Schiit Happened: The Story of the World's Most Improbable Start-Up
- Thread starter Jason Stoddard
- Start date
-
- Tags
- schiit audio schiit-audio
Jason Stoddard
Sponsor: Schiit Audio
2016, Chapter 11:
The Road to Jotunheim
So, if you've been following the book to date, you know that Ragnarok had a long and very painful gestation—that is was, in fact, one of the first products I wanted to make back in 2010. Introducing it in 2014 was a huge milestone for us.
But what you don't know is that there's one product that has arguably gone through a even more arduous journey on its way to production...complete with false starts, two complete product redefinitions, seven prototypes, and more "first article" metal than we've ever run for a single product.
And that's Jotunheim.
Since you're reading this chapter shortly after the intro of Jotunheim, you already know the gist of it: a do-all, balanced in/balanced out, single-ended in, single-ended out amp that also happens to be configurable with the use of an add-in card—and that this configurability is what makes it unique in its field: the only upgradable, updatable amp/DAC or amp/phono pre or just-an-amp on the planet.
Yes, we're seriously proud of Jotunheim, as we think it makes the entire concept of non-upgradable DAC/amps completely obsolete. Why get a combined product—especially one that costs much more than Jotunheim—knowing it will be obsolete in a couple of years, as digital audio continues its inevitable changes?
But I'm editorializing. I think Jotunheim is a big deal. But I'm biased. And the fact is, Jotunheim started for much less sweeping reason than "redefining a market segment."
It started, quite simply, because I was staring at boxes upon boxes of very expensive balanced volume pots.
An Inauspicious Beginning
Now, this was a long time ago—late 2012, in fact. We were still in the noisy, dusty, dirty, held-together-literally-by-tape-and-baling-wire Schiithole on 6th Street in Newhall. Moving into our current palatial (ha) Schiitbox wasn't even a dream. (Well, except maybe to Alex, because we were running out of space at the Schiithole.)
So, about those pots? Yeah. We'd started making Mjolnirs with a stock of 400 balanced pots, which I'd managed to find ahead of the production run. As the first Mjolnirs flew off the shelves, I placed an order for another 1000 pots. And then, since they were very long lead-time items, I placed a second order.
These two orders were probably the most expensive single-part orders that we'd done to date. They were also, in engineering terms, "NC/NR," or, "noncancellable, nonreturnable."
And, of course, I placed those two pot orders right before Mjolnir sales started to nosedive.
Yeah. There you go. It seems that not having a single-ended output on the original Mjolnir wasn't one of my better ideas.
So I was standing there, looking at these boxes and boxes of very expensive pots, wondering how we would ever sell them all, and I had a thought: Hey, if we made a less expensive balanced amp, we could use those pots in there, too.
Yes, I know. Not rocket science. And not the best reason to develop a new product. But it was either that, or sit on a pile of very expensive parts, like, forever.
So, in my spare time, I started working on a new amp design. I knew kinda-sorta what I wanted to do with it. Of course, what I wanted to do at the time isn't what Jotunheim turned out to be, but we'll get to that. If there had been a design brief, it would have read something like this:
Note on #4. This is the key to led us to an entirely new topology...one that Dave didn't even believe would work at all, and one that even I had trouble with accepting, even after extensive testing. This is our new "Pivot Point" differential current-feedback topology, which (as far as I know) is unique. More on this later.
Note on #5. This is what you call either "ego talking" or "straight out delusional." Of course, I didn't know this at the time. I just figured it would be a nifty cool thing to re-introduce the Hawksford-style linearization that we'd used at Sumo in a headphone amp. Hawksford linearization is a really neat trick that helps manage the transition between N-channel and P-channel MOSFETs, or NPN and PNP transistors, which, of course, are never truly complementary, and always have some transconductance linearity problems near zero-crossing.
Aside: if your eyes are crossing due to the heavy dose of engineeringese, look at it this way: the battle for better amps usually comes down to one thing--how the output stage behaves. This is why some designers choose to make hot, heavy, inconvenient, inefficient Class A amplifiers...because they sidestep the zero-crossing problem by never having the output turn off at all. Of course, this dramatically limits the kind of power output you can have without self-immolation (assuming a true Class A design, but that's a screed I've already written.) Using Hawksford linearization is a way to make the output stage behave.
And there you go. #4 and #5 are, together, why you should give yourself the time and freedom to explore new designs...and to kill your babies when they are really, really ugly.
At the Pivot Point
Staring at those boxes of pots led to me playing with breadboards, late in 2012. The first order of business was to nail down a new topology for the amp. Like I said, no circlotrons need apply. Given that I wanted this to be a balanced amp, I was rather limited on the choices I had from the list of "usual suspects:"
*Now, let's be clear: there's lots and lots and lots of "design reuse" going on in audio. More than 95% of the power amps out there are essentially Lin-based design. Magni and Magni 2 are. There's nothing wrong with that. The Lin topology is perfectly fine for a speaker amp or a headphone amp. It's been tested, busted, iterated, refined, explored, augmented, and examined every which way. But...I like to see what else is out there, when I can.
**Yes, I know, more engineeringese. Sorry about that. But let me try to explain. Sumo amps were what they called "fully complementary." The input stages used both NPN and PNP BJT transistors. Now, the problem with BJTs is that they have a tendency to rectify RF noise. Say, like from a cellphone sitting next to the amp. That's why Schiit tends to use JFET inputs. But finding a good complementary pair of JFETs these days is kinda like looking for a lost Rembrandt in your granny's attic--good luck with that. So our input stages are not fully complementary. Would this topology work on a noncomplementary input?
To make a short story short, I modified the never-used Sumo-era topology, pasting in a noncomplementary input and DC servos, and hooked it all up on a breadboard. I figured that it would probably:
But, surprise surprise: it worked. DC offset was low, distortion was very low, and the transistors didn't fly off the board.
I blinked a few times, re-ran the tests, verified that yes, it was working, and yes, it was all hooked up right.
And when it all checked out, it was time to go to layout.
But not before Dave's protests.
"It Won't Work."
I was feeling pretty full of myself with this new topology, especially after some internet searching didn't find anything like it. Hell, this might be unique, I thought. It might even be new enough to be patentable.
Aside: now, before the Internet lawyers get their panties in a twist, no, we don't patent stuff. Even if it was patentable. Because, before we pursued a patent, we'd have to do a much more thorough search. And even if we went through and obtained a patent, it would buy us nothing more than the license to sue people who were using it. It would also reveal the entirety of the topology to the world, including parts of the world where US patent protection has no meaning. So relax.
