A Vacuum Tube Hi Fi Primer
Here is a copy of the vacuum tube article I wrote back in 2003. Still a pretty good starting point for anyone looking to design vacuum tube amps. This info is useful for hi fi or guitar amps. The best part is that this describes an actual amp that you can build.
Vacuum tubes are an increasingly popular trend in underground DIY audio. Vacuum tubes offer simple circuit topologies, straightforward construction techniques without the need for heatsinks and printed circuit boards, and great sound. Of course the coolness factor of building something using 1930s technology that rivals expensive commercial equipment is undeniable. And DIY is far more rewarding than acquiring commercial gear because something you built always sounds better than something you bought :)
I would like to present here a straightforward process for designing and building a simple vacuum tube hi fi amplifier. I would like for the end result to be a design and parts list for an amplifier that can be built by anyone who can tell the hot end of the soldering iron from the cool end. With any luck that same novice solder slinger can extend the concepts illustrated here to their own designs.
Disclaimer: I am no expert! I only recently have learned enough to design this stuff and there is much I do not know. I really am completely unqualified to write about the topic, but my journey to reach this basic amount of knowledge has been fairly arduous, so my desire is to give the curious novice a boost up to the level of knowledge required to absorb the vast amount of information available online pertaining to vacuum tube hi fi.
The real disclaimer: Vacuum tubes like high voltages and the human body does not. Mix one with the other and you can quickly become dead. I am not going to tell you how to be safe! I am comfortable giving sketchy audio advice because all that can do is make your records sound bad — giving sketchy safety advice can get you dead an me sued :) So please learn proper safety procedures for working with high voltages.
And now down to business…
What is a Vacuum Tube?
The simplest vacuum tube — the triode — is made up of 4 parts — the plate, the grid, the cathode and the filament. (Sometimes the cathode IS the filament — this is called a Directly Heated Triode or DHT) In a typical amplifier like we will be building, the plate (a.k.a. Anode) has a positive voltage applied to it. Think of this as a serious lack of electrons. Electrons are like water in the ocean — if you scoop out some water, more water rushes in to fill the void. The cathode of the triode has a much lower voltage applied to it, and so has many more electrons than the plate. These electrons want to jump across the space between the cathode and the plate. The only thing stopping them is the grid (a.k.a. the screen — because it physically resembles a screen), which is negatively charged and so repels the electrons back to the cathode.
The filament just heats up the tube to get the electrons to an energy state where they are comfortable flying around inside the tube. Another name for the filament is the heater — how convenient!
So what happens when we vary the voltage applied to the grid? If we make it a little bit less negative, it allows electrons to rush from the cathode to the plate, and if we make it a little more negative, it lets fewer electrons flow between the cathode and the plate. To make this fantastically simple device into an amplifier, we simply set it up so that with no signal applied to the grid, there is a little bit of current flowing between the cathode and the plate. We then apply a small AC signal (music!) to the grid. When this AC signal makes the grid more positive, electrons leap to the plate causing a current proportional to the positive-ness of the signal. When the AC signal makes the grid a little more negative, the electrons are held back from the plate, causing proportionally less current to flow. This results in a large AC signal at the plate (loud music!)
Great! Now we know how to make music into loud music! Let’s build something that does it!
Operating Points and Load Lines
The only thing we really need to do is get the tube into this condition where just enough current is flowing between the cathode and the plate. This is called the bias current. The state of the tube when no AC signal is being amplified is called the quiescent point or bias point or operating point of the tube. The challenge is getting this point just right so that the tube amplifies the whole signal without running out of positiveness or negativeness. It can only go so far in each direction before it “clips”, meaning that the waveform gets it’s top or bottom chopped off because the tube has gone as far as it can go in that direction.
So how do we determine the operating point of the tube? As with most things in life, this can be done the easy way or the hard way. The easy way is to simply look up the tube in one of the many tube manuals, and use one of the published operating points. The hard way is to find a copy of the curve chart for the tube in question and use that to determine the operating point you want.
