[golfbravobravo]

As I recall PayPal recently changed so I am pretty cautious when using it to accept funds.

Do you have any details?

 

[Paladin79]

Do you have any details?

https://www.cnn.com/2021/11/09/success/payment-app-tax-reporting-feseries/index.html

 

[Ohno Schiitizen]

2023, Chapter 12
The Ultimate Heresy?


While I was tweaking on Magni+, trying to get the maximum performance out of its discrete stage, I hit a wall.

As in, it just wasn’t gonna get better. Not unless I had a bigger budget and more space to add the boosted power supply and linearized front end I knew I needed. And so, frustrated, dreaming of something with a bigger chassis and less cost constraints, I muttered, “Gawd, I wish we had something a bit more fancy so I can do this super-duper Magni.”

Tyler overheard. And looked confused. “What do you mean, more fancy?”

“I mean, I need more budget for the additional parts. And I need more space for a power supply. If I had that, I could make Magni all it could be.”

Again, Tyler looked at me like I was nuts.

“You mean…like Magnius?”

And in that moment, I sat stunned. I’d totally forgotten we made a Magnius. It was that…forgettable.

I burst out laughing.


Not Because it’s Hard

I mean, I did introduce Magnius with a chapter entitled, “Not Because it’s Hard,” as an inversion of JFK’s famous line about going to the moon not because it’s easy, but because it’s hard…and as a jab at everyone who thought a high-measuring balanced amp was difficult to do and had to cost a lot of money.

And Magnius was easy. I mean, it’s the same story as all high-measuring IC-based amps: pick some of the best measuring chips, put ‘em in a box, and away you go.

It was so easy that I didn’t want to do it.

And it was so easy, that just a few years later…I’d totally forgotten we made it!

Aside: now, this isn’t to say that it’s a bad amp. It’s a perfectly fine amp. But designing it was too easy, and it was a trip into a cul-de-sac of me-too design, and it didn’t make me super happy. So I retconned the whole thing out of my mind. ​

Sooooooo, getting back to the super-Magni I wanted to make: that was different. That was a fully discrete design. One that would be really interesting to see how far we could go, and still keep it super-affordable. That was interesting. And that, I realized, might make a perfect Magnius 2.

Except for one thing: Magnius was balanced.

This imagined super-Magni? Single-ended.

Hmmmmmm. Would that fly? I wasn’t so sure.


Balancing the Equation

Now, making the Magnius 2 a single-ended amp would have a lot of benefits, most notably:

• Much higher power output from the single-ended headphone jack
• Much higher SE performance (than taking one phase of a differential amp)

But it would also mean dropping the balanced inputs and outputs, as well as the balanced 4-pin Neutrik headphone jack.

I mean sure, we could do fake balanced. We could just leave the connectors on there and run them single-ended. But that’s a horribly cheap and lazy and cynical way to do things, and it could cause incompatibilities with other balanced gear.

Aside: I did write a chapter on this, FWIW.​

Alternately, we could drop the balanced connectors. But that would be pretty silly when you could stack this product on a Modius, which had balanced outputs. And it would change the chassis, both top and bottom, which would mean new metal, and we had stock of Magnius chassis, and it would be best to simply use it.

So the question became: could I retain the balanced I/O—and make it meaningful—on a single-ended power amp?

Now, yes, retaining the balanced analog input and preamp output really wasn’t a huge issue. We could sum and difference the outputs, just like any amount of pro (and high-end gear) does, the same that Lokius does, the same that Loki Max does. That would provide the benefits of balanced transmission, for real, no fooling.

But the 4-pin XLR for the headphone?

Yeah, some handwaving can be done about “separate ground returns,” but that’s handwaving. There’s not much “there” there.

But there are also other things you can do.

Like error correction, applied via the negative terminal of the balanced output. As in, you compare the input to the output and apply a corrective difference. Note that this isn’t feedback. This is I-O = E , where E is a difference signal, just the error component. Feedback feeds back the full O (output) signal to the input.

What’s more, error correction could potentially improve the performance of the amp. I was shooting for near state-of-the-art performance with the upgraded Magni single-ended discrete stage, but this had the potential to push performance even higher.

