2017, Chapter 10:
Failures and Flyers
I’ve been promising a chapter about “Schiit that didn’t make it.” AKA, designs we played with, but in the end decided not to put into production.
And with the release of Jil, and our tease of Sol, it seems like a great time to deliver.
Because Jil and Sol could have just as easily been shelved. They’re both diversions from what business-types would call our “core competencies.” Of course, seven years ago, we had no core competencies at all, and five years ago we were still just an amp company (no digital) and three years ago we hadn’t yet introduced any Multibit DACs. So maybe this idea of “core competencies” is a bit limiting—or at least it should be tested and expanded from time to time.
And that’s what we do. I’ve said, many times, that we play with a lot of designs, and not everything makes it to market. I’ve even covered one such design (which was actually 3 separate products, the Valkyrie series of portables.) You can read about that here:
https://www.head-fi.org/threads/sch...obable-start-up.701900/page-291#post-11137279
“But we want you to do portables!” someone says.
Yeah, and I hear you. I continue to play with stuff, but my heart really isn’t in it. Every time I think I find a unique angle, something stops me—cost, or size, or complexity, or any of a dozen other excuses. And when you find yourself finding excuses not to do something, you probably should take the hint.
Bottom line, if we ever do a portable product, it can’t be a me-too piece of gear. There are plenty of companies making great stuff in the portable world. If you need a DAP or an amp or a DAC/amp, there are plenty to choose from. What can we add to this market? We’d have to bring something unique and compelling in order for me to get excited about. The world doesn’t need yet another Schiit-branded product for the sake of having another Schiit-branded product. So, if you absolutely have to have portability, take a look at the great options already out there.
But this chapter isn’t about retelling the story of the three Valkyries. It’s about some more things we decided to let slide…and a couple of things we decided to take a flyer on.
First the Failures
Well, “failures” might be a bit strong. “Shelved products,” might be more accurate, or it might be a bit weak. What I’ll be talking about here are three projects that are currently dead, and should stay dead. I’ll go into the whys and wherefores of why they were left on the side of the road, and you can let me know why you think that was a great idea, or why I’m the biggest idiot on the planet.
Above: Gallarhorn conceptual design (yep, I know it says, "Gungnir." Ignore that.)
Fail the first:
Gjallarhorn.
Gjallarhorn was a high-power tube OTL headphone amp design. It would have used 6SN7 tubes for gain, and 6AS7 tubes for output. It would have had some really nifty features, like a choke-input high-voltage power supply. It would have also been as big as a Mjolnir 2, and dissipated nearly 100W of heat at idle, with four large and hot tubes sticking up out of the top of the product. We were shooting for 3.5W or more into 50 ohms (yes, from a tube OTL design.)
“Ah hell, that’s exactly what I want!” some of you are saying.
Yeah, that’s cool. And there are companies that make high-output tube OTL amps with those kind of tubes. And if we’d proceeded with the design, we would have been very near their price. Which put a big damper on my enthusiasm.
But that wasn’t the only thing that bothered me about the design. The transformers and chokes I’d specced would only just barely fit in a Mjolnir-sized chassis—there were a total of 6 pieces of iron in there. And they were at the limit of the cores. Which could mean that we’d have to move to a completely new and different chassis, and/or we could have ended up with too much heat in the case. Which would mean lots of new design work (and something that didn’t stack with any other product)…and the possibility that we’d end up having to use a chassis fan to keep the thing cool.
And it got worse. 6SN7 and 6AS7 tubes, even the Russian equivalents, were either (a) of questionable quality, (b) getting scarce, or (c) both. Which didn’t give me warm-and-fuzzy feelings about the future of the amp.
So, after doing the basic design work, including a schematic and a PCB layout, I put the Gjallarhorn on the shelf. After all, I had something more interesting to play with (at least to me): the early Jotunheim design. Which in itself was a multi-year odyssey.
But I did keep coming back to the design, once every couple of years when I went through the ritual of cleaning up my engineering files. And every time, I paused, looked at the completed PCB, and thought, “Hmm, maybe worth pursuing.” But each time, the thought was fainter and farther away. Because I’d remember that I’d never gotten the transformers and chokes, and that they might not work, and that was the key to this design. At the same time, our line got bigger and better, and the tubes I wanted to use got more scarce…and more expensive.
