How to Build a Portable Bluetooth Tube Amplifier

After seeing several builds of bluetooth speaker boxes I wanted to try one myself, but with a twist. It has to use vacuum tubes, but small, not a converted tube radio, a scratch build. And it has to use batteries, otherwise, where is the fun?

Before we start, be aware that we will use Li-Ion batteries, which if not handled correctly can catch fire and explode. We will also work with high voltages, which can be dangerous. To make things safe, we are going to use an existing and safe charging module, which disconnects the load while the battery is being charged. The high voltage will be supplied at small currents, through a small switched-mode-power-supply (SMPS).

In the next steps we will go trough most of the designing and building process. If want to skip the designing part, where we draw and simulate the circuit go you can jump to step 5, where we start with the wooden box.

For this build you will need the following tools:

• handsaw
• hacksaw
• hand drill and set of drill bits, including some drills of larger diameters, or something to punch 22 mm holes
• rotary tool
• screw drivers
• metal files
• sandpaper (180, 400, 600, 1200)
• dust mask and googles
• pencil
• ruler
• compass
• box cutter
• scissor
• pliers
• Soldering Iron
• Multimeter

The parts needed for this build listed for each different section:

Wooden box:

• 1000x100x5 mm wooden board
• 200x100x1.5 mm brass sheet
• 1000x5x5 mm wooden stick
• Wood glue

Tube amplifier:

• 2x9 pin tube sockets with set of screws
• 7-way tag strip
• 1x 10 W 100 v line transformer (or any push-pull transformer with a 10k or 20k primary, for example the Hammond 125A)
• Resistors:
  • 1x 1k ohm 1 W
  • 2x 1.5k ohms 1/4 W
  • 2x 3k ohms 1/4 W
  • 2x 68k ohms 1/4 W
  • 3x 100k ohms 1/4 W
  • 2x 150k ohms 1/4 W
  • 2x 220k ohms 1/4 W
  • 2x 1M ohm 1/4 W
• Capacitors:
  • 2x 100 nF 50 v
  • 3x 100 nF 400 v
  • 1x 10u F 16 v
  • 1x 10u F 350 v
  • 1x 22u F 350 v
  • 2x 100u F 50 v

Bluetooth module:

• MH-M18 module
• B0505S 1W isolated DC/DC converter
• 2x 10u F 16 v capacitors
• 1x 100n F 16 v ceramic capacitor

555 SMPS:

• 1x 555 IC
• 1x IRF644Pb mosfet or any other for 250 v or more and Rds_on less than 0.2 ohms.
• 1x BC547 transistor
• 1x 47u H 3A inductor low ESR
• 1x ES2G diode or any other ultra fast diode (UF4004)
• 1x 1k ohm trimmer
• Resistors:
  • 1x 470 ohms 1/4 W
  • 1x 1k ohm 1/4 W
  • 1x 10k ohm 1/4 W
  • 1x 2.2k ohms 1/4 W
  • 1x 56k ohms 1/4 W
  • 1x 240k ohms 1/4 W
• Capacitors:
  • 1x 100p F 50 v
  • 1x 2.2n F 50 v
  • 1x 4.7u F 350 v low ESR
  • 1x 330u F 25 v low ESR

Other parts:

• 1x TP4056 charger with separate pads for the load
• 2x 2.5" (65 mm) 88 dB speakers
• 2x oval (70x40 mm) passive bass radiator
• 1x circular (50 mm) passive bass radiator
• 4x 18650 batteries
• 4x DPDT 6 mm push buttons (of good quality)
• 3 mm LEDs: 1x red, 1x blue, 1x white (or desired colour)
• 1x 3 mm LED socket
• 1x 3.5 mm switched stereo audio jack
• wire
• shrinking tube

Defining the tubes

There are different ways of achieving the desired 1 W of power, depending on the choosen tubes. But only a few are not filament hungry. If one only considers vacuum tubes that are still in production, and easy to get, there are not many options. The chosen tubes are a 12AX7 for the preamp and a 12AU7 for the power amplifier. This way the total filament power is only 12 v 300 mA = 3.6 W.

Battery powered tubes use even less power, but they are not in production anymore, and require a special biasing approach.

