2019, Chapter 11: The Six and a Half Year Itch Okay. This is a big one. How big? This big: we’re abandoning Class A in Asgard. That’s right. The new Asgard 3 is a Continuity™ amp, not a Class A amp. And yeah, I know, this is a big change. Hell, the Asgard was our first product. It launched the company, over nine years ago. It’s one major update to the Asgard 2 was six and a half years ago. It’s been a good, solid, stable product (hell, I have an original Asgard that’s been on continuously pretty much those entire nine years.) It’s also been the only Class A amp in our line. Notice I don’t say “real” Class A, because, to us, Class A has a specific definition, and it doesn’t include any kind of fakery. To us, Class A means an amp that cannot go out of Class A, no way, no how. Asgard and Asgard 2 used simple single MOSFET output devices, biased with a current source. They could not go out of Class A, no way, nuh uh, not gonna happen. Similarly, our DAC and preamps with discrete output are biased in such a way that they never go out of Class A. Aside: this shouldn’t be surprising—“small signal” electronics like preamps and DACs have no real reason to be run in Class AB. They dissipate a relatively small amount of power, even running Class A. Most small signal stuff, unless it’s op-amp based, is going to run Class A. And even op-amps can be forced into Class A output, though the benefits of that are decreasing with the newer and more sophisticated audio op-amp designs out there. But I ramble. Let me get back to the story. Anyway, as I was saying, Asgard has always been the only Class A amp in our line. So changing it now is a huge break with tradition. Seems a bit crazy, right? Well, maybe not so much. Let’s look at the reasons Asgard 3 is Continuity, rather than Class A. Hitting the Ceiling Here’s the deal: Asgard and Asgard 2 were simply at their limits. You may have noticed that Asgard and Asgard 2 shared the same power output rating (1W into 32 ohms). As in, there was no changein power output from generation to generation. That might seem a little odd in a world where Magni went from 1W to 1.2W to 2W, from Gen 1 to 2 or 3, with no real difference in size and heat output. And, it might seem especially odd, given that Magni is a $99 amp, and Asgard 2 is $249. Based on numbers alone, Asgard and Asgard 2 seem a little...weak. So why didn’t we up the power output when we went to Asgard 2? It’s simple: we were hitting the ceiling on thermal limits. A Class A amp has to run full-out all the time. There’s no way around it. So we have to dissipate a ton of heat. No negotiation on that. Which is why Asgard and Asgard 2 both used our original U-shaped chassis as a 100% heat sink. It’s also why they both ran 45 degrees C or so in a normal 20 degree C room. To dissipate more power, we’d need to add heatsinks (costly) or a fan (costly, irritating, and a source of failure). And so, Asgard and Asgard 2 soldiered on with the same power rating. They also shared another limitation: coupling capacitors. The original Asgard had two—one at the output and one interstage. The Asgard 2 lost the output coupling cap, but still had the interstage. Aside: the interstage capacitor is less of a problem than the large output coupling cap, as it is relatively small and can be a high-quality polystyrene or polypropylene capacitor. But it’s still a capacitor, and I’d much rather have something DC-coupled from input to output. For a time, I played around with level-shifters as a way to get rid of the coupling capacitor. I built a few breadboards, but we never got to a PC-board level prototype. I simply couldn’t muster enough enthusiasm, when I knew that we’d simply be introducing another 1W-rated, hot-running amp. In a world where Magni was moving into a second generation with more power, this didn’t seem to be the right thing to do. Especially when it would have to live in the same U-shaped, relatively expensive chassis that was soon to be supplanted by Jotunheim. So, I sat back and waited. The Long Path to Asgard 3 But “waiting” for me might have a different definition than for most people. “Waiting” still means “wondering and playing with other ideas.” Because with the Jotunheim chassis coming, I started looking for a way to fit a Class A amplifier into it. And so, the first PC board prototype of an Asgard 3 was one that would fit into a Jotunheim chassis. It also had provision for Jotunheim’s optional modules. But beyond that, it was radically different. The first Asgard 3 was my first design to use a fully complementary current-feedback topology, with 100% bipolar transistors. It was also the first design to use a hunk of heatsink (the intention was to bolt this to the top chassis, and provide thermal dissipation similar to Asgard and Asgard 2. This was a huge change from the single-ended, JFET-input, MOSFET-output, chassis-is-the-heatsink Asgard 2. And, like many huge changes, it was a huge failure. That early prototype was where we learned about the special thermal considerations of current-feedback designs. As in, the first ones went up in smoke. It’s also where we learned that high-density heatsinks, while providing lots of dissipation area, were not so efficient in either radiation or convection, and were no so great at getting rid of heat. Still, a board revision later, we had an Asgard 3 prototype running with the outputs bolted to a hunk of heatsink. I biased