HAROLD - Retro Style Amplifier


Jacek "Jado" Domański (e-mail : )

The idea to make a small tube amplifier for home use walked with me for a long time. In the absence of spare time I moved the start of a work date indefinitely, but in the end it was "the happy moment" that I could begin construction work. One of the main goals of the project, was to build an amplifier using owned materials and components and bargain shopping  - in other words to bulid an amplifier at minimal cost, but with a significant contribution of my own "labor". Another assumption was, that the amplifier can not be a high energy consumption device - so that the moderate daily use does not cause excessive burden on the household budget. So it had to be something simple ...

Initially, I planned to build a small amplifier with two ECL86 (I had these tubes). Unfortunately, there was a problem with suitable output transformers. I had one transformer already done for testing purposes and worked very well, but I missed the second one. I found the core of the burnt transformer of the same shape and cross-section, but the differences in the thickness and construction of laminations,  rised concerns that the subsequent performance of the two transformers would be uneven.

Luckily I found three identical transformers (purchased of the flea market in Warsaw), and decided to use them as a base for the new output transformer (Fig. 1).


Fig. 1.

Core measurements showed cross-section of 9.25 cm2 - quite a bit, so I thought maybe I would use other output tubes.

I had two used but still efficient EBL21 tubes:  one of them worked in the old tube radio "Stolica" (produced in 1957), and the other was my purchase of later times to the same radio, because ... "It seemed to me that the old was worn". After purchasing a new one and swapping them I found that the radio basically worked the same. So all in all I had two good output tubes. And these tubes were in my collection ...

I also had a power supply (Fig. 2) providing anode and heater voltages, made a few years ago for testing purposes.


Fig. 2.

Power supply delivers the following outputs: +250 V/100mA and 2 x ~ 6.3 V / 2.5 A. It uses a full-wave rectifier 5U4G tube and choke as a smoothing filter. Previously, it was used in various kinds of experiments with tubes, and now I decided to use it to power my amplifier. The weak element in the power supply  is a transformer - during winding process (in 2006) there were problems with fitting the windings and I had to use a thinner wire for the anode winding. Hence, the 100 mA is a maximum value that can not be exceeded under any circumstances.

For this reason I assumed the anode current of EBL21 tube at 36 mA and anode resistance of 7000 ohms, accepting to reduce output power (which also contributed to the mentioned energy efficiency). Lower anode current resulting in a smaller diameter of wire in the output transformer. Thinner wire, in turn, means more free space to use, so you can attempt to increase the number of turns of the primary winding, resulting in increased inductance of the transformer and better handling of lower frequencies.

So - a small concession on output power may bring benefits in terms of better frequency parameters of the amplifier. Also less anode current means longer tubes life. With such assumptions I could proceed to calculate the transformer windings. I used the spreadsheet evaluated by Jasio : http://www.mif.pg.gda.pl/homełages/jasiu/eka/trafo.html1

Here are the results:

The primary winding is divided into two sections connected in series, and between these two sections, three secondary windings connected in parallel are inserted (first, secondary first winding, then half of the primary, then the second secondary winding, the half of the primary and the at the end the third secondary winding). I hope that this will reduce the leakage inductance and increase the upper frequency limit of the amplifier.

Acquired transformers (donors of cores) proved to be a "difficult case". Windings and cores were flooded by insulating coating, effectively prevented separation of cores. Fortunately, it turned out that the acetone can dissolve a lacquer, so after immersing the cores in a box filled with acetone for about half an hour, they could be separated with a knife.

Transformers were devoided of bobbins (windings wrapped with oilcloth strips), so it became necessary to build them from-scratch.

Fortunately, I had a thin plate (0.8 mm) of textolite, I decided to use it as material for transformer bobbin. You can use a little thicker textolite (1-1.5 mm), which gives greater strength of bobbin, but this reduces the area of the window where windings can be fitted. Instead of textolite, thin glass epoxy laminate (without copper) could be used. In the past, pressboard was also used. But let's get back to the description ...

Output transformer bobbin was done as follows:

I started by drawing bobbin's component parts (you can also do it on your computer, then print). You have to remember to take into account the textolite thickness, so that the internal and external bobbin's dimensions fits the EI lamination size. Indeed, this is a classic design often described in many books (Fig. 3).


Fig. 3.