But before we pick an operating point, it might make sense to pick a tube :) Let’s pick something common and easy to design around. I would like to use a triode like the 2a3 because it is so nice sounding and so popular with DIYers, but I do not want you to have to mess with the added complications involved with using a directly heated triode. I can’t think of a common indirectly heater triode. I have used the 6s4 in a couple designs, but they aren’t very common. Tubes that are common are those tubes that are used in guitar amplifiers. These tubes are usually pentodes (which means they have 3 grids — triodes have 3 elements, pentodes have 5… see a pattern?) but a pentode can be made to act like a triode by connecting one of the grids to the plate. This is called triode mode, or triode-strapping. The sound of a triode strapped pentode is characterized as being a bit more “forward” than that of a true triode, which seems to mean a more pronounced midrange. I have used triode strapped 6v6 tubes in projects and I think they sound just fine. The 6v6 is super common. It is what Fender uses is almost all of their guitar amplifiers. There are many old console systems that used 6v6s. Best of all, there are many 6v6s being sold today, both current production and NOS (New Old Stock) so prices are reasonable, and the chances of finding 6v6s at your local flea market, yard sale, or surplus shop are very good.
We will use the 6v6 in triode mode and basically just pretend from here on out that it is a triode. So where do we start? The easiest way to design an amplifier would be to simply look up the typical operating characteristics for the tube we ant to use. Below is that data for a 6v6 in triode mode.
The complete datasheet for the 6v6.
The operating point of a tube is determined by the voltage at the plate of the tube and the bias current flowing through the tube. This is known as plate voltage and plate current. Here we see that a 6v6 can be operated at 250V and 49.5 mA. Later we will learn how to build the circuit that will let this tube operate under these conditions, but first let’s look at how the operating point can be determined from the plate curves of a tube. This will give us a much better idea of exactly what is going on in the amplifier and we will be able to design amplifiers around tubes for which we only have the plate curves.
The plate curves for nearly all tubes should be available in one of the many tube manuals available. The best single resource for tube data is TDSL, the Tube DataSheet Locator, at duncanamps.com. The amount if information stored here is truly amazing.
So what are we looking at? The x axis is plate voltage and the y axis is plate current, so the operating point I mentioned above is literally a point on the graph. The curves are the voltage that the grid will be at for each combination of voltage and current. If we look at the intersection of 250V and 49.5mA, we will notice that it falls between the -10V and -15V curves. Notice that the data we were just looking at lists the “Grid №1 Voltage” as being -12.5V. This can be thought of as the grid-cathode voltage, meaning that the grid is at -12 relative to the cathode (12V lower voltage than the cathode).
Now let’s take the next step, which is to draw a load line on the plate curves. a load line is a way to visually represent what is happening when an AC signal is applied to the grid and amplified at the plate.
This is the 250V/49.5mA operating point and corresponding load line for a 5K output impedance. The Large dot is the operating point. The straight line that intersects the operating point is the load line. The slope of the load line is determined by the impedance of the load that the tube in working into — in this case the load will be a 5K output transformer. The slope of the load line is calculated by dividing the load impedance (5k ohms) into a chosen voltage — say 100V — to get the change in current — 20mA. (100 / 5000 = 0.02 = 20mA) This determines the slope of the load line. Slope is defined as the change in current divided by the change in voltage (rise-over-run as they say in trigonometry class). So with a slope of 5k ohms, for every 100V the load line drops 20mA.
The load line describes what happens to the plate voltage and current as you vary the grid voltage. At the quiescent point, the grid voltage is -12.5V, the plate voltage is 250V and the plate current is 49.5mA. If the grid voltage is increased to 0V, the plate voltage will be 160V and the plate current will be 68mA. If we decrease the grid voltage to -25V, the plate voltage will be 330V and the plate current will be 35mA. These are the endpoints of the load line because the grid cannot become positive without causing clipping, and the grid cannot become more negative than -25V because the other side of the waveform will be driving the grid positive. If we apply an AC signal to the grid, varying the grid voltage up and down along the load line, the plate voltage and current will vary accordingly.
Another thing we need to think about is plate dissipation, which is the amount of power that the tube can safely dissipate. Every tube has a maximum plate dissipation in Watts. This should be on the tube datasheet. The 6v6 has a maximum plate dissipation of 12W, and Power = Current x Voltage (P=IV). This limits the region in which our operating point can reside. If we plot the line at which the dissipation is 12W we will have a curve that our operating point cannot cross unless we want to risk melted and imploded tubes :) Many guitar amplifiers run tubes significantly above the rated plate dissipation, and many current production tubes will gladly take this abuse, but I would prefer not to risk it. The 12W plate dissipation curve is shown on the graph above as a crude dotted line. Notice that the operating point is right on the line — in fact if we were to calculate the plate dissipation for 250V and 49.5mA we would get about 12.4W. It would appear that this operating point is running the tube a little bit too hot, but for reasons we will discuss when we talk about the cathode resistor, the actual plate dissipation is just under 12W. For now, let’s treat it as an acceptable rough estimate.