Transformer and output stage

The output stage could be either single-ended and stereo or push-pull and mono. Both would give something close to 1 W in output (SE 2x0.5 W, PP 1x1 W). The main difference is that two SE output stages would require two output transformers. The SE transformer also has to have an air gap and is usually larger than the equivalent PP transformer. Because both speakers are in the same box, the stereo sound is not so different than when using a mono. This settled the matter to me. If you prefer, you can go the SE route, and use two transformers, just remember, they need to have an air gap. A good candidate is the fender reverb transformer I used in my previous build.

In this build I opted to go for a cheaper version of the output transformer: a 10W 100 v line transformer with 0, 10, 5, 2.5, 1.25 and 0.625 W taps and 4, 8 and 16 ohm connections. One can achieve 32k plate-to-plate when using the 0, 2.5 and 0.625 W taps with a 16 ohms speaker connected at the 8 ohms tap. The 2.5 W tap works as the center tap. This, obviously, is not perfect. One of the sides has more turns than the other. The resistance is also different due to the different wattage ratings. The result is still satisfactory, though. For a better result look for a transformer with a 20k primary.

Battery configuration

For it to be portable it must be battery powered, otherwise you will always be looking for a wall outlet. Here you are going to use 4 batteries to achieve a longer playing time while powering filaments and the SMPS. Since both tubes can be supplied with 12 V at the filaments 4 batteries is the right choice. 4x3200 mAh batteries will last longer than running 2x3200 mAh with the filaments at 6v.

An estimation of the power required can be calculated by considering the power of the filaments of 3.6 W plus the power required to supply the high voltage.

When using a SMPS an efficiency of around 70% is expected. If we run our output stage at 250 v and 10 mA

250 V x 10 mA x 100/70 = 3.57 W,

so that we can expect a total of 7.2 W.

Worst case scenario, with the batteries at 12 V it would give us

3200 mAh / 0.6 A = 5.3 hours of playing time.

This shows that, although tubes can be used with batteries, they have higher power requirements in comparison to other class D amplifiers of same wattage.

Speaker size and efficiency

When looking for a speaker, the smallest would save space, but would also have a lower efficiency. Between 80 dB and 90 dB the difference is perceptible, so that you will have to look for the smallest speaker with the highest efficiency you can find. That will also influence the size of the wooden box. The best fit I could find at a reasonable price was 88 dB and 66 mm (2.5") in size.

Cabinet size

There are several bluetooth speakers with roughly the same size. The goal here is to have something similar, but using tubes. To make things easier it is nice if one can check the size of the components and the positioning through drawings. Ideally the box size would be a bit larger than a hand, so that it can be easily carried, butt still packs enough output power for a living room.

Conclusion

In summary, It will be a mono tube amplifier using a 12AU7 in push-pull operation and a 12AX7 tube as a gain stage and the phase inverter, delivering almost 1W. It will be powered with 4 Li-Ion batteries, which supply the filaments and a switched mode power supply (SMPS). Everything will be stuffed inside a wooden box with 220x75x80 mm (8 5/8"x3"x3 1/8") made with 5 mm thick mahogany planks. It can be charged with any 5 v supply using a micro usb.

3D design

The use of CAD tools can accelerate the fitting and positioning of components. If one has access to those tools, through the university, for example, it can really help. In this case I used the available Solidworks, which is not free, but can be used at the university. There are some online repositories where some of the components, such as the 12AX7, can be downloaded and used in the final build. The other components can be quickly drawn, just to give an idea of the size of the part. If you don't have access to this software, Sketchup used to be free.

When using the software save each part (box, speaker, tubes, batteries, transformer) as a separate part. You can make a 2D sketch and then extrude it to have an idea of the used space inside the box. In the build package you can add the necessary parts and position them according to contact between planes or, concentric arrangement, as used for the speaker and the speaker hole. Here I also used some rendering to improve the appearance of the final result. The first figure is a mock up of how it could look like, the second and third one are plane cuts, that show the interior of the box and the chosen arrangement. If you are using a different tube, or two transformer, make sure that it will fit. Later you can make a drawing with the main lengths, that will help you when making the wooden box.

