“You’ll Never Do Any Upgrades Anyway.”
When we first introduced Bifrost as “the least expensive fully upgradable DAC on the planet” in 2011, we had some interesting responses.
Some of them went like this:
“Yeah, but it’s not like there’ll be any upgrades.”
“Not that you’re planning to do any upgrades.”
“I’ll believe it when I see it (with respect to the upgrades.)”
Why the doubt? Hell, I don’t know. Audiophiles can be a morose bunch. Maybe that’s the only reason. Or, maybe there was a manufacturer which promised upgrades and recently went out of business, or otherwise didn’t make good on the promise—and that was coloring their response.
But, needless to say, I was shocked at the amount of negative sentiment we received. Sure, there was plenty of positive press, but the opinion of “the audiophile on the street” was less thrilled.
I didn’t worry. I knew we’d have upgrades. It’s in Mike’s DNA to do upgrades.
I just didn’t know when they’d come.
First, Let’s Talk Theta
Mike’s upgrade DNA was implanted at the inception of Theta Digital, I believe. (Mike, correct me if I’m wrong, or if you think this story needs more bat testicles.)
Why? You have to consider the environment. When Mike started Theta, digital audio was in its infancy. Manufacturers were still trying to get the price of a good CD player under $500 (think $900+ in today’s dollars.) There were no standalone DACs. Zero. None. SPDIF, as a transmission standard, was brand new.
And Mike wasn’t just at the leading edge with Theta—he was bleeding edge. Literally. Before Theta, nobody had even considered making a standalone DAC. And nobody else would have started with a flagship $3000 product, using their own digital filter code running on megadollar Motorola DSP chips.
Aside: Think about that a bit. A flagship product that “only” cost about $5500 in today’s dollars? Insanity! Add more CNC-machined parts and a fancier chassis and some custom dampers and heebie-jeebie clocks, until it’s $100K+! Yeah, that’s the only way to go! But that’s a comment on how far we’ve fallen in the past 30 years. From a $5500 first-of-it’s kind product using the latest, greatest, very expensive chips of the time, designed and coded by a team that spanned university researchers, brilliant mathematicians, and top-of-their-field engineers, over a timespan measured in years, all packed in a relatively plain-jane chassis, to beautiful, overpriced audio jewelry powered by voodoo and obfuscation. (And, you think this is harsh? Ask Mike what he thinks.)
But, back to the subject at hand: digital audio was new. Nobody knew what was around the corner.
And things were changing. In contrast to today’s relatively stable market, new D/A converters and new digital filters appeared regularly—pushing from 16 bits to 18 to 20, from no oversampling to 2x and 4x and 8x. New technologies appeared as well—the first “one bit” D/A converters. And, these new chips (with the exception of the “one bit,” or delta-sigma” converters) usually offered significantly better measurable performance than their predecessors.
And, manufacturers were learning as well. Mike was the first to measure jitter, opine that it might have something to do with the sonic deficiencies of early digital audio, and devise ways to minimize it. He experimented with the best interfaces for SPDIF, using transformer-coupled coaxial, and later adding AT&T ST-optical glass fiber as an option.
In this constantly-changing environment, bringing a very pricey, first-of-its-kind product into this market with no ability to upgrade it would be suicide. Because, even if you won some first sales, how thrilled would your customers be when it fell behind the latest “latest and greatest?”
In short: they wouldn’t be.
Which is why Theta Digital was built, from the start, around upgradability. Theta owners could pay a relatively nominal amount to upgrade their DS Pro Gen 1 to 2, 3, and 5 (there was no 4) as the years passed. Same with Theta’s lesser gear. This way, they could keep pace with technology, without filling trash cans with their old DACs.
Fun fact: Theta’s first non-upgradable product was the Cobalt 307, which I designed—their first “disposable” DAC, at “only” $599 in 1992. Compare to today’s $99 Modi. Yeah. There is such a thing as progress.
Today’s “Stability” and Upgradability
“But…but today, digital audio isn’t stable!” some of you are protesting right now. “USB changes quite a bit, and there’s DSD, and standards for transmitting digital over WiFi and Bluetooth, and high-res music, and all that.”