Let’s dig just a bit deeper into the load line. Why are the ends of the load line at 0V and -25V grid voltage? In a class A amplifier, which is what this is, the grid cannot go positive without severe clipping of the waveform. If one half of the amplified signal is limited to 0V, the other half is essentially limited to -12.5V below the quiescent point, -25V. The load line can also tell us a bit about the relative linearity of the tube. The linearity of the amplified signal is determined by the evenness of the spacing of the load lines. If we had a perfectly linear tube, the length of the load line above the quiescent point would be equal to the length of the load line below that point. In our design it is clear that the load line is slightly longer in the 0V to -12.5V region than in the -12.5V to -25V region. This results in distortion. In many cases, a higher load impedance results in less distortion — meaning a flatter load line. In general it is best to keep the load line out of the extremely low current areas of the curve where the curves get very close together. Also, the power output of the amplifier is proportional to the area under the curve, so maximizing this area will maximize the power output.
We are operating the tube in what is called class A, which means that the grid of the tube is always negative and there is always current flowing through the tube. This is the most linear but least efficient way to operate a tube, and so there are several other classes of amplifier that attempt to gain efficiency without sacrificing linearity. It is arguable whether they are successful.
The Circuit
Cool. Now we are vacuum tube gurus! Let’s take our newfound knowledge and use it to design an actual circuit!
In a nutshell, the circuit we will be building is a very simple 2 stage amplifier. The signal will enter the amplifier from our source (CD player, phono stage, radio) and be attenuated by a volume control, then fed to the grid of a single triode from a 6sn7 (the 6sn7 is a dual triode). The 6sn7 will amplify the signal enough to drive the grid of the 6v6 power tube. The 6sn7 will be capacitor coupled to the 6v6. The 6v6 will be configured in a conventional cathode bias, transformer output configuration.
Here is a first cut at the schematic. Notice that the 6v6 has one of the grids connected to the plate with a resistor — this is triode-strapping. Also notice that the only difference between the topology of the driver section and the output section is that the driver section has a resistor connected to the plate of the 6sn7, and the output section has a transformer connected to the plate of the 6v6. The approach we use to design each of these sections will be very similar.
B+ is a convention used to indicate the high voltage DC supply for a circuit. I think the term dates back to when there were batteries powering each part of the circuit, with the batteries labeled A, B and C.
The 6v6 Output Stage
Let’s look at just the output section of the circuit. This configuration is known as the Grounded Cathode Amplifier. Basically, the cathode is grounded through a resistor, known as the cathode resistor. There is also a resistor that ties the grid to ground. This is called the grid resistor. The plate load in this case is not a resistor, but a transformer. That’s it. That’s all there is to it :)
So how do we determine the values for each part? I will break this down part by part…
Output transformer — The impedance of the primary winding of the output transformer determines the slope of the load line (remember?). In this design we drew the load line for a 5k ohm transformer, so that is what we will use. A lower impedance load would cause a steeper load line — which means more power and more distortion. American amplifiers tend to use lower output impedances in a quest for more power, while many Japanese designs tend to use higher impedances for lower distortion.
Triode strapping resistor — The sole purpose of this resistor is to insure that the second grid of the tube is never at a higher voltage than the plate. This keeps current from flowing from the plate to the second grid, which would be bad, but I am not sure why — see “not an expert” disclaimer above :) Typical values are between 100 and 200 ohms.
Grid resistor — Since there is no DC current flowing through the grid, the grid resistor simply references the grid to ground such that it is essentially at 0V. The grid resistor also acts as the load that the coupling capacitor sees, and so it is used to calculate the value of the coupling cap. Typical values are 240k ohms or 500k ohms.
