Rail-Rocket Design Notes

A Guitar Amplifier Design Narrative


7 February 2000



There are very few amps that I have felt couldn’t be improved upon. Maybe it’s an, “OK, it’s got the ‘Scooped Tone of Death’ thing for what that’s worth, but try to play anything else, and it’s as stale as day-old pancakes.” Or, “Shees, this thing makes me play and sound fantastic, but it flattens local flora and fauna within a hundred yard radius, doesn’t have reverb or anything, and any effects have to be shoved in before the grunge goes on.” Oh, and the thing weighs so much that the designer laughs just thinking of anyone actually trying to use that silly little handle on top.


Now, I’ve slowly built up a bit of tube amp knowledge and building skills, and with that I’m taking a swipe at building what I want into an amp. A design that hopefully borrows from and keeps the best of several amps, while hopefully addressing some of their limitations. And (fingers, toes, eyes crossed) sounds even better than my Frankenstein’s Combo prototype that is its predecessor.


Note that if you pick up any of the basics of tube design here, that’s a bonus. Really, this is intended as a next step after the basics, or, how you might design an amp. For something more along the lines of the basics of how things work, take a look at, The AX84 P1 Theory of Operation document, a wonderful, concise look at how the various parts of the amp work. If you’ve got some cash, I’d also recommend buying Bruce Rozenblit’s Beginner’s Guide to Tube Amplifier Design. Others recommend Kevin O’Connors The Ultimate Tone (I didn’t buy it - something about that pretentious title - but I’ve heard that it’s a great book).


In this document, I’ll try to lay out what went into putting the Rail-Rocket design together. After getting down a few of the basics, many guys wonder, “but what do you do to design your own?” Well, I think I know a bit about that, and I’ll give it my best shot at passing it along. Now, it’s just my take, and there are countless others, some of which are more disciplined, some more ‘seat-of-the-pants.’ Some of which are more scientifically driven, some more “mojo” driven. Hopefully, this will be a good place to begin when first starting to design your first amp. Note,  your own designing style will develop the more you design.



“What do I want in an amp?” It’s the first thing to decide upon, right at the outset. How well you achieve those goals is for later, but first you must have something to shoot for. Or else, you’ll get .. well, whatever. Let’s take a look at some amplifier traits that appeal to me, and see if I can quantify them a bit:


·         Great Tone. Well, duh. OK, great tone for what? Great tone for that Thin Lizzy, Humble Pie kind of organic, articulate, smooth, but ballsy distortion. Oh, and spanky clean for Tele-Castin’. And sweet sweaty ripping blues blowing. And very subtle yet warm harmonics for jazz comping and riffing. And ... hmmm, this is leading into the next goal:

·         Versatility. See all those great tones described above? I want all those from the same amp. Give me an authentic mix of preamp and power tube distortion, but unlike most amps let me choose the overall volume for it. Crush Bone at micro-wattage. Bleached-Whites-Clean at loud enough to stupefy that obnoxious drummer. And if it’s not too much hassle, even deeper into über-metal, jah, Dieter? (Hey, flailing about in “Dropped-D” can be cathartic.) Oh, yeah, reverb is a big plus, as “all turf and no surf makes Moon-Doggie a dull Ho-Dad.” All the way from a light-misting out to shipwreck-drench. Oh, and some tremolo, not the “bup-bup-bup-bup” stuff like some amps, but the “woh-am, woh-am, woh-am” stuff found in others. This is all leading into the next goal:

·         Footswitching. Something simple, so that even I don’t mess it up at a gig. But something that allows for changing between preset volumes, distortion levels, reverb levels, and tremolo levels at a “stomp.”

·         Reliability. Something I can gig with for years to come, and always rely on.

·         Quiet Operation. Such that any noise contributed by the amp is swamped by what the pickups and the guitar cord contribute.

·         Size and Weight. Something I can schlep around, guitar in one hand, with my music case and amp tailing along behind on one of those small airport luggage dollies. (Hint: extra bonus points for lightweight combos!)

·         Wattage. Being able to go as loud as a Plexi or a Dual Showman would cover just about anything.

·         Simplicity. Since I’m building this thing, it better be as simple as humanly possible. (But without giving up anything else, right? Sure.)

·         Inexpensive. Hey, since I’m building it myself it has to be less expensive than those amps at “Mars” and “Guitar Center,” right? (Heh. Heh, heh. Sure.)

