Schiit Happened: The Story of the World's Most Improbable Start-Up
- Thread starter Jason Stoddard
- Start date
-
- Tags
- schiit audio schiit-audio
kostaszag
100+ Head-Fier
- Joined
- Mar 22, 2014
- Posts
- 354
- Likes
- 62
Why dont you send it to cheaprecovery in Crete? Even with postage and all it will be cheaper than that. they are very nice guys, ask them for a quote.
Shini44
Headphoneus Supremus
- Joined
- Sep 22, 2012
- Posts
- 2,502
- Likes
- 604
.
SuperU
100+ Head-Fier
- Joined
- Dec 6, 2014
- Posts
- 327
- Likes
- 98
I'm in - late to the party but better late than never.
And one page in to the story and even more glad I bought my Schiit from you.
And one page in to the story and even more glad I bought my Schiit from you.
kostaszag
100+ Head-Fier
- Joined
- Mar 22, 2014
- Posts
- 354
- Likes
- 62
OK, just saw your location. I have been reading this thread on tapatalk and thought for some reason you were located in the UK. Sorry about that
BackToAnalogue
100+ Head-Fier
- Joined
- Oct 6, 2014
- Posts
- 106
- Likes
- 30
After growing up in the 60's, college and military in the 70's, and working for record and film companies and radio stations in the 80's I collected a rather large vinyl collection and many thousands of dollars worth of high-end turntable paraphernalia In 2005-2006 I sold or gave it all away and joined the digital revolution. Now my rather large music collection lives on a 9Tb hard drive array and I will never go back to the drudgery of vinyl. Go for it ye hipsters, and may ye find your holy grail. But me, I've been there done that and I ain't never going back! This was one wall...
Same here -- never replaced all of my albums though. Seeing Tubular Bells made me think of that...
"In 2005-2006 I sold or gave it all away"
You .......... did................what ????
Please reassure me that this was at the behest of some female member of our species, and not just an act of wanton vandalism.
If you haven't admitted to yourself already that you regret that, then you one day will do.
Or it could even come as a death-bed reconciliation, in much the same way that lapsed members of the Catholic faith sometimes rediscover their religion at this late time in their lives.
But the God of Catholicism is a far more forgiving God, than the God of Analogue.
Jason Stoddard
Sponsor: Schiit Audio
Chapter 41:
Completely Fulla Schiit
Okay, so now we get to Fulla.
But before I start on that, let me start by thanking everyone who’s read this far. I get a lot of email from people thanking me for writing this, but the reality is…this isn’t the easiest book to read. There’s a lot of dense tech and boring business, and it’s wrapped around a lot of stuff that may make you think we’re totally incompetent at this whole audio business thing.
But, like I’ve said before…nobody is perfect. What matters is making it right. Which will definitely weigh on the Fulla story, as you probably already know.
What you don’t know is all the other stuff wrapped around Fulla, which includes a host of dead-end products, Mike Moffat’s disapproval, and the whole grabbing-around-in-the-dark thing that comes with the first digital product I’ve designed since the Cobalt 307.
Yes, you read that right. This product is one I took to fruition. Although it’s heavily based on Mike’s Modi design, I made the decisions about what power supplies to use, what op-amps for gain and output, what form factor, what feature set, and so on.
Why? Mainly a difference in point of view. When it comes to digital, Mike feels that we should always look to lead. To stand apart. To set the standards. This is a very natural point of view for someone with Mike’s background and resume. And, to be very clear, he’s absolutely right. If we were in the business of simply taking the same D/A chip that everyone else was using, and marrying it to the same USB interface everyone else was using, the potential for real advancement would be much smaller.
So, I understand his point of view. But I also like to experiment. And, even in the most perfect world, even Mike will admit that there’s a price point below which multibit technology really isn’t viable.
And that’s the roundabout way I began working on Valkyrie.
Wait, Valkyrie? I Thought This Was About Fulla!
It is. But to get to Fulla, because to get to Fulla, we have to go through Valkyrie. And before I get to Valkyrie, let’s talk about the business case here. Specifically, why did I want to play around with a tiny DAC/amp combo, when we already had Magni and Modi?
It’s simple:
A $198 combo purchase is still a decent amount of money for a lot of people. Something that could sell for less than half the cost means we can reach a lot more people.
How many more? Based on our experience, halving the cost of a product squares to cubes the volume, depending on the product. So, half the cost means 4x to 8x higher sales. Guys, that’s huge.
“Oh, well, now you’re becoming a big evil company,” you say now. “Chasing volume, because that’s all that counts to you.”
Well, uh, no. I mentioned in the last chapter that every ongoing business must think about continuity. And, in the case of an audio business, we have to think about it more than many other businesses, because everything in audio is strongly seasonal. This is a fancy way of saying that your sales in December and January are probably 3-4x your sales in July and August. Products which sell consistently throughout the year are a big deal when you’re planning for a slow summer.
And guess which products are most consistent: inexpensive ones.
So adding even less expensive products is a very, very good idea—providing they fit with your company philosophy, and deliver a level of performance you’re happy with.
Stop. Go back and read that again. Simply designing to a price point never works. But, if you can deliver a great product at a low price, it can be a real winner. It can make a lot of people happy.
That was the big experiment with Valkyrie (and eventually, Fulla)—to see if we could produce a combined product that was fun, sounded good, and provided huge value.
But, as far as I’m concerned, there are also a couple of other reasons to pursue inexpensive products on the audiophile side of things:
“Okay, okay, I’ll agree with you if you just shut up,” you’re saying now. “Get to the story. We get that you like to experiment. What made those experiments end up at Fulla?”
