2022, Chapter 10
Troublesome Thoughts
Or, Curing Tube Nervosa, Part 2
Lyr+ started with a simple, passing thought:
can I put a relay ladder volume control into a desktop sized headphone amp?
To an engineer, passing thoughts can be troublesome thoughts. Because they might take you down some long, dark, and crazy roads. Especially when those thoughts are about a product that’s already, ah, volumetrically challenged.
I mean, hell, I already
ran out of space on the top of the board of Lyr 3—the output stage is all on the bottom. Where the heck would you put a relay ladder on it, even if it was only 6 or 7 relays?
And the microprocessor? Where would it go? Because you would need a microprocessor to run the ladder.
Aaaaand, another thing: adding a relay volume control to Lyr 3 is a neat thought, but there’s no pressing need for it. Lyr 3 already uses a very good Alps RK27 volume pot. It’s a very good, very versatile amp. Hell, I have one on my desk in Corpus Christi.
Aaaaaaaand that’s why that thought about tweaking Lyr 3 stayed idle for a good long time.
Until I had a very, very evil thought…
Crappy photo of Lyr+ at the Texas Audio Roundup. We'll get you some real photos shortly.
Where Is Your God?
If Lyr 3 has any issue, it’s only that it needs a good tube to run, and good tubes are getting harder to get.
I mean, Lyr 3 would be
pretty much perfect if it could
switch seamlessly from tube to solid state and back again. Not with anything external like LISST, but built-in. So it could be
a great solid state amp you could try tubes in—or a
great tube amp that always had a solid-state fallback mode…
…and with that thought, the earlier musing about doing a relay ladder hit hard, and suddenly everything was right there:
…well hell we could actually build in a tube sensing circuit and switch over to depletion mode MOSFETs automatically, and we could do a relay ladder if we didn’t have a card, and we didn’t really need a card because we had Modius, and since we had a relay ladder and microprocessor it could also have remote control, and holy moly this could be the most advanced desktop-sized headphone amp and preamp we ever did…
BOOM.
I mean, I literally saw everything in my head. It was easy!
Well, except the part where nothing fit. And except the other part about this being a super-crazy, super-complex idea. And except the other other part about there being a ton of details about putting tubes and MOSFETs together (but separately) that weren’t so obvious in the beginning, but became super painful as I got into the detailed development.
So. Yeah. Pretty much another crazy idea from us…that took, ah, a bit more development to realize.
But think about it.
I mean,
really think about it: a tube amp that would just
keep running if you pulled the tube out.
How fun would that be at shows? Pull the tube out, click, it keeps running, you grin maniacally and cackle. “Where is your god?” you ask. People gasp. It’s a whole new world.
Aside: in reality, people would probably just be confused. And think you were insane.
But still, the idea stuck:
a tube amp that never needed a tube—but still had pure tube gain when you wanted it.
That would be a real game-changer. While other companies had messed around with tube amps that allowed rolling different tube types, and we had made solid-state tube replacements in LISST, and while we’ve given people options for tube or solid state in Freya and Freya+ by literally swapping out the entire active stage, from tube to solid state buffer, and while some companies had played around with the same idea in other products, nobody had ever really conceived of a
truly protean, tube/solid-state stage
using the same topology…literally swapping out tubes for MOSFETs in real time.
At the time, I didn’t bother to wonder why.
After I got into the development of Lyr+, though, I quickly found out.
Tubes and MOSFETs are Not The Same, Durr
“Well, yeah, duh, why do you even have to say it,” the smarter engineers are asking.
Here’s why: so many lazy companies have fed the audio public the same oversimplifications for so long—you know, “MOSFETs are like tubes, JFETs are like tubes, ooooohhhhh, let’s imply you want to avoid the EEEEEEVVVVIIIL current-driven BJTs like the plague (even though they have significant advantages and MOSFETs and JFETs both have giant warty boil-like problems on their respective butts, just like everything else),” that now it’s common to hear the assertion that “MOSFETs and JFETs are like tubes” all over the jabbertubes, even though it’s 100% bull.
Am I salty? A bit.
Here’s the pure, unadulterated truth: MOSFETs, JFETs, BJTs, and tubes are all very different. All have their strengths and weaknesses. But they. Are. Not. The. Same. None of them. Period.
So swapping out a MOSFET for a tube, by intartube logic, is
EZPZ, no big deal, go for it, brah, ya gonna be fine.
In reality, it’s a
holy-schiit I really thought those were in any way similar? moment.
Let’s break it down, for those of you who have been drinking at the “JFETs and MOSFETs = tubes” firehose for wayyy to long:
Tubes. These really exist in their own world. They are the only device that literally has
no physical connection between their grid, anode, and cathode. I mean, like duh, these terminals are in a
literal vacuum. JFETs, MOSFETs, and BJTs are all made on the same substrate and have a physical connection. Tubes are also the only device that require separate “activation terminals,” AKA, the heater, to create an electron cloud around the cathode to enable current flow. But, even though tubes have no physical connections internally, that doesn’t mean they can’t crap out a ****ton of grid current when they are coming up to temp or shutting down.