So, late last year, as I was making the biggest change of engineering design software I’d undertaken, I reviewed the files one last time…and decided that yep, this design is dead.
Engineering Epitaph:
- Name: Gjallarhorn
- Description: High Power Tube OTL Headphone Amplifier
- Born: Late 2012
- Died: Late 2016
- Achievements:
- Conceptual design complete
- Schematic complete
- PCB laid out, but not produced
- Custom transformers and chokes not spec’d or ordered
- Causes of death:
- Too impractical overall
- Uncertainty about the magnetics/heat
- Tubes getting costly
- Price too close to other options
- Headphone amp line getting too big
- Always something else better to do
Above: early Sol (amp, not turntable.)
Fail the second:
Sol (the 1st).
Before the name “Sol” was applied to our upcoming turntable, it was the designation for a small desktop speaker amplifier. Think a 20W x 2 amp that you could also bridge and run as 60W monoblocks. Sol (the 1st) would be Asgard-2 sized, with a low-bias, linearized Class AB design that would allow for chassis heatsinking like the Asgard. It would also have a bypassable volume control on it, so you could use a single volume control in the bridged configuration.
“Well, holy hell, we asked you about this one!” someone says. “And you said, ‘No, probably not, probably too expensive, blah blah blah. Now you say you’ve already looked into it?”
More than that, I’m afraid—I actually built a working prototype. In 2013.
“Aieee! Why didn’t you make it? I’d buy it!” someone yells.
Lots of reasons. They start, again, with the competition. When you’re up against an Emotiva desktop amp with 3X the rated power and more flexibility for $199, you really start getting second thoughts when your best shot gets you a retail price of $299. That’s 50% more. That’s a lot of money. And you can’t really ignore that.
But there’s more. A lot more. Starting with simple physics. Although I was using a transconductance-linearized, MOSFET-output gain stage so I could run ultra-low bias (less than a Magni—
way less) and still have good measured performance and cool running at idle, as soon as you hooked up speakers and cranked the power, things got very hot, very fast. Worse, things got so hot so fast that the heat was localized, so the case didn’t work well enough as a heatsink. So I’d be looking at a complete thermal redesign, and maybe including a fan to boot.
And…the transformers we needed for that kind of power output didn’t quite fit in the chassis. This meant we’d be stuck with something weird like a 9 x 6.5” form factor. Not quite the same as our other products, so not easily stackable with them. Clumsy, and not ideal.
And more…the topology I was using didn’t thrill me, as it had in the past. It was complicated, unstable, hard to compensate, and, after doing a bunch of adventurous new things on the headphone amp side, it seemed too old, too hidebound, too Lin. There were more exciting designs I could play with.
That would be enough to kill it, but we had additional doubts. In a world where powered monitors were starting to proliferate, did anyone really need a low-powered desktop amp? Would it confuse everyone who already knew us, people who were just looking for headphone amps? We didn’t know, and it didn’t seem to be worth the chance.
Finally, it just didn’t seem time for a speaker amp. Or at least not this speaker amp. If we wanted to get into speaker amps, I figured we needed to make something that could run most speakers. And, after 4 years, we finally got there.
Engineering Epitaph:
- Name: Sol (the 1st)
- Description: 20+20W bridgeable desktop speaker power amp
- Born: Mid 2013
- Died: Late 2013
- Achievements:
- Conceptual design complete
- Schematic complete
- PCB laid out
- PCB prototype complete
- Custom transformers produced
- Working prototype complete
- Causes of death:
- Too expensive
- Way too much heat!
- Complex topology
- Line confusion
- Not time for a speaker amp
Above: an early Xap prototype.
Fail the third:
Xap.
Xap was an all-tube OTL electrostatic headphone amplifier. Yes, you read that right. Xap used 6SL7 tubes for gain and 6SN7 tubes for output, all running on +/-400V regulated rail voltages. The intention was to create a simple, relatively inexpensive 4-tube e-stat amp that could be sold for $999 or so. Unfortunately, the reality proved to be a little different—the reasons for which I’ll get to.