Yes. But it’s still much more stable than the 1980s and early 1990s, when everything was changing. Today, most DACs have settled down into a comfortable model: inexpensive delta-sigma D/A conversion from AKM, Analog Devices, Crystal, ESS, TI, or Wolfson, coupled with a USB interface solution from C-Media, TI, Tenor, or XMOS, plus (perhaps) a SPDIF interface from AKM or Crystal…plus, of course, a power supply, analog output stage, and associated interface electronics. Sure, there are some outliers, but they’re usually at the scary end of the price spectrum.
Beyond that, let’s go through the sources of instability today:
- USB. Much more mature, but still improving. When Bifrost was introduced, things were a lot less stable. Today, we have robust, relatively good-sounding solutions. But, we suspect they will get better. So, yes, the ability to upgrade the USB input would be a plus.
- DSD. Yeah, it’s there, but the floodgates have not opened. In our opinion, best to concentrate on the 99.999%—that is, PCM—rather than a possible dead-end excursion a la HDCD. Still, if it ever becomes more than 1% of the market, sure, it would be good to have upgradability that would allow DSD decoding.
- WiFi Digital Audio. Definitely changing, and may be promising in the future. The potential for uncompressed transmission is definitely there, but the challenge is in the interface (joining WiFi networks means a UI that is computer-like in its capability and configurability, unlike Bluetooth.) In our opinion, best to sit this out while it matures a bit. Jumping in now would be kinda like backing a digital transmission method before SPDIF was established as a standard—with the downside of extreme software prowess and support requirements.
- Bluetooth Digital Audio. Definitely changing, but too compromised to jump into for the sake of convenience, in our opinion. None to date is uncompressed (not even AptX.) Much of this is highly integrated, single-chip solutions that deliver analog output (not a digital datastream). Some can get I2S out. New Bluetooth standards with higher speed may allow uncompressed transmission. Probably too early to build a standard “Bluetooth compatibility module” into a DAC—just use an external solution until robust, uncompressed, I2S out versions are available.
So, what does this all mean?
When we started on the design of Bifrost, the latter three weren’t really a concern. We were worried about the USB input changing over time, and, to a lesser extent, the DAC and analog section.
Which is what drove our decisions—and our caveat. When we introduced Bifrost, we told potential purchasers, “We’re not going to have a DAC of the Month club or anything like that. When real, meaningful changes come to USB or the DAC/analog stage, we’ll release an upgrade.”
Of course, we didn’t know when those real, meaningful changes would happen.
But, after Gungnir development, Mike started wondering what the Gungnir analog stage would sound like in the Bifrost. That led to the first possible upgrade.
And, early in 2013, C-Media laid the second one on us: the CM6631A USB input receiver chip.
Which meant, as we went into 2013, we knew…upgrades would soon happen.
But First, Let’s Talk About the Way To Do Upgrades, and The Way Not To
“Real, meaningful changes.” That’s an important phrase for any manufacturer thinking about doing upgrades. That’s the phrase that keeps you from being the Burger King of audio.
Remember: it’s not about making everyone like you. It’s about making some people love you. And you won’t achieve that if you offer everything to everyone, with no position on what is best.
For example, it would be relatively easy to offer Bifrost with a half-dozen different DAC/analog sections. Just pick the latest chip from AKM, Analog Devices, Crystal, ESS, and Wolfson, re-do the analog section to meet their specific requirements, reprogram the motherboard, and you’re off and running. Everyone could have the DAC of their choice! Have it your way!
- Everyone would also argue about which one was best.
- If there was no consensus, nobody would have any idea what to buy.
- If there was consensus, we’d be stuck with a ton of DAC boards that were impossible to move.
- We’d have to stock 12 different versions of Bifrost (6 D/A options, plus with or without USB.) Total disaster.
- We’d have to spend 6x the engineering time in development, or shortchange one or more implementations.
- We’d have to maintain records for all variations, so they’d be serviceable in the future. And maintain the records so they were up to date.
- People would want to order two, or three, or four, or five, or all 6 and swap them, which would be a disaster, since the motherboard would have to be programmed, and Bifrost wasn’t designed to be hot-swappable. Say hello to unqualified people disassembling the product and possibly hurting themselves—then say hello to a huge lawsuit.*
*This is absolutely no joke. There’s a reason that “No User Serviceable Parts Inside” is printed on the back of virtually every electronic product, and why we say, “If you aren’t an electronics professional, have us upgrade your product.” That reason is: we want to stay in business. If we ever hint that it’s OK to take apart a product that’s powered by an AC line cord, believe me, that email or forum post will be dug up by a forensic attorney and used to hang us out to dry, when someone decides to take apart their Valhalla 2…while it’s plugged in…and while they’re taking a bath.