Cathode resistor — The cathode resistor is the heart of the grounded cathode amplifier. It is responsible for setting the operating point of the tube. The operating point we chose for the 6v6 requires the grid to be at -12.5V, but the grid resistor is causing the grid to sit at 0V. What do we do? We simply raise the whole tube to a potential of 12.5 volts by choosing a cathode resistor that will give us a voltage drop of 12.5V between the cathode and ground. The grid will now be at -12.5V relative to the cathode. To determine the value of the cathode resistor we simply have to use Ohm’s law (which is math, but it is easy math). We need this resistor to drop 12.5V and allow the bias current of the tube (49.5mA) to flow through it. R = V/I = 12.5V/49.5mA = 255ohms. whew!
It is likely that we will want to bypass the cathode resistor with a capacitor. This will basically remove the effect of the cathode resistor for AC signals, increasing gain and potentially reducing distortion. I plan to breadboard the circuit with and without this capacitor to see which is better.
Keep in mind during this process that tubes are very physical things — by that I mean that they are really just bits of metal and wire inside a glorified light bulb, and variation from tube to tube can be significant. Of course the characteristics of a tube can also change as the tube ages. Fortunately we do not need the circuit to be perfect to work satisfactorily. 10 percent variation in the value of any component would be easily tolerated by this circuit. In the heyday of tubes, carbon composition resistors were standard. Those resistors have not aged well, and many old carbon composition resistors can vary quite a bit from their stated value — yet most vintage amplifier will still work just fine (once you replace the failed paper capacitors).
Let me digress for a minute on the subject of engineering and perfectionism. I often find myself laboring over the details of a design and worrying that I am not able to achieve the accuracy necessary to obtain the prefect result. When I get too anally-retentive about things, I like to think about a vintage Jaguar my friend Marty has. Marty has had this Jaguar for many many years. I don’t know the exact year, but I would guess it was made in the 40s. At some point in this car’s life, one of the piston rings failed and scored the hell out of one of the cylinder walls. Instead of completely rebuilding the engine and boring all 6 cylinders, Marty just bored out the one bad cylinder, installed 1 larger piston and bolted that sucker back up. That damn Jaguar ran like that, with one piston significantly larger and heavier than the rest, for something like 40,000 miles. The lesson for me is that as long as you don’t do something stupid — and sometimes when you do — it will just plain work.
For a much more detailed and mathematical description of this type of amplifier configuration, see “Grounded Cathode Amplifier” in the 19 Jan 2003 edition of the Tube Cad Journal, which is a fantastic tube audio resource for those not afraid of math :) There is a really good discussion of load lines too.
The 6sn7 Driver Stage
I chose the 6sn7 for the driver tube for 3 reasons: It is octal, which makes it physically easier to work with compared to a 9 pin miniature tube; It is has a very good reputation for sound quality; And it is relatively common.
Since we used the curves to determine the operating point of the 6v6, let’s use another method to determine the operation point of the 6sn7. Let’s simply use a predefined circuit from a tube manual. Below is an excerpt from the 6sn7 datasheet.
The complete datasheet for the 6sn7.
Isn’t this convenient? The circuit diagram to the right of the chart shows how our circuit will look and all we have to do is pick a set of component values that will suit our project.
Let’s start with Rg1, the grid resistor. In our circuit we will be using a 100k volume pot instead of a resistor, so that narrows it down to one of the 0.10M rows. This 100k resistor will also set the input impedance of the amplifier — 100k is a fairly standard value for modern amplifiers and should work just fine with any modern source such as a CD player.
Next let’s look at Rs. This is an easy one because this is the grid resistor of the 6v6, which we decided to make 240k.
Looking at the charts, we see that there are 2 rows of values where the grid resistor (Rg1) is 100k and the load resistor (Rs) is 240k. For any of these combinations of values, we have our choice of several values of Ebb. Ebb is the voltage present at the plate resistor of the tube. Typically this is somewhat less than the B+ voltage being fed to the output stage because it is nice to add another bit of power supply filtering here (we will get to the power supply soon!). Let’s go with 180V.
So for the plate resistor Rp and the cathode resistor Rk, we look at the intersection of the column where Ebb=180V and the 2 rows where Rg1=100k and Rs=240k. We have our choice of either Rp=240k and Rk=5100 or Rp=100k and Rk=2700. Most amplifier schematics I have seen tend to run the 6sn7 with a fair amount of bias current, so let’s go with 100k/2700 because as we learned earlier, the smaller cathode resistor will allow more bias current to flow.