“An affordable e-stat amp from Schiit! I’d buy that!” say a whole lot of people.
“You’re crazy for doing any kind of amp like that, imagine the liability of people poking around inside of it,” say a whole lot of other people.
Sigh. Yeah.
The reality is, you’re both right. Good affordable e-stat amps are super-scarce. Especially when you start adding the word “tube” in there. And, yeah, anything that has 1000V of potential on the PCB before regulation—and the ability to swing 1200-1300V on the output—is plenty scary when you start looking at it from the POV of your typical risk-averse corporation.
But we decided to play with it anyway. After all, you won’t find any new roads if you never explore at all.
So, we dusted off an old, old design of Mike Moffat’s—one he did back in the 1970s for the Stax headphones he used at the time. It was a simple overall topology—6SL7s on the input, 6SN7s on the output, a total of 4 tubes for a stereo amp. It’s not a unique design—you’ll also see stuff like this also from John Broskie. Our implementation was fairly conventional, too: balanced and SE inputs leading to a 4-gang pot like we use on Mjolnir 2, then into the no-overall-feedback tube gain stage, with output via the typical Stax 5-pin connector. We did use a separate transformer for the 580V bias supply, and separate transformers for the HV rails, and heater/housekeeping functions. The only real innovation was an interesting differential servo, designed by Dave, that allowed us to keep the outputs centered around 0V without any potentiometer twiddling or other hand-wringing stuff that might drift with time.
So, would this design have challenged the ultra-high-end tube stat amps out there? Nope. That wasn’t the intent. The intent was something good-sounding for a 3-figure price tag.
We ended up building three complete prototype versions. The first was to get an idea of what we were getting ourselves into. The second was to correct lots of errors and tweak the servo design.
And that’s when things kinda went off the rails. Because, although we’d gotten the servos working well, and we had reasonable gain and output capability, the open-loop 6SN7 outputs were too high impedance to run the capacitance of a typical Stax headphone.
Meaning, we were 3dB down at 20kHz.
Yeah, I know, other designs have been down 3dB at 20k and still been accepted as “high end.” But we wanted something better. So the third prototype actually ended up being a Frankenstein combination of two of the previous prototype boards, with a set of 6SN7s added as an output buffer. That worked just fine, and was -3dB at 80kHz or so.
But, ouch. Now, instead of a simple 4-tube amp that fit neatly in a Mjolnir-sized chassis, we were looking at a much more complex, larger 6-tube amp that needed an additional transformer—or a bigger transformer—to run the additional current. An under-$1k price was out the window. Suddenly, the amp looked a whole lot less interesting.
And that’s when uncertainty started creeping in. Did we really want to make this amp? It ran voltages much, much higher than anything we’d ever done. Did we want the liability? Was this the right market for us?
So, yep. Xap, like the others, went on the shelf.
Engineering Epitaph:
- Name: Xap
- Description: all-tube OTL headphone amplifier
- Born: Early 2016
- Died: Late 2016
- Achievements:
- Conceptual design complete
- Schematic complete
- PCB laid out
- PCB prototype complete
- Custom transformers produced
- Working prototypes completed
- Causes of death:
- Too big
- Too expensive
- Too scary
So were these failures a complete waste of time? Of course not. I can’t even say we would have accelerated any other products if we’d avoided these cul-de-sacs. All of them built our knowledge base—Gjallarhorn on the design of high-powered tube amps, Sol on the right way to do a desktop amp (if we approached it now, the design would be very, very different), Xap on the niches we want to play in, and not (and a cool new servo.)
“But hey, there ain’t no digital products in this list,” an astute reader says.
And yep, you’re 100% correct. Mike and Dave have their own cul-de-sacs, but I’ll leave it to Mike to take you through them. When the time is right.
Or not…
On to the Flyers
And, in amongst this experimentation, sometimes we come up with a hit. Or at least something we think will be a hit. Or at least something that seems like it’s worth taking a flyer on.
Products like Jil…and Sol.
So why did Jil, an
analog to digital converter, and Sol, a
friggin turntable, make it to production, when Valkyrie, Gjallarhorn, Sol (the 1st) and Xap get the dirt nap?