Okay, fine. Because it’s a pretty good story. Because, when I floated the idea for the products that led to Fulla, Mike hated them.
The Beginnings of Valkyrie
Late in 2013, new dongle DACs were showing up seemingly by the week, chasing the success of the original Audioquest Dragonfly. This alone was enough to ensure that Mike would want nothing to do with a Schiit version.
But I still wondered…what exactly was the appeal of a dongle-DAC? They were inherently limited, in being powered by the USB port. If they didn’t use a switching regulator to create a negative rail, they really couldn’t deliver much more power than a laptop headphone jack. And many of them never really moved around very much, being stuck into work laptops that just sat on desks.
So, maybe it was time to look at a combined Schiit product. Something that could be powered by a USB port, for a single easy connection and no power cords. Something that could be one small box, for portability.
But I wasn’t thinking Fulla sized—at first, in fact, I figured that just putting everything in a Magni-sized box would do it. Just one USB cable in for data and power, volume control, and headphone out. Might as well throw line outs on the back as well. And, well, maybe an analog in would be nice as well, so you could use a source besides your computer with just the amp.
But that idea never got beyond the pencil sketch stage.*
*Fun fact: I still keep paper engineering notebooks…pencil on paper is how most stuff starts at Schiit, even though it ends up in tolerance 3D CAD these days. It’s how I figure out pretty much everything, from the slick way to mount MOSFETs in Asgard 2 and Lyr 2, to the Fulla chassis.
Why didn’t it go further? Because I thought I had an even better idea. Why not make the device smaller, and sell it with a little base so it could sit vertically on a desk. That way, it could be, well, about the same size as a portable amplifier.
This is what I called “Valkyrie.” It was a 2.5” x 4.5” x 0.8”, round-fronted design that sat vertically like a blade on a desk. It had USB in for power and data, plus analog in, plus a gain switch, preamp outs, and full-size headphone jack. It used a switching rail generator to create +/- 9V from the USB 5V in, and had a discrete output stage.
Or, in other words, I stuffed about as much Schiit as I could into it—high power, discrete, versatile, etc.
But this wasn’t enough. I also made two more versions—one using the tubes out of the Vali (and a wall-wart), and one that ran on a 3.7V lithium battery pack and had all the charge management/etc you need to keep batteries like that healthy and alive.
Yes, you’re hearing that correctly. Fulla began as three products, none of which was Fulla.
So where did Fulla come in? As a lark. When I was getting ready to send the Valkyrie boards out to prototype, I wondered, Just how small could we go with this? Could it be stripped down even further?
So, over the next day, I pared back the Valkyrie design to its bare minimum—the USB input receiver, a -5V rail generator, the DAC, and op-amps for gain and output. After staring at integrated volume controller datasheets and thinking, We ain’t got no room for a microprocessor in here, I found an 8mm Alps potentiometer that was silly small and stuffed it into the board. The end result was about 1 x 2.5”, far smaller than Valkyrie.
Feeling silly, I screened on the board, “Dingle Dongle, it’s a DAC.” No product name. No idea what I’d call it.
It would be funny, having a dongle-DAC with an actual volume pot, I remember thinking. But I didn’t think much about it, because I was really focused on Valkyrie. What would become Fulla was just, well, playing. It didn’t even have any mounting holes on the board.
We probably wouldn’t do anything with it…
All the Best Plans…Change
When we got the prototype boards for the three versions of Valkyrie and what would become Fulla, I laughed at how tiny they all were—three products that were all dwarfed by Magni and Modi boards.
I showed them to Mike. He grimaced and shook his head, still not thrilled with the whole grand “cheap experiment.” He picked up the dongle board and said, “So what are you gonna call this one? Dingleberry?”
I laughed. “We probably won’t call it anything,” I said. “I just threw it in to see what we could do with it.”
Mike eyed the board. “A volume pot?”
“Yeah.”
Finally a grin. “Now, that’s cool. What does it use for output?”
“ADA4610 for gain, and a DSL line driver for power.”
Mike laughed. “A DSL line driver?”
“It has good distortion specs, though,” I said.
“No, no, I love it,” Mike said. “I hope it sounds good. I’d love to say we’re using a DSL line driver for audio.”*
*Aside: we didn’t end up using the DSL line driver chip. It couldn’t swing the rails, so we were throwing away too much potential output. Hence the AD8397.
Mike was less thrilled with the Valkyries, except for the one with tubes. “Now, that’s cool, too,” he said. “But USB power?”
“Nope, not enough power for that. It needs a wall-wart.”
Mike squinted at the tiny markings on the bare boards. “Who’s going to build the protos?” he asked.
“Me.”
A laugh. “Good luck.”
And Mike was right. After an abortive attempt to assemble one of the boards, I gave up and ended sending them to our PCB assembly house. The 0402 resistors and 100 pin QFN got me. Yes, I’m lazy.
And, interestingly enough, it was a good learning experience. Having your board house do prototypes means you have to have the bill of materials worked out (good), which takes a lot more time than you expect (bad), and they will inevitably find some problems with the boards you missed (good), but it may take some back and forth to get all their questions answered, meaning even more time (bad.)
Let’s leave it at this: proto assembly by boardhouse can definitely end up extending your development time.
When the boards came back, the only one that just started up and ran was the Dingleberry. I think that should have clued me in what was going to happen right there.
And yep. The name had stuck. Dingleberry.