This drawing helped me in designing the size of the box. For example, two tubes sockets side by side require at least 62 x 30 mm. With the 2.5" speaker (66.5 mm width), separating walls and the sides the total necessary width is 214.4 mm, here rounded to 220 mm. Similarly, the necessary height is 66.5 mm plus top and bottom boards (5 mm), rounded to 80 mm. The depth was chosen based on the dimension of the tubes added to the size of the output transformer, front and back boards, which resulted in the 75 mm shown in the final drawing. Other dimensions were defined based on aesthetic aspects, such as the cutout for the displaying the tubes.

Metal chassis

To minimize the size of the walls around the tubes a metal chassis was used. The golden brass would also give a nice look and reflect the tube glow. Due to the height of the tubes the chassis needed to be as close to the bottom as possible. The tubes sockets, on the other hand, also have a considerable height that needs to be considered.

Here we are going to use a space saving measure, but at the same time also a mounting approach, is to just slide the chassis metal plate using a small groove in the separating wooden walls. Another brass sheet hides the output transformer and simultaneously isolates the tubes from the interior of the box.

Now that we have an idea of the size of the box, lets design the schematic.

Push-pull design

To calculate a good bias point , where we can get close to 1 W out from the 12AU7 in push-pull with not that much harmonic distortion we are going to use an online calculator. First we select the tube and change the load to reactive, since we are going to simulate a speaker. Then we can select the PP configuration and change the load to the value of our transformer. I will use a 32k primary, but you can also try it with 20k to 24k and see how the output power at max. g1 changes. At the current state the load line, in red, is crossing the dashed line, the max. dissipation line, meaning we are going to melt the tube even on idle. We can reduce the quiescent current or change the plate voltage, V+.

Before we chose, we have to consider that we want to use a SMPS and it will have to supply two times that current at the defined voltage. There will be losses at the inductor and the mosfet, resulting in extra heat. If the mosfet gets too hot or the inductor saturates the voltage will fall, and we do not want that. For voltages bellow 250 V the IRF644 Mosfet, with a resistance of only 0.2 ohms, can still be used in our SMPS. Higher voltages can be obtained with better (or larger) mosfets, but then at 300 V we are also very close to the maximum plate voltage of the 12AU7. Because of the hassle we are going to chose something below 250 V, so that we still can use the IRF644. If you know of another mosfet with a low on resistance, that does not cost much, and is easy enough to find, please, leave a comment. I changed the voltage to 245 v and now we can play with the quiescent current.

You will notice that hotter bias (higher current) not always means more output and the class B part of our curve, where the red curve just changes its inclination, gets smaller, meaning both tubes will be working hard for a longer time. I decided to go with 4 mA. It gives me a bias of -11.8 V, and will require only 8 mA of our SMPS, which is good, since close to 15 mA things start to get really hot and we do not have the space for a large heat sink. Colder bias increases the harmonic distortion by getting our load line close to the region where the grid curves are all bunched together.

You can check the harmonic distortion by entering a value at the Out. headroom box. In this calculator, though, it is a little counter-intuitive. Normally I would say that the headroom is the maximum grid swing necessary to make the tube clip. It can clip at the zero grid curve, on the left, or at the bottom, where plate current is theoretically zero. This value is the same as the grid bias. Here, on the other hand, you have to specify the high voltage, not the grid swing, which is more complicated to guess, since it changes with the bias. Ideally you can get half of our V+, at 125 V, but if you change the bias, the green area is now outside the working region, meaning that you have to correct it. Here, we will just use it to check how much harmonic distortion we have, so that it is at reasonable levels.

Using another load line calculator (paint_kit) the increase in the distortion can be directly plotted with the grid voltage swing. At 11.8 V the maximum distortion is below 6%. Which is nice. This value also dictates how much signal swing we need at the output stage grid to get the full power output. To verify which strategy is the best we are going to simulate the stage driving the output stage.

Simulation of output stage design

The whole amplifier can be simulated with LTSpice to further improve the circuit. I added the files, so that you can try it too. When opening the file there will be three different designs and some comments. The comments define which kind of simulation we are going to run. The transient simulation (.tran 0 52ms 50ms) gives us the sinus curves, where we can check for the maximum output of each stage, and if it is distorting. The frequency sweep mode (.ac dec 10 100 20k), where we define the frequency range to be tested we can verify the frequency response of the amplifier. To comment or to use the desired simulation you just need to change the comma by a dot. The program will then understand that you want to use that kind of analysis. When the simulation is done, just check for the signal at different positions of the schematic by clicking on it. So now that we know the basics, lets discuss the schematics.