No. Upgrades are not “have it your way.” They are a path. A path to a rational future where you’re helping mitigate the cost of buying a whole new product.
Which means, quite simply: pick your upgrades carefully, and keep them to a minimum.
That is sanity. The other path, less so.
On USB and Mohammed’s All-You-Can-Eat Sushi and Deli
When we got our first CM6631A USB receiver chips, we were both thrilled and cautious.
Thrilled, because there were some notable limitations with the older CM6631 that the original Bifrost USB board was based on—namely, lack of 24/176.4 support, and an extreme pickiness about the USB interface and cable quality. A tiny percentage of systems simply didn’t like the CM6631 interface—on the order of 0.3%—but, in the time-honored tradition of Murphy’s Law, of course 1500% of that 0.3% happened to be forum members who complained about the problem.
Cautious, because we knew that “new stuff” isn’t always “better stuff.” C-Media promised 24/176.4 support, as well as support for even higher bitrates (24/384, specifically), but were mum on the subject of there ever being a problem with the CM6631 with some USB interfaces.
So we sent a few out to our PCB assembly house to have built up on the current-generation USB board. The chips were almost entirely pin-compatible, so that was the easiest way to see what they’d do.
Or, more accurately, Mike and Dave did it. Dave may have even soldered some of those 100-pin QFNs himself. No thanks, not for me.
In any case, we soon had a handful of USB boards with the CM6631A on them, happily programmed and running away.
“Here you go,” Mike said, handing me a little foil bag one day. “Try it out.”
“What is it?”
“The new USB receiver.”
“How’s it sound?” I asked. The first-gen USB board was pretty darn good by USB standards, but it was never any great shakes by good SPDIF standards—hence all our trash-talking of USB when Bifrost first launched.
“You tell me,” Mike deadpanned.
(He does that a lot. He wants to know what you think, not get confirmation of his own notions.)
So, I went home, installed the board, and plugged it into my most notorious source—an older Apple MacBook that would sometimes not play nice with the older USB board. It fired right up, showed all the sampling rates, and played fine.
Pretty damn fine, actually, I thought, after a while.
And, it wasn’t glitching like the earlier board.
Hmm. Maybe we had something here.
I tried it on a couple of different computer sources, and they all worked fine. I played it for Rina. Her face lit up. “This USB actually sounds good.”
“I thought it sounded better,” I told her. “But I was more worried about the functional side.”
“No. This is good. Really good.”
Still probably not as good as SPDIF, I thought.
But after switching the Bifrost back to optical input, I had to scratch that thought. It wasn’t the same as SPDIF, but it was definitely not worse…black and white had been turned into shades of gray.
Aside: Mike still prefers SPDIF. I listen mainly via USB, though. To me, SPDIF still sounds a bit more natural, but slightly smeared. USB sounds more precise, but slightly crispy. Yes, I’ve never been good at audiophile adjectives. I’d be toast if I got a job marketing audio jewelry.
“So when will we be selling this?” I asked Mike, the next day.
“You liked it, huh?”
“Yeah, lots better!” I said. “I think it might be better than SPDIF in some ways.”
Mike recoiled in mock horror. Or maybe real horror. “Have you started listening to techno? Savoring Mohammed’s All-You-Can-Eat Barbecue Sushi and Deli? Drinking plastic-bottle Vons-brand tequila? Huffing Testors paint?”
“No, I’m serious.”
Mike shook his head. “Ohh…kay. Remind me to pick the restaurant next time we go out.”
“And make sure there’s some real music on the computer when we go to shows.”
“Hell, maybe I should just buy you some Beats right now. USB? Better? For audio?”
“It’s shades of gray,” I told Mike.
Mike shook his head again. I don’t think he’ll ever really like USB, but he’s largely stopped complaining about it. And that says a lot.
On the Voodoo of Analog
At the same time as the USB Gen 2 board, we were investigating another upgrade…this one pulled straight from Gungnir. The idea was simple: if the Gungnir sounded so much better than Bifrost, what would a Gungnir analog stage sound like in Bifrost?
This wasn’t as simple as dropping it in, however. Gungnir’s stage was designed for the high +/- 24V rails in Gungnir. Not for the +/- 15V rails in Bifrost. I had to change every value on the board to get it optimized for its new home.