What Happened to Valkyrie
So why didn’t we pursue Valkyrie? Well, we did, for a while. But in the end, all of them ended up as non-products.
Why? Depends on the product:
Valkyrie Tube: Holy crap heat, needed a wall-wart, too noisy for the intended application. No chance of it ever being entirely USB powered. So cable hell on the desktop, too.
Valkyrie: noisy-as-hell supply (which could have been worked through, mainly), discrete output not swinging rails, so that meant a redesign, pricey chassis to do it the way we wanted to, might end up costing more than Magni/Modi combo.
Valkyrie Battery: Same as above, plus plenty of good portables out there anyway, why do something similar? Better to shelve it. In the round file. I’ll repeat this here, because I’m sure we’ll be asked, but I’m unsure if we’ll ever do a portable. It would have to be very cool/different to get us excited about it.
Valkyrie was a great lesson in modesty. Because 0 for 3 is a pretty crappy batting average.
But then there was Dingleberry.
Snap Goes the Dingleberry
Like I said, the first Dingleberry just plugged into a USB port, programmed, and ran. Of course, it wasn’t perfect—a couple of clocks needed hacked, and I’d screwed up the output ground. But with less than a half an hour of work, we were hearing music.
And…it wasn’t bad. In fact, it had pretty good promise.
There was just one problem: I took it home, plugged it into my laptop’s USB port, sat down…and the captive USB connector promptly snapped right off the board.
“Well, that’s that,” I said. I’d seen plenty of dongles with the captive USB connector, but ours wouldn’t be one of them. The next design used a mini-USB receptacle.
But that wasn’t the extent of the changes. In the brief lifespan of the first prototype, I’d seen promise, so it was time to get serious. Serious enough to put mounting holes on the board, and figure out how this thing might actually be assembled.
But to do that, I had to figure out the chassis…and that was a challenge in itself.
Why? Because I knew from the preliminary bill of material that the make-or-break point of Fulla was the chassis. If we could get a chassis made inexpensively enough, it could be a $79 product. If the chassis was costly, it could easily balloon into a 3-figure product…and, at that point, why bother?
And, when you’re talking small chassis, you have a ton of potential choices…but very few that would fit with our budget, volume, and aesthetic:
Milled aluminum? Yeah, I know, this is the thing that everyone likes to do these days. Unfortunately, unless you’re making literally tens of millions of them like Apple, the cost will make you faint. And, looking at it dispassionately, this is a horrendously wasteful process, milling a solid block of aluminum into a pile of shavings. That’s why it’s costly.
So that’s what we went with—more folded metal. Fulla uses a tiny, custom aluminum and steel chassis, together with a custom aluminum knob. To further simplify assembly, it also has no captive fasteners at all—the PEM insets are on the PC board, and screws sandwich the two chassis pieces together with the PC board. It’s actually kinda amazing we got it to work in sheet metal so well.
Aside: actually, it’s kinda amazing how much unique or custom stuff went into Fulla—from the chassis pieces, to our most elaborate knob design (and also our first 3D-modeled part), to custom M1.4 x 10mm screws for the volume knob, to the SMD-mount 2-56 PEMs, to a 4-layer PC board—but I’m getting really geeky here, aren’t I?
With the chassis design in place, I made the necessary adjustments on the PCB and sent it out for another proto run—this time alone, without a Valkyrie in tow. Because this time we were going to do a real run—a full panel of 40 boards. This one was named “SAGA,” because I didn’t really want to commit to “Dingleberry.”
When the boards came back, we programmed them with Modi firmware and gave them out to all Schiit employees and friends who were interested, since I wanted to have them banged around for a while before we committed to production. This is when I started going around the house with a prototype Fulla hanging off the end of a pair of HD800s with balanced-to-single-ended converter and a ¼” to 1/8” converter that were much bigger than the board.
It sounded good enough that I really didn’t miss the big iron too much…but it did have problems. Most notable was a very noisy negative power supply rail—which, while the noise was at inaudible frequencies, was not ideal. Also, the DSL line driver was limiting the power output, because it couldn’t swing the output close enough to the rails.
After the addition of a couple of inductors, a lot of ground plane work, additional bypassing, and the swap of the DSL driver for an AD8397, we had something that sounded quite a bit better.
“Better than the laptop output,” Mike said. Faint praise, but he was at least grinning now. “And the volume pot is cool.”
“So we’re gonna do this?” I asked.
Mike shrugged. “For $79? If it sounds a lot better than a laptop, why not?”
“You still gonna call it the Dingleberry?”
I frowned. “Umm, well…”
“Don’t tell me you’re chickening out?” Mike said. “The intrepid marketing guy is afraid that it’s in too poor taste? For a company called Schiit?”
I didn’t say anything for a long time. Because that’s exactly how I felt. Yes, we were the iconoclasts in the business, yes, we had a crazy name, yeah, we made fun of ourselves…but I also thought there was a line. And I thought that “Dingleberry” was stepping over the line.
“So what do we call it?” Mike asked. “The ZX-01 Interoscillator? Thor’s Nuts?”
“I don’t know, but I’ll come up with something.”
“I still like Dingleberry.”
I gritted my teeth and said nothing. Mike liked the name. Rina liked the name. A lot of our employees liked the name.
But I couldn’t do it.
Eventually, I happened on Fulla, a Norse goddess. And it all fell into place: not Schiit Fulla, but Fulla Schiit. Mike grudgingly admitted he was okay with it. Rina accepted it.
Or, in other companies, it got the 56 signatures and legal clearances necessary to proceed. I got to working on the final board revs, thinking, Soon we’ll be Fulla Schiit.