So how full of myself was I feeling? Enough to propose it as the output topology for the then-still-in-early-design-phase Yggdrasil.
"But it won't work," Dave told me, when he first saw the topology.
I laughed. "That's what I told myself when I first came up with it."
"But...you're not telling me it works?" Dave asked.
"Yes. It does."
Dave frowned. "I don't see how it can. What does the front end do when the output's at 20 volts and you have a gain of 1? It's seeing all of that voltage!"
"Yes, and it rides up on it. It makes the front end operate much more like a constant-voltage stage."
Dave shook his head and sighed. "But...it can't work."
"Test it," I told him. "It measures better than anything we make."
"I don't see how," Dave said.
"Test it."
Dave muttered something and gave up. I'm not sure if he ever tested it, but we did use it in early Yggdrasil prototypes. The reason we didn't go with this topology for Yggy is simple—it didn't need it. It turned out that all it needed was a simple two-JFET buffer.
I was also confident enough to start laying out the first Jotunheim board. This board resembles the final product only vaguely. Like the final product, it's on a board that fits in our 6x9" chassis. Like the final product, it uses a C-core "flat pack" transformer, a first in a Schiit product. Like the final product, it has a Pivot Point topology gain stage.
But it doesn't have any provision for an optional DAC or phono card.
Aaannnd it has one thing the production product lacks: the Hawksford linearized gain stage.
So, why no optional DAC/phono card? Simple. We hadn't thought of it yet.
And why doesn't the production product have Hawksford linearization? Well, two reasons:
Yeah. Seriously on #2.
"Wait, what the hell are you talking about?" Someone is asking right now. "It welds the headphone plug to the jack? How the hell does it do that?"
Well, to understand why, you need to know two bits of engineering—one on the electrical side, one on the mechanical side. On the electrical side, the Hawksford linearization does exactly what it says: it linearizes the input and output. Aggressively. As in, any deviation between input and output engages a servo loop to correct it. As in, you can literally tune the circuit down to zero output impedance.
Aside: this is what allowed Sumo amps to deliver 200A (that is, two hundred AMPS) peak current for short periods of time. Like I said, aggressively.
On the mechanical side, you need to know how the common headphone plug works. It actually shorts the outputs as you insert it. This is why some amps say "turn it down before plugging in or unplugging headphones."
Engineers are already wincing at the "zero output impedance + shorted output" statement, but let me spell it out for the nontechnical audience in very simple terms:
With the first Jotunheim board, if you plugged in a headphone to the single-ended jack with the volume turned up, and there was a significant difference between left and right channels, it was possible to arc-weld the plug to the jack.
This, in turn:
Now, even in early 2013, when I was working with this first Jotunheim board, I knew that putting a "hey, turn the volume down" disclaimer on the Jotunheim would do exactly two things for its reliability and for customer satisfaction: jack and schiit.
And so there you go. Design revision #1: ditch the linearized output stage.
But that wasn't the only revision. Oh, no. Not by a loooooonnng shot.
Design Adventures, Cul-De-Sacs, and More Ego Talking
The second prototype of Jotunheim still wouldn’t be very recognizable to someone looking at the production version today. It had no space for an optional module, for one thing. It also still had a lot of parts on it that reflected my incomplete understanding of what the Pivot Point topology was. In short, I was still building this new topology like a conventional amp, in some ways.
But it was the first prototype that we actually spent a lot of time with, and for that reason, it was important. This was where the tough stuff got worked out: compensation, stability, stress testing, and initial measurements.
Stability was interesting in this case, because of the inherently “fast” nature of current-feedback amps, coupled with the overall complexity of the stage. Add in the fact that many current-feedback topologies don’t like to have a pole in the feedback network (if you’re an engineer, think about it—it’ll become obvious), and I had a number of adventures in major oscillation before everything got quieted down. Not quite as bad as Dave’s speaker power amp that took out Channel 28 on UHF (back when analog TV was a thing, yes, we’re old, deal with it, but also look up what frequency band UHF is and go a little pale).
But, once the amp was stabilized, we had our first working, playing Jotunheim. This was probably around summer of 2013. We started running it by some of our trusted listeners, and they were enthusiastic.
But they also asked questions like, “Yeah, but if you bring this out, what will it do to Mjolnir?” and “Hmm, it’d be a tough call on this or Lyr.”
And that got us wondering: hey, you know, we have a lot of amps in the line, so where does this one fit?
Call it a crisis of faith. But for good reason. We had Magni, Asgard, Valhalla, Lyr, and Mjolnir. Ragnarok was coming. I kept thinking about another amp, a more powerful tube amp, as well. That was a buttload of headphone amps. Especially if you threw Jotunheim in there as well.
Why is this important? Because most people don’t like choice. Give them too many choices, and they’ll choose “none of the above.” The amp line had already departed from the tried-and-true three-choice “good, better, best” scenario, so more complexity wasn’t going to help it.
And, at the same time, I knew that if we didn’t move forward, someone else would.
Aside: this is probably the core of any successful business that operates in the real world—how do you know when to “not fix schiit that ain’t broken,” and when to “kill your babies?” I don’t think we have any 100% answer for this.
In any case, those uneasy questions didn’t start me moving into a third prototype—this one intended to be the final one, the one that we could sell. Maybe by the end of the year. Yes, I know, stop laughing.
The third proto was pretty much a done deal, too—we’d pulled out the extraneous parts, added the compensation components, and it ran just fine. Except, of course, for the humming transformer and backwards relays.
I used that excuse to put it on the shelf and get back to other, more important things…like Ragnarok.
When it came off the shelf again, though, Mike and I had the first real conversation about where Jotunheim would fit in the line.
“It’s a whole new ball game,” Mike told me. “A new topology. And it sounds great. We should treat it like Theta—start at the top, work our way down.”
I shook my head. “Ragnarok is a done deal, it’s not going to go Pivot Point now. Unless you want to see it in 2016 or something.”
Mike nodded. “You have a point. How about as a Mjolnir replacement? Or Uber-Mjolnir?”
“This thing makes any Mjolnir replacement pretty much a stupid idea,” I told Mike. “The only Mjolnir replacement that makes any sense is one with tubes.”
