Separate names with a comma.
Yep...I wrote a similar post in the Jot thread...although a little less colorful.
Schiit makes solid products, for a fair price, in the USA -- all good! That said, there are plenty of other products out there worth considering the merits of.
I say this owning 4 pieces of Schiit gear but some other gear as well...which provided me with better value for the given use case.
Mono amps = better imaging than stereo is pure nonsense. The only thing a pair of mono amps can give you over an equally powered stereo amp is the possibility of better channel separation. Everything else is equally possible in a stereo configuration.
2017, Chapter 12:
Third Time’s the Charm
It’s time for a whole new Magni. And the world may never be the same.
Melodramatic much? Perhaps a bit. But Magni 3 is radically different—and radically better—than any Magni that has ever come before. Although it might look the same, it’s a totally different amplifier. It’s no longer “entry level” in any way, shape, or form.
What do I mean?
Well, if you could get in a time machine, go back 5 years to the time when we were finishing up the design of the original Magni (for December release), and tell me about Magni 3, I’d say you were totally, 100%, barking-mad crazy.
I mean, you’d be telling me, “In 5 years, even with 5 years more inflation, you’ll be selling a Magni with 2X the power, 1/4 the distortion, 1/5 the noise, a gain switch, preamp outputs, an aluminum top, and an exotic, all-discrete, current-feedback gain stage with 6MHz bandwidth…for the same price.”
Yeah. Like, no. Likefergetit.com.
Hell, I didn’t think we could afford even gain switching on the original Magni (which is why it didn’t have a gain switch.)
Hell, Magni 2 was so razor-edge it lost out on the better gain stage, bigger transformer, preamp outputs, and aluminum top of Uber.
Hell, I never thought I’d be able to do a real exotic, current-feedback topology in Asgard…much less Magni.
But times change. And we learn things. And we get bigger, and better at making things.
And then Magni 3 happens.
Improving on Twos
I’ve mentioned before that improving on our second-generation products is really, really tough. The first generation of Magni, Asgard, Valhalla, and Lyr had some operational limitations that were easy to address (usually related to the lack of a gain switch, and/or preamp outputs). And, while were were addressing those limitations, it was easy to go back in and tweak the basic idea for higher performance.
And that worked, in general, really well. Better than I thought, actually. In fact, I have yet to figure out how to do a meaningful Asgard 3, Valhalla 3, or Lyr 3 without completely changing what the amp is. Each of these amps is pretty much at the limit of its topology, thermal, or budget envelope (or all 3). Making a more powerful Asgard 3 isn’t going to happen without fans and a bigger price tag (or a totally different chassis). Making a more powerful Valhalla 3 simply ain’t gonna happen, unless it becomes a hybrid…and then it’s Lyr. You see how this goes.
Of course, another way to do an Asgard 3 would be to make it a Class AB amplifier. That would allow higher output power, without the increase in heat.
But it would be an entirely different amplifier. One of the beauties of Asgard 2 is its purity. It’s a real, no-excuses, no-BS Class A power amp. And if you’ve fallen in love with Class A, it’s possible that only Class A may do.
Still, I thought, it might be worth trying to develop a Class AB Asgard 3. That would give me a chance to experiment with a new fully-complementary current-feedback topology. It might be interesting if it was something like 2.5 watts or so, and if we could move it to the new Jotunheim-style chassis (and maybe take Jotunheim cards, hmm…)
And that’s how Magni 3 started. As Asgard 3.
And yep, I have an Asgard 3 prototype. We took it all the way through two iterations, and even got a custom heatsink that sat on the PCB and would attach to the top chassis. This would allow us to run high current through the output stage and keep some claim to “Class A-ishness.”
Still, it didn’t feel right.
And the early prototypes didn’t even want to work. They were really, really squirrely. Not a huge surprise when you’re looking at gain stages that can amplify 10s of megahertz (current feedback is inherently insanely “fast.”) But the problem was less oscillation and more thermal. Without careful positioning of the biasing diodes, and the right value of emitter resistors in the diamond front end, the amp was a meltdown waiting to happen. Current would climb…and climb…and climb some more. What started at 40mA output stage current would go to 80, and 160, and 300, and then, well, boom.
But even that paled in comparison to the big problem. Because, even after a PCB rev that addressed the thermal instability, even after tweaking output current to 150mA and above, the Asgard 3 simply didn’t sound like a Class A amp.