Soon is Relative: The Problems You Didn’t See
When I got to work on the final board, it was early summer 2014. I figured we’d be shipping Fullas by late August, maybe early September. No problem, the design was done, the chassis was ordered, there really wasn’t a lot to work out.
Except for a few things:
So, by the time we started shipping Fullas, we figured, Heck, after all of the easy products we’d launched, we were due for a bad one.
We just didn’t know how bad.
The Oscillator Problem
Okay, all of you out there who want to run your own electronics business, take notes: even after a lot of pain, you can still get bitten. And two days after launching Fulla, we got bit, bad: multiple owners were reporting that the 48k and 96k sample rates weren’t working on their Fullas—specifically, they were getting distorted output. 44.1 and 88.2 were fine.
Uh, oh, I thought, when I heard about the second case. Because that meant only one thing: the oscillator for the 48k sample rate multiples wasn’t working for some reason.
Aside: I will use this to note that Fulla does have separate crystal oscillators for both 44.1 and 48k clock multiples, just like the “big boys.” It does not derive both from the 12MHz USB crystal—an insanely high-end approach for a $79 product.
I took a look at the Fulla boards we had in stock. The 49.152MHz oscillators were all in place, and the solder looked good.
But they also didn’t look like the 49.152MHz oscillators we used for everything else.
Some more investigation revealed what had happened: a different part had been swapped into the run at some point, most likely from a “taped reel.” (We found out later it was swapped in very early—there were very few working Fullas.)
Now, before you get worked up about how the PCB assembly house messed up, the reality was that they didn’t do anything wrong with the assembly. Taped reels are common. You’re usually just splicing in the same part. Or one that was substantially similar. This was a case of the latter—they used an alternate part we said was OK (it was still a 49.152MHz oscillator) but really wasn’t. It was the first time we’d used them, so we got bitten.
So how did this make it through to production and shipping, you ask? Here’s how: first article approval and boardhouse programming and test.
You see, before the PCB assembly house does a full run, it usually does a “first article” for approval. These first articles (usually 5-10 boards) are run with the same parts and processes as the whole run. So, if they work, the whole run works.
As usual, we received the first articles and ran them through the full battery of tests, including all sample rates on the Stanford analyzers. And they were fine. They all worked, on both 44.1 and 48k clock multiples.
So, with that assurance in place, we gave the assembly house the OK to build everything…and to program and test them as well. This was the first time we’d done programming and testing out of house, but at the volume we’re running Fulla at, it made sense.
Except…except they didn’t test them at all the sample rates.
With the approval of the first articles, this shouldn’t have been a problem. But clearly the first articles used the last of the good oscillators, whereas the majority of the production run used the bad oscillators.
Boom. Big problem.
Making it Good
So what do you do when confronted with a problem like this? When you’ve already shipped a few hundred products, and you have orders for hundreds more?
In our case, we shut down ordering on Friday, and were only able to address the problem on Monday at the PCB assembly house. Then, it took a couple of days to determine how many good Fullas we had (pretty much none) and how best to swap the oscillator on the bad ones. It was Wednesday before we had that all down, Friday before we had good stock, and now, on Monday, we’re shipping the replacements and getting ready to open up new orders.
Helluva week. Especially when we knew we had to launch the next-gen Magni and Modi, too…
Completely Fulla Schiit
Okay, so now we get to Fulla.
But before I start on that, let me start by thanking everyone who’s read this far. I get a lot of email from people thanking me for writing this, but the reality is…this isn’t the easiest book to read. There’s a lot of dense tech and boring business, and it’s wrapped around a lot of stuff that may make you think we’re totally incompetent at this whole audio business thing.
But, like I’ve said before…nobody is perfect. What matters is making it right. Which will definitely weigh on the Fulla story, as you probably already know.
What you don’t know is all the other stuff wrapped around Fulla, which includes a host of dead-end products, Mike Moffat’s disapproval, and the whole grabbing-around-in-the-dark thing that comes with the first digital product I’ve designed since the Cobalt 307.
Yes, you read that right. This product is one I took to fruition. Although it’s heavily based on Mike’s Modi design, I made the decisions about what power supplies to use, what op-amps for gain and output, what form factor, what feature set, and so on.
Why? Mainly a difference in point of view. When it comes to digital, Mike feels that we should always look to lead. To stand apart. To set the standards. This is a very natural point of view for someone with Mike’s background and resume. And, to be very clear, he’s absolutely right. If we were in the business of simply taking the same D/A chip that everyone else was using, and marrying it to the same USB interface everyone else was using, the potential for real advancement would be much smaller.
So, I understand his point of view. But I also like to experiment. And, even in the most perfect world, even Mike will admit that there’s a price point below which multibit technology really isn’t viable.
And that’s the roundabout way I began working on Valkyrie.
Wait, Valkyrie? I Thought This Was About Fulla!
It is. But to get to Fulla, because to get to Fulla, we have to go through Valkyrie. And before I get to Valkyrie, let’s talk about the business case here. Specifically, why did I want to play around with a tiny DAC/amp combo, when we already had Magni and Modi?
It’s simple:
A $198 combo purchase is still a decent amount of money for a lot of people. Something that could sell for less than half the cost means we can reach a lot more people.
How many more? Based on our experience, halving the cost of a product squares to cubes the volume, depending on the product. So, half the cost means 4x to 8x higher sales. Guys, that’s huge.
“Oh, well, now you’re becoming a big evil company,” you say now. “Chasing volume, because that’s all that counts to you.”