And then I sat back, stunned silent. Because that was the only thing that made any sense—a tube hybrid Mjolnir. Which would be pretty cool…
“So put it above Mjolnir, below Ragnarok,” Mike said.
I shook my head. “We already have too many amps.”
Mike groaned and shook his head. He knew I was right. And then he said, “Put a DAC in it.”
“What?” I asked, not sure if I’d heard him right.
“Make it an amp/DAC,” Mike said.
“But…amp/DACs suck,” I said. “They’re obsolescent as soon as digital technology changes, and they have digital noise running around in the same chassis…”
“So make it modular,” Mike said, and stopped cold. We both had a moment when we just looked at each other.
Because, holy schiit, that was it. If Jotunheim was modular, the obsolescence issue was moot. And if I ran the card vertically, with parts only on one side, with decoupling from the main power supply and ground…the noise might be a non-issue.
Or, even better, look at the whole thing like a power DAC. No different than a Bifrost discrete output stage. But with more power. Run the D/As with passive filtering, pass it through to the gain stage, and do the ultrasonic filtering throughout. And, if the USB powered the DAC, it wouldn’t be running at all unless plugged in. That gave had the potential to be a zero-compromise amp and zero-compromise DAC, all in one.
I explained that to Mike.
He nodded. “That’s a game-changer.”
And that’s how we got to the fourth prototype. That version would look familiar to you today. It introduced the idea of the vertical DAC card, with pin-header connector and relay switching. It added a three-position switch so you could select the output of the DAC, the balanced inputs, or the single-ended inputs.
The timeline? Middle of 2014.
“So why didn’t you introduce it then, ya buttheads?” someone is likely asking. “It’s middle of 2016 now. What, you sit on things for two years after they’re done?
Hey, bite me. I have two words for you: Ragnarok and Yggdrasil.
Yeah, the flagships put a dent in the timeline for Jotunheim, too. Both of them were a ton of work. Mike and Dave were full-bore on Yggdrasil, so they didn’t have time to lay out a DAC card for Jotunheim. I was full-bore on Ragnarok, so the whole thing went back on the shelf for a while.
Well, except for the time I took to do the layout for the DAC card.
Fun fact: the DAC card is what led to Fulla. In the process of doing a very compact layout, I wondered how small I could go…and if it could be made into a complete DAC and headphone amp. And that’s how we got Fulla. I just couldn’t say that at the time. Oh yeah, and the first Jotunheim DAC card was based on the AK4396, just like the Fulla.
Ironically, that first DAC card was never built. It used two AK4396 D/A converters, and we’d just started hearing about how those were end of life. So I held off sending that prototype to the board house, while we went completely nuts.
Nuts?
Yes. The 5th prototype was completely insane.
Why? Because I’d just gotten the Ragnarok working, and it was very cool. And I was happy and giddy and a bit full of myself. And because of this, I had a very stupid idea. That stupid idea was this:
Hey, wouldn’t it be cool if Jotunheim used the same microprocessor management system as Ragnarok?
Yeah. I am an idiot.
But there’s no force in nature greater than idiocy, so I convinced myself that this would be a great idea, because hey, we already had a DAC in there, and because hey, eliminating the DC servo would make it sound better, and because hey, we could claim direct lineage with Ragnarok, and because hey, I am a big dummy.
And yes, we actually designed and built this thing. This was in early 2015. It was exactly as nuts as you’d think. We had to go to smaller 0603 parts, we had to add a ton of power supplies and bypassing, we had to add a ton of stuff to interface the microprocessor with the gain stage and the relays, including a bunch of flyspeck diode arrays.
But we did it, and it worked.
Now, it never worked right. The microprocessor put the current draw on the low-voltage supplies over the edge, the switching noise got everywhere, and making measurements and managing DC offset never got good enough for preamp and headphone use.
Yeah, there’s the cul-de-sac. And the ego.
Aside: Know when to walk away from a failed design. You don’t get points for persistence.
Which meant the sixth prototype was similar to the 5th, except for a few details. We were still playing with a couple of different gain-switching designs, so that prototype got both of them, with a switch on the board so we could compare how they sounded. The intent was to listen to both, then do a final production-ready board with only the winner.
The 6th prototype was also where we first got to plug in a DAC. With the re-layout done for the AK4490, and a different power supply arrangement (we decided to take the power supply from the main rails, which can provide amps of current), we finally had a complete, production-ready product.
By this time, it was well into 2015. Too far in to think of this as a Christmas product. And hell, we already had plenty of plans for 2015. So Jotunheim got back-burnered yet again. We figured it would be a 2016 product, and that was that.
And that’s when I had a very interesting idea. It seems simple in retrospect.
Because you’ll notice, up until now, I always referred to Jotunheim as an amp/DAC. Of course, it could be used as “just an amp” too.
But…it had a modular card slot in it. And that slot could be filled with things other than a DAC.
Which then made it an entirely different kind of product. Now, it was a configurable desktop amp, or modular desktop control center, or protean desktop widget…hell, we didn’t really have the words to describe it.
Because, in addition to a DAC, we could do a phono input. Or maybe a wireless card. Or maybe just another input.
And that was completely new.
Now I was excited.
So, as I did the small revisions for the seventh (and final) prototype board—really just a board to confirm that production was 100%—I also started designing an alternate module…a phono board.
“Phono?” you may be asking. “Why phono?”
To which I say, Why not? There are a ton of people out there with turntables. There are also a ton of people who already have our multibit DACs, and won’t want to give them up. And there are a ton of people who will want turntable input, DAC input, and preamp outputs…as well as headphone outputs.
Aside: the design of the phono board is a variant of Mani, minus Mani’s extreme flexibility. It has a single gain and single load, appropriate for most MM cartridges. MC carts will need a step-up transformer or pre-pre gain stage. Ah well, you can’t have everything. However, in the process of simplification, we have gained some significant advantages as well. The Jotunheim phono input loses a gain stage and coupling caps. Although the RIAA is passive, it uses only two gain stages, and a DC servo, unlike Mani, which uses three gain stages and is capacitor-coupled. We can argue the sonic benefits of either approach…
And, for both the 7th prototype, and for the phono board, there was little drama. The 7th started up and ran, as expected, since it was just a minor change from the 6th. The phono board worked just fine, and sounded very good.
Electronically, everything seemed set to go for mid-2016…
A New Chassis and Last-Minute Complications
Sharp-eyed readers will notice that the Jotunheim chassis is a pretty significant departure from what we’ve done in the past. What’s with the L-bend (rather than the U-bend? What’s with the lack of fasteners on top? And what’s with the press-in feet? (HOLYCRAPWHUT?)
