This page describes my first guitar amplifier build in over thirty years. A day-by-day narrative of the building and tweaking process is documented separately.
I grew up on vacuum-tube electronics. I became an electronics hobbyist in my pre-teen years. I had a couple of mentors: my next-door neighbor's dad who was a ham operator, and one of my dad's acquaintances who was himself a hobbyist and builder of his own hi-fi gear. I built a number of tube-based projects from scratch: a radio, an FM transmitter, and several amplifiers of my own design. I modded guitar amps: two extra gain stages in a Silvertone 1485 in the late `60s. My summer jobs in high school were mowing lawns and doing bench work in a TV repair shop. Then transistors came along. I struggled with the books on the subject and, after a while, learned to understand transistors. I could service transistorized gear, and eventually learned to design and build solid-state circuits. I even built my own solid-state guitar amps. But it just wasn't the same. Construction techniques were shifting to printed circuit boards (which I also learned to design), and the circuits seemed insubstantial. Transistors were (and still are) fragile. They have no real thermal mass and can be destroyed in milliseconds by an unexpected short or surge. And with entire circuits being fabricated on silicon, the element of design was superseded by a Lego-like approach to assembling circuits. During college and for four years afterward I made my living designing, building and servicing analog and digital control circuits (mostly solid-state) from scratch. Eventually I got tired of hardware; it had lost the wonder and romance that I had experienced years earlier. I switched to software (not a huge leap, as I was already doing microcontroller firmware) and took a succession of jobs in that industry.
Six years ago, I rekindled my interest in tube guitar amps. At first, I just wanted to play. But in 2001 I got the itch to build an amp for myself. At first, I thought I'd try my hand at a very-low power amp. I started doing my research. I gathered tutorial materials to refresh my memory (I wasn't about to try to build an amp based upon the little I still remembered after thirty years), found suppliers for the components that are now specialty items, located an excellent reference book from 1957, reviewed classic and modern amp designs, and even ran simulations of tone stacks to help me understand amp voicing. All of this took a while to gel. I sketched some designs and wasn't happy with any of them, so I set the project aside for a while. Meanwhile, my guitar playing evolved and I came to realize that a very-low power amp wouldn't be suitable for me. I started considering new designs. While on vacation in Europe in 2002, I sketched a half-dozen new concepts. And then those sat for a year while I pondered some more...
During the winter of 2003, I finally settled on one design concept to execute for my first tube-amp build in over thirty years. This is my own design, based upon several concepts not found in classic amps. I finished the initial on-paper design in January 2004, then set the project aside once more. In January 2005 I finally decided to go ahead and build my amp.
This amp has a pentode input stage, a low-loss tone stack driven by a cathode-follower and a single-ended pentode driver stage which is transformer-coupled to a push-pull 6V6 ultralinear output stage. The power supply is solid-state rectified. There's no reverb, tremolo or effects loop. This design is all about trying out some new concepts in their simplest possible form. I've left out everything that doesn't contribute directly to making the amp work and making it simple to use. There's just one input and one output. No channel-switching or option knobs. The tone control has just one knob. Simple, simple, simple...
Tonally, I'm going for a warm low- to mid-gain tone that can be cleaned up with the guitar's volume knob and the player's picking intensity. The feel should be on the loose side without a significant loss of impact or clarity.
The tube complement is 6267/EF86, 6BM8/ECL82 and a pair of 6V6s. Although the first two are not as common as 12AX7s and EL84s, they're still available as both NOS and new-production parts. I'm not worrying at this point about manufacturing the amp, and I'm sure I can find enough spares to be set for life.
Pentodes are interesting. They have more gain than triodes, and that gain can be altered by controlling the screen grid voltage. Not only that, but the screen current increases with signal. You can get some compression out of a pentode stage simply by choosing appropriate values for the screen resistor and screen bypass capacitor.
The self-compression behavior of a pentode stage presents an opportunity to tune an amp's compression independently of its distortion. In amps based upon triode gain stages, compression is achieved through a combination of design techniques in the preamp, power stage, power supply and speaker choice. Preamp compression always comes with increased distortion. Some designers use gain-staging techniques in an attempt to separate the two, but this comes at the expense of signal clarity. The power stage and power supply are closely coupled. Most power-stage compression comes from a drop (sag) in the B+ rail under load, although some can be obtained through nonlinearity in the power stage. The effect of B+ sag is more pronounced with tube rectifiers, but can also be obtained with solid-state rectification through choice of a higher secondary resistance in the power transformer or through introduction of an extra dropping resistor in the power supply. The problem with compression in the power section is that it is inevitably coupled to increased ripple and resulting ghost notes, so I'll be using a relatively stiff power supply.