And, I realized, this error correction would only be applied through the XLR output, because it needed separate negative terminals. The ¼” TRS would have no error correction, due to its shared grounds. So you could compare a “standard” and “error corrected” output.

All of that was pretty exciting, so I proceeded with a prototype…

…and immediately ran into problems.


Desperation is the Mother of Invention

Here’s the thing. A lot of what I wanted to do was fairly straightforward. The main goal was to improve the performance of Magni+’s discrete stage significantly, without increasing loop gain. As in, it would stay a low-feedback design. It would just get more linear.

I had two strategies for making it more linear:

• Massive boost supply. Magni runs on +/-16V rails. Which is pretty solid. But it does run into issues with the input signal being a significant proportion of those rails. As the voltage across the input devices varies, the gain of the devices changes. Which causes distortion. In fact, that’s the primary cause of distortion in the Magni+ stage. However, if I ran it on +/-30V rails, that would extend the linear region of operation and lower distortion. So I segmented the supply so the output stage still ran on lower rails, and the rest of it ran on this “boost” supply. Nicely, the “boost” came from simple linear supply tricks—no switchers here!
• Current-sourced front end. Another way to linearize the front end without increasing the loop gain was to replace its resistor bias with current-source bias. Current sources are much more stable with varying voltage, which results in a much more linear stage. In application, the current sources resulted in almost 10dB improvement in THD.

Those two changes, I figured, would get us close to state-of-the-art performance, or pretty close to the IC-based Magnius. And in the end, it got us almost there. But I’m getting ahead of myself. More on this later.

Back to the problems.

Here was the big one: the error correction didn’t have to swing a lot of volts, but it had to be able to swing the entire output current. Which is a lot. I mean, it’s substituting for the ground!

At first, I figured I might be able to get away with a single, stout op-amp. But no, it fell to its knees. It couldn’t swing the current.

Aside: though that prototype did prove that the improvements to the Magni+ stage worked well—and sounded good! Tyler immediately stole it to use with his work system, replacing a Magnius.​

So I tried a bigger op-amp—the output chip used in Magnius, actually, a design capable of delivering 700mA. That would be fine, I figured.

Well, except it’s a kinda squirrely, high-bandwidth, current-feedback design. It didn’t like playing second fiddle to the whole high-bandwidth, current-feedback discrete stage that I was developing.

After a couple of meltdowns, I was discouraged. Because the paths ahead were pretty ugly:

• I could do a discrete error correction stage, but this would blow the budget.
• I could go with multiple stacked op-amps for error correction, but that would blow the budget.
• I could remove the balanced connectors and call it a day, but it would be weird to pair with our balanced DACs.

What I needed was something that made the 4-pin XLR headphone output meaningful…without the instability and flames, without the complexity, and without breaking the bank.

Eeeeeek…

I mean, there had to be something. Something that needed a direct connection to the transducer, something that would actually be important in the context of the design. I mean, years ago I remember seeing a thing for mixed-mode feedback for loudspeakers that attempted to put the transducer into the feedback network (and get a kind of motion feedback for back EMF)…

…which was relatively easy, and kinda interesting…

…but would it work?

Hmm and argh. Well, it was easy enough to do. I took a stab at calculating the resistor values and built the network in the air on top of the dead error correction IC, then gave it to Tyler and told him to tell me what he thought.

He was back in 5 minutes.

“What the hell did you do to this?” he asked.

“So it sounds different?”

“Through balanced, hell yeah,” Tyler said. “Actually I think the balanced is one of the best things you’ve done.”

I shook my head. “’One of the best things’ is pretty marketing-weaselly,” I told him.

“Yeah yeah you know what I mean. This is really really good! What did you do?”

“I tried mixed-mode feedback, kind of a derpy motion feedback thing. Maybe. I mean, if the headphone drivers even flop around that much. Because it’s dependent on back EMF. I mean, we don’t have access to a separate voice coil for real motion feedback, and it might only matter, like, with a 15” woofer in a ported box, but for that woofer you may even be able to get some better measurements, due to damping of the resonance frequency—”

Tyler held up his hands. “Wait a sec. Are you saying this might get you better measurements—”

“Not likely with headphones,” I interrupted. “Maybe not likely at all. If it worked at all, it would work best for floppy things, like ported speakers—”

“But if it worked…and it made the acoustic performance better…”

Tyler just trailed off, and we sat staring at each other for a while, not saying anything.