In short, yes, the chassis is a story in itself. But, as this is an epic chapter already, I’ll keep it brief.
The chassis started with the most noble and pure of motives, kinda like the original design of the Jotunheim itself.
Or, in other words, the chassis was intended to reduce cost.
Or, put bluntly: to be cheaper.
Yes. From the start, Jotunheim was intended to be an inexpensive product. Unlike many high-end audio companies, we don’t see any problem with this. In fact, we think the chassis should be as inexpensive as possible, so we can deliver higher value on the inside. After all, it’s the electronics that make the sound, not the box.
Aside: and yes, I understand, some are gonna argue with me on this one, but I don’t think anyone can argue with wanting to put the money in, say, a $150 transformer, rather than a $150 front panel.
Of course, there are limitations. I wouldn’t want to do an ugly box, or a “typical” box. You know typical: a front panel on a steel chassis, or front and back caps on a standard extrusion.
But, bottom line: an L-bend aluminum piece is less expensive to do than the U-bend we’ve been doing. And it still looks really good—it carries the front panel visual seamlessly to the top of the product. So this evolution shouldn’t be too shocking.
The invisible fasteners…that was a lark. I wanted to see if we could do it. We actually did prototypes that had both screw-on tops and fastener-free tops. The fastener-free tops won…but we had no idea that they would work at first.
You see, the key to invisible fasteners is a blind PEM…a fastener that fits a mating keyhole slot on the chassis. After our chassis supplier proved they could insert the blind PEMs without marking the chassis, we had these custom-made for us, from a nonstandard alloy (so they could be anodized with the chassis), and for a nonstandard chassis thickness.
Of course, no story is complete without a last-minute complication, and Jotunheim is no exception. When we got the first boards in from the assembly house, they were shockingly hard to assemble—as in, almost impossible to get into these new chassis. A quick measurement confirmed that the volume pot had been moved over by 0.050”. Why? I don’t know. I don’t remember doing it. But what was done was done. Luckily, we had enough time to scrap the boards and re-order correct ones…and the timeline was preserved.
And the feet? That was intentional. Mike doesn’t like stick-on feet (not that I blame him, it made sense when you could choose to run an Asgard vertically or horizontally, but it doesn’t make much sense for the line now.) So I expect you’ll be seeing a lot more press-in feet as time goes on.
And…that’s the story of Jotunheim.
Consider this our third generation. Or maybe even a little beyond.
Now, you get to let us know what you think.
The Road to Jotunheim
So, if you've been following the book to date, you know that Ragnarok had a long and very painful gestation—that is was, in fact, one of the first products I wanted to make back in 2010. Introducing it in 2014 was a huge milestone for us.
But what you don't know is that there's one product that has arguably gone through a even more arduous journey on its way to production...complete with false starts, two complete product redefinitions, seven prototypes, and more "first article" metal than we've ever run for a single product.
And that's Jotunheim.
Since you're reading this chapter shortly after the intro of Jotunheim, you already know the gist of it: a do-all, balanced in/balanced out, single-ended in, single-ended out amp that also happens to be configurable with the use of an add-in card—and that this configurability is what makes it unique in its field: the only upgradable, updatable amp/DAC or amp/phono pre or just-an-amp on the planet.
Yes, we're seriously proud of Jotunheim, as we think it makes the entire concept of non-upgradable DAC/amps completely obsolete. Why get a combined product—especially one that costs much more than Jotunheim—knowing it will be obsolete in a couple of years, as digital audio continues its inevitable changes?
But I'm editorializing. I think Jotunheim is a big deal. But I'm biased. And the fact is, Jotunheim started for much less sweeping reason than "redefining a market segment."
It started, quite simply, because I was staring at boxes upon boxes of very expensive balanced volume pots.
An Inauspicious Beginning
Now, this was a long time ago—late 2012, in fact. We were still in the noisy, dusty, dirty, held-together-literally-by-tape-and-baling-wire Schiithole on 6th Street in Newhall. Moving into our current palatial (ha) Schiitbox wasn't even a dream. (Well, except maybe to Alex, because we were running out of space at the Schiithole.)
So, about those pots? Yeah. We'd started making Mjolnirs with a stock of 400 balanced pots, which I'd managed to find ahead of the production run. As the first Mjolnirs flew off the shelves, I placed an order for another 1000 pots. And then, since they were very long lead-time items, I placed a second order.
These two orders were probably the most expensive single-part orders that we'd done to date. They were also, in engineering terms, "NC/NR," or, "noncancellable, nonreturnable."
And, of course, I placed those two pot orders right before Mjolnir sales started to nosedive.
Yeah. There you go. It seems that not having a single-ended output on the original Mjolnir wasn't one of my better ideas.
So I was standing there, looking at these boxes and boxes of very expensive pots, wondering how we would ever sell them all, and I had a thought: Hey, if we made a less expensive balanced amp, we could use those pots in there, too.
Yes, I know. Not rocket science. And not the best reason to develop a new product. But it was either that, or sit on a pile of very expensive parts, like, forever.
So, in my spare time, I started working on a new amp design. I knew kinda-sorta what I wanted to do with it. Of course, what I wanted to do at the time isn't what Jotunheim turned out to be, but we'll get to that. If there had been a design brief, it would have read something like this:
- A cheaper balanced amp!
- Oh yeah, and it has to have single-ended outputs too, because it's already clear that wasn't the hottest idea for Mjolnir
- If it's gonna be cheaper, it should fit our smaller chassis size, the same as Asgard and Valhalla and Lyr, because I don't want a million different sized chassis
- Not a circlotron, because those require expensive, complex transformers and don't have an easy way to get single-ended output
- And while we're at it, let's go for a real, linearized, zero-output-impedance style amplifier, to really drive the heck out of those popular, power-hungry planers (at the time)
Note on #4. This is the key to led us to an entirely new topology...one that Dave didn't even believe would work at all, and one that even I had trouble with accepting, even after extensive testing. This is our new "Pivot Point" differential current-feedback topology, which (as far as I know) is unique. More on this later.