it up to run about the same standing current as the older Asgard twins, and we were off to the races with a new Class A amp. Aside: however, this Class A amp wasn’t like the old Asgard Class A—it was complementary and push-pull, which meant it could come out of Class A. So, even if it was biased for 1W of Class A, it could put out about 3.5W of power in Class AB—a huge increase from Magni. But, “off to the races” was maybe an overstatement. I took the first prototype Asgard 3 to a small, local meet. People heard it. And, pretty universally, they said, “Meh.” I don’t blame them. It was a fairly “meh” amp. It was clean enough, but it didn’t sound like anything special. Especially if you liked the warm, rich sound of the previous Asgards. Ah well. That was just my cue for more waiting. In this case, though, “waiting” translated to, “Let’s get this damn heatsink out of the amp.” Because, by this time, I was working on Lyr 3. And Lyr 3 had similar heat dissipation needs to Asgard 3. So, if I solved one, I solved both. Lyr 3 ended up using an aluminum chassis and a big thermal pad (known as a Gap Pad) to transfer heat from bottom-of-the-PCB-mounted output transistors. Aside: and now, it’s actually moving to a steel bottom chassis, because we actually can get equivalent performance with a larger thermal pad and steel. Anyway, I figured I could use the same trick for Asgard 3, so I put together a new prototype with the output stage on the bottom. Maybe the different transistors would improve the sound, I figured. They were much more linear (spec-wise) than the ones I’d used in the first prototype. Ahhhh...nope. Still a kinda “meh” amp. Sigh. More waiting. During this waiting, I developed the final Lyr 3 prototype, the one that used the constant-transconductance Continuity output stage for the first time. That output stage had a huge effect on the Lyr—it took it to a different level. For an amp biased into 1/2W of Class A, that result was a shock. 500mW of Class A should be enough for pretty much any headphone to run in Class A all of the time. But Continuity sounded better. Huh. Why not try it in Asgard 3? But, in Asgard 3, we had additional challenges. If we put the output transistors on the back of the board like Lyr 3, that would make for costly assembly. Since I wanted to actually lowerthe price of Asgard 3, that could be a make-or-break decision. However, we also had some new paired output devices that were rated for higher temperatures. Could we use those? I did one more prototype with the matched pairs placed on the top of the board, with large copper lands under their power pads to dissipate the heat, and crossed my fingers. In short: nope. They still ran wayyyy too hot. It looked like we were stuck with putting the output transistors on the back of the board. Well, until Naomi, one of our technicians, asked me a fateful question. “Could we just use the thermal pad on the back of the board?” “You mean, with the output transistors on the front?” I asked. “Yeah, and pull heat through the board.” I frowned. Maybe. Maybe not. But it was easy enough to try. I stacked up a couple of thermal pads so they hit the back of the board, turned it back on, and took a look at it with the Flir. “Holy crap,” I told her. “It works!” Putting the thermal pad on the back brought the output temperature down more than 20 degrees C. Which put the operating temperature of the devices at only about half their continuous rating. Totally safe. Aside: I’m somewhat downplaying the impact of the matched pair outputs on the design of this amp, because, between matched pair outputs and matched pair inputs, a new compensation scheme, and a new power supply, the Asgard 3 is quite a different amp than anything we’ve made before. I learned a lot of tricks in its design, stuff we’ll be moving forward into other products. We let the new prototype Asgard 3 run for a few days on the burn-in rack, checked the temperatures again, then had a listen. Sad trombone. Sorry, still kinda meh. I mean, it sounded good. It sounded extremely clean, and it had a nice, expansive soundstage. But it was still a little...soulless. Naomi agreed. I took it home and tried it for a bit longer against the Lyr 3. Nah. Still not good enough. However, Continuity allows much more opportunity for tweaking the way an amp works. Continuity can compensate for transconductance droop—or it can overcompensate. Overcompensating gives an output transfer function that is more like a square-law device. Or, in other words, more like a tube. So, I swapped the Continuity resistors, to adjust the output to slightly overcompensate for the transconductance droop. It only took a few minutes. I was busy with something, so I gave it to Naomi to listen. She came back immediately. “What the heck did you do to this?” “Why? Is it broken?” Because I hadn’t checked it, I figured it still worked. “No. It just sounds...glorious.” “Huh?” I took it back to my desk and tried it with some familiar headphones. And, she was right. It was a completely different amp. Warmer, fuller, more harmonically complex...the timbre was much improved. It sounded...well, like a Class A amp. Finally. Still, I worried. It was pretty warm. Maybe a bit too much of a good thing? In the end, we decided not. Better this amp be warm and happy than cold and clinical. End of story, right? No, my education was just starting. Because Asgard 3 was really, really