Well, uh, no. I mentioned in the last chapter that every ongoing business must think about continuity. And, in the case of an audio business, we have to think about it more than many other businesses, because everything in audio is strongly seasonal. This is a fancy way of saying that your sales in December and January are probably 3-4x your sales in July and August. Products which sell consistently throughout the year are a big deal when you’re planning for a slow summer.
And guess which products are most consistent: inexpensive ones.
So adding even less expensive products is a very, very good idea—providing they fit with your company philosophy, and deliver a level of performance you’re happy with.
Stop. Go back and read that again. Simply designing to a price point never works. But, if you can deliver a great product at a low price, it can be a real winner. It can make a lot of people happy.
That was the big experiment with Valkyrie (and eventually, Fulla)—to see if we could produce a combined product that was fun, sounded good, and provided huge value.
But, as far as I’m concerned, there are also a couple of other reasons to pursue inexpensive products on the audiophile side of things:
- Affordable audiophile products are very thin on the ground. Seems like everyone wants to do $5,000 amps and $25,000 speakers, but when you ask for something that might get your kids interested, they shuffle their feet and look uncomfortable. We need more entry-level products that deliver great performance.
- It can teach you things that will make all of your products better. Seeing how much performance you can squeeze out of cost- or space-constrained designs is a real challenge—and, in doing so, you may discover even better ways of doing things, period. This is one reason we say, “we prefer to do things one way,” but challenge our assumptions on this from time to time.
“Okay, okay, I’ll agree with you if you just shut up,” you’re saying now. “Get to the story. We get that you like to experiment. What made those experiments end up at Fulla?”
Okay, fine. Because it’s a pretty good story. Because, when I floated the idea for the products that led to Fulla, Mike hated them.
The Beginnings of Valkyrie
Late in 2013, new dongle DACs were showing up seemingly by the week, chasing the success of the original Audioquest Dragonfly. This alone was enough to ensure that Mike would want nothing to do with a Schiit version.
But I still wondered…what exactly was the appeal of a dongle-DAC? They were inherently limited, in being powered by the USB port. If they didn’t use a switching regulator to create a negative rail, they really couldn’t deliver much more power than a laptop headphone jack. And many of them never really moved around very much, being stuck into work laptops that just sat on desks.
So, maybe it was time to look at a combined Schiit product. Something that could be powered by a USB port, for a single easy connection and no power cords. Something that could be one small box, for portability.
But I wasn’t thinking Fulla sized—at first, in fact, I figured that just putting everything in a Magni-sized box would do it. Just one USB cable in for data and power, volume control, and headphone out. Might as well throw line outs on the back as well. And, well, maybe an analog in would be nice as well, so you could use a source besides your computer with just the amp.
But that idea never got beyond the pencil sketch stage.*
*Fun fact: I still keep paper engineering notebooks…pencil on paper is how most stuff starts at Schiit, even though it ends up in tolerance 3D CAD these days. It’s how I figure out pretty much everything, from the slick way to mount MOSFETs in Asgard 2 and Lyr 2, to the Fulla chassis.
Why didn’t it go further? Because I thought I had an even better idea. Why not make the device smaller, and sell it with a little base so it could sit vertically on a desk. That way, it could be, well, about the same size as a portable amplifier.
This is what I called “Valkyrie.” It was a 2.5” x 4.5” x 0.8”, round-fronted design that sat vertically like a blade on a desk. It had USB in for power and data, plus analog in, plus a gain switch, preamp outs, and full-size headphone jack. It used a switching rail generator to create +/- 9V from the USB 5V in, and had a discrete output stage.
Or, in other words, I stuffed about as much Schiit as I could into it—high power, discrete, versatile, etc.
But this wasn’t enough. I also made two more versions—one using the tubes out of the Vali (and a wall-wart), and one that ran on a 3.7V lithium battery pack and had all the charge management/etc you need to keep batteries like that healthy and alive.
Yes, you’re hearing that correctly. Fulla began as three products, none of which was Fulla.
So where did Fulla come in? As a lark. When I was getting ready to send the Valkyrie boards out to prototype, I wondered, Just how small could we go with this? Could it be stripped down even further?
So, over the next day, I pared back the Valkyrie design to its bare minimum—the USB input receiver, a -5V rail generator, the DAC, and op-amps for gain and output. After staring at integrated volume controller datasheets and thinking, We ain’t got no room for a microprocessor in here, I found an 8mm Alps potentiometer that was silly small and stuffed it into the board. The end result was about 1 x 2.5”, far smaller than Valkyrie.
Feeling silly, I screened on the board, “Dingle Dongle, it’s a DAC.” No product name. No idea what I’d call it.
It would be funny, having a dongle-DAC with an actual volume pot, I remember thinking. But I didn’t think much about it, because I was really focused on Valkyrie. What would become Fulla was just, well, playing. It didn’t even have any mounting holes on the board.
We probably wouldn’t do anything with it…
All the Best Plans…Change
When we got the prototype boards for the three versions of Valkyrie and what would become Fulla, I laughed at how tiny they all were—three products that were all dwarfed by Magni and Modi boards.
I showed them to Mike. He grimaced and shook his head, still not thrilled with the whole grand “cheap experiment.” He picked up the dongle board and said, “So what are you gonna call this one? Dingleberry?”
I laughed. “We probably won’t call it anything,” I said. “I just threw it in to see what we could do with it.”
Mike eyed the board. “A volume pot?”
“Yeah.”
Finally a grin. “Now, that’s cool. What does it use for output?”
“ADA4610 for gain, and a DSL line driver for power.”
