By December 5, Sasa reported that "I revised the power supply concept for the Kaivalya. I tried the amp with a tube rectifier but in the powerful sections of certain songs, I somehow got momentary compression as you'd expect from guitar amplifiers. That's because this is a push/pull amp. It wants a lower impedance supply than a single-ended circuit would. This is a really interesting game." Far from frustrated, Sasa sounded like a pioneer who explored novel country with a keen sense of adventure. One never knows what lies behind the next bend.


"This compression is a very interesting observation. I built guitar amps before and know this effect very well. It is good for guitars but not our cause. With solid-state rectification you can lower the impedance. Incidentally, this circuit working without negative feedback causes some other challenges but I'm using a 10H choke right now—usually designers use 2H or 3H for push/pull circuits especially for an EL84 amp—and some other tricks. I will try a small paper-in-oil capacitor of about 4-10mF in parallel with a big electrolytic in the power supply and create a very stable low impedance for the supply voltages. It is of course very important for the bass range but the vocal and treble bands perhaps need it even more. You cannot believe how directly this paper-in-oil capacitor influences the sound. I must find the right value and match it perfectly to the IT and OPT."


A few days later, these first photos arrived. "In the beginning it looks ugly, no? The final amp won't use toroidal mains transformers of course but double C-cores based on a similar principle as our Classic 500 conditioner. I have already listened but only to the output stage. I used my Experience Line One preamplifier with an added output transformer as phase splitter. The first step is to make a good and stable output stage with a superior output transformer. When I know the results from the output stage, I can very easily tailor the sound from the interstage transformer. That must remain the only and final variable 'X'. The IT will be the big mystery."

A few more days passed when Sasa started detailing cosmetics in his head: "What do you think if instead of nickel-plated transformer covers, we paint them white? I believe that we will end up with three separate covers, perhaps four. The wood frame will be white gloss lacquer, the transformer covers matte white and the rest will be aluminum - screw-less top plate, a small circular dress plate around the power knob as well as the knob itself."


This I didn't entirely agree with so we quickly settled on a pure two-tone scheme - the same white gloss lacquer for the chassis and transformer cans, nickel-plated aluminum for the top plate, circular dress plate on the front and the master knob. This would give us just two textures and colors, not two different whites and aluminum. The connectors would mount directly to the lacquered wooden chassis to eliminate a metal back plate. Sasa then forwarded this sketch of the general layout which captured my ideas perfectly. At this point, we were up to four transformer/choke casings on deck while the drawing indicated that both tube rectifiers had been eliminated. Apparently Sasa had also settled on the ECC81 driver over the ECC99.


"Yes, I eliminated the tube rectifiers from the design. For this push/pull amp, tube rectification ended up sounding far too vintage. Below deck will be three separate power supplies, one for the EL84s, one for the ECC81 and one for the G2 of the EL84s. There will be a total of 10 electrolytic capacitors, three chokes—one 20H and two 3H—and something more just for the power supply.


On first sonic impressions: "I'm building some WLM Minutas for Hannes right now and compared one to the Kaivalya prototype. I could be biased of course even though both are my creations but I think this new amp will easily be twice as good. The Minueta of course was very aggressively designed to hit the price Hannes wanted to pay. I would not be surprised if the Kaivalya ended up becoming the best amps I've ever built. The EL84 really is my favorite output tube.
"The trick I told you about for the interstage transformer works amazingly well. The effect it has on the sound is fantastic. And it's also a problem. Winding it this way makes it not symmetrical across the whole layer but it must be. So this is not the end of the research. This trick is a miracle and you cannot believe how it influences the sound. I will play more games with this, some more windings, some less. I actually never saw this done before anywhere else. I could be wrong but it'll be a very rare detail in any case."


The same day, this arrived: "Regarding the color scheme, I am in complete agreement but I am afraid the whole amp will become too sparkly with this nickel top plate. The second problem is that I can't engrave in a nickel finish. I think the amp will look beautiful but not glitzy with an aluminum top plate. Unlike WLM's brushed finish, we'd go for very fine sand-blasting, then clear anodize. This will tone down the otherwise overly glossy effect. I apologize for my changes of ideas but that's R&D." Which is precisely why this reportage doesn't edit the process. Readers inspired to approach Sasa—or any other designer—for their own custom project should be cognizant of possible zigzags involved. I could of course have put my foot down on certain finish items but Sasa's mounting excitement had already clarified that this would become a production amplifier. Thus its cosmetics had to work broadly, not just appeal to one customer. They had to be something its maker could fully stand behind and which lent itself to future models in an expanding line.


