High-Voltage Entertainment in the Home



Louder is Better (Amplifiers)


March 15, 2009



I have built four amplifiers out of tubes, but only two of them exist today. One is a single-ended amp, built around the venerable 6AS7/6080 twin triode, which was designed for use as a series pass tube in voltage regulators, and has the virtues of being sturdy and cheap.  Purists frown because it is not so linear.  The 6AS7 is well-suited for low-voltage, high current applications.  The second amplifier is a push-pull unit built around the 813, a beam-power tube much beloved by ham radio enthusiasts.  It requires a more exotic (1 kV) power supply to achieve its potential.  My ability to develop that potential is severely limited by my day job.



1.  The Single-Ended 6AS7 Power Amp


One Channel of the Amplifier

The Tannoys are bi-amped by a quad set of SE 6080 amps (above, supplied by the scuzzy-looking power supplies, below) each of which produces about 9W of Class A triode audio power.  While not as sexy looking as, say a 211 or 845, the 6AS7/6080 has three big advantages:  it employs a high current, low-voltage (290 VDC) power supply, which is easier to build and safer to be around than the1kV unit a 211, say, would require; it has a low plate resistance, which means lower primary impedance on the output transformer, which is consequently wider in bandwith; it is orders of magnitude cheaper (NOS GE 6080s go for $10 apiece.  Compare that to a NOS GE 845).  In the version pictured above, each channel has a parallel 6SN7 VA stage with a constant current source load, which drives the identical woofer and tweeter power stages through parallel 0.33uF foil-in-oil caps.  Each power stage consists of three sections of 6080, for a total of 180 mA drawn through 1.5K output transformers. The original transformers were Audio Note models that are no longer made, but Brian Sowter makes an equivalent model (SE-15)--and Brian is so much easier to deal with.  90V of cathode bias is produced by 560R 50W Xicon resistors bypassed by 470uF 250V Nichicon caps.  I tried not bypassing the cathode resistors, and the result seemed harsh and fuzzy on the high end. There is no feedback of any sort in this amp.   The glow eminating from beneath the amps is caused by the red LEDs in the constant current sources.  


AMp on top, PS on bottom


Each 290V B+ supply has a power transformer (which you can't see) lifted by Rick Francis from, I believe, an organ amp, followed by a 5V3 rectifier.  The power supplies in the picture above, which includes those huge blue cans (2200uF @ 400VDC), are the previous cLCRC configurations, where the c was 5uF to goose the voltage a little, and the Cs were those big cans (and the L a 4H Hammond unit you can't see in the picture).  Since, then, I have replaced the big blue caps with a swinging choke (2 to 0.2 H, as the current goes from 90mA to 900mA) and a 220+220@500V Elna Cerafine cap.  This is just as quiet as the old version, and maybe lower impedance. Because the 5V3 turns on very quickly, the initial voltage still gets up to 560V (!) until the 6080s heat up.  Future plans for modification include replacement of the 5V3 with a pair of HexFreds, followed by a 6CJ3 for slow turn-on. Everybody likes the slow turn-on, right?  Another planned modification is replacement of all the electrolytics with polypropylene caps, because I know many of you find electrolytic caps déclassé.   OK, maybe a bank of 8uF oil caps would be even cooler-looking, I'll put that on my list of things to do (it would certainly raise the Dunker factor), along with trying a slightly higher voltage and a regulator, or, maybe a capacitance multiplier.   Maybe not, I gotta look at the B+ on the scope while the music's playing, apparently, the capacitance multiplier works best in constant current applications.  Here's a schematic of the previous version of the amp, you can figure out a PS easily enough:

Bi-amp Power Amp Schematic

Here is a printable copy of the power amp schematic.  

After the experience with the V-caps in the pre-amp (go look at the pre-amp discussion), I started to see if I could replace the paper-in-oil caps throughout the rig.  I started by replacing the 470µF cathode bypass caps on the output stage with 24µF Aeon polypropylenes.  That worked OK.  Then, I started to think about the coupling caps after the first stage.  Although, of course, I would just love to spend $70 apiece for 0.22µF TFTFs, I decided to search for alternative solutions first, since, if Nancy decided to kill me, $70 coupling caps  might very well constitute justification for homicide.  I mean, would you blame her?  Anyhow, it turned out there was a pair of Lundahl LL1635 5mA interstage transformers in the basement.  I have an interesting basement.  Initial experiments showed that the 6SN7 had too high a plate resistance to drive the 1635 properly, but that the 6BX7 looked good on 20kHz square waves on the O-scope (it has been my experience that anything that looks bad on square waves sounds bad, but two configurations that both pass the square wave test can still sound different.  So the square wave test is necessary but not sufficient.  And, the square waves themselves sound terrible).  Listening tests demonstrated that the 6BX7 (µ=10) didn't have enough gain, so I swapped the 12B4s (µ=5) in the pre-amp for 5687s (µ=15) , and the 6BX7s for 6BL7s  (µ=15).  That provided enough gain, but the mid-range seemed congested.  I realized that two sides of a 6BL7 (which I hadn't inspected on the O-scope) was way too much DC for the Lundahls, so I disconnected one side.  That opened up the midrange.  So, for now, that's where these babies stand.