So even with no physical connection, it can behave as if it’s connected. Yes. Mind blown up.
JFETs. These have a high input impedance thanks to their inherent reverse-biased PN junction, similar to a tube’s theoretically infinite input impedance (other than Miller capacitance, yeah, and the previously mentioned transitory grid current, and other real-world, weirdo/not-ideal stuff, eesh), and they start running current as soon as you apply a voltage across their drain and source, but they are really quite different. JFETs can only handle a piddly few volts when compared to tubes (50V for the best of the current crop, vs hundreds of volts for many tubes). They also don’t have heaters. They also have very pentode-like, rather than triode-like, curves. Soooooo very different—not a solid state tube at all. Sorry you’ve been lied to—er, I mean, oversimplified to—your entire life.
MOSFETs. Like tubes and JFETs, MOSFETs also have a high input impedance, thanks to the gate being insulated from the rest of the semiconductor structure. But they also have significant input capacitance, also due to the same topology. Also like tubes, its common to find MOSFETs that can take hundreds of volts. However, MOSFETs are divided into two types—enhancement mode and depletion mode. Enhancement means you need several volts at the gate to get any current to flow, which is completely unlike a tube. Depletion means it’ll flow current immediately with no volts on the gate, similar to a tube. Sounds most similar to a tube? Yeah, except for that input capacitance, pentode-like curves, need for gate protection (too many volts = boom), completely different transconductance, and other MOSFET vagaries. For the TL;DR crowd: NO STILL NOT LIKE TOOBS.
BJTs. The first solid-state gain devices were bipolar junction transistors, which live in a world of their own. A tube Upside Down, so to speak. Unlike tubes, they need current to run, they need more than 0.6V at the base to turn them on, and can have very low effective input impedance, especially for power devices run hard into the beta droop region. Sounds terrible? Well, they are the only predictable device out there…you know when it will turn on, PNP complements are readily available (and kinda-sorta-similar-ish to the NPNs), it has very high inherent gain, and for many applications you can treat it like a voltage-run device…which is why they are still popular today.
Why the dissertation? Because Lyr+ is trying to sub out a tube for a MOSFET
in real time, and it
also uses BJTs in the Coherence and output stages. And to head off any questions about why we don’t use JFETs: because at 200V, they go boom.
Aside: I love JFETs. We use tons of them. You just aren’t gonna replace a tube with a currently available JFET. Period. Sorry. It’s MOSFETs for high voltage stuff, at least in this universe.
And the difference between tubes and depletion MOSFETs led to the first issue in Lyr+ development: different gain structure, different operating point.
Simple Concepts, Messy Execution (AKA “Argh”)
Let’s dive right into the development here. The concepts behind Lyr+—elevate the top end of the mid-size, desktop-friendly product line with relay volume, and make it a seamless tube hybrid or solid-state-only amp—are not super complicated.
But, like many things, it’s getting the ideas to work that’s painful.
The relay ladder, sure, yeah, sounds easy. And if you throw out the card, you have the space, right? Well, yeah, except then you need a microprocessor to run the relays. Oh yeah, and you need a microprocessor with ADC so you can monitor whether or not a tube is connected and switch to SS if not. And then you need more relays to switch the MOSFETs and tube gain. And protection for the MOSFETs. And DC sensing and protection in case there’s a glitch in the matrix when everything switches over. And power supplies for all of this, including new power supplies for the microprocessor and the relays.
Aside: yes, a separate dedicated, regulated supply for the relays. Because unregulated DC, when used to power a relay, can couple to the signal, degrading the signal to noise ratio. Yes. Seriously.
All that above is a buttload of stuff. Even without the card in place, the early layouts of Lyr+ were painful. Lots of readjustment had to be made. Especially if we wanted to keep the over-the-top double-transformer arrangement with a 24VA transformer just for the high voltage rails, and a 48VA transformer for the output stage. (And yes, we did want to keep that in place, and we did. That’s why Lyr+ weighs a ton, just like Lyr 3.)
But space was only the start of the problem. The big problem is that
MOSFETs aren’t tubes.
As in, when everything switched, things got weird. Much weirder than I expected. Now, the switching itself is bad enough, a carefully-orchestrated sequence of events to make sure the tube, the MOSFETs, the outputs, and the relays are all happy. But the problems persisted after the switch.
As in, the entire operating point of the amp changed. Oh, and the gain was significantly different as well.
What the above two statements mean, in English, is that we would have had to choose one ideal operating point, and let the other be, ahem, less than ideal. As in, in tube mode it would run nicely, with the correct output current, etc, but in MOSFET mode it would run cold. Or vice-versa. In addition, it would be louder when it switched to MOSFET mode.
Argh.
With a major unspoken goal of the design being “let’s keep the tube and MOSFET modes as similar as we can,” this just wasn’t gonna work.
So why were things not matching up? Several reasons:
- The way Lyr+ switches between low and high gain changed the operating point significantly. This is an oddity of our Coherence™ gain stage. It wasn’t a problem in Lyr 3, because we compromised by using a coupling capacitor in the feedback network. I wanted to eliminate that cap in Lyr+, which caused the operating point change.
