often measuring the output of the amp, I also added 8ohm thick film resistor which can be switched in to provide the speaker load without a speaker blasting out ear piercing test tones.  Here it is with one channel completed:

Choosing an Output Tube

Choice of an output tube dictates the power supply requirements and the type of output transformer which can be driven.  I've changed my idea of the perfect output tube 3 times now.  My first choice was an EL84 pentode because that's what I had used for my guitar amp that I built in 2003.  My first prototype of a hi-fi using an EL84 were somewhat disappointing though.  It at least made sound, but it included some upper mid distortion and wasn't very loud.

I decided a 6V6 might be a better choice: higher output and easier to find NOS (New Old Stock) tubes around.  That was a bit louder, but still not enough to "crank".  The more I read, the more I started to think the sound I was after was really a large bottle directly heated Triode instead of a Pentode like the EL84 and 6V6.

Now I was on a quest to find the perfect Triode.  I've always loved the 300B, created by Western Electric in 1937 to amplify telephone signals.  It can belt out 7 Watts and was used in most every movie theater before the takeover of the transistor in the 70's.  The problem with the 300B is that it has a 5volt heater supply (my power transformers have 6.3v windings) and even new production 300B's cost well over $100.  That led me to consider the 2A3 and better yet, it's sister tube, the 6B4G which has a 6.3volt heater, Octal base, and still puts out 3.5 watts.  It's specs fell within my main design goals.  I scored a "matched" pair of NOS (New Old Stock) Sylvania 6B4G on ebay for $67.

 Those will be the output tubes for my first version.  For my second round, I might design in a 5volt heater supply so I can run a pair of 300B's.