The plates presented in Fig. 3 have shapes allowing assembling the bobbin without a single drop of a glue, as a self-supporting structure.

Having the finished  project, I had to move it from paper to the textolite plate. Since we are dealing with right angles, the method proved to be enough is to "prick" - every corner of the drawing with something sharp, for example with stylus or needle.


Fig. 4.

After piercing textolite surface looked like this:


Fig. 5.

Then I connected the points with lines traced by means of stylus.


Fig. 6.

In this way, I transferred the drawing of bobbin's elements from paper to textolite surface:


Fig. 7.

Then I started to cut elements - along traced lines. For the outer edges I used sheet metal shears, for inner egges, where I could not get the scissors I had to use a coping saw.


Fig. 8.

Using these methods, and additionally using of needle files to align the crooked edges , I cut out all the elements of the bobbin.


Fig. 9.

In places where there were narrow longitudinal slots provided on the soldering tips and ends of  windings wires, I drilled holes - densely side by side, along each slot.


Fig. 10.

And then I connected the holes by cutting the material between them with a miniature cutter driven by a "Dremel" drill.


Fig. 11.

Having all prefabricated elements I temporarily assembled the bobbin to see if everything fitted. Where there were some mismatches, I made patches using needle files.The final version of a bobbin looked like this:


Fig. 12.

Finally, the solder connectors remained to be done. I used thin copper plates with a thickness of 0.3 mm, which I just had - you can use a brass or similar material, as long as it is not too hard. I cut strips of width matching the slots in the bobbin.


Fig. 13.

I rounded one end of each strip...


Fig. 14.

... and I bent them so as to be able to slip them through the slots in the bobbin.


Fig. 15.

After putting the plates through the holes, I straightened the, rounded the other end - and eventually formed the tips.

In the end - after making sure that everything fits into the core and that I did not any mistake, I let a few drops of "Super Glue" between plates - so that all tied together. The result was both bobbins:


Fig. 16.

Now I was able to wind two output transformers. Fortunately, I possessed the necessary winding wires:


Fig. 17.

At the flea market I bought missing insulation oilcloth. In this way I had a full set of materials for insulating spacers for transformers (oilcloth, 0.05mm waxed paper, 0.02 mm condenser paper):


Fig. 18.

After cutting them into strips of desired width and cutting edges every 3 mm, I got ready insulating spacers:


Fig. 19.

The old, simple winder fulfill the necessary set of tools for winding transformers.


Fig. 20.

I prepared a wooden block which had dimensions that matched the internal cross-section of the bobbin and I was ready for winding transformers.


Fig. 21.

An insulating spacer:


Fig. 22.

If the wire was severed during winding, I reconnected it and insulated with paper tabs.


Fig. 23.

The ends of the wire windings, inserted into insulating tubes were soldered to the connectors.


Fig. 24.

By doing so, I did all the windings of both output transformers. Despite the complexity of the work, it was not so very time consuming - it can be done within a few hours. At the end I set up one final outer layer of insulation with a width equal to the width of the bobbin and placed stickers with descriptions and parameters of transformer windings.

Since the transformers had to be designed to work in SE amplifier it must have a slot. So, another work waited for me. I had to make strips fixing the core along E and I shaped laminates. Earlier in the book written by Cykin I read that the strips made of magnetic material shunted the transformers gap, so I used a non-magnetic material - stainless steel. This material is a little heavy in machining, but after polishing takes on a beautiful luster, and we do not have to worry about rust.

The last step was to make "feet" mounting transformers to the chassis - here I used aluminum angles purchased at supermarket. After placing all the laminates, inserting the paper spacer into the gap between "E" and "I" packs and painting cores with black paint I obtained ready to use output  transformers:


Fig. 25.

The first stage of the work on the amplifier was completed. Now, I could think about the concept of the amplifier and detailed schematic.

After reviewing of several books (Meluzin etc.) and searching the Internet (Trioda, Electroda, etc.) in terms of schematics applying EF22 and EBL21 tubes, I decided to make an amplifier based on the principle proposed by Studi (nick of Trioda member):

And according to him, I decided to build an amplifier, increasing - only slightly - the values of coupling and cathode capacitors, as output transformers were calculated assuming low frequency range of 30 Hz, and so in the end I wanted to obtain this range.