Note on #5. This is what you call either "ego talking" or "straight out delusional." Of course, I didn't know this at the time. I just figured it would be a nifty cool thing to re-introduce the Hawksford-style linearization that we'd used at Sumo in a headphone amp. Hawksford linearization is a really neat trick that helps manage the transition between N-channel and P-channel MOSFETs, or NPN and PNP transistors, which, of course, are never truly complementary, and always have some transconductance linearity problems near zero-crossing.
Aside: if your eyes are crossing due to the heavy dose of engineeringese, look at it this way: the battle for better amps usually comes down to one thing--how the output stage behaves. This is why some designers choose to make hot, heavy, inconvenient, inefficient Class A amplifiers...because they sidestep the zero-crossing problem by never having the output turn off at all. Of course, this dramatically limits the kind of power output you can have without self-immolation (assuming a true Class A design, but that's a screed I've already written.) Using Hawksford linearization is a way to make the output stage behave.
And there you go. #4 and #5 are, together, why you should give yourself the time and freedom to explore new designs...and to kill your babies when they are really, really ugly.
At the Pivot Point
Staring at those boxes of pots led to me playing with breadboards, late in 2012. The first order of business was to nail down a new topology for the amp. Like I said, no circlotrons need apply. Given that I wanted this to be a balanced amp, I was rather limited on the choices I had from the list of "usual suspects:"
- Supersymmetry. Nelson Pass' patent had just expired for supersymmetry. I could use that idea. It was a neat way to create an inherently differential topology, and it had the potential for using a single output phase as the single-ended output. So why didn't I use it? Well, while I'm less "not invented here" than Mike Moffat, I still like to think I can contribute to the art. A direct rip-off of Nelson's ideas didn't sit right.*
- I could use totally separate gain stages for each phase, as some balanced amps do. Of course, this would mean that a single-ended input would not produce balanced output. Not ideal.
- I could use a balanced-bridged topology, like Sumo used in the Andromeda II. But these topologies are kludgy and typically don't sound very good. Andromeda II was an exception. But I didn't want to be shooting for an exception.
- Finally, I could start looking for a true differential topology that sidestepped the problems with the rest—one that had high-impedance inputs, plus good places to hook the feedback, plus low distortion and high performance. And to do this, I went back to look at something I was never really able to deploy for Sumo. In the late days of Sumo, I was playing around with a differential topology that used current feedback. It had all the hallmarks of what I needed, but my notes from the era were vague and inconclusive. I didn't know if I'd ever built it. And, even if I had, I didn't know if it would work in a topology that wasn't fully complementary, like we used to do at Sumo.**
*Now, let's be clear: there's lots and lots and lots of "design reuse" going on in audio. More than 95% of the power amps out there are essentially Lin-based design. Magni and Magni 2 are. There's nothing wrong with that. The Lin topology is perfectly fine for a speaker amp or a headphone amp. It's been tested, busted, iterated, refined, explored, augmented, and examined every which way. But...I like to see what else is out there, when I can.
**Yes, I know, more engineeringese. Sorry about that. But let me try to explain. Sumo amps were what they called "fully complementary." The input stages used both NPN and PNP BJT transistors. Now, the problem with BJTs is that they have a tendency to rectify RF noise. Say, like from a cellphone sitting next to the amp. That's why Schiit tends to use JFET inputs. But finding a good complementary pair of JFETs these days is kinda like looking for a lost Rembrandt in your granny's attic--good luck with that. So our input stages are not fully complementary. Would this topology work on a noncomplementary input?
To make a short story short, I modified the never-used Sumo-era topology, pasting in a noncomplementary input and DC servos, and hooked it all up on a breadboard. I figured that it would probably:
- Have huge DC offset
- Have huge problems with distortion due to noncomplementary nonlinear inputs
- Blow the input transistors right off the board at high output
But, surprise surprise: it worked. DC offset was low, distortion was very low, and the transistors didn't fly off the board.
I blinked a few times, re-ran the tests, verified that yes, it was working, and yes, it was all hooked up right.
And when it all checked out, it was time to go to layout.
But not before Dave's protests.
"It Won't Work."
I was feeling pretty full of myself with this new topology, especially after some internet searching didn't find anything like it. Hell, this might be unique, I thought. It might even be new enough to be patentable.
Aside: now, before the Internet lawyers get their panties in a twist, no, we don't patent stuff. Even if it was patentable. Because, before we pursued a patent, we'd have to do a much more thorough search. And even if we went through and obtained a patent, it would buy us nothing more than the license to sue people who were using it. It would also reveal the entirety of the topology to the world, including parts of the world where US patent protection has no meaning. So relax.
So how full of myself was I feeling? Enough to propose it as the output topology for the then-still-in-early-design-phase Yggdrasil.
"But it won't work," Dave told me, when he first saw the topology.
I laughed. "That's what I told myself when I first came up with it."
"But...you're not telling me it works?" Dave asked.
"Yes. It does."
Dave frowned. "I don't see how it can. What does the front end do when the output's at 20 volts and you have a gain of 1? It's seeing all of that voltage!"
"Yes, and it rides up on it. It makes the front end operate much more like a constant-voltage stage."
Dave shook his head and sighed. "But...it can't work."
"Test it," I told him. "It measures better than anything we make."
"I don't see how," Dave said.
"Test it."
Dave muttered something and gave up. I'm not sure if he ever tested it, but we did use it in early Yggdrasil prototypes. The reason we didn't go with this topology for Yggy is simple—it didn't need it. It turned out that all it needed was a simple two-JFET buffer.
I was also confident enough to start laying out the first Jotunheim board. This board resembles the final product only vaguely. Like the final product, it's on a board that fits in our 6x9" chassis. Like the final product, it uses a C-core "flat pack" transformer, a first in a Schiit product. Like the final product, it has a Pivot Point topology gain stage.
But it doesn't have any provision for an optional DAC or phono card.
Aaannnd it has one thing the production product lacks: the Hawksford linearized gain stage.
So, why no optional DAC/phono card? Simple. We hadn't thought of it yet.
And why doesn't the production product have Hawksford linearization? Well, two reasons:
- It uses a ton of parts. It made the gain stage so complex, it barely fit on the board. The layout was pretty compromised, especially since the early boards were 2-layer. (Production boards are 4-layer).
- It can weld single-ended headphone plugs to the Neutrik jacks.
Yeah. Seriously on #2.
"Wait, what the hell are you talking about?" Someone is asking right now. "It welds the headphone plug to the jack? How the hell does it do that?"