Now, it sounded good, damn good. It had more power than Jotunheim’s single-ended outputs, and an overall warmer character. I liked it. Most everyone who heard it liked it.
But people who liked Class A didn’t think it was all that hot.
Yeah. Dead end.
I prepared to put the PCB on the shelf and forget about it. Because, for all my engineering angst, Asgard 2 was still selling just fine. It didn’t really need to be changed. This was, maybe, change for the sake of change, and therefore violating the “if it ain’t broke, don’t fix it,” mantra.
But then I paused. The gain stage worked. It measured really good. It sounded really, really good. It was, essentially, the single-ended Jotunheim that everyone would be asking for.
And…the gain stage actually had less parts than the current Magni 2.
So what if the single-ended Jotunheim everyone wanted turned out not to be $299, but $99?
When Jeep told its engineers they needed a new Wranger, I imagine they gave them some huge document full of specs to improve and features to implement, but I also imagine it was disbursed by a grizzled old DGAF engineer who glared at them and said, “Don’t even look at that ****, look at me. And listen to this. These are your orders: You Can’t **** This Up.”
When something’s working great, when it’s selling well, when people like it, that’s always the first order of business: You Can’t **** This Up.
And that’s how the Magni 3 project got a little scary.
Because, even though Magni 3 looks the same, it’s a fundamentally different product. All previous Magnis had been based on the well-known Lin amplifier topology. It’s literally the most-used discrete topology in the world. Its foibles are well-documented. Our implementation was rock-solid. Magnis were Maytag-repairman-style reliable. Nobody was asking for a new Magni.
In short, we had no burning reason to change. And the new current-feedback gain stage was, well, possibly a big bag of problems.
First was its thermal stability. Although that had been sorted in the Asgard 3 prototype, the Magni’s chassis had a lot less room for error.
Second was its inherently wide bandwidth. Earlier Magnis got away with a very simple input filter. How complex would the current-feedback Magni’s filter need to be?
Third was its bipolar inputs. Earlier Magnis used JFET inputs. These are less likely to rectify RF (like, from nearby cellphones and routers). The current-feedback stage used bipolar inputs, which could be more sensitive to RF in the environment.
And, on top of that, we’d only tested the current-feedback gain stage with an output stage biased deeply into Class A. What would it sound like, running rational Class AB standing current?
Aside: “Biased deeply into Class A” is a misnomer in more ways than one. Real Class A can’t go out of Class A. It runs maximum current all the time. It’s literally incapable of sourcing more current than its standing current. That’s how Asgard 2 works—it biases a single MOSFET into Class A. There’s no worries about matching devices or crossover regions, because there aren’t any at all. A high-bias Class AB amp isn’t really Class A, because it can go out of Class A…and because it inherently uses both NPN and PNP devices, which are not perfect matches. So there can be nonlinearities around the crossover region, as well as a significant nonlinearity when it passes out of Class A’s transconductance-doubled region.
There was only one answer to all these questions: build a prototype and see how it worked.
And that’s what I did. Dropping the Asgard 3 current-feedback stage into the Magni was a piece of cake. Like I said, it actually used less parts than the current Magni 2, let alone the Magni 2 Uber—even with the addition of a full input “T” filter. I started with the Magni 2 Uber architecture, with preamp outputs and adjustable voltage regulators, because at that point I didn’t know if this was going to be a replacement for Magni 2 Uber alone, or both Magnis.
Aside: I’d already started hoping the Magni “CFA” would be a replacement for both Magnis, both to simplify the line, and to really up the value equation. There’s no way we’d be able to price a single replacement higher than $99—that would be suicide. It remained to be seen, though, if we could fit everything in for that price. It turned out we had to lose only one thing. More on that later.
And the first prototypes…just worked.
Yep, they still required close coupling of biasing diodes to the outputs and careful selection of operating points on both the front-end diamond and the voltage gain stage. But they settled down…they were stable.
And, in an amazing turn, the Magni CFA prototype actually sounded better than the Asgard 3 prototype!
In fact, the Magni CFA prototype was damn near ideal, especially for an entry-level amp. Subjectively, it was a warmer-than-neutral amplifier with good dynamics and very good detail. I could see a whole heck of a lot of people being happy with it.