Now we get to talk about capacitors! A capacitor is basically 2 sheets of conductive material separated by a sheet of insulating material. A capacitor acts like an open circuit to DC current, but will allow AC through. The capacitor across the cathode resistor basically causes the cathode of the tube to be grounded as far as AC is concerned, while not affecting the DC bias current. This increases the gain of the amplifier section. I do not completely understand cathode bypass resistors, and the chart’s recommendation of an “adequate” value is of little help, so I will just use other schematics I have seen as a guide and make this a 100uF capacitor. (uF stands for micro Farads, which is the unit of measure for capacitance).
There are 2 coupling capacitors shown in the schematic on our tube chart. A coupling capacitor serves to isolate the DC bias current flowing through one tube from the next stage, while permitting the AC signal to pass. We will not be using a capacitor on the input of our amplifier because we will trust our source not to be feeding us any DC voltage — usually a safe assumption :)
To determine the value for the coupling capacitor between the driver stage and the output stage, we get to do more math. When you have a capacitor followed by a resistor connected to ground, you end up with a frequency dependent voltage divider. Those of you familiar with speakers will recognize this as a high-pass filter. Just as with the high pass filter element of a speaker crossover, there are 2 things that determine the frequency at which the filter begins to attenuate the signal — the value of the capacitor and the value of the load. In a speaker crossover this load is the tweeter, but in this circuit the load is the grid resistor of the 6v6. We want to pick a value that begins to attenuate the signal at a low frequency so that we will not notice the attenuation — somewhere around 20Hz.
The formula for calculating the value of the coupling capacitor is 1 / (2 * pi * F * R) where F is the frequency at which the signal will be attenuated by 3 dB and R is the value of the load resistor. Substituting 20Hz and 240k we get a value of 0.03uF. This is a minimum value, anything larger will simply push the frequency below 20 Hz. A commonly used value is 0.1uF.
So now we have the whole circuit figured out. Hooray! Next we design the power supply and then we will be ready to think about construction.
The Power Supply
The circuit we have developed above needs 2 DC voltages to power it — 250V for 6v6 and 180V for the 6sn7. The power supply merely has to convert the 120V AC mains voltage to these DC voltages. This is done with a transformer, which converts the 120V mains voltage into the higher voltages we need. Imagine AC as a sine wave. To convert AC to DC, you first need to rectify it. This means that you only allow current to flow in one direction. A rectified sine wave has the negative portion of the waveform moved to the positive side of the 0V axis. A rectifier is nothing more than a pair of diodes or a rectifier tube like the 5y3, which is simply a pair of vacuum diodes.
A rectified sine wave is a series of peaks and valleys, and so the next thing we need to do is smooth this into a constant DC voltage. This can be done with a combination of resistors, capacitors, and inductors (also known as chokes). The most common form of filter circuit used to smooth rectified AC into DC is called the pi filter, named after the Greek symbol that it somewhat resembles. The pi filter is a capacitor in parallel between the rectified AC signal and ground, followed by a choke or resistor in series with the signal, and then another capacitor between signal and ground. This is also called a CLC or CRC circuit depending on whether a resistor or a choke is used. The capacitors keep the signal from dropping in voltage between the peaks of the waveform so that we get a steady DC voltage. Additional chokes and capacitors can be added to further smooth the signal.
Designing power supplies can be frustrating because it can be very difficult to get exactly the B+ voltage you want. A CLC filter will result in a DC voltage roughly equal to the peak value of the AC voltage it is fed. We want 250V so we will need to feed the CLC filter with about 250V peak voltage. Transformers are specified by RMS voltage, which is roughly .707 X peak voltage, so we want to see about 175V at the output of the rectifier. Keep in mind that any resistance will cause a voltage drop when current flows through it. This is true of all the elements in our power supply, including the rectifier. We can account for these drops in voltage by increasing the voltage supplied by the power transformer. So let’s look at a CLC power supply using a 20uF capacitor, then a 7H choke, then a 40uF capacitor. This is very similar to the power supply described in the 5y3 datasheet.
This shows that we want about 275V RMS from the power transformer to get roughly 250V out of the power supply. We might be dropping a little more voltage across the choke, and the Hammond transformers we will be using are typically underrated so all we can hope for is that we will be in the ballpark of the correct voltage, which will just have to be good enough.