With the 12AX7 there are different ways to drive the output tubes. Since we are running the tubes in push-pull it also requires a phase inverter (PI). There are two configurations of driving stages that can be used, as shown in the second figure, either using a gain stage followed by a cathodyne or a long-tail-pair. Since in the long-tail-pair both triodes are used for the phase inversion it has a slightly lower driving capability, or, in other words, less gain. If this were a guitar amplifier, that would reduce the output stage distortion, which is not always desired. Since this is a bluetooth speaker this is not a concern, it just need to have enough gain to drive the output stage to full output (at 11.8 V). The cathodyne, when compared to the long-tail-pair, is not as good balanced, so that when the output stage starts to distort (positive grid) is loses its balance resulting in an annoying distortion. You can verify it by increasing the voltage at the input produced by the two AC sources. I was not sure of the output of the bluetooth module and the required filtering between bluetooth module and tube amplifier, so that I opted for the cathodyne with the extra gain stage. (With the amplifier built I discovered that I could also have use the long-tail-pair approach). The curves shown in the third figure illustrate the output at the speaker, here simulated as a resistive load. It is interesting to see that at the same input the extra gain stage in front of the cathodyne gives an extra output. Just remember that the input is only 0.25 V here.

In case someone is wondering what would happen if a bluetooth module with differential outputs would be used, the schematic in the fourth figure shows an attempt of modelling a differential driver stage. Each triode drives one side of the output stage. The advantages of the differential design is normally a stronger output coming from the bluetooth module and good noise rejection, since noise is picked by both sides, which will cancel each other later, when they are amplified by the output stage. Now the input of the differential amplifier is driven by a 0.5 V sine curve, to account for the extra signal of the symmetric input. The results show that it can produce the same output than the long-tail-pair, which was driven by just 0.25 V, but now with probably less noise.

Summing amplifier

To convert our stereo signal to mono a summing network (100k resistors) in front of the first gain stage is used. More complex bluetooth modules allow the use of a mono output, if necessary, only by some settings in the memory of the chip. Here, however, a simpler module was used, so that the summing network is necessary.

WIth this we conclude the simulating stage, and now have a nice schematic as a base for our build. Before we start to build however, we need to solve the battery charging problem.

As said before, the batteries need to be in series to have enough voltage to heat up the filaments and to power the SMPS. On the other hand, they also need to be charged at some point. Charging cells in series is not as easy and safe as charging the cells in parallel without a proper charger. So that the batteries do not need to be removed from the box for charging and to minimize the required space for the charger, charging them in parallel is the best option. Here we are going to use a normal 5 V charger, easily found online, that would even allow charging through the USB of your computer (but at very slow rates).

A different problem is how to switch the 4 batteries from series to parallel?

It can be done with relays, but they require power, or with transistors, but that would require another schematic, much more complicated. The easiest solution is using a 8PDT or a 6PDT and switching only 3 of the batteries, while the first one is always connected to the charger. The charger will disconnect the first battery from the load using the onboard IC (remember to get the right one, with the extra outputs).

The connections used here are shown in the schematic, where the 6PDT is illustrated by separate SPDTs. BAT+, BAT-, LOAD+ and LOAD- are the connections of the charger, while AMP_GND are the grounds of the DC-DC isolation for the bluetooth module, the tube amplifier and the SMPS.

Unfortunately the 6PDT is not a cheap switch. It also requires some extra space, that must be considered. Good that we have a 3D drawing were we can check for the necessary space, which is between the batteries and the output transformer.

Now, we can start building!

We can print our drawing of the box and start cutting the boards the proper size.

Bottom and top boards need to be cut to 220 mm by 75 mm, while front and back boards need to be cut to 210 mm by 70 mm. Sides are 75 mm by 70 mm. Be sure to cut two of each. Use a ruler to mark where you have to cut and cut a little outside of this line if you are using a hand saw, so that the mark is still there when you have finished cutting. Later, you can use this extra millimetre to adjust possible mistakes and sand everything flat.

