Chapter 31:
R&D Sometimes Means, “Try It, See If It Works”
Sometimes R&D is extremely focused. You know exactly what you’re shooting for, and you apply the collective smarts you have in a concerted effort to hit—or exceed—the mark.
This was certainly the case with the previous two products I talked about (Ragnarok and Asgard 2.) We knew (pretty much) what we wanted, and set about to do achieve it. In the case of Ragnarok, it was an all-in deal with me, Mike, and Dave all contributing—and a long and winding road to the end game. In the case of Asgard 2, it was just me—and, as I said before, R&D-wise, the product was a gimme.
But I strongly believe that R&D shouldn’t always be so focused. There’s value in making sure your engineering staff has time to play with crazy ideas.
How much value? Consider this:
Without play-time, Mike would have never put together the micro-DAC that became the first Modi. Which also would have meant:
- There’d be no reason to develop Magni, ever
- We may never have investigated low-cost products at all
- We may never have discovered how cost-efficient all-steel chassis could be
- Which would mean that we also wouldn’t have Vali, Sys, Loki, or Wyrd
- Our lowest-cost product today might still be Asgard (because why bother updating Asgard in the face of rising sales?)
- Valhalla 2 and Lyr 2 might not exist, either, for the same reason
So what would Schiit look like today, if Mike hadn’t had the wild butthair to develop a DAC orders of magnitude less expensive than any he’d ever done?
Well, we’d certainly be a lot smaller. Magni, Modi, and the sub-$150 products are the majority of our sales.
But at the same time, ironically, we probably wouldn’t be much farther along, if any, on Ragnarok and Yggdrasil—the low-cost product line barely impacted Dave’s software development time at all.
So, it would easily be possible for us to be, say, half the size. Maybe still at the Schiithole in Newhall. And still not have Ragnarok and Yggdrasil out.
Give your engineers some playtime. It pays off.
Those Tempting Tubes
Sometimes it’s funny what sets off those “I wanna play” moments. In the case of what would become Vali, it was eBay.
Yes, eBay.
In this case, it was my fault. I keep an eye open for bulk tube deals. We kinda have to. Valhalla 2 and Lyr 2 production chew through an incredible amount of tubes. And good tubes are getting scarcer.
Why? Simply because they aren’t making any more 6N1P or 6BZ7 tubes, just to name a couple of NOS tubes we use. And, while there are some good new-production tubes, they tend to be eye-wateringly expensive. So, we prefer to work with NOS tubes, at least as long as we can.
Note: I’m not worried about Valhalla 2 and Lyr 2 tubes, at least for a couple of years. We have some good suppliers we’re working with, and we have yet to be held up in production for lack of tubes.
Anyway, my search for bulk tubes sometimes takes me to tube resellers in Russia, sometimes to surplus warehouses where piles of tubes are forgotten for decades, and sometimes, yes, to eBay.
And what came up—what started the whole Vali thing—was an incredible bulk of NOS Jan Raytheon 6088 tubes, at a very attractive price. I mean, so many tubes that it would take us through half a decade of production, even if the amp sold like Magni. Truly crazy numbers.
And those numbers—and that attractive price—got the wheels turning. What could we do with this? Could we make an amp with them?
An amp, maybe, at a near-Magni price?
There was only a whole buttload of problems with this scenario:
- 6088s were pentodes. Using pentodes in pentode mode for gain is pretty barfy. And you don’t know what the triode-strapped curves will look like until you run them on a curve tracer. Would they be linear enough to use?
- We’d never used subminiature tubes, so we had no idea how best to use them, nor any specifics of their care and feeding.
- Even if the thing worked, how would we run the high-voltage supply (because, as Mike says, running a tube from a low-voltage supply is for “children and amateurs.”)
- And, what about the output stage? What exactly could we bolt this tube to?
And it’s doubts like that which kept me from simply clicking “buy it now.” Because if I just jumped on it, we might end up with a whole bunch of useless tubes.