Prioritizing the ‘Wannas’


OK. Here’s the secret of design: Plastics - Wait, that was the advice from The Graduate. The real secret of any design is: Compromise. You see that grocery-list up above? We’re going to have to figure out where to compromise conflicting “wannas” there. For instance, an OT that makes your jaw drop it sounds that good but costs so much that your wallet goes into hysterics. Or, finally having that last decibel over the drummer, but starting to seriously consider a steroids and weight lifting program - so you can lift that new amp. Those are conflicting wants that must be compromised between.


Drawing the line for compromise is done either intuitively or formally. Intuitively you would say, “I’d really like to blast my drummer off his stool, but I’ll always be a skinny-runt guitarist, so I won’t be able to lift that 300-watt Marshall-Moose.” Formally you would identify the pertinent factors in a trade-off, identify the pros and cons of each factors in a several likely “candidate compromises,” give each “pro” of a given factor a given “plus” weighting, and each “con” a given “minus” weighting. After all that, you identify your best compromise by finding a candidate compromise with the largest positive tally.


Comparing the two methods, the beauty of intuition is that it feels right, and it’s fast. The danger of intuition is that lots of little “cons” can outweigh a big “pro,” but your brain ignores them as background noise (the corollary is also true there). If this were my livelihood I would crank-out a formal trade: it’s a tiresome but responsible thing to do. I get enough of tiresome activities at my work, and this is only a hobby for me, so I opted to trust my intuition for finding the best compromises quickly and easily.


Here then are my own compromises, intuitively arrived at, with as much rationale as I can muster for supporting them. Note that they’re (likely) the best compromises for me, you’ll likely find that other compromises work better for you:


·         Great Tone. Well, there goes solid-state right out the window; I don’t care if it is lighter-weight. Ancient Greeks had the word Muse that denoted a spirit-like entity bestowing inspiration upon an artist. I’ve never seen a Muse captured in silicon, but they live and breathe in many tube-amps.  So, the heavier solution gets the nod. Oh bother - the better sounding speakers usually weigh more too.  Hopefully we can loose a bit of weight by keeping the power to only what is needed.

·         Versatility. Versatility is at odds with Simplicity. The more things we want this amp to be good at, the more things we’re going to need to add to it. And the more complexity we add, the more likely reliability could suffer. Well, let’s see, we need to be able to footswitch between two channels, each with independent drive, verb, and trem levels. Oh, and effects send/return and direct-out. Whew, boy, this isn’t a little-old Champ there, Sparky. The only thing that will keep this at all simple is extreme cleverness. (Spinal Tap’s Nigel Tufnel: “There’s a fine-line between clever and stupid.” )

·         Footswitching. Covered above.

·         Reliability. Reliability will have to be traded-off in many areas, for instance the wattage-to-weight compromise  (can you get by with smaller trannies pushing the same watts?). Or consider the price-reliability compromise (those same smaller trannies are no-doubt less expensive, too). Even tone and reliability have to be traded, running the output tubes hotter usually sounds better - but they wear-out faster.

·         Quiet Operation. Quiet operation is traded against high-gain (high-gain amps often have Hissy-Fits: that constant ‘hssssss’ in the background), and traded against complexity (e.g., D.C. heaters would be nice). This may not be a trade that can be estimated up front: “How much gain can be added before the hiss becomes a nuisance?” I’d be willing to sacrifice the “super-metal” regions of gain if it makes the difference between nuisance or not. As far as the DC heaters go, any hum from a well-executed amp with AC heaters is swamped by guitar cord and pickup buzzz, so that’s one bit of extra complexity that won’t be designed-in. Well, OK, if we’ve got 12.6Vdc already there for the channel-switching relays, we’ll make good use of that to heat a few of the “crucial” heaters.

·         Size and Weight. Ehm well, as this is going to be hauled around on a luggage-dolly it can’t be so heavy that the dolly wheels bend inward after just two gigs. Since either is so heavy, a Twin or Marshall 2 X 12” combo is almost definitely out of the question. Those things weigh a ton. And they’re so wide that they’d keep slipping off one side or the other of that dolly at the slightest bump. The Deluxe Reverb seems to be the biggest combo that can comfortably fit on there (I’ve had experience with these). Since the Deluxe Reverb houses a single twelve-inch speaker, or (with a non-stock baffle change) two ten-inch speakers, that also tells me: one 12” speaker maximum, or two 10” speakers, or maybe one 12” and one 10”, but that’s really pushing it.

·         Wattage. Prior experience tells me that my 22 watt Deluxe Reverb is the bare minimum for keeping up at the big-band gigs. My 30 watt Marshall Artist (4203) is a bit more comfortable, as the Deluxe starts breaking up for cranked solos, most of the time that’s great but not always. So, maybe 30 watts, and no more than 40 watts because then the weight starts to get unreasonable.