Mike laughed. “A DSL line driver?”
“It has good distortion specs, though,” I said.
“No, no, I love it,” Mike said. “I hope it sounds good. I’d love to say we’re using a DSL line driver for audio.”*
*Aside: we didn’t end up using the DSL line driver chip. It couldn’t swing the rails, so we were throwing away too much potential output. Hence the AD8397.
Mike was less thrilled with the Valkyries, except for the one with tubes. “Now, that’s cool, too,” he said. “But USB power?”
“Nope, not enough power for that. It needs a wall-wart.”
Mike squinted at the tiny markings on the bare boards. “Who’s going to build the protos?” he asked.
“Me.”
A laugh. “Good luck.”
And Mike was right. After an abortive attempt to assemble one of the boards, I gave up and ended sending them to our PCB assembly house. The 0402 resistors and 100 pin QFN got me. Yes, I’m lazy.
And, interestingly enough, it was a good learning experience. Having your board house do prototypes means you have to have the bill of materials worked out (good), which takes a lot more time than you expect (bad), and they will inevitably find some problems with the boards you missed (good), but it may take some back and forth to get all their questions answered, meaning even more time (bad.)
Let’s leave it at this: proto assembly by boardhouse can definitely end up extending your development time.
When the boards came back, the only one that just started up and ran was the Dingleberry. I think that should have clued me in what was going to happen right there.
And yep. The name had stuck. Dingleberry.
What Happened to Valkyrie
So why didn’t we pursue Valkyrie? Well, we did, for a while. But in the end, all of them ended up as non-products.
Why? Depends on the product:
Valkyrie Tube: Holy crap heat, needed a wall-wart, too noisy for the intended application. No chance of it ever being entirely USB powered. So cable hell on the desktop, too.
Valkyrie: noisy-as-hell supply (which could have been worked through, mainly), discrete output not swinging rails, so that meant a redesign, pricey chassis to do it the way we wanted to, might end up costing more than Magni/Modi combo.
Valkyrie Battery: Same as above, plus plenty of good portables out there anyway, why do something similar? Better to shelve it. In the round file. I’ll repeat this here, because I’m sure we’ll be asked, but I’m unsure if we’ll ever do a portable. It would have to be very cool/different to get us excited about it.
Valkyrie was a great lesson in modesty. Because 0 for 3 is a pretty crappy batting average.
But then there was Dingleberry.
Snap Goes the Dingleberry
Like I said, the first Dingleberry just plugged into a USB port, programmed, and ran. Of course, it wasn’t perfect—a couple of clocks needed hacked, and I’d screwed up the output ground. But with less than a half an hour of work, we were hearing music.
And…it wasn’t bad. In fact, it had pretty good promise.
There was just one problem: I took it home, plugged it into my laptop’s USB port, sat down…and the captive USB connector promptly snapped right off the board.
“Well, that’s that,” I said. I’d seen plenty of dongles with the captive USB connector, but ours wouldn’t be one of them. The next design used a mini-USB receptacle.
But that wasn’t the extent of the changes. In the brief lifespan of the first prototype, I’d seen promise, so it was time to get serious. Serious enough to put mounting holes on the board, and figure out how this thing might actually be assembled.
But to do that, I had to figure out the chassis…and that was a challenge in itself.
Why? Because I knew from the preliminary bill of material that the make-or-break point of Fulla was the chassis. If we could get a chassis made inexpensively enough, it could be a $79 product. If the chassis was costly, it could easily balloon into a 3-figure product…and, at that point, why bother?
And, when you’re talking small chassis, you have a ton of potential choices…but very few that would fit with our budget, volume, and aesthetic:
Milled aluminum? Yeah, I know, this is the thing that everyone likes to do these days. Unfortunately, unless you’re making literally tens of millions of them like Apple, the cost will make you faint. And, looking at it dispassionately, this is a horrendously wasteful process, milling a solid block of aluminum into a pile of shavings. That’s why it’s costly.
- Plastic injection molding? Sure, that’s cheap in reasonable quantities, but…it’s plastic. And you still have to meet FCC somehow. So the metal shields inside might send the budget over the target. So, nope.
- Metal injection molding? Seriously considered, but it’s really better at higher volumes, since the tooling costs have to be amortized into it.
- Extrusions? Sure, but they are thick and clunky, and I wanted this to be thin. Still, we did seriously look at going this way.
- Folded sheetmetal? At first glance, this looked like the least likely option. Fulla is tiny, and requires precise bends. But we lobbed a sketch at our sheetmetal guys, and they turned a quote that was well-within our price range. Suddenly, a $79 product could be an actual reality.
So that’s what we went with—more folded metal. Fulla uses a tiny, custom aluminum and steel chassis, together with a custom aluminum knob. To further simplify assembly, it also has no captive fasteners at all—the PEM insets are on the PC board, and screws sandwich the two chassis pieces together with the PC board. It’s actually kinda amazing we got it to work in sheet metal so well.
Aside: actually, it’s kinda amazing how much unique or custom stuff went into Fulla—from the chassis pieces, to our most elaborate knob design (and also our first 3D-modeled part), to custom M1.4 x 10mm screws for the volume knob, to the SMD-mount 2-56 PEMs, to a 4-layer PC board—but I’m getting really geeky here, aren’t I?
With the chassis design in place, I made the necessary adjustments on the PCB and sent it out for another proto run—this time alone, without a Valkyrie in tow. Because this time we were going to do a real run—a full panel of 40 boards. This one was named “SAGA,” because I didn’t really want to commit to “Dingleberry.”