Because if we could create better performance at the transducer, holy crap that was the holy grail. Everything blown up. Because a -110dB and a -120dB amp into a -50dB transducer both give you -50dB for the system.

But if one amp bumped that -50dB to -55 or -60…

Holy schiit!

That would literally be a game-changer.


Like Questing for the Grail

But it really couldn’t be that easy, could it? I mean, this wasn’t real motion feedback, like they do with subwoofers, this was dependent on resonance and floppy drivers and (most likely) ports and stuff like that, all of which is alien to headphones, and especially alien to planar headphones, which, although they may have resonance peaks, they don’t have impedance changes at those peaks, and who knows what kind of back EMF they have, because that’s something that isn’t really studied.

Aside: and why isn’t that studied? Same reason as for most of audio: no money in it. Things get grants and R&D investment because they’re hoping to make billions (or maybe even trillions). A solid audio result moves millions at best. And that’s at best, because there are still going to be endless arguments on each side about the validity of the data, the design of the experiment, etc. Sigh. Audio research is so derpified that there are tons of companies willing to cite another company’s standard, just because they did some measurements, not thinking about the reality of, “Hey, we’re citing another company’s research, not an independent entity, not something that’s been extensively confirmed.” Argh. But that’s the reality. It’s not going to change anytime soon.​

To cut to the chase: no, it isn’t really that easy.

In fact, getting reproducible results from an acoustic transducer, even with isolation chambers, even working around known resonance points, is super difficult. Maybe impossible.

Why “maybe impossible,” and not “impossible?”

Simple: we don’t have infinite time. Our early results (done by a third party) showed a potential improvement in THD in the low frequencies when run through our mixed-mode feedback topology. Later results were mixed or null. There’s not enough data for us to claim improvement, but there’s also not enough experimentation to say there’s no ‘there’ there.

So, here’s an idea: anyone who wants to try proving a difference, and who has an acoustic measurement rig for testing headphones, contact us. I’m sure we’ll be happy to get you a Midgard to play with. And yes, we promise to report all data, including nulls.

Which, after all, is what I expect: nulls. Because we’re only measuring a tiny corner of the audio universe—steady-state, typically single-sine or low-amount-of-sines for test signals, etc. Transient response, complex signals…we need to look a lot deeper if we expect to find any correlation with subjective results.

Aaaaaannnd…there still may be no correlation with subjective perception. Or unpredictable results. Because everyone has different preferences, and there are certainly differences in auditory capability.

I feel like I hammer on this a lot, but I’m not in the mood to codify it into the Three Laws of Audio Measurements. Let’s just leave it at this: there is no One Number to Rule Them. And there never will be.

Anywhoo, back to Magnius 2/Midgard:

• Most people who listened to the prototypes noted a significant difference between the balanced and single-ended outputs, with all favoring the balanced outputs. The only difference between these outputs was the mixed-mode feedback schema, which makes the balanced outputs meaningful, in the context of a single-ended amp stage.
• We performed experiments to see if we could show a measured difference in the transducer output between the single-ended and balanced outputs, but results were mixed or null, so we’re not claiming any measurable benefit.

“Wait a sec, what is this ‘Midgard’ thing,” you ask. “I thought this was Magnius 2.”

Yes. It started as Magnius 2, and then evolved to Midgard. Why? Mainly because we are falling out of love with “ius” names, as we feel it minimizes the significance of products bearing those names. Is Modius just a big Modi? No, it’s significantly different, right down to the DAC chip. Is Magnius just a big Magni? No, it’s totally different.

There was also the discussion about this product and Asgard. Asgard 3 is its own thing, and produces its own sound—a sweeter, mellower take on absolute accuracy. It’ll continue on as it is. The prototype Magnius 2, however, s a significantly different overall topology even if it was still single-ended, a focus on excellent measurements, and a new take with mixed-mode feedback.

All in all, the prototype Magnius 2, we realized, was perhaps our most universal amplifier—delivering high power without balanced connectors, delivering full performance thru a standard TRS jack, and offering a unique feedback arrangement that included the driver in the loop.