But beyond that, you guessed it—the Uber Analog board is, pretty much exactly a Gungnir board. It has a much more sophisticated discrete topology, and a DC servo to eliminate the coupling caps in the original Bifrost analog stage.
But it’s the same DAC—the AK4399.
Why not a different DAC? Well, we hadn’t found any we liked better than the 4399. It’s that simple. And, to this day, we still like the AK4399 and AK4396 better than the newcomers—they are unique in their implementation of switched-capacitor filtering to lower high-frequency noise, which seems to give them a more natural quality than other DAC options.
(Or it could be all in our heads. The voices, the voices!)
If you’d told me that our first DAC/analog upgrade would use exactly the same DAC chip as our standard board, I would have laughed…
…until I heard the new Analog board.
It was a big step up. At least as big as the USB upgrade. Bifrost was no slouch in standard form, but with the new Uber board, it was in a different class.
From there, the hard part began.
So, why “USB Gen 2” and “Bifrost Uber?” Simple: to differentiate two very different products.
“USB Gen 2” was a product that could be used in both Bifrost and Gungnir. And it was only the first of a line of USB upgrades, we expected (yes, someday there will be a USB Gen 3…or maybe even something else to plug in there…but as to when, no idea…)
Bifrost Uber, because it really took Bifrost up in performance, and we wanted a way to differentiate it, without going to a whole new model. When, some day, there’s a new DAC/analog board, we’ll figure out what to call it then. Uber2? Who knows? But again, no idea on timing.
And…that’s also why you won’t be seeing a Bifrost 2 or Gungnir 2…perhaps ever. They’re so modular that they can keep going, with relevant updates, pretty much indefinitely. Or at least for a very, very long time.
The Logistics of Upgrades
You know how, when a new car model comes out, a whole bunch of people have to have it RIGHT NOW?
Yeah. That’s the problem with upgrades. When you announce an upgrade, and you already have a significant number of products in the field, the first months are gonna be crazy. Everyone’s going to want it, and want it RIGHT NOW. That could easily bring your service department to its knees—especially if everything all comes in at once. Then you’ll also have the fun of trying to explain to customers why it’s taking two weeks to turn it around.
Luckily, we already knew this from the Theta days. Unluckily, we also knew that our current systems would never survive the onslaught.
Which is one reason we completely rebuilt the website—throwing out it’s creaky old taped-together platform for a new, custom-from-the-ground-up development that was tailored exactly to our needs.
This change allowed us to put in place a queuing system to help manage the updates. With the Schiit Upgrade Queuing System, customers could buy the upgrades, and we could tell them when to send them in so we could guarantee fast turnaround. It also kept everyone in the loop—sending automated emails when the product was received, and when it was shipped out again.
But then, there was the question of self-upgrades. As I mentioned before, the “No User Serviceable Parts Inside” disclaimer was no joke. We seriously considered not allowing self-upgrades.
But that would inconvenience people who really could do it themselves—and it would especially inconvenience international customers, who would have to hope their distributor would be able to do it.
So, a week or so before we announced the upgrades, we reached a compromise: purchasers could choose to self-install, but only if they said they were a “professional electronics technician.” The copy is still in place to this day on the upgrade product pages.
“They’re gonna break them,” Alex said, before we launched.
“We’ll see,” I told him.
“They’re gonna screw up their Bifrosts, and we’ll have to fix them.”
“Let’s see how it goes.” I reassured.
Alex shook his head, but agreed, yes, it was worth a try.
So how did it work?
Well, it was a damn good thing we had the queuing system in place—early orders could have easily swamped us, and even with the system in place, we were sometimes doing 20 upgrades a day.
And, as far as self-install goes, it turned out to be a non-problem. Maybe one or two boards were destroyed by careless individuals, or maybe they just died in transit—yes, even after testing, it could happen. But there wasn’t a flood of bad boards, or bad Bifrosts. So, our compromise ended up being the right thing, in the end.
And those early comments about, “Well, you’ll probably never do upgrades anyway?” Yeah, those ended.
Instead, some people said, “Wait, that’s not fair! You mean you’re gonna milk us every year, and have us re-buy the DAC we just bought?”
Yeah. As if throwing away an entire DAC is a better solution. And as if an upgrade suddenly makes their current Bifrost or Gungnir unusable.
What was I saying about some audiophiles being a morose bunch?