·         Simplicity. With the kitchen-sink being thrown-in, this is going to be tricky. I guess I’m willing to trade some simplicity for the cool features mentioned above. But again, I’m going to use every bit of clever I can to keep things simple. The other thing going for this is that it will be constructed in phases, and so with a bit of pre-planning of space for future expansion, things can start off simply.

·         Cost. Those who’ve built from scratch know that it isn’t cheap. I’d say a combo like this will end-up costing maybe $350 to $500 (US) to build, depending on the speakers, trannies, etc., chosen. Even though you’d think the labor savings (since you’re providing the sweat) would be substantial, it’s completely off-set by the higher cost of buying components in the “ones and twos.” The same PT you pay $70 for is had at the 1000-piece price of $32.76 by the “big boys.” So, hang the cost. You might as well just go buy a production amp or used amp if cost is your biggest concern. On the other hand ... you can at least start with only pieces you need now. This is a project to be spread over a few months, and buying over a period of a few months “eases the pain” a bit. Where practical, I’ll try to give some low/high price options, so that folks can make the trade of cost vs. “some feature” as they see fit.

Of Output Tubes, Amplification Class, and Biasing


OK, so this thing’s going to be in the 30 to 40 watt range. There are lots of different ways to come up with that: four EL84s, four 6V6s, two 6L6s, two KT66s, two EL34s, maybe two 6550/KT88s. Those four by EL84 and four by 6V6 designs can sure be sweet playing. But they take up more room in the chassis than do the “two by KT66/6L6/EL34/6550/KT88” solutions. With the all the preamp tubes and the up-top capacitors we’ll need for the various stages to support all these features, I’ll opt for the lesser “real-estate” solution: the “two by KT66/6L6/EL34/6550/KT88” thing. It’s a bit less wiring, also.

Another factor when considering the PT requirements: What class of output stage is the amp? Class A push-pull? Would be nice and fat, but we’d be limiting ourselves to 20 or 25 watts before clipping. With 6550s or KT88s in push-pull class A, that could get closer to 30 watts, but that’s just too close to marginal, given the stated goals. Switching the power tube types but not the output-tranny (OT) would mean even lower outputs due to the non-optimal plate loading. So, marginal power (given the stated power requirements) would be achieved from a reduced tube selection. Oh, well.

Class AB it is, then. But grid or cathode biased? With grid-biased we’d get maybe 20% more output power than cathode-biased for the same PT. Sifting through the tube data sheets and, interpolating, extrapolating, and divining what power we can get from them, maybe around 45 watts with a 380Vdc plate for grid biased. Cathode-bias may just buy us a bit of extra warmth for the price of an output power drop, so let’s take a look. The tube data sheets indicate delivery of roughly 35 watts from a 380Vdc plate supply with cathode biasing from two 6L6s (actually, I found that in the 7027 data sheets, but it’s the same internally as the 6L6) or two EL34s. The KT66 will deliver 30 watts from 390Vdc plate and 275Vdc screen with cathode biasing. The 6550 will deliver 41 watts from a 400Vdc plate and 300Vdc screen with cathode biasing. That’s well within our stated goals for power. So, for a bit less complexity and perhaps a warmer tone, we’ll go with Class AB cathode-biased. It’s simpler as well: no negative rail components or bias-pot twiddling.

Of Output Trannies, Plate/Screen Voltages, and Cathode Resistors


The output transformer (OT) needs to handle about 35 watts. The EL34 with a 375Vdc plate and screen supply wants 3400 ohms per the data sheets. The 6L6 for cathode-biased AB operation and about 380Vdc on plate and screen wants about 4500 ohms. The KT66 wants 8000 ohms for that 30 watt delivery (I’d bet that wattage would jump up some with a load of about 5000 ohms, though). The 6550/KT88 would like to see about 4500 ohms to deliver 41 watts with 400Vdc on the plate and 300Vdc on the screen. Hehm. Since I’d like to be able to use any of these tubes, maybe split the difference and go with something in the middle: 4500 ohms or so. And shoot for 380Vdc at the plate, that seems to be the right voltage for around 35 watts.

Next comes the fun, trying to find that in a decently priced OT with decent performance. After much searching and giving a bit on the desired impedance to get a good price, the following two seemed to be good options.