But I kept coming back to
all those tubes, at
that price.
After a couple of weeks, nobody else had jumped on them, probably because they had the same doubts.
I did some research—had anyone used the 6088 for audio? A couple of DIY projects popped up, but they were the most basic and simplistic things imaginable—nothing that would be able to drive a wide range of headphones, nothing that could be sold as a commercial product.
But still, those tubes…
Screw it. I contacted the seller and purchased 100 tubes just to play around with. In the process, they confirmed that they had 5x the amount of tubes they had listed on eBay actually available—a truly eye-popping number.
Which meant if I could do something with them, then we could have a real winner on our hands.
The Road to Vali
The tubes came in a couple of days, and I set up a quick breadboard circuit to see how they performed. I was just interested in the basics:
- What plate voltage did they run best at?
- What plate load did they like for lowest distortion?
- What was the distortion like in triode mode?
- What did the operating current and heater current look like?
Why these basics? Because, based on these measurements, I could make a go/no-go decision on purchasing the tubes. Or so I thought.
It was interesting, running the early tests on those tubes. They were different than any others we had ever worked with. How so?
- Maximum plate voltage was only 60V—most tubes are in the hundreds of volts
- Plate current was 750uA-1.5mA for typical loads—much lower than most tubes
- The cathode was directly heated, rather than indirectly, like most tubes we’ve worked with
- It only needed 1.25V heater voltage and 20mA heater current—again, far lower than the 6V and hundreds or thousands of mA for most tubes (wonder why tubes run hot? Simple—they have a heater in them that usually dissipates a few watts of power.)
So, how’d they do?
Not so hot, at first. Nearly 1.5% distortion at the first plate voltage and load I tried. But, by tweaking the plate voltage and load, I was able to chart where the tube was the happiest. THD was still high, by our current standards—0.3% or so—but it was mainly second-harmonic distortion, and the distortion profile was nice, with 3
rd 20dB down and 4
th almost at the noise floor.
But—0.3% was still pretty high. How would it do, with an output stage bolted to it? What would it actually sound like?
That was beyond what the breadboard would tell me. I needed to lay out a board, and see how the amp really would do.
Laying it out on a Magni-sized board would be easy…except for the fact that it was a totally different amplifier, with radically different voltage requirements. (Remember I mentioned those high voltages for the tube? 60V is pretty low in the tube world, but it’s still a far cry from the +/-15V we were running in Magni.
At the same time, I wouldn’t want to run a solid-state output stage at 60V. That’s pretty, ah, adventurous, especially since the standard TRS headphone jack shorts the output every time it’s connected or disconnected.
And, we needed a regulated heater voltage, too, at 1.25V.
Oh yeah, and 5V for the relay.
And it would be ideal if we could get all those voltages from Magni’s standard 16VAC wall-wart.
Sounds impossible, right? Actually, far from it. With AC input, you can run a voltage-quadrupler and easily get 60V after regulation for the tube. Half of that circuit gives you DC voltages that can be regulated to 30V. And the 1.25V and 5V requirements are low enough that you can bring them down from the standard rectified output.
But (you knew this was coming, right?)…
But there’s always a but. Voltage quaduplers also aren’t very good for high-current output, and have significant ripple. But the half-wave rectification we’d used in Magni (effectively a voltage doubler) worked well enough. For a couple of mA going to tubes, run through a voltage regulator, it should be fine. Or at least that’s what I told myself.
I laid out the board in a couple of evenings. Everything fit really easily, including the tubes and an output stage that kinda started as something out of a Magni, but morphed into a pretty cool design that used a phase flipper to level-shift the output of the tube for DC coupling from the front-end, plus LED biasing for the output devices, which ran in Class AB.
Then it was just…send out the boards, and wait a week.