“What about Midgard?” Tyler said, finally.

“As a name for Magnius 2?” I asked.

“Yep.”

“As in, Earth. Our realm.” I said.

“Yeah, it makes sense—” Tyler began. But I got it. Earth was here, earth was us, earth was grounded. Midgard made sense for the first amp with this new toplogy. It was, in effect, grounded.

So you get Midgard.

And yes, I know. So now we have Midgard and Asgard and which one will sell better aren’t we worried and all of that?

But again, we’re getting ahead of the plot. We’ll break down Midgard vs Asgard 3 a bit later on. For the moment, let’s talk a bit more about dev and buzzword bingo.

“Buzzword bingo?” you ask.

Yep. Because, as you know, we have a habit of naming things we consider important: Unison USB™, Nexus™, things like that. This new mixed-mode semi-motion-feedback topology really deserved a name, but what? I mean, any number of “completing the circle” or “including the transducer” single-word things come to mind, but many of them crash into our existing names.

And, to be frank, the new topology promised great things—literally holy grail things—but we weren’t able to reproduce good enough results, so it might not deliver anything.

After agonizing for a while, something struck: Halo.

Why “Halo?” Because it’s a continuous ring, as in, completing the circle with the transducer. Also because, if anyone proves actual transducer-level measurement improvement, it is a holy grail. And also because, if nobody proves better measurements at the transducer, it’s still a shiny object, gold leaf adornment on a fancy painting.

Aside: and yeah, I know lots of people think these names are silly, but we try to keep them kinda-sorta in balance. As in, there are companies that don’t name anything, and then spend 50 or 500 or 5000 words trying to explain why they have some unique juju, and there are other companies that name nearly every damn thing, minimizing the importance of each individual name. We’re trying to name only the really important stuff. And we’re not above simply not using a lame name, like Equipoise. I mean, I use it here, but we don’t make a big marketing deal about it.​

So yeah. Now you have Midgard. The first Halo topology headphone amp.




Practical Considerations

Okay, okay. So now you have a ton of questions. Let me guess a couple of them.

• I’m sure you’re wondering why, if this Halo thing is such a great idea, why stick it on a cheap amp rather than a super-expensive one?
• And you’re wondering if it really works, why did we go with incomplete and inconclusive data, instead of being really truly sure?

Actually, these two questions are connected. So let’s talk a bit about how companies develop products, and the different ways those products can reach the market.

Many companies like to develop and test and measure and test again and benchmark and test and redevelop and test and measure and redevelop and test and maybe their endless development and test is symbolic of internal doubt, so maybe better to kill the product through nervosa rather than admitting you’re scared of your own idea.

Aside: this is frequently known by the fantastic phrase “paralysis by analysis.”​

Now, this is a great approach in some cases. Want to get something on Mars and only have one shot? Yeah, best to be certain.

Aside: and even then it doesn’t always work out, does it?​

In other cases, all the testing and redevelopment may mean the market has moved on by the time you get there, which ain’t so great. Or maybe, if the idea is truly bad, best not to put the effort into production ever, so maybe it also works in that case.

You also have companies that decide to try to charge you for their R&D time, because they spent a ton of person-years developing it, and hey you know that’s worth something, right? Well, yeah, maybe, unless someone else has also been developing the same thing and they know R&D is best seen as sunk cost and it’s best to win market share with lower prices.

Now, we’ve never tried to capitalize on the insane amount of person-years we’ve put into True Multibit, or Unison USB, or Autonomy, or Nexus, or any other of our innovations. We don’t do that. In addition, our focus is on affordable gear, so we’re not going to try to pull what-the-market-will-bear pricing.

And, at the same time, we’ve been willing to experiment and see what people really want, rather than tweak and redevelop forever internally—see our product Thunderdomes, for instance.

So it really makes sense we decided to start with Halo on an inexpensive product, and without full data: it fits our ethos. It’s a choice as valid as spending 5 years getting data, applying for a patent, crowing to the moon about how this is the greatestest invention everer, and putting it in a $12,000 machined-from-a-solid-block-of-titanium coffin.