The low-price option: from Angela, a “Blues Deluxe” OT, which according to their blurb, was “Good for ... 6L6/5881 tubes running in the 20 - 35 watt range.” OK, this is probably a 5K ohm tranny with just enough low-end for guitar, nothing spectacular, but at $28, just the thing for the wallet. With this tranny, the 6550/KT88 might just be too much, so if you build with this one, verify safe OT conditions when actually built.

The higher-priced option: again from Angela (they just seem to have decent prices), a Hammond 1645, 30 watt, 5K ohm tranny for $54. With a guitar, we won’t be putting all that bass that the Hammond is rated for through it, so we can figure that it will probably handle 50 watts of guitar with ease.

So, let’s figure out what that 5K primary does to the output power. All the following were from spec sheets for each tube, with cathode biasing and as close to 380Vdc for the plate and screen as I could find. For the 6L6, that difference from what the spec sheets lists (4500 ohms) is not enough to even worry about. I suspect the KT66 will perform similarly to the 6L6, I’m just not too worried about it. The 6550 is seeing very close to that 4500 ohm load and so it should kick-out about 40 watts per the specs. But then there’s the EL34 that wants to see 3400 ohms ... There’s no other way around it, it’s time to do some load-line calculations for the expected power there.

I found a very complete data set on the 6CA7/EL34 from Philips, via a link on Duncan Munro’s tube search engine. In there was a plate voltage vs. current graph for Vg2 = 360Vdc, pentode operation. Close enough to 380Vdc for horseshoes and howitzers! I had recently poured over load-line primers, tutorials, and articles from Randall Aiken’s site, Steve Bench’s site, and some recent Glass-Audio articles, and I was raring to try this stuff out. So, I got two copies of that graph, took one and flipped it on its head, and met that with a “right-side up” copy where the Vak was 358Vdc, and the idle plate current was 75mA.

“Whoah, where’d you even get those operating points,” you ask? Looking back in the Philips data sheets on page four, there’s the figures for cathode-biased, class AB at 375Vdc plate. They list a cathode resistor there of 130 ohms, that times the sum of the plate and grid currents at idle (2 X 75mA, 2 X 11.5mA respectively) gave me the 22Vdc for Vk, subtract that from our 380Vdc B+ and get the 358Vdc Vak, and the idle plate current for one tube alone is 75mA. And there you go. I include this part of the calculation because I didn’t see cathode-bias operation handled in any of the load-line stuff. Other than the above calculation, it’s all straight from the literature though, so get that and go to it.

So, after lining up those two graphs and re-labeling the axes of the composite graph, I did the partly graphical, partly numerical power calculations of the expected power output before clipping. First I wanted to check my numbers agains the known data-sheet figures. With a 3400 ohm load-line, I came up with about 40 watts expected. The data sheet says 35 watts. A five watt discrepancy between the two. After some head scratching, I think I know why: at full power with the cathode-biasing, the bias voltage is pushed colder, and that shaves-off those five watts. That seems to make sense, and may even be the real reason for the discrepancy. Well, at least we can calibrate our numbers: “Shave-off 5 watts from what you get with the load-line for this instance.”

With that calibratory fudge-factor in-hand, the 5K load line was laid-out and the analysis re-worked. Thirty-five watts was the “paper and calculator” preliminary answer. Shaving-off those 5 watts then, I’ll expect to see 30 watts from a pair of EL34s cathode biased with a 130 ohm resistor (or two 260 ohm resistors) , a Va of 380Vdc, and a 5K primary load. Not great, but still within the design goals, and so on with the rest of the design.

We can pick some notional cathode resistances for the 380Vdc plate and screen operation. We might just have to adjust these a bit after getting it built and fired-up, but we should be close. From the data sheets for cathode biased, and somewhere near 380Vdc supply voltage comes the following. Where a single cathode resistor was listed, I converted to two separate cathode resistors, that allows for less-well matched sets to be used:

·         6L6: 360 ohms each cathode, 380V plate and screen.

·         KT66: Closest published is 500 ohms each cathode with the 390V plate and 275V screen, for 30 watts. Let’s make that 360 ohms, 380V plate and 380V screen, and lump it together with the 6L6 for simplicity. After it’s built, we’ll verify safe operation for the KT66.

·         EL34: 260 ohms each cathode, 375V supply.