For the holes on the front and back plates use a hole saw. If you do not have one, you can use a small drill bit, as shown in the third figure and round the shape with a rasp. Round the corner to your liking. For a better finish use sanding paper, starting with the rough one and moving to the finer one. Before moving from the 600 to the 1200 be sure to that you removed the marks all the marks from the 300 one.

The boards that will be the outer structure of the box can now be glued together with wood glue (top, bottom and sides, not the front and back!). I used the small 5 x 5 mm stick at the inner corners. Cut it to 60 mm pieces and make sure that they are almost centered when gluing the boards together. At the front side the speaker might have to fit between it and the front plate (just if you bought a speaker with a square frame), so check if you need this space before you glue it.

Chassis

While the glue is drying we can work on the chassis plate. Make the markings to the correct size, here a bit more than the gap between the wooden walls enclosing the tubes, resulting in 65 mm by 65 mm. To cut the brass sheet use a hacksaw or, if you have one or access to one, a band saw.

Now lets check the ideal positioning of the sockets. To use the least amount of space position the screwing rings inclined 45 degrees with the 65 mm front, otherwise they will not fit (see picture). My sockets were 26 mm by 35 mm so that one of the sides always was larger than the drawn space. Since between the wooden walls we only have 62 mm they will not fit with the screws aligned. Side by side the depth is to large, because the walls in our drawing only have 30 mm, because the output transformer needs to fit behind the tubes. Rotated by 45 degrees (diagonal) they kind of fit closely together.

Before making the markings for the holes draw a line from the front to the back (both sides with 65 mm) and make sure that the sockets touch exactly at this line, so that the tubes will be centered. Put the socket rings as close together, and as close to the front edge as possible and mark the centers for drilling.

To drill the holes I used a step drill bit, but it was dangerous. The brass sheet easily gets stuck to the drill, rotating the brass sheet with it. If it is clamped it will deform, and we do not want that. A hole drill works better in this case, even when it was made for wood.

After the holes for the sockets are done, check the right position of the screws that keep the socket in position. Ideally there will be one close to the center of the metal plate, where we are going to fix the tag strip. Measure 30 mm from the front and position the transformer. Another brass plate will be positioned there. It has to be cut to the exact size, so it is better to cut it after the walls are glued (next step of the wooden box).

This plate will have approximately 70 mm in width, so that it rests against the smaller wooden walls, and 70-12.5-1.5 = 56 mm in height. Now make the markings were you will drill the holes for the transformer screws and the holes where the wire can go trough.

Making a chamber for the tubes

With the sides glued, we can now make the walls that will isolate the tubes from the interior. If you were really good with the saw you can even use the 30 mm by 210 mm piece left over from the front and back.

Cut it to 70 mm and make a mark at one of the sizes at 12.5 mm from the edge.

We are going to cut the groove for the chassis. I used the hack saw for a tighter fit, but the hand saw also will do. Cut it at least 2 mm deep.

To make sure that it will be glued correctly use the chassis and slide them in the box. If you used the pencil trick, it might be a little more than 70 mm and the wooden wall will not slide in the wooden box, so that you will have to sand it with the rasp until it fits. A tight fit is ideal, so that you almost do not need to glue it in place.

Position it with the chassis in the box and with one of the sides on a table use the speaker to check if it is correctly centered. With everything centered glue it in place. You can remove the chassis later.

When the glue dried we will use the speaker again to check were they end and a little further make a marking on the center of one of the sides. This will be the position of our charger.

With a small drill bit make aligned holes, perpendicular to the bottom, so that the charger will be parallel to the back of the speaker.

Use a file (yes, only a metal file is small enough) to sand the holes in a long slit where the charger will be connected. At this point the thickness of the side is more than the protruding USB connector.

On the inside use a larger drill to remove some material. Because there is not enough access it will look goofy. If you do this before you glue the box it will probably be better, but since there was no marking for it before I did it at this stage. With the holes for the LEDs it ended up looking like a sad face.

The holes for the input jack, on LED and 6PDT can also be drilled now. They are not exactly at the center of the depth of the box, because of the depth of the speaker used.

Measure that you have enough space for the 6PDT, which is roughly 10 mm from the back of the speaker and the brass plate that goes between the transformer and tubes. I used less, and cursed it once a day. In my defense, my first DIY switch was smaller, but it was not reliable enough.