Aside: why would a Class AB implementation sound better than a high-bias “Class A” implementation? Well, Magni 3 is run closer to the “ideal Class AB operating point” than the high-bias Asgard 3 prototype, so perhaps that has something to do with it. But it could be a number of factors, including the fact that Magni 3 uses different output devices than the Asgard 3 prototype.
Fun Fact: the Asgard 3 outputs are now the drivers on Vidar.
Fun Fact 2: Magni 3 and Vidar are HIGHLY related. In fact, they’re the same basic topology. Vidar’s just a lot bigger. With microprocessor oversight.
But great sound and flawless operation are not 100% correlated. So, the next step was obvious: make more prototypes. I made several, for both internal and external listeners, and told everyone, “Use it with your cellphone sitting on top of it, use it near a router, stack books on it, let me know if it does anything weird or heats up or whatever.”
And, surprisingly, we had no operational issues. The input filter appeared to be working. The bipolar inputs weren’t a problem. I took the prototype to a couple of small meets and let people beat on it. It worked without a hiccup.
Yeah. Sometimes you get lucky. We’d just changed literally everything about the Magni, and it had gone without a hitch. But we were still in the “handful of prototypes” phase. How would it go in production?
And, even more importantly, what would it cost?
Alex and Nick vs the World
If you gave Alex (our Director of Operations) his wish, it would always be for less products and less options. We already have a broad product line. When you start throwing in variations like Ubers and Multibits and black chassis and B-stock, it can be overwhelming. Splitting the Magni into Magni 2 and Magni 2 Uber gave Alex a bit of a headache, because now he had to forecast the demand for both products and keep both in stock. He also had B-stock variations on two products, and the possibility of black chassis popping up from time to time. That’s a heck of a lot of variations to deal with. If there could be a single Magni 3, Alex would be very happy.
Similarly, Nick had to deal with the questions. “Which should I get? Magni 2 or Uber?” Even though we tried to make it as clear as possible on the website, and even though some potential customers had a clear need for Uber (like, if they had powered monitors and headphones), we’d still get asked this question all the time. Again, if Nick had a single Magni 3 to recommend, then his life would be a lot easier.
But easy isn’t always feasible. And that’s where the cost of Magni 3 got critical.
Because, with literally no stress, it would be simple to replace Magni 2 Uber with Magni 3 at $149, and leave Magni 2 alone at $99. Easy. Done deal. But…we’d still be stuck with two products. We’d still have to stock all the variations. We’d still have to answer questions.
So the question became: could we get the cost of production down to the point where we could have a single product—one with all the features of Uber, but at the previous non-Uber price?
In some ways, this was pretty simple. The cost of the extra jack for the preamp output, for example, is minimal. The circuitry to switch the preamp output with headphone insertion was literally no cost at all.
But the larger transformer (21VA vs 8VA) was more expensive, and we had to consider that. But even that could be absorbed, if we could do something about the elephant in the room: the chassis.
Sigh. It always seems to come down to the chassis.
Here’s the problem: the aluminum top on the Magni 2 Uber cost more than the painted top on the Magni 2. Worse, the machined aluminum knob was stupidly expensive for such a simple part (plus, it had to be attached with a screw, slowing production.) And, to add insult to injury, the overall cost of the chassis had been rising.
And that’s when I discovered stamping.
Previously, our chassis had been done with CNC machining. This is fine when you’re doing relatively low quantities, but we weren’t seeing much cost savings as the runs got bigger. I talked to a local stamper—the same one who saved our butts on the Vidar heatsink clip—and discovered that for our quantities, it would be significantly less expensive to tool up and stamp everything.
Still, we’d been burned by bad chassis before, and I didn’t want to take the plunge on an important new product.
So, I played it safe: I had the stamper tool up the Fulla 2, while we also placed an order for the conventional CNC Fulla 2 chassis, just in case. After a glitch with the silkscreening, which the stamper took care of posthaste, the Fulla 2 chassis from them looked better than the CNC ones. Not surprising, since tooled products can actually hold tighter tolerances.
With that experience under our belts, we decided to go ahead and move the Magni 3 (and most of our small products) to stamping. This is why we can do the Magni 3 at $99 and add an aluminum top to the Modi 2 without changing the price. This is why we can add new features like pressed-in feet. This is why we can make all the small products fit better and look better.
Of course, there’s no free lunch, so we had to lose one thing: the machined aluminum volume knob.