For this design, I've chosen a pentode input stage and a pentode driver. That gives me two opportunities to tune the screen circuits for control of the compression. In addition, both the driver and the output stage are cathode-biased, which also (depending upon choice of bypass capacitors) introduces some compression.
The original tone stack was a simple bridged-T circuit driven by a cathode follower (the 6BM8 triode section). The single tone control alters the resistive leg of the bridged-T to ground, which in turn changes not only the relative amounts of bass and treble but also the center frequency of the mid dip. The dip frequency should range from about 500 Hz with an emphasis of the bass end of the spectrum to about 250 Hz with a slight emphasis on the treble and a shallower dip. The amount of bass and treble compensation is relatively small. The tonal changes are dominated by the shift of the mid dip frequency. The depth of the dip is relatively mild (about 6 to 9 dB). This tone control will be used more for fine-tuning the critical midrange response of the amp, with the guitar (and player) providing the majority of tone control through pickup selection, guitar tone and volume control settings, and picking technique.
As of Rev 1.2, the tone stack changed to use a Big Muff topology, but with different values. This tone stack topology consists of a high-pass filter and a low-pass filter plus a passive mixer. As implemented, the stack, with the control centered, has an insertion loss of about 6 dB (at 80 Hz and 7 KHz) with a 20 dB scoop centered at about 500 Hz. (I added a Mids switch which mixes a portion of the input to reduce the depth of the midrange scoop to about 12 dB.) As the control turns CCW, the scoop frequency rises as the treble response drops dramatically and the bass response rises slightly. The situation is reversed as the control turns CW: the scoop frequency drops, the bass response drops dramatically, and the treble response rises slightly. Near the center of the control range the effect is similar to the original bridged-T tone control design. However, this new tone stack allows for much more dramatic tone changes.
The driver uses the power pentode section of the 6BM8 to drive an interstage transformer which splits the signal to the output stage grids. Because the driver is single-ended, any second-order harmonics generated in that stage will be amplified by the output stage. The driver is designed to be able to source the grid current necessary to push the output stage into class AB2 for higher power. Pushing the the output tubes this way should not only yield a bit more power, but also force them to go a bit further into nonlinearity than they would in an AB1 design. Of course, the output stage's contribution will be richer in third-order harmonics, as is the case in other push-pull amplifiers.
The front panel is kind of barren. There's the input jack, volume control, tone control and power switch. (A second front-panel switch was added in later revisions for basic tone-shaping options.) There's no power indicator or standby switch. Why? Because they're non-essential convenience items and potential points of failure. Why do you need a power light? If you play your guitar and sound comes out of the amp, then it's getting power and is turned on and warmed up. If you don't get sound, check the obvious followed by a quick peek around the back of the amp to see whether the tube filaments are glowing. Want to double-check the amp before you leave the room? The switch is on in the up position, just like the light switch on the wall. A pilot lamp never prevented anyone from absent-mindedly leaving the amp turned on overnight in the practice room.
The standby switch is a questionable design feature in most guitar amplifiers. Kevin O'Connor of London Power claims that the standby switch was adopted unthinklingly when transmitter designers turned to audio amps. Cathode stripping is a real concern in transmitters but not, Kevin claims, in tube gear where the plate voltage is below 700 volts. I grew up with tube gear of all kinds -- radio, hi-fi, television, ham radio, even process control -- and have never seen a standby switch except on guitar amps and radio transmitters. Standby switches introduce an unnecessary point of potential failure. Toggle switches normally used for that function are not commonly rated to switch three hundred or more volts of DC current. They're rated for AC operation, and for low-voltage DC. At high DC voltages these switches will arc, pit, carbonize and eventually fail. Also, in an amp that has a cathode-biased output stage the sudden application of DC to the circuit after the tubes have warmed-up causes a sudden large surge of current through the output tubes as the cathode bypass cap charges up to its resting potential; that can't be any better for the tubes than a few seconds of DC applied while the tube warms up. Yes, this is all debatable. But in the absence of empirical evidence supporting the requirement for a standby switch, I'm leaving it out. Want to mute your guitar during set breaks? Turn down the volume...