The LED can go on top of the transformer, but the input jack and the 6PDT will not fit there. so they have to go at least 10 mm away from the transformer. Be sure to use the correct transformer dimensions while adjusting this at the drawing.

Now we will make a knob for the 6PDT, so that there is no difference in color between the button and the box. You will need a thicker piece of wood, or you can glue scrap pieces together .

Cut a 15 mm x 15 mm piece and make a knob out of it. I used a rotary tool as a lathe and a file to sand it down.

Before you use the rotary tool you will have to make it rounder, otherwise it will break pieces every time it hits the file.

Battery holder

Plastic battery holders are normally larger than the dimensions we chose for this build, with a height of 80 mm. Instead we are going to use a wooden battery holder. To make it we will again use a scrap piece of the boards we bought.

The internal height of the box is 70 mm, and every wood piece has a thickness of 5 mm. The batteries have 65 mm in length, so that we will have to remove 2.5 mm of material of each side, and the batteries plus wooden holders will slide perfectly into the box. A tighter fit is better in this case.

Use a drill, with a 18 mm flat wooden drill bit, to make a recess where one of the sides of the battery has to fit. I used a smaller drill bit (16 mm), so that I had to enlarge the recess with a hobby knife.

With a small drill (1.5 to 2.5 mm) drill a hole from one of the sides for each circle for the wiring.

Each piece will only hold two batteries, so that we can distribute them symmetrically in the box, as in our drawing.

With the wooden pieces cut and glued you can paint them with your preferred kind of finish. Here I used Teak oil for a natural look.

First fix the tube sockets to the brass sheet like in the first picture, from the top, where the 9th pin is close to the edge of the chassis.

This way once the circuit is soldered there is no way of removing the sockets from the chassis, but you just gained some extra millimetres between socket and the bottom of the box. At the screw close to the center of the chassis fix the 7-way tag strip. I also tried adding a small one near the front of the chassis, close to the 9th pin, but later I decided to take a different approach. You can cut the 9th pin off of the socket, since we are not going to use it (it is used for the 6.3 v filament supply configuration), or protect it with shrinking tube, as I did. Bend the other pins so that they are flat with the socket, otherwise they will touch the bottom of the wooden box when slide in.

Now solder the filament wires from one socket to the other.

After the filaments start soldering the resistors from the first to the last stage, starting on the right of the layout. The input resistors, for the summing network will be hanging from the socket, so that it is a good idea to use the shrinking tube to protect them and avoid any shorts. Then move to the capacitors, the film capacitors first and then finish by soldering the electrolytic capacitors. Bend the capacitors so that they can lay down parallel to the chassis. I attached the point-to-point layout and figures to give a better idea of the positioning.

You will see that, instead of using the smaller tag board I just used the leg of the capacitor to anchor the resistors between the sockets at the top of the picture (front of the chassis). The capacitor is considerably large, since it is for 400 V, where 250 V would also work, so that by avoiding the tag boar at this place we save some space.

Solder the input capacitors and 1M resistors directly to the input jack, or to a perf board. under the input jack.

Filament voltage regulator

The filaments require 12.6 V, but the batteries fully charged will deliver around 16.8V, which is way more than the accepted voltage. To reduce this voltage we will use a voltage regulator. I used the LM317, but the LM7812 will also work. Instead of etching a board we are going to use a perf board cut to 5 x 6 islands with a hacksaw. Solder the LM317 first and then the resistors and capacitors, to have a better idea if the spacing. To join the islands use the cut component legs. The layout is shown in the 6th figure. The 240 ohms and 2.2k ohms divider sets the output voltage to 12.7 V. this part can be directly soldered to the sockets, just make sure that the mosfet will not touch the chassis. The 9th figure shows where I fixed it in my build, at the space behind the 100 nF capacitor of the output stage.

SMPS

For the SMPS there is a PDF with the schematic, image for etching and the position of the components. After etching the PCB do not forget the jumper under the 555 IC. The mosfet will be the last part to be soldered, and with the heat sink already screwed. Therefore the legs of the mosfet need to the longer. Bend in over the 555, so that is uses less space. Make sure that it is not touching the resistors, though. Depending on the size of the heat sink the high voltage capacitor will not fit. In my case I soldered it laying on the board, instead of standing. This is visible in the picture.