When the boards came back, we programmed them with Modi firmware and gave them out to all Schiit employees and friends who were interested, since I wanted to have them banged around for a while before we committed to production. This is when I started going around the house with a prototype Fulla hanging off the end of a pair of HD800s with balanced-to-single-ended converter and a ¼” to 1/8” converter that were much bigger than the board.
It sounded good enough that I really didn’t miss the big iron too much…but it did have problems. Most notable was a very noisy negative power supply rail—which, while the noise was at inaudible frequencies, was not ideal. Also, the DSL line driver was limiting the power output, because it couldn’t swing the output close enough to the rails.
After the addition of a couple of inductors, a lot of ground plane work, additional bypassing, and the swap of the DSL driver for an AD8397, we had something that sounded quite a bit better.
“Better than the laptop output,” Mike said. Faint praise, but he was at least grinning now. “And the volume pot is cool.”
“So we’re gonna do this?” I asked.
Mike shrugged. “For $79? If it sounds a lot better than a laptop, why not?”
“You still gonna call it the Dingleberry?”
I frowned. “Umm, well…”
“Don’t tell me you’re chickening out?” Mike said. “The intrepid marketing guy is afraid that it’s in too poor taste? For a company called Schiit?”
I didn’t say anything for a long time. Because that’s exactly how I felt. Yes, we were the iconoclasts in the business, yes, we had a crazy name, yeah, we made fun of ourselves…but I also thought there was a line. And I thought that “Dingleberry” was stepping over the line.
“So what do we call it?” Mike asked. “The ZX-01 Interoscillator? Thor’s Nuts?”
“I don’t know, but I’ll come up with something.”
“I still like Dingleberry.”
I gritted my teeth and said nothing. Mike liked the name. Rina liked the name. A lot of our employees liked the name.
But I couldn’t do it.
Eventually, I happened on Fulla, a Norse goddess. And it all fell into place: not Schiit Fulla, but Fulla Schiit. Mike grudgingly admitted he was okay with it. Rina accepted it.
Or, in other companies, it got the 56 signatures and legal clearances necessary to proceed. I got to working on the final board revs, thinking, Soon we’ll be Fulla Schiit.
Soon is Relative: The Problems You Didn’t See
When I got to work on the final board, it was early summer 2014. I figured we’d be shipping Fullas by late August, maybe early September. No problem, the design was done, the chassis was ordered, there really wasn’t a lot to work out.
Except for a few things:
- The first articles for the chassis were much too thin. We’d originally spec’d all aluminum for the chassis, which flexed too much. Back to the drawing board.
- The second articles for the chassis were still too flexy. Time to go to a steel inner chassis. Also time to wait some more.
- We needed a second run of prototypes to qualify the “real” board. This took longer than expected, with a boardhouse underwater in new products, including the Ragnarok boards.
- The knobs were late.
- The chassis were late.
- The original PEM nuts to hold the whole thing together broke the boards when inserted. Time to find new SMD versions.
- About 80% of the original USB mini cables were junk. Time to find a new supplier.
- Our original chassis insulation plan didn’t work. Time for a new custom part to ensure the sparky parts of the board didn’t meet the aluminum chassis in bad new ways.
- The original 2-56 screws were too short for easy assembly. Yeah, not a huge deal, but still a delay we didn’t need.
So, by the time we started shipping Fullas, we figured, Heck, after all of the easy products we’d launched, we were due for a bad one.
We just didn’t know how bad.
The Oscillator Problem
Okay, all of you out there who want to run your own electronics business, take notes: even after a lot of pain, you can still get bitten. And two days after launching Fulla, we got bit, bad: multiple owners were reporting that the 48k and 96k sample rates weren’t working on their Fullas—specifically, they were getting distorted output. 44.1 and 88.2 were fine.
Uh, oh, I thought, when I heard about the second case. Because that meant only one thing: the oscillator for the 48k sample rate multiples wasn’t working for some reason.
Aside: I will use this to note that Fulla does have separate crystal oscillators for both 44.1 and 48k clock multiples, just like the “big boys.” It does not derive both from the 12MHz USB crystal—an insanely high-end approach for a $79 product.
I took a look at the Fulla boards we had in stock. The 49.152MHz oscillators were all in place, and the solder looked good.
But they also didn’t look like the 49.152MHz oscillators we used for everything else.
Some more investigation revealed what had happened: a different part had been swapped into the run at some point, most likely from a “taped reel.” (We found out later it was swapped in very early—there were very few working Fullas.)
Now, before you get worked up about how the PCB assembly house messed up, the reality was that they didn’t do anything wrong with the assembly. Taped reels are common. You’re usually just splicing in the same part. Or one that was substantially similar. This was a case of the latter—they used an alternate part we said was OK (it was still a 49.152MHz oscillator) but really wasn’t. It was the first time we’d used them, so we got bitten.
So how did this make it through to production and shipping, you ask? Here’s how: first article approval and boardhouse programming and test.
You see, before the PCB assembly house does a full run, it usually does a “first article” for approval. These first articles (usually 5-10 boards) are run with the same parts and processes as the whole run. So, if they work, the whole run works.
As usual, we received the first articles and ran them through the full battery of tests, including all sample rates on the Stanford analyzers. And they were fine. They all worked, on both 44.1 and 48k clock multiples.
So, with that assurance in place, we gave the assembly house the OK to build everything…and to program and test them as well. This was the first time we’d done programming and testing out of house, but at the volume we’re running Fulla at, it made sense.
Except…except they didn’t test them at all the sample rates.