Well, to understand why, you need to know two bits of engineering—one on the electrical side, one on the mechanical side. On the electrical side, the Hawksford linearization does exactly what it says: it linearizes the input and output. Aggressively. As in, any deviation between input and output engages a servo loop to correct it. As in, you can literally tune the circuit down to zero output impedance.
Aside: this is what allowed Sumo amps to deliver 200A (that is, two hundred AMPS) peak current for short periods of time. Like I said, aggressively.
On the mechanical side, you need to know how the common headphone plug works. It actually shorts the outputs as you insert it. This is why some amps say "turn it down before plugging in or unplugging headphones."
Engineers are already wincing at the "zero output impedance + shorted output" statement, but let me spell it out for the nontechnical audience in very simple terms:
With the first Jotunheim board, if you plugged in a headphone to the single-ended jack with the volume turned up, and there was a significant difference between left and right channels, it was possible to arc-weld the plug to the jack.
This, in turn:
- Destroyed the jack.
- Usually made it so you couldn't pull the headphone plug out of the jack--it was welded in that well.
- Typically blew up the output stage.
Now, even in early 2013, when I was working with this first Jotunheim board, I knew that putting a "hey, turn the volume down" disclaimer on the Jotunheim would do exactly two things for its reliability and for customer satisfaction: jack and schiit.
And so there you go. Design revision #1: ditch the linearized output stage.
But that wasn't the only revision. Oh, no. Not by a loooooonnng shot.
Design Adventures, Cul-De-Sacs, and More Ego Talking
The second prototype of Jotunheim still wouldn’t be very recognizable to someone looking at the production version today. It had no space for an optional module, for one thing. It also still had a lot of parts on it that reflected my incomplete understanding of what the Pivot Point topology was. In short, I was still building this new topology like a conventional amp, in some ways.
But it was the first prototype that we actually spent a lot of time with, and for that reason, it was important. This was where the tough stuff got worked out: compensation, stability, stress testing, and initial measurements.
Stability was interesting in this case, because of the inherently “fast” nature of current-feedback amps, coupled with the overall complexity of the stage. Add in the fact that many current-feedback topologies don’t like to have a pole in the feedback network (if you’re an engineer, think about it—it’ll become obvious), and I had a number of adventures in major oscillation before everything got quieted down. Not quite as bad as Dave’s speaker power amp that took out Channel 28 on UHF (back when analog TV was a thing, yes, we’re old, deal with it, but also look up what frequency band UHF is and go a little pale).
But, once the amp was stabilized, we had our first working, playing Jotunheim. This was probably around summer of 2013. We started running it by some of our trusted listeners, and they were enthusiastic.
But they also asked questions like, “Yeah, but if you bring this out, what will it do to Mjolnir?” and “Hmm, it’d be a tough call on this or Lyr.”
And that got us wondering: hey, you know, we have a lot of amps in the line, so where does this one fit?
Call it a crisis of faith. But for good reason. We had Magni, Asgard, Valhalla, Lyr, and Mjolnir. Ragnarok was coming. I kept thinking about another amp, a more powerful tube amp, as well. That was a buttload of headphone amps. Especially if you threw Jotunheim in there as well.
Why is this important? Because most people don’t like choice. Give them too many choices, and they’ll choose “none of the above.” The amp line had already departed from the tried-and-true three-choice “good, better, best” scenario, so more complexity wasn’t going to help it.
And, at the same time, I knew that if we didn’t move forward, someone else would.
Aside: this is probably the core of any successful business that operates in the real world—how do you know when to “not fix schiit that ain’t broken,” and when to “kill your babies?” I don’t think we have any 100% answer for this.
In any case, those uneasy questions didn’t start me moving into a third prototype—this one intended to be the final one, the one that we could sell. Maybe by the end of the year. Yes, I know, stop laughing.
The third proto was pretty much a done deal, too—we’d pulled out the extraneous parts, added the compensation components, and it ran just fine. Except, of course, for the humming transformer and backwards relays.
I used that excuse to put it on the shelf and get back to other, more important things…like Ragnarok.
When it came off the shelf again, though, Mike and I had the first real conversation about where Jotunheim would fit in the line.
“It’s a whole new ball game,” Mike told me. “A new topology. And it sounds great. We should treat it like Theta—start at the top, work our way down.”
I shook my head. “Ragnarok is a done deal, it’s not going to go Pivot Point now. Unless you want to see it in 2016 or something.”
Mike nodded. “You have a point. How about as a Mjolnir replacement? Or Uber-Mjolnir?”
“This thing makes any Mjolnir replacement pretty much a stupid idea,” I told Mike. “The only Mjolnir replacement that makes any sense is one with tubes.”
And then I sat back, stunned silent. Because that was the only thing that made any sense—a tube hybrid Mjolnir. Which would be pretty cool…
“So put it above Mjolnir, below Ragnarok,” Mike said.
I shook my head. “We already have too many amps.”
Mike groaned and shook his head. He knew I was right. And then he said, “Put a DAC in it.”
“What?” I asked, not sure if I’d heard him right.
“Make it an amp/DAC,” Mike said.
“But…amp/DACs suck,” I said. “They’re obsolescent as soon as digital technology changes, and they have digital noise running around in the same chassis…”
“So make it modular,” Mike said, and stopped cold. We both had a moment when we just looked at each other.
Because, holy schiit, that was it. If Jotunheim was modular, the obsolescence issue was moot. And if I ran the card vertically, with parts only on one side, with decoupling from the main power supply and ground…the noise might be a non-issue.
Or, even better, look at the whole thing like a power DAC. No different than a Bifrost discrete output stage. But with more power. Run the D/As with passive filtering, pass it through to the gain stage, and do the ultrasonic filtering throughout. And, if the USB powered the DAC, it wouldn’t be running at all unless plugged in. That gave had the potential to be a zero-compromise amp and zero-compromise DAC, all in one.
I explained that to Mike.
He nodded. “That’s a game-changer.”
And that’s how we got to the fourth prototype. That version would look familiar to you today. It introduced the idea of the vertical DAC card, with pin-header connector and relay switching. It added a three-position switch so you could select the output of the DAC, the balanced inputs, or the single-ended inputs.
The timeline? Middle of 2014.
“So why didn’t you introduce it then, ya buttheads?” someone is likely asking. “It’s middle of 2016 now. What, you sit on things for two years after they’re done?
Hey, bite me. I have two words for you: Ragnarok and Yggdrasil.