Aside: because inexpensive products are super-important for getting people into this crazy audio thing, and because, even if we had a stack of data, it still might not be the greatestestest thing evereverever—or not for everyone!​

Now let’s go back to the question of Asgard 3 and Midgard—how are they different, which is better, and what works for you?

Let’s break it down:

	Asgard 3	Midgard
Output Power	3.5W/32 ohms	4.8W/32 ohms
THD	-103dB at 2V RMS/300 ohms	-118db at 4V RMS/300 ohms
SNR	115dB ref 2V A weighted	128dB ref 4V A weighted
Inputs	RCA SE, plus optional DAC cards	XLR balanced and RCA SE
Outputs	¼” TRS headphone RCA preamp outs	4-pin XLR, Halo headphone ¼” TRS headphone RCA preamp outs XLR preamp outs
Modularity	Yes, via optional DAC, phono, or RCA input cards	No
Power supply	Internal transformer with integrated stacked, dual-regulated power supply, +/-24V and +/-18V rails, plus +/-15V for servos	External wall-wart with internal stacked, dual-regulated power supply , +/-30V and +/-20V, plus +/-15V for servos
Topology	Fully discrete, current feedback, with Continuity™ output stage	Fully discrete, current feedback
Price	$279	$219

“Wait a sec,” you might be saying. “Seems like Midgard offers a lot more, and better measurements, for a lower price. How does that work?

How it works is simple: Asgard 3 is in a heavier metal chassis, with an internal transformer. No wall-wart. Plug the AC cord right into the chassis. And that chassis and transformer cost a lot more than Midgard. So Midgard can pack in additional features at a lower price.

Midgard also gives you the ability to compare Halo and non-Halo topologies. Halo requires a 4-pin connector with separate negative returns in order to include the transducer in its feedback. It can’t be done on a 3-pin TRS jack.

So try it: connect via 4-pin, then try the TRS. See what you think. You might be surprised.

With Midgard, you can also easily integrate into a balanced system. Have a balanced-out DAC, or need balanced outs for powered monitors? No problem. And yes, the balanced inputs are summed to SE, and the balanced outs are derived from SE, because Midgard is an inherently single-ended amp, but those inputs and outputs are there—and they will work flawlessly with balanced I/O.

And remember—if you have an acoustic measurement kit and want to play with Midgard, contact us. Maybe there’s something to this after all. Or not.


The Future?

With this hot new topology, you’re probably wondering what our plans are with respect to other products.

Weeeeeeeeelllllll…it’s murky.

Here’s the thing: implementing Halo on actual balanced products is, ah, more interesting. And not a ton of fun. So we’re really talking “do we plan to use Halo on other single-ended products?”

In short: yeah, but…

Yeah, as in it’s interesting enough to try in another format. So another product will be launching eventually.

But, as in, it ain’t proven, and it doesn’t need to be on everything until it’s more than a shiny maybe-something.

So, I hope you enjoy Midgard. And the other Halo product.

As to the rest…we’ll see!

Can I use an adaptor to convert the TRS headphone to enjoy the XLR output of the Midgard Halo? That said, connect to the Sleeve of the TRS to both grounds of the 4pin XLR?

 

[Jason Stoddard]

Can I use an adaptor to convert the TRS headphone to enjoy the XLR output of the Midgard Halo? That said, connect the Sleeve of he TRS to both grounds of the 4pin XLR?

Nope, it doesn't work that way. Nor do I know of any adapters that work that way, since it will cause huge issues with a balanced-output amp.

 

[Paladin79]

Nope, it doesn't work that way. Nor do I know of any adapters that work that way, since it will cause huge issues with a balanced-output amp.

You beat me to it lol. If you have dual entry headphones you should be able to replace a TRS cable with balanced. I have also rewired headphones that started out as single entry.

 

[Ohno Schiitizen]

Midgard can be another (cheaper) option as a preamp for running mono blocks, say Vidar 2.

 

[Ohno Schiitizen]

Nope, it doesn't work that way. Nor do I know of any adapters that work that way, since it will cause huge issues with a balanced-output amp.

Thank you Jason, that said, same as Jot2, cannot run SE at the 4 pin XLR, understood.

 

[Jason Stoddard]

Now if only I had a balanced turntable…

Teac TN5-BB is what we ended up with. Pro-ject also makes one, but it has a nonstandard connector and will need an adapter cable to standard XLRs. Beyond that, I think it's gigadollar range.

