In high resolution audio circuits, the power supply is of great importance.   I've been learning this by painful experience for several years.   Here is, approximately, what I've figured out; starting from the AC in (left), and working to the DC outs (on the right).   I'll link my vendors the first time I mention them, and again at the end.

There are actually two possible power supply configurations at this time, one using solid state rectifiers, and the other a 5AR4 vacuum rectifier.  The one with the solid state rectifiers is ridiculously economical, the B+ transformer coming from Angela for the princely sum of $17, and the complete iron cost for the amp is less than $70.  To me, the solid state powered version sounds just as good as the vacuum one, but there's about a one in three chance that that B+ transformer will have a slight mechanical hum, so I don't want to use them in a commercial piece; thus the vacuum rectified one, which is also simpler to operate and prettier.  I'll list the parts common to both versions (most of them, actually) the first time through, and then the changes for the vacuum version..

Line Cord   I've tried cheesy zip, great big AC - clothes dryer zip, boutique hi-end; and the biggest bang for the buck is 16ga 'appliance cord'; the round black one you can get in Home Depot.   Zip, large or small, is terrible.   Some of the boutique cords are a little better than the appliance cord, some substantially worse.   I have taken to mounting an IEC320 connector on the back of the piece, so that the power cord can be disconnected. I've heard it both ways about the sonic hazards of an extra connector in the power incoming circuit.   I hear no difference at all, having no line cord trailing from the piece during construction is a terrific convenience, and it looks nice.  It is more of a job to inlet nicely into the back wall of the chassis, however. I use the kind that has the fuse holder integral, and an additional benefit for me is that the fuse holder no longer sticks an inch and a half into the crowded chassis. It's Digi-Key part Q201-ND.

Switches and fuses   The fuse is a 5mm × 20mm euro-style, because that's the way the fuseholders are, 1.5 amp slo-blo.   The use of a slo-blo fuse allows the fuse to be rated more closely to the draw of the circuit for better protection, and still withstand the current surge at turn-on into all those empty capacitors.   There will be a green (safety ground) wire in the line cord, take this to a convenient point nearby on the chassis.  In my case, there are four unconnected (electrically) plates, so I go to a transformer mounting bolt and from there to another over on the audio side of the chassis.  I connect the bottom plates to the these with copper foil across the wooden chassis wall.
        The use of two power switches for separate switching of the tube heaters and the high-voltage circuits is known in DIY audio as a Human Powered Time Delay; and is a feature made neccessary by the use of the sand (silicon, solid state) diodes for rectification.   The tube heaters take a rated 11 seconds to bring the cathodes to operating temperature, while the diodes will apply power instantly to the HV main circuit, which will bring the tubes' plate-cathode circuit up to voltage in a second or two.   Two things will then happen.   Since there's no current draw in this circuit yet (the cathodes not being up to temperature, and unable to emit electrons), the voltage can rise rather higher than its steady-state value in the operating circuit, stressing components.   Also, the application of high voltage to a cold tube can shorten tube life somewhat.   There are a few methods of addressing this problem; separate switching of the heaters and the HV circuits is the simplest.  Both switches are 6A 250V SPST miniature toggles from Digi-Key, part EG2367-ND.   AC live (black wire) from the fuse goes to one side of switch 1, and the primary of the heater transformer and a wire over to switch 2 to the wiper (center) pole.  Then, the other transformer primaries (main and negative) go to the wiper of switch 2.  In this way, the high voltage switch is in series with the heater switch, and the HV can't be turned on unless the heaters are on.  Did I explain that okay?

Transformers   The heater transformer is a Hammond 6.3V 4A, number 166N6.  The high voltage main transformer is a Hammond 167G120, 120V 500mA.  The choke is a 5H 150mA Hammond 158Q; all these from Angela Instruments.  The transformer for the bias and tail negative supplies is a little 10VA Amveco toroid, Digi-Key part TE62045-ND.  I wire the thing unconventionally to get the two negative supplies:

You see that I use one of the two 115V primaries as the primary, the other 115V primary in series with one of the 22V secondaries to get 137Vac for my -160Vdc tail supply, and the other 22Vac secondary for the -28Vdc bias supply.  This is rather unconventional, so I use a 10VA transformer to supply two circuits that burn a little less than two watts.  It doesn't even get warm.