The complete datasheet for the 5y3.
The 180V supply is attained by simply adding a resistor to drop the voltage to 180V and a third capacitor to isolate this resistor from the plate resistor and further filter the supply since the preamp section will be more sensitive to power supply noise.
The as yet unexplained 1M resistor in the power supply circuit is a bleeder resistor that allows the circuit to discharge when the power is turned off. This is a good thing if we ever plan to work on the amp after it has been powered on. Without this resistor, the power supply capacitors can stay charged to 250V for a very long time. Keep this in mind when poking around inside old tube amps. I personally have a 100k ohm 10W resistor with clipleads on bothe ends of it that I use to discharge power supllies before touching the insides of amplifiers. I suggest not waiting until you get a nice jolt from a charged capacitor to build such a simple safety device. If nothing else, please be sure to measue the voltage across the power supply EVERY time you open up an amplifier.
Parts
Here is a rough parts list. I have not built this amp yet, so this is very preliminary. Obviously the list is incomplete, but it should give a fairly good idea of how much this whole thing will cost. If the prices for tubes look a little high, especially for the 6sn7, you can usually find some good deals on eBay. Fortunately we only need a single 6sn7 so we don’t have to look for matching pairs.
Stuff from Angela Instruments:
(1) power transformer: Hammond 270HX 176VA, sec. 275–0–275, DC ma 200, 5.0v @ 3a ct, 6.3v @ 6.0a ct.$45
(1) choke: Hammond 159Q 7H, 150ma, 100 ohms, 500VDC. $19
(2) output transformers: Hammond 125ESE $70/pair
(1) power supply multisection capacitor: 40uF+20uF+20uF+20uF/500VDC $10
(1) Mounting clamp for above capacitor $1.50
(4) octal sockets $6.40 for 4
(2) 0.1uF 400V orange drop or paper in oil
(1) 5y3 $5
(2) 6v6 $24/pair
(1) 6sn7 $32
total: about $150 without tubes, $210 with tubes
stuff from mouser:
(1) 1M 71-RN65D-F-1M
(2) 2.7k 71-RN65D-F-2.74k
(2) 100k 71-RN65D-F-100k
(2) 240k 71-RN65D-F-249k
(2) 255 (5W) — This gets pretty warm and so should be at least 5W. It wouldn’t hurt for it to be even higher power, or even chassis mounted.
(2) 100 71-RN65D-F-100
(2) ps dropper (3W) 3.5k-5.8k 71-RS2B-3.5k, 71-RS2B-4.0k, 71-RS2B-4.5k, 71-RS2B-5.0k, 71-RS2B-5.6k
(4) 100uF 50v sprague atom 75-TVA1310
(2) black speaker binding posts 164–3201
(6) red speaker binding posts (for 4, 8, and 16 ohm taps) 164–3205
(1) IEC socket 161–3516
total: about $25
stuff from Radio Shack (or mouser if you prefer):
(1) power switch
(1) fuse holder
(1) 1/2A fuse
(2) RCA input jacks — insulated are best
(1) 100k stereo log taper potentiometer
(1) knob for potentiometer
(1) selection of hookup wire
total: about $10
Cost of all parts should be about $250 with nice NOS tubes, $200 with cheap tubes. Of course it is fine to substitue lower or higher quality parts. The parts I have listed should give good results for just a bit more money that the cheapest parts possible.
Angela Instruments has a really great selection of Hammond transformers and various tube parts that are hard to find otherwise. Mouser has the Sprague Atom electrolytic capacitors for a little less money than Angela, and they have those pretty nice Sprague milspec resistors I like to use since they are only slightly more expensive that plain old metal film resistors. Mouser’s online ordering system can be cumbersome, so I usually just call the 1–800 number to order.
Construction
Most DIY tube equipment is either built on a commercial aluminum chassis or on an aluminum top plate attached to a wooden base. I have done both and find that each has advantages. Building on a complete chassis allows controls to easily be mounted on the front and rear of the chassis, which I find very conveient. This is likely why most commercial gear is this way. Your typical audio consumer is likely to expect all inputs and outputs to be on the rear of the smplifier and all controls to be onthe front. The drawback to the aluminum chassis is that it can be difficult to find quality chassis of unusual dimensions, and the sides can make it more difficult to reach components near the edges.