Two Big Surprises
The first Vali boards that came back weren’t perfect, but they weren’t bad. They needed a couple of hacks to smooth the input to the 30V regulator, and additional bypassing for the 1.25V regulator, but that was about it. After a couple of small tweaks, they were up and running, with the LED biasing simulating the glow of a traditional tube heater.*
*Subminiature tubes with 1.25V heaters don’t glow. They hardly run hotter than room temperature, in fact.
What was even better was the fact that it was running on the standard Magni wall-wart, without any signs of strain. Vali does draw a bit more current than Magni, so this was a welcome sign.**
**Thermal design really is the starting point for any amp. Get that wrong, and you’re in a world of hurt.
So, what did it sound like?
I took the prototype from the garage (where I still did most of the design and tweaking) back to the listening couch, where I kept Mjolnir and Gungnir.
Okay, I’ll admit it—I used Gungnir for those early first listens. Overkill, yes…
On first listen, I was a little surprised. Vali didn’t sound like I expected it to. I thought it’d be more tube-y, with the more typical euphonic colorations of an inexpensive tube design (rolled off highs, syrupy midrange, tubby bass—that kind of thing.) I had every reason to expect it to sound this way. The distortion profile suggested it. The single-supply output stage suggested it. The fact we were using (horrors!) coupling capacitors for the output suggested it.
But it didn’t.
In fact, it sounded pretty neutral and transparent. Maybe even a little bright. And, it sounded pretty darn good. I sat there for a while and just listened, which is usually a very good sign.
But was I hearing reality? Or was I just full of it? That’s always the designer’s dilemma—being too close to something, and losing perspective.
I gave it to Rina to have a listen. Her eyes widened. “Wow,” she said. “When do I get one?”
I also gave it to Mike. He listened for, like, 5 seconds in the shop, then picked up the prototype and put it in his bag.
“Hey!” I cried. “That’s the only one!”
“So build another,” Mike said.
“I will, but…I was listening to it!”
“Don’t be lazy,” Mike said, and left the building with it.
A few days later, he called me. “How many of these tubes are there?”
“A metric buttload,” I told him.
“Get them all. This is good. Really good.”
That’s what I’d thought, but it is good to have some confirmation. I contacted the eBay seller and cleaned up on the tubes.
Now, the only problem would be telling Alex he had to find more space—again—for the pallets of tubes that would be coming in. He was a lot more happy about it when I gave him a prototype Vali to play with.
And the early accolades kept coming. At a big head-fi meet, several of our golden-eared friends (including some who have given us, well, brutally honest feedback) proclaimed that it was better than 95% of the tube amps there. And people really flipped when we showed it at Can-Jam.
So, everyone’s happy, right?
The Catch
Nope. Of course not. There’s always a catch. And in Vali’s case, the catch is directly related to those great-sounding tubes: tube microphonics.
What are tube microphonics? They’re noise that’s generated from tapping or jarring a tube. Some tubes are pretty non-microphonic (especially the 6N1P we use in Valhalla 2), and some are very microphonic (like the tubes we use in Vali.)
In the case of Vali, microphonics sound like a little “ting” sound that takes a long time to decay. It’s like a delicate silver bell. It’s actually a neat sound…
…that is, if it isn’t interfering with your music.
And these tubes were microphonic enough to ring when you first turned the amp on, and when you plugged or unplugged headphones. We warned everyone about this, of course, but it wasn’t enough. It turned out that some amps were microphonic enough to be set off by typing on a keyboard, or simply rang all the time.
So what did we do? We started doing an extended burn-in on the Valis, and checking them when they were still warm, to weed out the self-ringers and over-ringers. We also had input from a very helpful customer, who came up with one idea we hadn’t thought of for reducing microphonics (specifically, damping the PC board itself, as well as using sorbothane pads under the tubes.) Those two changes have brought down Vali failures to fractions of a percent.
So What Did We Learn?
Take time to play. Even if that play comes from seeing a pile of tubes, and wondering, “What can we do with these?”