I proceeded to collect the necessary components.

Initially, purchased secondhand bakelite loctal tube sockets, did not raised my confidence because of the poor quality of the silver on the contacts (worn in areas where tube feet touched the connectors). The better idea is to buy some of the new old stock (NOS) sockets or some new, although you need to watch out for poor elasticity of contacts in some types - they lose contact after a few insertions of tubes.


Fig. 26.

Therefore, on the advice of fellows from TRIODA I looked for other sockets - ceramic, recycled Soviet military. Initially, I wanted to use aluminum clamp of the socket as fixing element:


Fig. 27.


Fig. 28.


Fig. 29.

However, thin-walled aluminum cup turned out to be too soft to hold the base firmly, so I had to abandon this idea.

When I completely removed the remains of the aluminum cup I had no other choice as to make the clamps that would push the sockets base from the bottom to the chassis by myself:


Fig. 30.

Now it was necessary to consider the concept of amplifier housing. For some time I thought about the idea to do the retro-style cabinet - as at the beginning of the tubes era, in the twenties - the housing of wood. So, I searched the Internet for pictures of that era equipment to pry the characteristic elements of the cabinets and the materials that were used:

http://historiaradia.neostrada.pl/Lata1925-1929.html

Then I had to think about the dimensions of the cabinet, so as to fit all the elements (the largest overall dimensions had output transformers and they determined the size of the housing). I found that the width of the cabinet could be tha same as the width of the power supply (27.5 cm). Left so just went to the carpenter and order the appropriate pieces of wood.


Fig. 31.

I planned to use at least the side walls made of real wood (not plywood), because their edges have to be visible from the front. In the twenties, plywood was probably still not known (at least all the images of machines from that period I have watched had cacings of pure wood), and it would be worthwhile to get the same effect. So I went to the carpenter armed with a paper drawing and dimensions of the walls of the cabinet. When I asked about the walls of the wood I heard that "maybe next week if all goes well a glued pine board will be accessible". It turns out that a carpenter today generally do not use real wood (except for some battens and beams), only particle board, plywood... and more and more plastic imitations of wood. What to do then?

Luckily, I got the idea to ask a carpenter for veneer. It turned out that on the windowsill there was a pile of oak veneer covered of thick dust - unless NOBODY long has looked into it. One piece of veneer costed 15 PLN - acceptable - in particular that could be useful for the next projects. So I decided to use veneered plywood.

Yet another unpleasant surprise was waiting for me - a walls made of 4mm plywood were wavy. Carpenter argued that "So thin plywood will always bend." The purchase has costed a total of more than 40 PLN (along with a bottle of glue).

After returning home, I sat down to arrange positions of tube sockets, connectors and controls on each of the walls. Before I began gluing the box, I had to cut out all the holes with coping saw. Fortunately, there were not a lot of them, and plywood was easy to handle, so the job was not too difficult.

Then I proceed to glue the segments of the housing. Fortunately, plywood side panels were smooth and properly cutl, and provided the basis for "stretching" wrinkled segments of the cabinet.

At first things looked good ...


Fig. 32.

...but then I had to use more force to straighten out "unruly" segments ...


Fig. 33.

.. and even my leather strap.


Fig. 34.

In this way the cabinet was finally completed:


Fig. 35.

Not everything came out perfectly, but it would be aligned at a later stage of the project.


Fig. 36.


Fig. 37.

I added a nice decorative strip at the bottom of the housing.


Fig. 38.


Fig. 39.

Then it was necessary to align housing ... and it was ready for veneering.

Veneering is a bit like wrapping a box with paper - I just had to be a little more careful when cutting veneer (it likes to split along the grain, not along traced straight lines), even after gluing the edges and generally to be careful not to contaminate it with glue or damage in another way. But the end result was quite nice.


Fig. 40.


Fig. 41.

I had prepared mahogany mordant and I did not hesitate to use it.


Fig. 42.


Fig. 43.

After the first layer of nitrocellulose clearcoat...


Fig. 44.

... and after the second.


Fig. 45.


Fig. 46.

I settled for two layers, in order to preserve the wood grain texture, so I obtained a more severe case, the original design, which I wanted. As you can see,  the 0.5 mm front plate of bakelite is mounted on the test.

Therefore, the case was more or less complete and I could proceed to the next phase of construction.