A good week later, R&D had taken another zag. "By the way, some changes are coming - a new IT and ECC82 drive. It's a similar configuration to the first but the IT has grown by about 30% for extended bandwidth and because the ECC81 had too high an amplification factor, the ECC82 turned out to be more appropriate. Input sensitivity will be 2Vrms, ideal for monos. In this assembly, the prototype already outperforms all of my OEM amplifiers and is similar to my own Reference 300B monos but I will see what more can be done. As you know, I went through eight iterations of the IT with the 300B monos. I hope I can hit this one after three."


Another week passed: "We're into our third IT now and very close to perfection but it must still ace a 10kHz square wave.
This IT is so different from the one in our 300B monos that the prior experience was no help at all." Next time I checked in, Sasa was on fifth base. "Yesterday I finished the 5th version, now I have capriced to go to the end of all possibilities. The 4th one was excellent but I found that I could still improve it a bit with a slightly different geometry of the core that's changed over the first four. This was a nice opportunity to see what happens with different ways of winding an IT when you already have a superior OT. I conclude that the IT must be at minimum 30% better than the OT to predictably produce a superior sounding amplifier. This really maxes out the sound of the EL84. By now I like it more than any 300B/2A3 amplifier I've heard - and I listened to it CD direct without preamp. I'm surprised and you will be too I think."

By January 24, "I am just now drawing up a new type of IT which is actually a new approach to the problem - and hopefully even better if I can trust my mathematical predictions. The problem is in the ultrasonic band. Everything is fine up to 70kHz but beyond it I have 90° phase inversion between the left and right sides of the push/pull outputs of the transformer and 180° rotation at 100kHz. It's not a real concern since this is well out of band but now is a good chance to make this IT perfect since the market offers nothing that's good enough for my tastes."


Three days later Sasa sent this photo of a prototype 'body' on which to wind the latest iteration of the interstage transformer. "Leakage induction should be three times lower than with the earlier design." Then, "I am on the right track with the new winding system. I now have perfectly symmetrical halves of secondaries but a new problem, albeit smaller than the symmetry of secondaries. It's endless work but will be very good in the end." The final email on this topic: "Bingo! Finally we have got it right. In these photos you can see how it looks. The final one has 9 separate horizontal windings for 38 separate windings total (the one in these photos still only had six). The sound is unbelievable for an EL84 construction. I am happy."

If Sasa was happy, I was. It appeared that the project was approaching the first serious auditions.
On a snowy day which had his Trafomatic Audio van under a white crown, Sasa then copied me on an e-mail which the article had stirred up at this stage. "I was very much intrigued by the various reviews written by Srajan of 6moons on Trafomatic products. In Singapore, there is no distributor for Trafomatic except for the Experience Head One which I had opportunity to audition and instantly fell in love with it. I'm deeply convinced that a great transformer properly built and voiced makes or breaks any tube design. Like Srajan, I'm excited about the Kaivalya. Your project spurred me to ask about the viability of another custom amp based on the almost forgotten PX25 tube. I've only heard the Art Audio PX-25 but given its almost decade-old circuit design and being built with circuit boards, I'd like to know whether you plan on a future PX-25 SET to better what Art Audio has done 10 years ago? Otherwise, if I were to request a custom design, would it be possible to keep it essential SET but achieve between 10-15wpc instead of the rated 6wpc? My intention is to drive a pair of DeVore Nines rated at 91dB sensitivity with a Wyred4Sound STP-SE preamp."
 

Jack Rock meanwhile had his bread-boarded 250TL hooked up to Sasa's transformers. "Note the off-angle on the 10K:10 bifilar as best orientation for hum. Wait until I make the 450TL monsters and nicer photos. These sound amazing already."


By March, Sasa was ready to test deep engraving and decide on the size and spacing of the lettering. Because the router bit diameter would limit just how narrow the lines could be while preventing them from coming to a perfect point (all the ends would be round), Sasa strategically picked a font that lent itself perfectly to this application. The very first router pass into a scrap piece still had a spelling error Sasa caught after he'd already programmed and run the machine.

By now the elusive IT was locked in and a Russian Marlboro cigarette pack serves as a useful size reference.

The partially obscured handwritten chart hints at the complexity of windings hidden beneath the tape wraps. For all the complications involved in arriving at the final design, this part turned out very compact.