The second option is to build the amplifier on a simple aluminum plate and then attach that plate to a wooden base. This offers the advantage of being able to easily reach all components, and build the chassis any size you want. The disadvantage is that all inputs and outputs must be located on top, which can be unsettling for those accustomed to commercial equipment. There is also the problem of what to do with the power cord. You can make it top mounted, and perhaps use a cord with a right-angle connector; You can make it hardwired with the cord running out through a hole in the wooden base; Or you can mount an IEC socket to the wooden base. I like to have the whole assembly attached to the top plate, so I usually choose a top mounted IEC socket.
Many DIY projects, mine included, have no bottom cover — and therefore rely on being placed on a flat surface to protect the high voltages inside them. This is a very bad idea, and really has no excuse other than laziness. It is not difficult to imagine a situation where someone unfamiliar with what they are handling could pick up a powered or not fully discharged amplifier, curl a finger under the chassis, and recieve an unmistakable indication that there are high voltages under there. Unless you live in a household of electrical engineers and trained DIYers, with no children or pets with opposable thumbs (A monkey maybe? Do Pandas count?), I would recommend a bottom cover. In fact I think I have just convinced myself to build future projects with bottom covers. Perforated “tempered” pegboard works well, and is very cheap and easy to work with.
In the spirit of top mounting everything possible, I chose to use a multisection capacitor for the power supply. Any other style of capacitor can be substituted of course.
The first thing I like to do when I get all the parts for an amp I am building is lay them out on the workbench or chassis to attempt to optimize the layout. The 2 things to keep in mind are interference between components, and short signal path.
Transformers create magnetic fields that can couple to other transformers if they are oriented in the same direction, so it is best to mount transformers at right angles to each other whenever possible. Any AC wiring will also create a megnetic field that can affect nearby wiring. This can be minimized by always twisting any wiring that is carrying AC. Be sure to twist the filament wiring, and the high voltage wiring between the power transformer and the rectifier.
The part of the circuit most susceptible to noise is the very low voltage signal wiring between the input jacks and the driver stage. This wiring is commonly shielded, but I find that it is often good enough to simply twist the signal wire with a ground wire and attempt to locate the input jacks near the volume control and driver tube. Of course these components should be located as far as possible from the power supply.
Once you get the parts laid out the way you want them, the next step is to mark and drill the chassis or top plate. I like to start with a small pilot drill bit and drill all the holes on the chassis. Then I drill the holes from smallest to largest, drilling all holes with each bit of increasing size. I find it makes a much nicer hole if it is drilled with a sequence of larger bits rather than with a single large drill bit. Using a large bit tends to create a slightly triangular hole because of the mation of the bit as the hole is drilled. I usually use a 1–3/16" hole saw to drill the large tube socket holes, so for these I only drill the initial pilot hole and then use the hole saw. Be sure to use a grommet anywhere wires pass through a hole in metal. Once all the hole are drilled, be sure to remove any burrs and sharp edges on the holes, perhaps using a countersink bit and small round file for larger holes.
You will notice that I didn’t use the 4 output taps for various impedance speakers. I didn’t want to crowd the binding posts and so I just wired it for 8 ohms.
Once everything is mounted to the chassis it is just a matter of wiring up the circuit. This becomes an excercise in minimization of the signal path, but I prefer to sacrifice a little signal path length to make things tidy and sturdy. If you have a chance to look at the construction of old military tube equipment or a tubed tektronix scope, it can be an education in reliable construction. I like to use zip ties to bundle wires together (never bundle signal wire to filament wire!) and nylon clips to attach wires to the chassis where possible. Always be sure that all components are securely attached, and nearby components are insulated from each other.
I usually start by wiring the power supply and filaments, then wire all transformer leads that have a place to go just to get all the attached leads out of the way, then move on to signal wiring. I use the little upright terminal strips that Radio Shack sells to build the wiring around the tube sockets. There are 2 ways to do the signal wiring. You can attach all of the components (resistors and coupling capacitors mostly) to some sort of terminal strps or circuit board and then attach each component to the tube socket with hookup wire, or you can attempt to attach the component leads directly to the tube socket. I like to do th elatter because it it usually possible without too much clutter, and results in far more compact wiring with fewer solder joints. Component leads that need to run close to each other can be insulated with heat shrink or the insulation from hookup wire.