·         6550/KT88: For the KT88 I scared-up some cathode biased, ultra-linear data with 375V plate and screen. Rk at 400 ohms for each cathode. We can use those for the idle conditions of non-ultra-linear. As the 6550 can handle 35 watts at idle (less than the KT88) let’s see if the power dissipation at idle is something the 6550 can take. The Ia+Ig2 was 87mA for each cathode, so Vak would be [375 - (400 * .087)] = 340V. Figuring about 80mA of that cathode current is from the plate, that’s an idle plate dissipation of 340V * .08 = 27 watts. Taking 70% of the max 35 watts at idle for a 6550 (70% being a good class AB1 idle de-rating from maximum plate dissipation) gives 25 watts. Frankly, I’m not worried about 2 watts at extra idle on the plate of a 6550, they’re very tough. Regarding the audio power, even with ultra-linear operation (an option with that Hammond OT, “UL” can be great sound) and our 5k load, we’d still get about 30 watts according to the specs. Without the ultra-linear I suspect close to 40 watts output, given other data I found for the 6550.

Of Tube Rectifiers, Power Transformers, and Chokes


In coming up with the power tranny (PT) specifications, the type of rectification (tube or solid-state) is a factor. Tube rectification can be very nice. At first I had the lay-in on the schematics for tube-rectification. What they have over silicon rectification is that they’re naturally quiet from hash noise, they do slow-start-up beautifully (watch those electrolytics last forever), and they can add considerable “touch” to the play when the amp breaks a sweat. But now the down side: they increase the size, weight, and cost of the required power-tranny (PT) by roughly 20%, to make-up for that expensive voltage drop across the rectifier. With $20 for a 5AR4, $20 extra for the PT, and the 2 extra PT lbs. hanging around, it’s starting to add-up. And it means one more tube, and this amp is going to be tube-happy enough as-is (just wait and see). Also, I’m hoping to get most of that great sound from a “midstage,” so the tube rectification may not be as important for tone. And so, with a bit of nudging from an un-named guy who runs a great “Do-Yerself-In Guitar-Amp” website, the 5AR4 was terminated. In with the “Standby Switch,” in with the UF4007s (at least they’ve less hash than the 1N4007s), and keep your fingers crossed. I’ll always wonder, “Would that extra bit of touch been worth it?” But, the decision has been made to the best of our judgement, and so let’s go onward.

So, with silicon rectification and stiff filter caps, we can figure about Vac * 1.3 will be the B+. Since we’re looking for about 380Vdc on the B+, that means we want about 290 - 0 - 290. Checking for cathode-biased, class AB, (roughly) 380Vdc operations of the 6L6/7027, EL34, and 6550/KT88, the idle current sunk by the power tubes will be in the 150mA to 175mA range. Full-tilt, the power tubes might be sinking 220mA.  If we choose a 250mA PT, that would have enough left over for the preamp, at least 30mA. Note, we could spec a 200mA PT, and when really cranking we’d go over but since we won’t be operating at continuous maximum power real-world guitar signals have lower duty-cycles than a sine wave that should be fine. Let’s see what we find. The other thing we have to work out it heater draw. Let’s figure worst case: two 6550s/KT88s at 1.6A each, and (looking ahead) six 12AX7, a pilot lamp, and a 6SN7. That’s about 6A needed from the 6.3Vac winding.

Well, it turns out that Angela is offering a great Hammond 272JX with the 250mA of 300V - 0 - 300V, 8A of 6.3V, and 4A of 5V for a very reasonable $62 (Angela just seemed to have the best price on iron, all the way around. Sure makes ordering a no-brainer). Using silicon rectification, the 272JX should be able to kick out 380Vdc at full tilt, very much in the range that we were looking for. And we didn’t even have to stretch the maximum current limit on that PT, it’s a perfect 250mA. We’ll be using that 5V winding also, not for a tube rectifier, but with a voltage-doubler type rectifier to provide 2A of 12.6Vdc for channel switching relays, and maybe a few selected preamp tube heaters.

Lastly for this section is a quick look at that funky choke that usually ends-up in these designs. I want the extra preamp and output screens B+ filtering that a choke can give. The more henries the better, but that has to be tempered with the current needed, cost, weight, and size. The choke needs to supply the screen grids (20mA for the 6L6, 45mA for the EL34, 40mA for the 6550) and various preamp stages. For those preamp stages, let’s estimate 1.25mA to bias each triode, and take a peek ahead to see that 6 triodes will draw through that choke. So, a total of 45mA (worst case screen draw) and 7.5mA, or 52mA. Let’s bump that up a bit so that any unforseens are likely covered. A few chokes pop-out: from Angela, a Hammond 158L 15H @ 75mA for $16 (open-frame, no end-bells), or  also from Angela, a Hammond 193B 12H @ 100mA for $24 (with end-bells). That 193B would look spiffy with it’s black end-bells matching the Hammond PT and OT. But more importantly, those end-bells may just end-up being a bit quieter for hum. Either would do the job, however.