Rectifier Diodes   The ones I used in the beginning were regular 3A 1KV (1N5408) types from Radio Shack; they were the most mongo high-voltage diodes in the place, and were about two bucks or so for three.   I notice that RS now no longer stocks this diode on the wall, but they do have a 2.5A 1KV diode that will work fine.   A better deal is the Diodes, Inc. UF1007, which is a 1000V 1A ultrafast switching diode, Digi-Key part UF1007DICT-ND.   These are fifty cents if you buy one, forty if you buy ten; I get them a hundred at a time for 29 cents.   They're quieter and have a more easily filtered turn-off transient (the wart on the sand diode's nose) than regular 1NXX-type diodes.   The RatShack parts worked okay for me for ten years; take yer pick.   A prime reason that I use sand diodes in my circuits (besides that they're cheap, don't require heater supplies, and sound good), is that they lend themselves easily to the circuit illustrated. This is called voltage doubler rectification. There are shortcomings associated with it, primarily poor load regulation; but in these days of limited availability of high-voltage transformers, it's a very useful technique.   It takes the 120Vac from the trans secondary and produces approximately 320Vdc or so, depending on the filter circuit and load.

Capacitors   Capacitors are one of the main items of discussion in Rec Audio.  You ask a hundred experts, you're going to get ninety nine different opinions; here are mine.
        The first cap, you will notice, is a small-value (.01µF to .25µF is shown) film-foil or NPO ceramic capacitor between the diode stack and the transformer, and between the diodes and the PS circuit, in all these schems.   This cap shunts the sand diode turn-off transient burst to ground, so it doesn't exite high frequency resonances in the tranny windings or appear in the power supply.   I recently started using these, and noticed a slight but definite blackening of the background in the amp.   They're optional - the first place I put them was right after the diodes, before the choke.   The rest help, too; and they're cheap medicine.   The guys on the Bottlehead list like 1KV 10,000pF (.01µF) ceramics in this application.   Angela has the Sprague 716P's up to .47µF (the ceramics only come up to .01µF), and I'm always ordering from them; they work fine.   Again, take yer pick.
        For the larger-value caps, those marked (with a '+') as being polarized, there are several alternatives.   The first is a common electrolytic, available from the bulk vendors.  These are sort of a grab-bag, as there are so many brands and types, some of them very good indeed, most probably pretty indifferent.  Then, there are the boutique caps; Elna Cerafine, Rubycon Black gate, maybe others.  Some guys really like motor-run film caps; I haven't tried them because they're huge and won't fit on my chassis.  In fact, my power supply cap of choice for all my gear nowadays is the Panasonic TSHA / TSHB from Digi-Key. To me, these are quieter than the Cerafines I used to use, just as quiet as Black Gates, available, and economical.  I recently made a Mœbius for a gentleman who specified all Black Gates in his power supply, and all Vishay and Caddock bulk film resistors in the audio circuit. I tell you, I heard no difference whatever from TSHAs and Rikens.  Mind you, it didn't sound any worse.  The TSHAs come in ratings up to 400 volts, I use them by preference; and the TSHBs come up to 450 volts, I use them in the EL34 amp where the voltage is getting up there near 400Vdc on the output supply. I actually can't hear any difference between them, but the TSHA is a direct recommendation of my man J.C. Morrison, so I use them out of respect.  There is a small implementational headache in that these are a snap-in configuration, with little prongies instead of leads, but fortunately the prongs are just about the same distance apart as the terminals on a terminal strip. Just be careful how you route everything, because it's tighter, and keep your iron clean and wet, because you'll be soldering close to the cap body.  I haven't fried one yet.  I use Nichicon Muse for the bias supply, and other supplies below their max 100Vac rating, just because they have leads and so are a little easier to implement.  Michael Percy has them.