I think the layout would have been cleaner if I had put the rectifier where the filter cap is and vice-versa, but I wanted the tubes to be symmetric for aesthetic reasons. It is also nice for the hot rectifier to be away from the power switch and any interconnects.
Closeup of the tube sockets. It would have been better to mount the 6sn7 (center) 180 degrees from how it is here, because the routing of the filament wiring would have been kept away from the input wiring. Notice that the cathode resistors (the big ones) are mounted as far away from other components as is practical so that they can have enough airflow around them to stay nice and cool. Also note that both cathode resistors are bypassed. I think this is a matter of personal preference in sound.
Finally we have a complete amplifier! Hooray! And it even works :) The base is boxjointed Poplar. I really like the color of the wood and the really subtle grain. Home Depot sells Poplar in finished boards for less than the cost of Oak.
I decided to cover the bottom of the amp with this 1/4" pegboard. Once stained it looks pretty neat and is a nice contrast to the light bases.
Tweaks and Improvements
One question is whether or not to bypass the cathode resistors. I tried it both ways and the sound was definitely different, but not overwhelmingly so. Unbypassed, the sound seemed a bit less dynamic and more closed in, which I supposed could be called “sweet”. In the end I chose to bypass the cathode resistors on the driver stage and the power stage with 100uF Sprague Atom capacitors. This might result is just a bit more gain as well.
It is difficult to read about DIY amplifiers without being hit with extensive discussion of capacitors at every turn. I personally like paper in oil caps very much, and tend not to like Solens very much. I used the cheapest paper in oil capacitors I can find for the coupling capacitor in this amp.
The size of the coupling capacitor can be adjusted to vary the low end frequency response of the amplifier. There is a design school that says that the low end response of the amplifier circuit is ultimately determined by the low end response of the output transformer, and so there is no sense using a large coupling capacitor that allows low frequencies down to well under 40Hz if the output transformer is going to roll off at 40Hz anyway. Why force the output tube to waste power amplifying these low frequencies when they will not make it to the speakers anyway? I chose not to size the coupling cap to limit the low end because I figure that even if the output transformer is going to be rolling off the low end, it is better to roll off the full bandwidth instead of adding an additional 6db/octave slope to the low end attenuation. My thinking is that when it comes to output below 40 or 50Hz, the more the better since it is an uphill battle to produce these frequencies anyway. I may be underestimating the power loss in the output stage, but if this were about high power and efficiency, we would be doing push-pull class B or something.
Increased fidelity can be had by decoupling the left and right channels, ultimately by building monoblocks. Steps can be taken towards complete decoupling by using a seperate 6sn7 for each channel, and by using twin power supply filtering sections.
—
I hope you have found this information useful, and are encouraged to pursue vacuum tubes further. All I ask is that you do not drive the NOS tube prices any higher than they already are :) Have fun and be safe :)
Feedback can be sent to jsn at boozhoundlabs dot com. I welcome corrections, elaborations, suggestions, and whatever you feel like sending.
Thank you to everyone who has sent me mail about this page. I have gotten tons of feedback, corrections, and questions. I am happy to get all of it. Here is a bit of what I am getting (book jacket style):
“Wow thanks jsn… I have printed out your primer to read carefully. Thanks very much!!” — Ale (audioasylum post)
“As a newbie designer I found this article to be one of the most straight forward, easy to understand how to’s I have ever read. Having recently cobbled together a preamp design of my own, I was glad to find this BEFORE I started my next project.” — Rod (www.dortoh.ca)
“I’ve been reading all I can in the last three months, trying to understand tube audio & guitar amplification. There is a lot of information on the net, as you know. Your primer though is the clearest, most straightforward explanation I have found. It was the first explanation which explains exactly what each component is used for, and why the value selected was selected. This is great, thanks.” — Steven Husting
“Just dropping you a note to say how helpful the hi-fi primer is. After a few years of making my own interconnects and speakers, I got the itch to try my hand at making an amplifier (though I’ll probably do a Foreplay preamp first). At least now I have answers to those simple questions like “why the heck does this wire run to this pin socket, and why is there a resistor between _these_ two pin sockets, and… Anyway, thanks.” — Doug Asherman, Oakland, CA