Well, sorry to disappoint you, but there wasn’t a whole lot of market research, focus groups, or grand strategic alignment going on. A lot of what we do is by gut feel. And Jil and Sol felt right. I’ll get into more of the reasons for that below, but consider:
- Jil fit into a standard-sized chassis. It used many, many common parts, including our standard USB interface. Making a thousand Jils to see how they go isn’t a big deal. If they go, they go. If they don’t, we get another cake, as with Loki.
- Sol was a huge departure, but we kept coming back to it. And it was really, really different—a chance to shake things up. And after the initial investment for two expensive casting tools, it really wasn’t all that costly or complex. If Sol falls flat, yep, we’re out the investment in the tooling, but we’ve absorbed bigger jolts without a hitch.
Jil is the simpler of the two, so let’s talk about the details of that product first. Jil is Schiit’s first Analog to Digital Converter, or ADC. The name is a play on “Jack the Ripper,” which is what Mike first referred to the project as. (“Ripper,” as in “ripping,” or digitizing, a vinyl record. Or a tape source. Or, well, whatever you want.)
Mike’s last ADC was a reference-quality piece for Mobile Fidelity Sound Labs, co-designed with Nelson Pass. For those of you who blanch at the cost of the Yggy’s DACs, consider that this 16-bit design featured something like $1,500 of monolithic analog-to-digital converters—per channel. And that’s just the cost of the conversion chips. It also used a digital filter based on the same basic closed-form math as Yggdrasil, running on multiple Motorola DSPs. The resulting system, known as GAIN, took up many, many full-size rack spaces.
How good was it? For fun, look up the cost of CDs mastered with MFSL’s GAIN system (not GAIN 2). Yeah. There you go.
(Also for Schiit Multibit DAC owners, consider that the GAIN system digital filter is complementary to our current digital filters—a hint of a what a closed-loop megacomboburrito system could be.)
Jil is the complete opposite of the GAIN system. Jil is small, affordable, and based on a delta-sigma AKM ADC. Mike developed it largely on a whim, then discovered it actually sounded quite good. After some Moffat-style tweaking, Jil provides a very high-performance way for customers to digitize virtually any line-level signal. It’s also a neat complement to the Mani phono preamp, for those who want to do vinyl archiving.
“So where’s all this going? Are we looking at a step-up ‘Jak’ multibit ADC?” someone asks.
Well…we don’t know. We’ll see how Jil goes. This is a flyer. Jil isn’t really your standard professional mic interface with preamps and ADC functionality. Nor is it an ultra-expensive, reference-level piece. But neither are Yggy and Gungnir your typical pro DACs…but we know plenty of pros who use them.
So we’ll see.
Cross your fingers.
Above: an early Sol prototype.
Sol…Sol is something Mike has been playing with for over 3 years. He’s had wood-plinth versions up and running at his house almost that entire time. When he first told me about it, I was skeptical. A Schiit turntable? We’d never done anything remotely like that.
Still, watching the crude thing spin on his dining room table, and listening to the result…it sounded damn good.
But I wasn’t a turntable expert, so a dug a little deeper into what was available on the market. And I found that most affordable turntables followed a similar template: simple MDF plinth, MDF or acrylic platter, motor bolted via isolators to the plinth, short arm of conventional or unipivot design.
That’s a monoculture, I thought.
Monocultures are vulnerable for disruption.
Now, this isn’t the place for a whole chapter on Sol. That’ll come later. But suffice to say, Mike took the design through several iterations, from the first T-shaped wood plinth and Delrin platter to the die-cast aluminum platter and skeletal plinth Sol has today. The tonearm has remained the same 12” carbon fiber unipivot design throughout. The same inverted bearing has been retained all along. The motor has always lived in a separate pod.
And now, it’s playing at the Schiitr, as we get things ready to move Sol into production. Die cast tooling is purchased, details are being worked out, and we’ll update the model at the Schiitr as we make changes to it.
But, there’s still a long road to Sol. Lots of details to work out. Pricing hasn’t been set. I’m currently finishing a design for a bolt-on bracket that has space for the tonearm plug-in, RCA outs, power I/O, and power switch (on top this time!)
I’ll keep you updated as things progress. I hope you’ll enjoy the journey.