- Further compounding the operating point change was the inherent difference between MOSFETs and tubes. Tube curves and MOSFET curves don’t match. So the voltage at the cathode didn’t necessarily match the voltage at the source. And the higher gain of the MOSFET meant it needed a bit more resistance in the source to match up better.
- Because of the above two differences, the amount of current running through the tube and the MOSFET was significantly different, so the amount of current running through the BJT part of the Coherence stage was different…so the amount of current running through the output stage was different. We needed to match the operational points much better.
So how did we solve it? Different MOSFETs, additional switchable source resistance, and a couple more relays. In the end, Lyr+ runs very similarly in all modes and at all gains.
But that wasn’t the end. One of the things I wanted to do was to be able to pull out the tube and have the amp switch seamlessly over to solid state. That would be quite a trick, right?
Yes. It would be. Except, no matter how we did the firmware, we never got it quite tuned up enough to catch the transition seamlessly. In retrospect, this is a “duh,” moment, because it’s really hard to detect the absence of a tube that quickly.
So, yeah, I’m sorry to report that we can’t recommend pulling the tube out of Lyr+ while it’s operating…you know, like every other tube amp out there.
But, unlike every other tube amp, you can turn it off, yank the tube, turn it back on…and continue happily to listen to your music!
Naming, Cosmetics and Other Frivolities
Some of you are skipping right over all the engineer-ese above and are reacting to the name of the product, the name of the tech, the cosmetics, or the fact we’ve included a remote control. So let’s dive into those.
The name. Why Lyr+ and not Lyr 4? It’s simple. We don’t like the number 4. Mike has never liked it. He changed the Theta Gen 3 to Gen 5 on purpose. I thought he was kinda nuts, until we had all the trouble with the Fulla 4. Boardhouse issues, end-of-the-line AKM issues, delays, price increases, firmware glitches…you name it, it had it. You notice it went to Fulla E real fast? Yeah. Nothing we make is ever gonna have a 4 on it ever again.
The technology name, aka Fusion Architecture™. Why did we inflict another buzzword bingo thing on you guys? Same reason we did Autonomy Architecture™ on Bifrost 2. Because it needs to be called out. This is the
only seamless-switching, same-topology, tube-or-solid-state amp on the planet. It’s also our most advanced desktop-sized product, ever. So yeah. The seamless fusion of tube and solid state needed a name.
The cosmetics. Big change, isn’t it? Yes. Lyr+ is our first amp designed to be used as tube or solid state—and to be comfortable being solid state forever, if that’s what you want. Hence the new multi-piece chassis design that includes a deep “heat funnel” to keep tube heat out of the chassis, and a “tube cap” to finish off the design (and keep dust out) if you’re not using a tube. It’s also the first desktop headphone amp and preamp to use soft-touch aluminum pushbuttons, just like Loki Max.
The remote. Why a remote on a desktop product? For some—including me—it summons images of 1980s douchebags using their remote control on their DIN-sized Alpine car stereo, by holding it a few inches away from the faceplate. Well, Lyr+ is not just a desktop product. It’s also a great preamp. And we include a remote with our Saga preamps, which are the same size as Lyr+. Aaaaaaaand…once we have a microprocessor and relay attenuator in there, adding a remote costs very little, so why not?
What Does This Mean for the End of Tube World As We Know It?
A smart reader is stroking their chin and going, “Hmm, interesting…what does this Fusion Architecture mean for other tube products? Are we looking at a whole line of protean tube and solid-state gear from Schiit to help solve the tube crisis?
Well…
we’ll see.
I’m being cagey, not because I want to be an asshole, but because this whole Fusion Architecture thing gets a lot more complicated on multi-tube products. Especially at high rail voltages, and especially if you want to switch high volt rails and heaters and stuff. (Ask me how I know.)
I mean, you may have noticed that we aren’t trying to shut off the tube heaters or high voltage rail in Lyr+. After all, it’s only a single tube, and if you want to preserve its lifespan, you can simply remove it when you’re not using it. Heck, we gave you a cover for that exact purpose.
Which means Fusion Architecture, as it is, isn’t just a cut-and-paste to take it across to, say, Freya+. Hell, getting relays that are rated to switch 300V rails is nearly impossible. And if you add in, say, switching heaters (like, hey, all of them, or only the voltage gain, or only the current gain, or either, or what), and you add in the additional bizarrity of cascaded operating points (between two tube sections), things get pretty hairy pretty fast.
And, to be clear, we have a great solution already with Freya+ and Freya Noval—a tube or solid state solution that only turns on the tubes when you use them. It may not be as conceptually elegant as Fusion Architecture, but it’s well-proven and very reliable. And you can use LISST in Freya+. So I think we have your “With or without tubes” covered for the Freya line.
So what’s coming? Sorry, I don’t know.
What I do know is that Lyr+ is the most advanced desktop-sized headphone amp we’ve ever done—with relay volume for perfect channel matching, plus all-new, completely unique Fusion Architecture so you can use tubes or solid state seamlessly. Pair it up with Modius or Bifrost 2/64 and you’re set for an insane, insane desktop.
I hope you enjoy!