Yeah, the flagships put a dent in the timeline for Jotunheim, too. Both of them were a ton of work. Mike and Dave were full-bore on Yggdrasil, so they didn’t have time to lay out a DAC card for Jotunheim. I was full-bore on Ragnarok, so the whole thing went back on the shelf for a while.
Well, except for the time I took to do the layout for the DAC card.
Fun fact: the DAC card is what led to Fulla. In the process of doing a very compact layout, I wondered how small I could go…and if it could be made into a complete DAC and headphone amp. And that’s how we got Fulla. I just couldn’t say that at the time. Oh yeah, and the first Jotunheim DAC card was based on the AK4396, just like the Fulla.
Ironically, that first DAC card was never built. It used two AK4396 D/A converters, and we’d just started hearing about how those were end of life. So I held off sending that prototype to the board house, while we went completely nuts.
Nuts?
Yes. The 5th prototype was completely insane.
Why? Because I’d just gotten the Ragnarok working, and it was very cool. And I was happy and giddy and a bit full of myself. And because of this, I had a very stupid idea. That stupid idea was this:
Hey, wouldn’t it be cool if Jotunheim used the same microprocessor management system as Ragnarok?
Yeah. I am an idiot.
But there’s no force in nature greater than idiocy, so I convinced myself that this would be a great idea, because hey, we already had a DAC in there, and because hey, eliminating the DC servo would make it sound better, and because hey, we could claim direct lineage with Ragnarok, and because hey, I am a big dummy.
And yes, we actually designed and built this thing. This was in early 2015. It was exactly as nuts as you’d think. We had to go to smaller 0603 parts, we had to add a ton of power supplies and bypassing, we had to add a ton of stuff to interface the microprocessor with the gain stage and the relays, including a bunch of flyspeck diode arrays.
But we did it, and it worked.
Now, it never worked right. The microprocessor put the current draw on the low-voltage supplies over the edge, the switching noise got everywhere, and making measurements and managing DC offset never got good enough for preamp and headphone use.
Yeah, there’s the cul-de-sac. And the ego.
Aside: Know when to walk away from a failed design. You don’t get points for persistence.
Which meant the sixth prototype was similar to the 5th, except for a few details. We were still playing with a couple of different gain-switching designs, so that prototype got both of them, with a switch on the board so we could compare how they sounded. The intent was to listen to both, then do a final production-ready board with only the winner.
The 6th prototype was also where we first got to plug in a DAC. With the re-layout done for the AK4490, and a different power supply arrangement (we decided to take the power supply from the main rails, which can provide amps of current), we finally had a complete, production-ready product.
By this time, it was well into 2015. Too far in to think of this as a Christmas product. And hell, we already had plenty of plans for 2015. So Jotunheim got back-burnered yet again. We figured it would be a 2016 product, and that was that.
And that’s when I had a very interesting idea. It seems simple in retrospect.
Because you’ll notice, up until now, I always referred to Jotunheim as an amp/DAC. Of course, it could be used as “just an amp” too.
But…it had a modular card slot in it. And that slot could be filled with things other than a DAC.
Which then made it an entirely different kind of product. Now, it was a configurable desktop amp, or modular desktop control center, or protean desktop widget…hell, we didn’t really have the words to describe it.
Because, in addition to a DAC, we could do a phono input. Or maybe a wireless card. Or maybe just another input.
And that was completely new.
Now I was excited.
So, as I did the small revisions for the seventh (and final) prototype board—really just a board to confirm that production was 100%—I also started designing an alternate module…a phono board.
“Phono?” you may be asking. “Why phono?”
To which I say, Why not? There are a ton of people out there with turntables. There are also a ton of people who already have our multibit DACs, and won’t want to give them up. And there are a ton of people who will want turntable input, DAC input, and preamp outputs…as well as headphone outputs.
Aside: the design of the phono board is a variant of Mani, minus Mani’s extreme flexibility. It has a single gain and single load, appropriate for most MM cartridges. MC carts will need a step-up transformer or pre-pre gain stage. Ah well, you can’t have everything. However, in the process of simplification, we have gained some significant advantages as well. The Jotunheim phono input loses a gain stage and coupling caps. Although the RIAA is passive, it uses only two gain stages, and a DC servo, unlike Mani, which uses three gain stages and is capacitor-coupled. We can argue the sonic benefits of either approach…
And, for both the 7th prototype, and for the phono board, there was little drama. The 7th started up and ran, as expected, since it was just a minor change from the 6th. The phono board worked just fine, and sounded very good.
Electronically, everything seemed set to go for mid-2016…
A New Chassis and Last-Minute Complications
Sharp-eyed readers will notice that the Jotunheim chassis is a pretty significant departure from what we’ve done in the past. What’s with the L-bend (rather than the U-bend? What’s with the lack of fasteners on top? And what’s with the press-in feet? (HOLYCRAPWHUT?)
In short, yes, the chassis is a story in itself. But, as this is an epic chapter already, I’ll keep it brief.
The chassis started with the most noble and pure of motives, kinda like the original design of the Jotunheim itself.
Or, in other words, the chassis was intended to reduce cost.
Or, put bluntly: to be cheaper.
Yes. From the start, Jotunheim was intended to be an inexpensive product. Unlike many high-end audio companies, we don’t see any problem with this. In fact, we think the chassis should be as inexpensive as possible, so we can deliver higher value on the inside. After all, it’s the electronics that make the sound, not the box.
Aside: and yes, I understand, some are gonna argue with me on this one, but I don’t think anyone can argue with wanting to put the money in, say, a $150 transformer, rather than a $150 front panel.
Of course, there are limitations. I wouldn’t want to do an ugly box, or a “typical” box. You know typical: a front panel on a steel chassis, or front and back caps on a standard extrusion.
But, bottom line: an L-bend aluminum piece is less expensive to do than the U-bend we’ve been doing. And it still looks really good—it carries the front panel visual seamlessly to the top of the product. So this evolution shouldn’t be too shocking.
The invisible fasteners…that was a lark. I wanted to see if we could do it. We actually did prototypes that had both screw-on tops and fastener-free tops. The fastener-free tops won…but we had no idea that they would work at first.
You see, the key to invisible fasteners is a blind PEM…a fastener that fits a mating keyhole slot on the chassis. After our chassis supplier proved they could insert the blind PEMs without marking the chassis, we had these custom-made for us, from a nonstandard alloy (so they could be anodized with the chassis), and for a nonstandard chassis thickness.