2.  Adventures in Higher Voltages:  The 813 Push-Pull Amps

Desirous of more power and the milk-white glow of a big transmitting tube, and somewhat humiliated in a bake-off between the 6080 amps and Larry Moore's 845 SE amp, I decided I needed to focus and get the transmitting tube push-pull amp that has been on the drawing board for six years fired up, by replacing the SV572-10 A2 output stage for a pair of Class A1 813s, triode strapped.  The 813 is a beam power tube introduced in 1938 by RCA with a thoriated tungsten plate, although the plates in the later versions I use are probably zirconium-coated nickel. The 813 has a constant maximum disappation of 125 watts, although the ultimate application here will not run them quite that hot.

813 Exploded View


Moving from A2 to A1 operation simplifies the driver stage significantly.  Given the sonic implications of an operating point  near 0V bias, I doubt I will go back.  Anybody want a quad set of lightly-used Svetlana SV 572-10s?  Make me an offer. Here are the details at this time:

Main Power Supply (B+)  For the proof of principle amp, I simplified the power supply by reducing the B+ from 1200V to 600V, allowing a reduction in the size of the power supply, so it is now only as big as a Japanese pickup truck (soon, these will be the only kind):  vintage UTC transformer in a battleship grey can, HexFred half-wave rectification (with ceramic snubbers), 6CJ3 slow turn-on damper diode, CLCLC filter (Triad and Hammond, 10H each, with a 12uF Solen, followed by 2x400uF of electrolytics).  But, Christmas is coming and it may be time to jack the voltage up to where it longs to be by switching from full-wave to bridge, although this requires the bias regulator, (below).  

Filament Supply  This was alarmingly hard to get right.  Let me give you a hint here--for quiet, cool operation, always de-rate the current through Hammond transformers by 1/3 (i.e., specify 50% more current than you need).  God love 'em, Hammond continues to wind the widest line of current-production power iron (I get mine from Angela Instruments, Yesterday's Technology Today!) , but the ratings they quote are for the old ham shack, where you don't care about the noise and, sometimes, you don't care about the smell.  At rated current, the voltage will sag, the transformer will get hot, and you will hear the windings vibrate from your Fortress Of Solitude, er, your sofa.  For quiet listening room operation, multiply by 1.5.  To light up two 813s, which requires 10A@10V, I use a 15A@15V Hammond toroid, followed by an encapsulated 35A bridge rectifier, a 20mH choke (at 10A, but, empirically, the de-rating appears not to be required for Hammond chokes) and 3600uF@20V of high-ripple Panasonic electrolytics (the Panasonics from Digi-Key are not as cheap as some of the Mouser alternative, but, IMHO, offer the best value). Because I use fixed bias (below), the cathodes are at ground.

Driver Stage  I have gone through a fabulous number of drivers, because I am a fool for iron.  I find the use of various choke- and transformer-based phase splitting schemes, currently represented by folks such as Lynn Olson in his Amity designs (but going back to the 1920s), very appealing for a number of reasons:  simplicity; low-level detail; graceful recovery from overload.  However, all of the schemes I have tried have fallen short, with excessive phase shift and ringing in the middle-high end (5 kHz and up), which results in an unnatural brightness that initially sounds lively, but soon induces the dreaded listening fatigue.  I have tried a variety of triodes with low plate resistance as drivers including the  5687, 6CK4, 6EA7, 5998 and 6BX7 (you thought I was exaggerating, didn't you).  I have tried center-tapped chokes from Electra-Print, the Lundahl 1660S and a long-tailed pair with a Lundahl 1660PP as a load.  I have tried cathode followers.  None of them work as well in the phase-splitting department as a long-tailed pair with a constant current sink on the cathodes and resistors on the anodes.  Alas, this means cap coupling to the output stage grids, but, given that the voltage swing will be highest at the grids and I will ransack the kids' college accounts for V-caps, this will be OK.  Right now, I am experimenting with dual 6DJ8 cascodes, followed by a 6SN7 cathode follower.