I decided to build this amp using "true" point-to-point wiring techniques. This is uncommon because it's more difficult to implement and maintain than the more common tagboard or turret-board wiring that is also called "point-to-point". I'll wire all the components directly to the lugs of the tube sockets and to chassis-mounted terminal strips. There are advantages and disadvantages to this approach. On the plus side, you minimize the amount of extra wire that you have to run in the chassis and you get to lay out components in three, rather than two, dimensions. Lead lengths are short, and signals go directly from source to destination. There are fewer opportunities for stray coupling because there are very few signal-carrying wires and because the extra dimension gives you a better chance of orienting parts away from each other.
My strategy in laying out the components was to go with a left-to-right (facing the chassis bottom) signal flow; a terminal strip to the left of a socket handles the connections to the input components, while a terminal strip to the right handles the output connections (which become the inputs of the succeeding stage). Because the input and output components head off in opposite directions, there's little chance for coupling. This strategy was aided by the use of a pentode input stage; if I had used a traditional 12AX7 input I would have had a harder time making that work. The driver stage layout was done with a slightly modified approach which groups the output and bypass caps on one side, away from the heat-generating plate and cathode resistors. The driver layout was complicated by the fact that there are two separate stages in one tube: a cathode-follower to drive the tone stack and a power pentode to drive the interstage coupling transformer. Fortunately, these stages run at low gains, so coupling is less of an issue. The output stage is simplicity itself, with a transformer driving the cathode-biased push-pull outputs. There's a load resistor across the driver transformer secondary to help reduce the effects of the discontinuous load caused by grid current changes in AB2 operation. There are no screen resistors; I used the ultralinear taps on the output transformer.
The down side of point-to-point construction on terminal strips is that serviceability may be compromised. The components have a firm mechanical connection to the terminal lugs in a properly-executed point-to-point design. You can shake an unsoldered build, and nothing will come loose. That, of course, means more work for component replacement. Also, it may be more difficult to reach the connections because of the three-dimensional layout. I tried to arrange my layout so that this wouldn't be an issue; I'll find out soon how well I did.
A cabinet was the last hurdle to putting together a finished prototype. I wanted to do this myself despite not having any real experience in woodworking. I actually spent several days sketching cabinet designs before coming up with the final candidate. My objectives were:
Here's the end result from day 4 of the project. (A toggle switch was added just below the tone control in a later revision.):
|Front panel, from top: input, volume, tone, power|
|Speaker: Jensen P15N|
Tubes, from top: EF86, 6BM8, 6V6, 6V6
The overall size (excluding feet and handle) is 18.5" high by 26.25" wide by 10" deep.
Because this is a prototype cabinet, I tried to work with materials on hand or locally available. Most of the wood was cut from scraps left behind by the previous owner of our home. The cabinet sides are half-inch textured exterior siding, giving the amp kind of a "rural" look. The speaker baffle was another piece of half-inch plywood cut down from the remains of my pedal board. The rest of the materials were purchased from the local hardware store.
Tools were the biggest issue for this part of the project. We have a creaky old benchtop table saw that was just barely up to the task. The table was small enough that Mary-Suzanne had to make most cuts without a fence. My friend Stephen loaned me his router to cut the speaker opening. For the final prototype cabinet, I may farm out the cutting and do the assembly myself. Further down the road, who knows... ?
The cabinet is assembled using butt joints and cleats, held together using fine-pitch drywall screws and carpenter's glue. Originally I had cleats only in the corners and for the baffle board, but I added one more along the top rear edge of the cabinet for strength. Now it's rigid enough that I can stand on the cabinet.
The handle is a heavy-duty door pull. It fits well with the look of the prototype, and it's comfortable.
I didn't want to spend a lot of money on grille cloth for what is essentially a disposable cabinet, so I used black fiberglass door screen at a fraction of the cost. I painted the baffle board flat black after installing the T-nuts, then stretched and stapled the screen just as if it was grille cloth.
I stapled a strip of double-sided Velcro to the bottom of the cabinet to hold the power cord during transportation. The final cabinet will have an L-shaped trim panel on the back to hide the tubes and to keep the unplugged power cord inside the cabinet.
The weight of the completed prototype is about thirty-six pounds. I haven't yet decided whether to try another cab made of poplar plywood; that should shave another four or five pounds from the total weight. As is, this amp is a joy to carry at half the weight of my Vibro-King.