Chokes (Inductors)   Having a choke in the power supply (supplies) really blackens the background and increases the effectiveness of the little RF-shunt capacitors.   The one I used in my original example was from my fixed-to-death Dyna ST-70, 1.5H.   Angela sells the equivalent modern Hammond part.   I now use 5H, the Hammond 158Q, even more blacker and delicious.  I don't currently use chokes in the bias and tail supplies, due to my usual space constraints, and because they're not directly adjacent to the signal path.  I bet they wouldn't hurt anything, though.  If you're going for the complete catastrophe, DIY style, get as mongo with the Henries as you like, can afford, and have room for.

Assorted Stuff   Resistors, pots, wire are thankfully not as critical, quality-wise, in the power supply circuit as in the audio signal chain.   In fact, very high-bandwidth series parts may be considered detrimental in this application.   Low noise is, however, a consideration.   I use regular PVC hookup wire from wherever I find it, Radio Shack pots, and regular-grade metal-film resistors.   Don't use metal oxides for that last resistor before the power takeoff; they're a little noisy.   I have used reclaimed wirewounds from old gear to good effect; it's probably the only thing other than the transformers (and big air adjustable capacitors) that I look for when salvaging.  A note:  You will see in my AC heater supplies a resistor labeled 'RDROP', preceded by a nonpolar cap.  A lot (most, actually) of heater transformers/windings will produce more than the 6.3 specified volts in service.  I saw a test report yonks ago (I think it was in an early 'Glass Audio') that showed that this shortens tube life by tens of percent, statistically.  Conversely, operating them just a tenth or two of a volt low statistically lengthened their life similarly. I have a bunch of 2W inductive wirewound resistors of 0.1Ω and 0.25Ω, and I take the voltage in service to just about 6.1 or 6.2 volts.  This affects tube operation not at all, as far as I have tested for, and insures a long, long life for these valuable components.  That cap is a filter, it gets all the high frequency rubbish off the heater AC, so it can't get into the audio circuit.  It makes a difference.

You will see that for the vacuum rectified supply there are just a few changes.  The main power transformer is now a Hammond 372FX, which contains the bias and heater (and rectifier heater) windings.  Actually, the last two examples have used the 370FX, which is 275-0-275Vac on the HT secondary, but I couldn't get quite the voltage I want from them, so now I use the 300-0-300Vac version and a little smaller first capacitor, a 10µF SCR metallzed polypropylene, to keep the voltage down at 320Vdc or so.  The rest of the B+ filter is as above.

The bias supply is now half-wave rectified from the 50Vac bias tap on the main winding, the only change being the reduction of the final voltage divider resistor from 15K to 4.7K, since the raw supply is 50Vac rather than the 22Vac from the toroid winding.  The tail transformer is now an Amveco 62042, same type toroid but with 12Vac secondaries, and wired a little differently:

The List   Again, here are the guys I use; in no particular order of preference, but approximate frequency of use.

• Angela Instruments   Tubes, sockets, resistors, capacitors, transformers, chokes, other stuff. Nice folks.
• Michael Percy   Wire, solder, switches and pots, connectors, capacitors, resistors, the fancier stuff.
• Antique Electronic Supply   Tubes, sockets, interesting transformers, capacitors, guitar amp and old radio parts.

• Mouser Electronics   Who I actually don't use, but a lot of guys like them best of all.
• Digi-Key   My fave general vendor; lots of stuff, good prices.
• Allied Electronics   The biggest of the general vendors. The best source for transformers.
• Electra-Print   Very good transformers. Jack specializes in custom winding, if you get that far.

That's all I can think of.   If I've left anything out, or you have other questions, drop me a post at .   As I said, I'm working on making this amp buildable from the page, and I welcome suggestions.

Poinz