But you don't need a balanced turntable (and an associated MC cartridge) to use Skoll. You don't even need to use the balanced outputs. It works fine with anything.

 

[Paladin79]

Earlier someone mentioned The Dark Side of the Moon Redux album, I thought this Pro-Ject turntable interesting as well since it is the 50th anniversary.

https://imboldn.com/the-dark-side-of-the-moon-special-edition-turntable/

 

[Erratic Gravy]

A high-end phono preamp from Schiit? The dream product I’ve been wanting since the Mani was my first-ever phono preamp all the way back in 2017 (but I soon upgraded to more advanced offerings from Pro-Ject)? Can this really be happening?

I think a Day 1 buy is in order.

Now if only I had a balanced turntable…

Pro - ject are now offering TT's with a balanced output

 

[adydula]

2023, Chapter 12
The Ultimate Heresy?


While I was tweaking on Magni+, trying to get the maximum performance out of its discrete stage, I hit a wall.

As in, it just wasn’t gonna get better. Not unless I had a bigger budget and more space to add the boosted power supply and linearized front end I knew I needed. And so, frustrated, dreaming of something with a bigger chassis and less cost constraints, I muttered, “Gawd, I wish we had something a bit more fancy so I can do this super-duper Magni.”

Tyler overheard. And looked confused. “What do you mean, more fancy?”

“I mean, I need more budget for the additional parts. And I need more space for a power supply. If I had that, I could make Magni all it could be.”

Again, Tyler looked at me like I was nuts.

“You mean…like Magnius?”

And in that moment, I sat stunned. I’d totally forgotten we made a Magnius. It was that…forgettable.

I burst out laughing.


Not Because it’s Hard

I mean, I did introduce Magnius with a chapter entitled, “Not Because it’s Hard,” as an inversion of JFK’s famous line about going to the moon not because it’s easy, but because it’s hard…and as a jab at everyone who thought a high-measuring balanced amp was difficult to do and had to cost a lot of money.

And Magnius was easy. I mean, it’s the same story as all high-measuring IC-based amps: pick some of the best measuring chips, put ‘em in a box, and away you go.

It was so easy that I didn’t want to do it.

And it was so easy, that just a few years later…I’d totally forgotten we made it!

Aside: now, this isn’t to say that it’s a bad amp. It’s a perfectly fine amp. But designing it was too easy, and it was a trip into a cul-de-sac of me-too design, and it didn’t make me super happy. So I retconned the whole thing out of my mind. ​

Sooooooo, getting back to the super-Magni I wanted to make: that was different. That was a fully discrete design. One that would be really interesting to see how far we could go, and still keep it super-affordable. That was interesting. And that, I realized, might make a perfect Magnius 2.

Except for one thing: Magnius was balanced.

This imagined super-Magni? Single-ended.

Hmmmmmm. Would that fly? I wasn’t so sure.


Balancing the Equation

Now, making the Magnius 2 a single-ended amp would have a lot of benefits, most notably:

• Much higher power output from the single-ended headphone jack
• Much higher SE performance (than taking one phase of a differential amp)

But it would also mean dropping the balanced inputs and outputs, as well as the balanced 4-pin Neutrik headphone jack.

I mean sure, we could do fake balanced. We could just leave the connectors on there and run them single-ended. But that’s a horribly cheap and lazy and cynical way to do things, and it could cause incompatibilities with other balanced gear.

Aside: I did write a chapter on this, FWIW.​

Alternately, we could drop the balanced connectors. But that would be pretty silly when you could stack this product on a Modius, which had balanced outputs. And it would change the chassis, both top and bottom, which would mean new metal, and we had stock of Magnius chassis, and it would be best to simply use it.

So the question became: could I retain the balanced I/O—and make it meaningful—on a single-ended power amp?

Now, yes, retaining the balanced analog input and preamp output really wasn’t a huge issue. We could sum and difference the outputs, just like any amount of pro (and high-end gear) does, the same that Lokius does, the same that Loki Max does. That would provide the benefits of balanced transmission, for real, no fooling.

But the 4-pin XLR for the headphone?