fat. As in, really too expensive to make and hit the $199 price target. We’d already thrown a bunch of tricks at it to bring cost down—going to stamped chassis most prominently—but it was still too expensive. I eyed the power supply. The new power supply was probably more complex than we needed. We probably didn’t need the pre-regulators. We could probably just get away with cap multipliers to smooth out the rails. So, I jumpered across the pre-regulators and ran all the standard amp tests. No real increase in noise, nothing really changed. So probably good to go, right? Wrong. The sound, the “glorious” sound, as Naomi put it, collapsed. The soundstage shrunk. Even the tonal richness ratcheted down a couple of notches. It was pleasant enough....but not great. Aside: this is 100%, totally, completely astounding to me. It makes no sense. It makes as little sense as Unison USB sounding better than, well, any other USB implementation. It makes as little sense as amps that measure below the limit of human perception sounding different. This is the kind of thing that drives people crazy. And yes, I know, there are plenty of people who will say that Naomi and I are just fooling ourselves, that there’s no real difference, it’s just expectation bias. And I have my days when I believe them. “Well, maybe if I use simple regulators,” I said, trying to salvage some cost savings. I pasted in the parts for the simple regulators. Wah-waaahh. Better. But still not good enough. “So what do we do?” Alex asked, when I told him about the cost. “We say **** it,” I told him. “We use the two-stage regulation, and we do it right.” “And sell it for—“ he trailed off. “And sell it for $199.” Alex winced. “Yeah, I know,” I told him. “But it’s the right thing to do.” And, you know what? It’ll be fine. If Asgard 3 is as reliable as Asgard 2, all will be well. If we find a bit more savings in larger production quantities, we’re good. It’ll work itself out. The Road to Production The original title of this chapter was “the six year itch.” You’ll note that it’s now “six and a half.” That’s a hint that we didn’t quite meet our timeline. No huge surprise there, but it may be interesting to go through all the little things that got in our way. Obstacle #1: Longer testing phase. Asgard 3 was treated to a short production run before we went into mass production, so we could really see if they met the standard for a product that had to be dead-reliable. We threw them in junk Lyr 3 chassis and shipped them around a bunch of places. Getting the feedback from that and tweaking the design took a bit longer than usual. Obstacle #2: New stamping machine. Asgard 3 is produced on a new 70-ton press at our sheetmetal partner here in Valencia, CA. It’s a bigger, thicker chassis than they’ve made for us, and they decided to upgrade the tooling for a brand new machine they were getting in. The teething problems of getting a new machine installed and debugged put the chassis behind a bit. Obstacle #3: New production house. We’re using a new PCB assembler in Nevada for some of our products, and Asgard 3 is one of the ones that went there. They’re doing fine, but again, new partnerships take some time to spool up, so it took longer than we expected to get boards rolling. Obstacle #4: Selling the last of the Asgard 2s. Timing of new products is always interesting and usually wrong. Ironically, once we thought we finished the Asgard 2s, we found a bunch more finished boards with no chassis. So we’ll probably do a short run of chassis if you still want a real, honest-to-goodness Class A amp. They’ll all be black, to simplify production. And mentioning black, you’ll notice that Asgard 3 is available in black from the get-go. It’s also using new black screws on the black chassis, so It looks better, too. Aside: yep, it gets a screw-on top to save costs. It also gets no front LED, only an internal LED that shines up through the top perf. We’ve tried to take cost out wherever it doesn’t matter. Even then, don’t be surprised if we can’t hold $199 for long. Aside to the aside: don’t be surprised if we play around with colors a bit as we try to simplify the color choices. What do you think about a graphite metallic top and a black bottom as the only color scheme? So what’s the bottom line? Asgard 3 is a big step forward from Asgard 2: 3.5W Continuity with 500mW Class A bias vs 1W Class A—3.5x the power output Ability to take one optional input card, vs “just an amp.” Ability to be a 3.5W, $299 DAC/amp with 4490 card, vs no card Gain switch on front, vs on back Input switch on front, vs no input 48VA transformer, vs 36VA $199 vs $249 price (amp only) Plus, Asgard 3 can be: A $299, 3.5W DAC/amp with the addition of the AK4490 card A $399, 3.5W True Multibit DAC/amp with the addition of the Multibit card Upgraded when new cards become available And, regarding the price…yes, I know, I mentioned this in an aside, and in development, and probably a couple other places, but Asgard 3 is pretty thin. I don’t know that we’ll be able to hold the $199 price. It may soon be $229 or even higher. This is literally the first product I’ve issued a price warning on. I’ll do everything I can to keep it at $199, but we’ll see how it goes. I always wanted Asgard to be $199…literally from the first day we started. We’ve never been able to do it. Now, we’re going to do it, as long as we can. I really, really hope you enjoy it!