With the approval of the first articles, this shouldn’t have been a problem. But clearly the first articles used the last of the good oscillators, whereas the majority of the production run used the bad oscillators.
Boom. Big problem.
Making it Good
So what do you do when confronted with a problem like this? When you’ve already shipped a few hundred products, and you have orders for hundreds more?
- Well, first you shut down ordering, so the problem doesn’t get any bigger.
- Then, you set a policy to swap out every problematic Fulla, at no cost to the owner—or issue a refund, if they’d like.
- And then you go to the PCB assembly house and see how deep the mess is.
In our case, we shut down ordering on Friday, and were only able to address the problem on Monday at the PCB assembly house. Then, it took a couple of days to determine how many good Fullas we had (pretty much none) and how best to swap the oscillator on the bad ones. It was Wednesday before we had that all down, Friday before we had good stock, and now, on Monday, we’re shipping the replacements and getting ready to open up new orders.
Helluva week. Especially when we knew we had to launch the next-gen Magni and Modi, too…
rmullins08
100+ Head-Fier
- Joined
- May 12, 2014
- Posts
- 469
- Likes
- 98
That's some pretty impressive turnaround on the damage control. My fulla was in FedEx's possession when the problem was discovered. Laura was able to request a return before it made it to my doorstep, and I will have a replacement (I don't even know if I can call it a replacement, since I hadn't even received the bad one) in hand on Friday.
cizx
Headphoneus Supremus
- Joined
- Jul 31, 2008
- Posts
- 1,582
- Likes
- 209
What are you going to do when you run out the Norse pantheon for names?
superjawes
500+ Head-Fier
- Joined
- Apr 2, 2014
- Posts
- 860
- Likes
- 157
Well just going by gods and goddesses, they have more than 50 options, so it will take some time to reach that many. On top of that, many of the current products don't even use those names. Mjolnir and Gungnir are weapons, Asgard and Valhalla are places, and Bifrost is a bridge.
So if they "run out" of Norse names, I think they'll just retire with their mountains of money.
cizx
Headphoneus Supremus
- Joined
- Jul 31, 2008
- Posts
- 1,582
- Likes
- 209
Right, replace pantheon with mythology in my original question.
tonykaz
500+ Head-Fier
- Joined
- Feb 2, 2014
- Posts
- 778
- Likes
- 388
Ever see those "Quality Checked" little stickers on everything you buy?
"Got-to" check everything before it goes out, even those under $100 little Canon Cameras in Bubble-Gum colors.
Ideally it should also have the checker's initials hand written on the sticker.
Ok, it adds cost but your little "Made in USA" american flag sticker and 100% Quality Checked sticker will add a 25% bump to Quarterly Sales numbers. People love these stickers!, even bananas come with stickers.
Wooda, shuda, cuda, Alex shuda caught this!
Not to make too much a point about all this but the last people on any Auto Assembly Line are a Team of Quality people that "Must" sign-off on each car. The Germans are the biggest proponents of this , watch their videos to see how careful they are.
Everybody loves you guys, we love your stuff, keep up the good work.
Tony in Michigan
ps. I don't think it applies in your case but if you build with re-cycled materials (such as the steel in your chassis) you can honestly say that you use re-cycled materials: allowing you to use a re-cycled materials sticker which will add an additional 15% to Q sales
"Got-to" check everything before it goes out, even those under $100 little Canon Cameras in Bubble-Gum colors.
Ideally it should also have the checker's initials hand written on the sticker.
Ok, it adds cost but your little "Made in USA" american flag sticker and 100% Quality Checked sticker will add a 25% bump to Quarterly Sales numbers. People love these stickers!, even bananas come with stickers.
Wooda, shuda, cuda, Alex shuda caught this!
Not to make too much a point about all this but the last people on any Auto Assembly Line are a Team of Quality people that "Must" sign-off on each car. The Germans are the biggest proponents of this , watch their videos to see how careful they are.
Everybody loves you guys, we love your stuff, keep up the good work.
Tony in Michigan
ps. I don't think it applies in your case but if you build with re-cycled materials (such as the steel in your chassis) you can honestly say that you use re-cycled materials: allowing you to use a re-cycled materials sticker which will add an additional 15% to Q sales
Puma Cat
100+ Head-Fier
- Joined
- Jan 6, 2012
- Posts
- 339
- Likes
- 90
Same for me here. Laura very quickly intercepted mine and had it returned for the new oscillator. I actually received an email on SUNDAY that mine was going back out on Monday. It will be delivered today. The folks at Schiit were working OVER THE WEEKEND to get stuff back out to customers as quickly as possible.
That's what I call dedication.
Great chapter, BTW. I was going ask Jason why not a tube-based DAC along the lines of Valli, but now I know the answer.
derek1ee
New Head-Fier
- Joined
- Dec 10, 2014
- Posts
- 14
- Likes
- 10
Hi Jason -
Really enjoy reading this, I have sort of a random question -
What's the benefit of using a potentiometer, instead of encoder to control the volume?
Thanks,
Derek
Really enjoy reading this, I have sort of a random question -
What's the benefit of using a potentiometer, instead of encoder to control the volume?
Thanks,
Derek
rivon
New Head-Fier
- Joined
- Jul 25, 2013
- Posts
- 6
- Likes
- 11
What?
Magni 2? Modi 2? Uber option?
Man, it's official, you've got uber-bananas...
Best regards,
Roberto
Magni 2? Modi 2? Uber option?
Man, it's official, you've got uber-bananas...
Best regards,
Roberto