Here I have a little back in time, I forgot to mention two things that I have already done: "banana" sockets and bakelite knob.

Modern connectors did not fit my concept of retro housing. I decided to do something by myself that would be the most "in style". The recipe is quite simple indeed. One takes a piece of metal sheet and routes to the appropriate lines...


Fig. 47.

... and then has to cut plates according to the drawing ...


Fig. 48.

... then roll on the rod with a diameter of 4 mm (which might be a drill) ...


Fig. 49.


Fig. 50.

... and something like this is created.


Fig. 51.

Now solder the edges to close the tube.


Fig. 52.

Finally, we add a wire ring, which rests on the edge of bakelite, in which our tube will be inserted into. This is the finished tube, ready for assembly.


Fig. 53.

After drilling holes in the prepared bakelite plate, I inserted tubes, put rings of wire and soldered them which permanently fixed tubes into the holes. You can even add a drop of glue to secure the tube ends. The finished "jacks" are presented in Fig. 54.


Fig. 54.

I have done them with silver, but you can use ready-made brass tubes with an internal diameter of 4 mm and avoid soldering.

The knob was a bit easier to make.

I purchased an old bakelite knob for only 2 PLN. It served as a base.


Fig. 55.

After lightly sanding the edges (bakelite can be polished after cutting and sawing, so that we get a black, shiny surface) and glued together with some fabricated pieces I received the knob presented in Fig. 56 and Fig. 57.


Fig. 56.


Fig. 57.

Now, one more thing left to do before installing components. Lining the inside of the case with metal sheet, which shields the amplifier from the impact of the whole "foreign electromagnetic fields." I planned to use a sheet of iron (steel) zinc, but in the shop seller said that "in Warsaw you will not get any such distributions." What could I do - so I chose 0.8 mm aluminum sheet.

It had to be cut and bent according to the inner walls of the cabinet...


Fig. 58.

... and then pasted it into the housing.


Fig. 59.

I sealed the edges with self-adhesive copper tape.


Fig. 60.

Such prepared casing is almost ready for mounting.

Before installation I had to do one more thing. I was able to buy a nice solder connectors and decided to use them.


Fig. 61.

From the piece of 10 mm textolite I cut out the "pillars" and after fixing the soldering tips in varying amounts I received cool "solder bars."


Fig. 62.

One can adjust the number of necessary terminals and connectors to receive even quadruple connectors - depending on the needs.


Fig. 63.

I glued bars to the chassis using the two-component resin in places that are naturally (as a result of the scheme) united elements as the shortest paths from the terminal blocks in the tube sockets, etc. To make the work easier the connections can be noted on the diagram.

To get a good, low impedance ground path I used the fat (2.5 mm) copper wire, which I bent and soldered to solder terminals.


Fig. 64.

Then I placed the wires leading to the tube heaters (with a rather thick, twisted wire in insulation) trying to lead them away from the other tube pins to avoid possible effects of interference (hum).


Fig. 65.

Tube sockets were properly oriented so that the EF22 anode outputs are the opposite the EBL21 tube pins. This resulted in the shortest signal path from one tube to the other.

After pasting strips the result is presented in Fig. 66.


Fig. 66.

Now there was nothing left as to solder the rest of the elements. Larger electrolytic capacitors were glued with neutral silicon to immobilize them.


Fig. 67.

Then there was a start-up phase - I temporarily connected the output transformer, which was located above EBL21 tube. After switching voltages - silence. Did something not work? ... But after touching the banana jack I heard a hum ... it meant the amplifier had a zero humming noise. How nice ;-). But my happiness was still premature, because after connecting the signal source, I discovered that the sound was distorted. Has been an error in the diagram? In the meantime I connected the output transformer into the second channel and SURPRISE!!! - clean, undistorted sound.

So there's a defective element in the first channel. Initially I suspected that one of the electrolytic capacitors in the cathodes was broken - it were NOS capacitors produced by ELWA in the 80's.

However, measurements of voltages on EF22 pins showed that the second grid voltage was 0V. The culprit was the MIFLEX capacitor (0,1 uF/630 V).


Fig. 68.

After replacing the damaged capacitor, both channels played properly and without distortions.

After the first listening, and the general impression that the sound is very clean, I noticed, however, that there was too little bass.