Yep. Sorry. It’s a goner. The knob costs too much and slows down production. So it’s gone. The new press-on knob is, however, larger in diameter than the old one, which contributes to an overall better feel.
We also re-curved the custom Alps potentiometer for better tracking at low levels and made a bunch of tiny tweaks to the circuit and chassis, from channel-independent DC servos (not shared on a dual opamp, like before) to optimized layout for both thermal and electrical characteristics.
We even took the time to hold back on shipping for extended production testing. Although we could have announced the Magni 3 in July, I decided that we should build at least a thousand of them and see how they went through sound check. We gave out some more for extended listening tests. And the Magni 3s performed without a hitch.
So, now we’re ready. And now, we’re really, really confident that we’ve reset the bar on entry-level headphone amps—to a level where they’re not really entry-level anymore.
Think about it. Now, you can get a fully-discrete, exotic-topology, current-feedback, all-linear-power-supply, 2-year-warranty, headphone amp and preamp that delivers 2W RMS per channel into 32 ohms for a 2-figure price tag.
Crazy. Or at least crazy value. And that is what gets us excited.
Rewind and Reflection
When we were getting ready to start testing the first production Magni 3s, I found a sealed plastic bag with the first production Magni in it—serial number 000001.
This was a product that was made in our old SchiitHole shop in Newhall in December 2012. It was amazing for its time—a powerful discrete headphone amp for $99. But it was only a headphone amp. Not a preamp. It didn’t even have a gain switch. The quality of the painted top looks laughable today. The Bourns potentiometer has noticeably worse tracking than the Alps pot we use today.
And the sound?
We handed Magni #000001 to Asa, one of our sound test technicians, who was listening to a stack of Magni 3s. He looked at it doubtfully, pursing his lips at the painted top and shaking his head at the lack of a gain switch. Still, gamely, he plugged it in and had a listen. His pained expression said it all.
And that’s when it struck me: the Magni 3 is a completely different product than the original Magni. About the only thing it shares is its exterior dimensions and PC board size. The topology is different. The feature set is different. The way it is made is different. The power supply is different. The power output is different. The noise and distortion specs are different. Although it may look the same, it’s in no way, shape or form a Magni.
So, like I said: third generation products aren’t evolutionary. They’re totally different.
In fact, you can probably think of Jotunheim as our first “third generation” product. It was a radical re-think of that form factor, and introduced the idea of a modular architecture. If you ever see an Asgard 3, Valhalla 3, or Lyr 3, it will probably be in the mold of Jotunheim.
Aside: before you get too excited, consider the huge problems in moving these products to the Jotunheim modular form-factor. Asgard 2 needs the U-shaped chassis to dissipate heat—there’s no way around it. Valhalla 2 has literally no power supplies compatible with Jotunheim’s drop-in cards. Lyr 2 is at its thermal limit, too, and its output stage means it has no room for cards. Things are never as simple as they seem.
And now we have Magni 3, our second “third gen” product. And we had to literally change everything about the Magni 2 in order to get there.
Of course, some of you are already asking, “What’s in store for the fourth generation?”
To which I ask, “After you change everything, where do you go?”
You take a different path, of course. But where it leads remains to be seen.
Excellent !! .. would never have guessed a Magni 3 ... quite amazing ! Congratulations !
Agree completely. I read somewhere that something less than 30db of stereo separation is all you need for a proper stereo effect. Everything over that is just specsmanship.
@Jason Stoddard your products always seem to fall right into impulse buy territory for me. Well, I just ordered one of what will be many multitudes of Magni 3s sold. Very impressive release, I'm sure impressive sounding too
Best Schiit I've heard all day! Any internal pics?
Great release. Well done.
A great start for everyone in this sport and a huge lift for affordable high-end.
Up to the 26th and maybe something for us old farts who have Mjolnir 2s, Ragnaroks and Yggdrassils
This is great news for patient customers like me. I was going to buy Vali 2 but the wait was worth!
Yep, as soon as I get them. I'm awaiting internals, stacks, and beauty shots. We need a new stack shot, since Modi 2 now has an aluminum top as well.
Well, my Magni 2 Uber isn't even 7 months old, but I just ordered a Magni 3. Why not?
I thought there was an opinion in this thread that the world doesn't need more headphone amps...
I created the Magni 3 thread so people can come and leave their impressions, reviews, comparisons and else
That's why there's one less now.