Transformer

The last step is to screw the transformer and solder the correct taps. The center tap, also connected to the high voltage pin at the tag board, is the 2.5 W pin in my case. For the Hammond transformer check the correct number. The other two connections come from the plates of the tubes and go to the 0 and 0.625 W taps of the line transformer. On the other side we will connect the speakers in series, resulting in 16 ohms, at the 0 and 8 ohms tap. This way we are multiplying the 16k plate-to-plate impedance of the primary to achieve the 32k we need.

Now we can test the circuits.

To check the circuits we are going to solder the batteries in series, without using the battery holder. To test each circuit we are going to connect the batteries directly to the circuit. Alligator clips can also be used, to make connecting and disconnecting faster.

Test the filament voltage regulator by connecting the end of the batteries to the positive and ground inputs and measuring the output with a multimeter in the voltage setting at 20 V.

It should read around 12 V between the output and ground.

Test the SMPS output. Before connecting the input make sure that the trimmer is in the middle position. Without load the output will be a little higher.

Connect the positive and negative pins and measure the output with the 1000 V (anything higher than 200 V, depending on the multimeter) setting of the multimeter.

It should be in the order of 200 V. If it is higher than 250 V carefully adjust the trimmer to 245 V. If it is lower, but still higher than 200 V there is no problem, we will adjust it later. If it is only about 16 V there is a problem with the SMPS, and a deeper analysis with the multimeter is necessary. Turn it off and check the continuity of the components. and the respective tracks.

Testing the tube circuit
To test the tube circuit we will solder the input jack, the speakers, the filament regulator and the SMPS.

When everything is connected add the batteries. Check the voltage that goes from the SMPS to the amplifier. It should reduce a little bit as the tubes heat up and stabilize. When it is stabilized you can adjust the trimmer once more to obtain the 245 V.

For testing we are going to use the input jack and play something through it. If there is sound coming through, congratulations, you did everything right. If there is no sound, you need to check some voltages with the multimeter and make sure that every stage is working correctly.

• First, check if the filaments are lit, and measure if you have 12.7 V between them.
• Second, check the high voltage tap (2.5 W) at the output transformer, and at the reservoir capacitor in the tube circuit (22 uF) both should read around 245 V.
• Third, check bias at the output stage. There are two 3k ohms resistors, one for each triode, so that both should read something similar, around 12 V at the cathode.
• Check if all the grounds are connected together, for example, one of the filament wires connects directly to the ground wire.

Checking the bias

In the schematic and in the build there are two resistors biasing the output stage. Instead of using a single resistor with half the resistance, I opted for separated resistors, so that I could check and correct any difference between the triodes. This occurs because the triodes do not match perfectly. One will have a higher current than the other. To avoid that one triode does all the work, they are biased separately. Here, luckily the difference was not significant, so that the same resistance can be used for both tubes.

Check the bias of your triodes and adjust the resistance if necessary.

If you remember the bias we had with the load line calculator, you will also see, that the resistor I used is larger than expected, because:

11.8 V / 4 mA = 2.95k ohms, with the nearest resistor available at 3k ohms.

When I measured the bias voltage it was not matching the load line, because the 3k resistor resulted in a 10.4 to 10.6 V bias, which only gives:

10.4 V / 3 k ohms = 3.5 mA for this resistor.

Decreasing the resistance, to adjust the current also reduces the cathode voltage, as expected.

After checking the circuit I could not find the reason behind the deviation between the expected and the measured value. The filament is at 12.7 V, so that it is not caused by a cold tube. It is either a characteristic of the tube or the circuit. Maybe someone has a better idea of what is happening.

Adjusting the imbalance in the output stage

Because I used a line transformer, there might be a little difference in the voltages measured at the output plates. The reason for that is a different number of turns for each side and different wire gauges for each power. Both will affect the resistance and result in a different voltage drop across each winding. This is not so bad as it sounds though. Checking the output with an oscilloscope shows only a small imbalance.

Another aspect causing a slight difference in the combined signals is the difference in gain of the used tubes.

If you have access to an oscilloscope testing the best triode to transformer tap combination can reduce this imbalance.