Here are some recent (Rev 1.3) chassis photos:
|Power Supply, Bottom View|
|Power Supply, Side View|
|Output, Bottom View|
|Output, Side View|
|Preamp, Driver & Tone Stack, Bottom View|
|Preamp/Driver, Side View|
|Tone Stack, Side View|
|Value||Idle (no input)||Load (half-volume)||Full-clip (full-volume)|
|output stage B+||392 V||387 V||375 V|
|output stage cathode||29 V||30 V||48 V|
|driver B+||331 V||327 V||318 V|
|driver screen||157 V||154 V||135 V|
|driver plate||127 V||127 V||137 V|
|driver cathode||12 V||12 V||11 V|
|preamp B+||266 V||263 V||256 V|
|preamp screen||66 V||64 V||63 V|
|preamp plate||94 V||95 V||93 V|
|preamp cathode||1.3 V||1.3 V||1.3 V|
|Value||Half Volume||Full Volume|
|input||400 mv p-p||400 mv p-p|
|cathode follower||70 V p-p||68 V p-p|
|volume hot||10 V p-p||9.6 V p-p|
|volume wiper||1.4 V p-p||9.6 V p-p|
|driver plate||40 V p-p||150 V p-p|
|power tube grids||28 V p-p||100 V p-p|
|output||12 V p-p||30 V p-p|
In general, the simplicity of a guitar amp's design belies its behavioral complexity. A guitar amp has distortion that not only varies with input level and frequency, but also with the recent history of its inputs. An amp also has a frequency response which varies with tone and volume settings, and also with output power. The combination of these details gives an amp its signature sound and behavior. This particular amp combines a number of interesting design features to produce its unique sound and behavior.
The pentode input stage generates predominantly second-order distortion until pushed hard, at which point some higher-order harmonics appear. The appearance of distortion increases gradually with the input signal. The driver stage introduces additional second-order harmonics which vary in magnitude depending upon the input signal. Additionally, the gain of the driver decreases with increasing input signal. These driver variations are nonlinear: the change is less pronounced for smaller signals. There's also a time constant associated with these driver response changes: initial attack transients get through the driver with less distortion and compression.
It's interesting to note, in comparing the measurements for Rev 0.7 to Rev 1.3, that the amount of driver screen compression is reduced in the latter. This is because the value of R14, the driver transformer load resistor, was increased in Rev 1.3. The higher reflected impedance loads the driver less, which in turn causes less screen current to flow.
The output stage, primarily as a result of its cathode bias circuit, also exhibits behaviors which depend upon signal level and recent signal history. Distortion increases with level, with some crossover (odd-order) distortion appearing at mid-level inputs and additional higher-order distortion when pushed even harder. Decaying inputs are subject to distortion that may not decrease as rapidly as the signal, giving the note decay some additional harmonic "motion".
The power supply has a small amount of sag with accompanying ripple. This is apparent as a small amount of IM distortion, particularly when playing distorted single-note lines in the upper register. It's probably exacerbated to some degree by the power supply rejection inherent in the ultralinear output connection. I think it would be possible to eliminate this IM with a larger first power supply filter, but I decided to retain it for "vintage" character it lends to screaming lead lines.
Here are some demos. For consistency, I've recorded them all the same way. I used an AKG C1000S condenser mic at the amp's grille about half-way between the voice coil and the edge of the cone. The recorder was a Fostex VM200 digital board into a Fostex D108 digital recorder. No EQ, limiting or compression was used. A small amount of reverb was added during transfer to an Alesis Masterlink recorder. The Masterlink was used only to normalize the track level to a full 0dB (the recording peaked at about -3dB). No EQ, limiting or compression was added during mastering. The CD was ripped at 192kb/S using iTunes on a Macintosh and the resulting MP3 file uploaded here.
Here's a quick demo of the amp (Rev 0.2) recorded on 2005-02-21 using my Koll DL Thinline guitar. The Koll is a 24.75"-scale maple-bodied semi-hollow guitar with a mahogany neck, ebony fingerboard and Seymour Duncan pickups (Jazz neck and Pearly Gates bridge). The guitar was plugged straight into the amp using a 10' Planet Waves instrument cable. I picked using my fingers. As you listen to this, remember that there are no channel switches or pedals.
Amp 1 Demo 1 (5:29)
|2005-06-20||Split the original document into this, the "Reader's Digest" version, plus a separate account of the step-by-step construction and tweaking details.|