Of course, no story is complete without a last-minute complication, and Jotunheim is no exception. When we got the first boards in from the assembly house, they were shockingly hard to assemble—as in, almost impossible to get into these new chassis. A quick measurement confirmed that the volume pot had been moved over by 0.050”. Why? I don’t know. I don’t remember doing it. But what was done was done. Luckily, we had enough time to scrap the boards and re-order correct ones…and the timeline was preserved.
And the feet? That was intentional. Mike doesn’t like stick-on feet (not that I blame him, it made sense when you could choose to run an Asgard vertically or horizontally, but it doesn’t make much sense for the line now.) So I expect you’ll be seeing a lot more press-in feet as time goes on.
And…that’s the story of Jotunheim.
Consider this our third generation. Or maybe even a little beyond.
Now, you get to let us know what you think.
Vigrith
500+ Head-Fier
- Joined
- Feb 21, 2016
- Posts
- 961
- Likes
- 300
AW YISS. Time to read, best of luck with the Show Jason, you'll do fantastic as always!
jfoxvol
500+ Head-Fier
My first CFA I built in undergrad was essentially and ugly looking oscillator. I feel your pain. We were designing For much higher bandwidths than here in audio, I think a few hundred MHz and built dead bug. Fun project.
artur9
Headphoneus Supremus
- Joined
- Apr 25, 2015
- Posts
- 5,474
- Likes
- 18,934
Interesting thought: Publish the spec on the slot so that someone can make an rPI that fits in there.
Vigrith
500+ Head-Fier
- Joined
- Feb 21, 2016
- Posts
- 961
- Likes
- 300
@Jason Stoddard I noticed the last question on your FAQ regarding the new product is regarding its name (Jotunheim) - if you don't mind my asking, whose idea was it to call the device Jotunheim? This may already have been referenced in this thread (or elsewhere) but are you and Mike both Norse Mythology aficionados?
I'm curious because as I said in the dedicated Jotunheim thread last night right after the announcement I find the name to be incredibly fitting - it's much more a domain than it is a weapon or a God, Norse Mythology was my favourite study subject for many years as a young kid (still is) and was one of the minors I took (two levels) in university so I'd be interested in knowing if there are any more intricate basis for your naming of the Jotunheim. For example, would I be correct in assuming you took consideration that in the Thrymskvitha, within the Codex Regius, Thrym (king of the Jötnar) stole Mjolnir whilst Thor was distracted during the feast the giant had prepared for him and Loki?
This then lead to Asgard and the Aesir being put into big danger due to the fact they weren't in possession of their be all end all means of defence against the Ice Giants. Interestingly enough, your Heimdall desktop system poses a direct "threat" (read: competition) to your Asgard (2) by incorporating part of the technology/functionalities of your Mjolnir 1/2.
There are so many cool ideas and so many interesting concepts one could come up with related to your range of devices now due to how wide its become (and will continue to expand in the future). It's one of the many reasons I started being mindful of your company years ago - my online handle for the past 20 or so years, Vigrith, is the field upon which the armies of Aesir and Surtr will battle as part of the Ragnarok.
I'm curious because as I said in the dedicated Jotunheim thread last night right after the announcement I find the name to be incredibly fitting - it's much more a domain than it is a weapon or a God, Norse Mythology was my favourite study subject for many years as a young kid (still is) and was one of the minors I took (two levels) in university so I'd be interested in knowing if there are any more intricate basis for your naming of the Jotunheim. For example, would I be correct in assuming you took consideration that in the Thrymskvitha, within the Codex Regius, Thrym (king of the Jötnar) stole Mjolnir whilst Thor was distracted during the feast the giant had prepared for him and Loki?
This then lead to Asgard and the Aesir being put into big danger due to the fact they weren't in possession of their be all end all means of defence against the Ice Giants. Interestingly enough, your Heimdall desktop system poses a direct "threat" (read: competition) to your Asgard (2) by incorporating part of the technology/functionalities of your Mjolnir 1/2.
There are so many cool ideas and so many interesting concepts one could come up with related to your range of devices now due to how wide its become (and will continue to expand in the future). It's one of the many reasons I started being mindful of your company years ago - my online handle for the past 20 or so years, Vigrith, is the field upon which the armies of Aesir and Surtr will battle as part of the Ragnarok.
Letmebefrank
500+ Head-Fier
- Joined
- Apr 18, 2015
- Posts
- 970
- Likes
- 643
Ordered a Jotunheim, no dac because I have a Momby, no phono because I don't have any vinyl. Looking forward to balanced on my hd650s.
Vigrith
500+ Head-Fier
- Joined
- Feb 21, 2016
- Posts
- 961
- Likes
- 300
Stole his Thunder, huh? Sorry, just couldn't let that one go...
Cheers!
-HK sends
Haha, quite literally yes!
bretemm
500+ Head-Fier
- Joined
- Feb 23, 2015
- Posts
- 811
- Likes
- 36
Would it make sense to get the Jotunheim with the phono card, because I already have a Bifrost? (I have a Valhalla2 I use with it.) I'm wanting to get into Vinyl and switch between more sources. I'm also looking at powered monitors in the future.
rsnblmn
Head-Fier
- Joined
- Feb 10, 2016
- Posts
- 86
- Likes
- 80
Tentatively, I would say "yes"...that is exactly the type of use case they designed it for, I think. There is another question in the impressions thread about whether the SE inputs can be used to drive the Balanced output, so I am waiting for someone from Schiit to confirm that...but I'm guessing they're kinda busy at the moment.
jfoxvol
500+ Head-Fier
The amp is inherently balanced. It doesn't use phase splitter to go se-balanced on inputs. Doesn't have to with this topology. What you put in will always come out balanced but can be se as well.
bosiemoncrieff
Headphoneus Supremus
You already have an awesome amp, why get another? Get Mani, upgrade to Bimby, throw in SYS if plugging and unplugging cables is too fiddly for you, and you're still at $71 less than Phonoheim.
Tentatively, I would say "yes"...that is exactly the type of use case they designed it for, I think. There is another question in the impressions thread about whether the SE inputs can be used to drive the Balanced output, so I am waiting for someone from Schiit to confirm that...but I'm guessing they're kinda busy at the moment.
jacal01
You've just been blocked by Mr. Conviviality!
Jason's stories are getting to be every bit as engaging as the products they're releasing.
Users who are viewing this thread