The bluetooth receiver I used presents a stereo output with common ground. The same ground is also used for the power ground of the module. A first test showed a nasty switching noise, which was successfully removed by using the suggested 5 V DC/DC isolator. It works by breaking the ground loop created when the bluetooth module is directly connected to the power amplifier.

Solder the DC/DC converter to a small perf board, add the two bypass capacitors, as suggested here. Add an extra 100n F capacitor at the output, for filtering the high frequency noise.

Instead of using the V+ pin of the bluetooth module, use the pin on the side, which bypasses the diode. According to the datasheet the V+ pin is for 5 V, while the lateral pin is for battery (3.7 V) operation.

The right and left outputs, shown in purple in the layout, connect directly to the switched jack. Because there is a single ground on this bluetooth module instead of connecting the signal ground directly to the input jack, and shorting the DC/DC converter use a large (10u F) capacitor, which only allows the signal to pass, without creating a DC path.

With most of the circuit built, it is time to put everything together.

• Glue the speakers to the front plate using wood glue.
• Glue the bass radiators to the back plate with the same glue
• Solder wires at the speaker terminals, later it will be difficult to access this region with the soldering iron.
• Fit the front plate in the wooden box and glue it in position.
• Glue a piece as large as the battery charger near to the hole for the charger and wait for it to dry.
• Glue the charger in position using the recently glued support. A small screw or nail trough the wire holes can be used for an extra support. They will be fixed to the support block.
• If you drilled the holes for the LEDs remove the LEDS from the charger and solder your own in position. Make sure to check the polarity on the charger before you desolder the LEDS.
• Insert the tubes in the sockets, position the brass plate between the tubes and the transformer and slide the whole circuit in.
• Cut the speakers wires to the necessary length and solder the speakers in series and to the output transformer. Try to keep them as short as possible, while also keeping them out of way.
• Screw the input jack, and solder the connections from the tube amplifier.
• Screw the LED holder and solder the negative terminal to the input jack ground. Add a resistor between the positive terminal of the LED and the input of the filament voltage regulator. Use anything starting from 2.2k, according with the desired brightness.

To fix the 6PDT and achieve the right spacing use a small piece of the mahogany board. It will add some extra depth to the hole, so that the plastic shaft of the switch will not stick out. I also cut it to half and drilled a hole in the under side of the wooden knob, so that even with the amplifier turned off the plastic shaft is hidden.

• Fix the switch to the small wooden piece with screws. Make sure to remove enough material so that the screws are flush with the wood. If the screws stick of the wood they will avoid a good superficial contact of the wooden piece with the inner surface of the wall and might get lose.
• Bend and cut one of the sides of the terminals of the 6DPT switch.

The 6PDT comes with wire terminals and PCB terminals, with the wiring terminals turned to the back of the box the PCB terminal have to be cut, so that they do not touch the speaker and the brass plate. I also insulated that side with a plastic card.

• Glue and screw the wooden piece with the 6PDT to the inner side of the box.
• Solder the necessary wires to the switch, as illustrated by the layout in the first.
• Solder the wires coming from the tube amplifier and filament regulator to the LOAD- terminal of the charger.

Since only the negative terminal of the charger is switched by the charger, we are not going to use the positive load terminal of the charger. This terminal is at 3.7 V, while our LOAD+ will be at roughly 16 V, using the pin indicated in the layout for the switch.

We can test the connections without the batteries using the multimeter to check the continuity. With the switch unpressed all battery positive terminals should be shorted to the BAT+ terminal. The same should occur for the negative terminals and the BAT- of the charger. To check continuity when the switch is pressed we need to check between the positive terminal of a battery and the negative terminal of the next battery. If everything is ok we can position the batteries in the wooden holder and slide them in the wooden box.

• Solder the bluetooth module signal terminals to the input jack (See figure).
• Solder the supply to the bluetooth chip from the DC/DC isolator to the 3.7 V point, at the bottom of the second battery and the negative terminal to the LOAD- terminal of the charger.
• Solder the SMPS output to the high voltage wire coming from the tube amplifier, the negative, or ground, terminal to LOAD- at the battery charger and the positive terminal to the LOAD+ terminal of the switch
• Before you position the SMPS cut a plastic piece and bend it over the switch, to avoid any short of the switch with the SMPS. Ideally fix the SMPS to the box with a wooden screw.