Triode Strapping  The 813s are power beam tubes (tetrodes), but run as triodes.  To do this, the screen has to be assigned AC and DC voltages. Traditionally, this has been done by putting a 100 ohm resistor between the plate and the screen, which causes the screen to track the AC variation in the plate, at a DC voltage just a little below the plate.  Triode-strapped tetrodes and pentodes have a reputation for being bright (indicating 3rd-order harmonics), but not as bright as in their usual application with a fixed DC voltage on the screen (where the distortion is ameliorated in audio applications by negative feedback).  Bill Perkins has determined through principles theoretical and empirical an alternative strapping that AC-shorts the screen to the plate, while putting the screen 30 DC volts lower than the plate: 

Bill Perkins' Triode Strapping


The string of five 6.2V Zener diodes keeps the screen 31DCV below the plate.  The reason for the five diodes is that 6.2V is the low-noise sweet spot for Zener diodes (this is explained, or at least mentioned, in Horowitz & Hill).  Meanwhile, the OS-CON caps (manufactured by Vishay, they can be sourced through Mouser, and are not cheap but are also not hair-raising) have exceptionally low AC impedance, providing an AC short between the screen and the plate.  The ferrite bead at the screen is supposed to suppress potential oscillations, but, on my amp, I found it had exactly the opposite effect, producing oscillations in the FM band that actually allowed me to achieve a Theremin-like effect through the tuner, which I could modulate by means of a clip lead connected between the (ostensibly grounded) "chassis" (see below) and Andrew's tongue. Moving his head varied both the pitch and the volume. He didn't like it, so I substituted one of the dogs.  Omitting the bead solved the problem, making both Andrew and the dog much happier.  YMMV; ferrite beads are cheap, check it out yourself.  The 10 ohm resistor and 1N914 diode suppress potentially catastrophic oscillations that could, conceivably, cause the screen to end up above the plate and thereby self-destruct.  I have not experienced this so far!  But then, I have included the resistor and the diode.  Most audio tweaks leave me cold--which doesn't stop me from trying them, don't get me wrong--but this actually made a big difference, eliminating haze around a reasonably late version of the test amp, and I highly recommend it.

Output Stage  The output transformer is a Sowter 8963 that was custom-built for the SV572-10s, which wanted higher current than Sowter's stock 10K was able to handle;  this version is rated at 300mA for two tubes (PP).  Bias is fixed, supplied by a voltage doubler around a PC-mount toroid and controlled by a pot, but I am aiming to replace this with the solid state bias VR described in great detail by Morgan Jones in his wonderful Valve Amplifiers.  

Below is the very first prototype, with a Lundahl 1620 PP output transformer that couldn't really handle the current.  The driver is a 6BX7 long tail pair, with a CCS cathode and the Lundahl 1660PP on the left as an anode load.  Since I don't really have the environment for clip-lead amps, they were fixed, if somewhat exposed, on 6x12" aluminum plates:

New 813 Amp, One Channel, No PS

 
The gain was a little low, because I was taking advantage of the pre-amp's high output, but still, at maximum volume, it could have been louder. They sounded OK.   But, the 1620 OPTs were a problem; in the older datasheets I have, Lundahl was coy about the amount of current they could actually handle, but the American distributior of Lundahl products, K&K Audio, pointed me to new datasheets that indicate the maximum total current the 1620PP can tolerate is 150mA, 75mA/tube, and, at 540VDC, I was unable to bias the tubes low enough to sufficiently reduce the current to keep the transformers from running hot.  So, the cute Swedish 1620PPs are history (make me an offer), and the Sowters, which were designed for this application (and purchased when the dollar was strong and I had some) don't even break a sweat.  I swapped these much larger transformers in, set the bias at -65V, and left everything else the same, resulting in the second, even uglier, proof of principle amps:

Really Ugly 813 PP amp

These also sound excellent, with a coherence to the soundstage (there, I said it) that is missing from the 6080 bi-amp set-up, the high-end  is OK (I was worried about this using the high Rp tubes), the tubes glow that wonderful color, and the transformers run cool to the touch.  I think this one's a keeper, but, given that the 813 is designed to run with up to 2kV on the plates, I think I really do need to replace the full-wave rectifier in the PS with a hexfred bridge, and thereby get the B+ up to 1100V (and, yes, I will enclose the chassis) to enjoy the full benefit of the 813s, yet remain in Deep Class A.  On the version I'm using now, the output transformer is actually bolted to the chassis.   I will post the schematic once I get a driver I really like. 


Return to the rig


Return to main page






© 2009 Daniel Normolle