So, I pulled out the function generator, oscilloscope, 4 Ohm resistor and started to measure the frequency response. While the upper frequency limit was satisfactory - 25 kHz, the bottom was too weak - only 65 Hz ... So I decided to do a little "odd trick" and increased the EF22 cathode capacitance from 47 uF to 220 uF. Well, it turned out to be a good option - lower frequency limit is now shifted to 30 Hz, which is just as it was assumed for the output transformer.

After this modification I mounted output transformers permanently inside the housing.


Fig. 69.

At the end I made 1 mm bakelite bottom wall and cables connecting amplifier to the power supply.

The amplifier has gained following appearance:


Fig. 70.


Fig. 71.


Fig. 72.


Fig. 73.

One more thing remained to do - descriptions of sockets, knobs, tubes, etc. ...

I decided to do all inscriptions, scales, emblems, etc. using a photographic technique (positiv) and etching techniques to consolidate and highlight the  inscriptions.

First, I started by reviewing images of the original receivers and equipment from that period, to be able to take a look and feel "as it was done" and on this basis design something by myself. Since the whole project was to be made using photographic technique, I had to develop a computer model.

Inkscape proved to be a helpful program here. After the initial training in its options and possibilities I could proceed with the design. The result was a postscript file, which when transferred to imagesetters, returned to me in the form of a film with exposed subtitles, emblems, etc. There were several options, not all were used in the final project.


Fig. 74.

Now, I could proceed to an appropriate action. First, I prepared sheet brass, which was to be the base for most badges.


Fig. 75.

I transfered the image from the film on the board  using UV irradiation.


Fig. 76.

After etching plates, I got something like this:


Fig. 77.

Then I used a coping saw to cut out all the pieces. This was quite a bit of precise work.


Fig. 78.


Fig. 79.

After the cut, finishing and dyeing treatment I received ready for assembly badges.


Fig. 80.

In addition, I did the nameplate with the parameters and name of the amplifier, which was to be placed on the front panel. It was not easy to find an appropriate name for the amplifier, because most typical names were already taken. In the end I got the happy idea that I wouldl use the name of Harold, which reminds me of actor Harold Lloyd, comedian occurring in silent films from the period of the 20's. This time I used aluminum and silver as starting materials for the implementation of the emblems.

After fixing all items in place the amplifier was ready:


Fig. 81.


Fig. 82.


Fig. 83.


Fig. 84.


Fig. 85.


Fig. 86.


Fig. 87.

Finally, a view of the amplifier on my "TV table".


Fig. 88.


Fig. 89.

Maybe there's a little space, but at this point there is always more or less the darkness, so you can see glowing tubes heaters and a blue glow on the inside of the EBL21 anodes.

After taking detailed measurements of the amplifier frequency response I received the following graph:

Summary of the construction of the amplifier

Amplifier construction - from acquisition of cores for output transformers, through all the described stages, lasted about 7 weeks. Of course, not non-stop, but in a spare time, but it happened a few days of more intense work, where I sat for a few hours a day. There were also a few days break.

Self-winding of output transformers, turned out to be crucial moment and allowed me to choose EBL21 tubes as output tubes. Output transformer manufacturers do not offer transformers dedicated to EBL21.

Construction of the amplifier is in most a mechanical work - the transformers, the chassis and finish. The electrical assembling of the elements is already a relatively small part of the whole work.

The use of sheet metal liner inside a wooden case proved to be a great idea - amplifier is not sensitive to interference fields. Mains hum level measured at 4 ohms of resistance is within 55 mVpp - regardless of the position of the volume potentiometer.

EF22 tubes in this system are not characterized by any excessive microphonic. Instead EF22 tubes, EF21 tubes can be used. You can then expect even better linearity and smaller distortion.

Power consumption of the amplifier is around 60 W (as planned).

When it comes to general conclusions for future designers of amplifiers, I would like to recommend not to be hurry and well prepare all the materials for the construction of the amplifier (wood, metal and electronic elements, etc.) before starting the assembling.

As for the sonic impressions - all the songs that I have and know, sound for me differently. Especially those, performed on a classical guitar instruments, wind instruments, etc. appear to be more expressive and rich in details. So now, there is nothing else to do then listen to the amplifier.

Written by Jacek "Jado" Domański (e-mail : )