Yeah, some handwaving can be done about “separate ground returns,” but that’s handwaving. There’s not much “there” there.

But there are also other things you can do.

Like error correction, applied via the negative terminal of the balanced output. As in, you compare the input to the output and apply a corrective difference. Note that this isn’t feedback. This is I-O = E , where E is a difference signal, just the error component. Feedback feeds back the full O (output) signal to the input.

What’s more, error correction could potentially improve the performance of the amp. I was shooting for near state-of-the-art performance with the upgraded Magni single-ended discrete stage, but this had the potential to push performance even higher.

And, I realized, this error correction would only be applied through the XLR output, because it needed separate negative terminals. The ¼” TRS would have no error correction, due to its shared grounds. So you could compare a “standard” and “error corrected” output.

All of that was pretty exciting, so I proceeded with a prototype…

…and immediately ran into problems.


Desperation is the Mother of Invention

Here’s the thing. A lot of what I wanted to do was fairly straightforward. The main goal was to improve the performance of Magni+’s discrete stage significantly, without increasing loop gain. As in, it would stay a low-feedback design. It would just get more linear.

I had two strategies for making it more linear:

• Massive boost supply. Magni runs on +/-16V rails. Which is pretty solid. But it does run into issues with the input signal being a significant proportion of those rails. As the voltage across the input devices varies, the gain of the devices changes. Which causes distortion. In fact, that’s the primary cause of distortion in the Magni+ stage. However, if I ran it on +/-30V rails, that would extend the linear region of operation and lower distortion. So I segmented the supply so the output stage still ran on lower rails, and the rest of it ran on this “boost” supply. Nicely, the “boost” came from simple linear supply tricks—no switchers here!
• Current-sourced front end. Another way to linearize the front end without increasing the loop gain was to replace its resistor bias with current-source bias. Current sources are much more stable with varying voltage, which results in a much more linear stage. In application, the current sources resulted in almost 10dB improvement in THD.

Those two changes, I figured, would get us close to state-of-the-art performance, or pretty close to the IC-based Magnius. And in the end, it got us almost there. But I’m getting ahead of myself. More on this later.

Back to the problems.

Here was the big one: the error correction didn’t have to swing a lot of volts, but it had to be able to swing the entire output current. Which is a lot. I mean, it’s substituting for the ground!

At first, I figured I might be able to get away with a single, stout op-amp. But no, it fell to its knees. It couldn’t swing the current.

Aside: though that prototype did prove that the improvements to the Magni+ stage worked well—and sounded good! Tyler immediately stole it to use with his work system, replacing a Magnius.​

So I tried a bigger op-amp—the output chip used in Magnius, actually, a design capable of delivering 700mA. That would be fine, I figured.

Well, except it’s a kinda squirrely, high-bandwidth, current-feedback design. It didn’t like playing second fiddle to the whole high-bandwidth, current-feedback discrete stage that I was developing.

After a couple of meltdowns, I was discouraged. Because the paths ahead were pretty ugly:

• I could do a discrete error correction stage, but this would blow the budget.
• I could go with multiple stacked op-amps for error correction, but that would blow the budget.
• I could remove the balanced connectors and call it a day, but it would be weird to pair with our balanced DACs.

What I needed was something that made the 4-pin XLR headphone output meaningful…without the instability and flames, without the complexity, and without breaking the bank.

Eeeeeek…

I mean, there had to be something. Something that needed a direct connection to the transducer, something that would actually be important in the context of the design. I mean, years ago I remember seeing a thing for mixed-mode feedback for loudspeakers that attempted to put the transducer into the feedback network (and get a kind of motion feedback for back EMF)…

…which was relatively easy, and kinda interesting…

…but would it work?

Hmm and argh. Well, it was easy enough to do. I took a stab at calculating the resistor values and built the network in the air on top of the dead error correction IC, then gave it to Tyler and told him to tell me what he thought.

He was back in 5 minutes.

“What the hell did you do to this?” he asked.

“So it sounds different?”

“Through balanced, hell yeah,” Tyler said. “Actually I think the balanced is one of the best things you’ve done.”

I shook my head. “’One of the best things’ is pretty marketing-weaselly,” I told him.

“Yeah yeah you know what I mean. This is really really good! What did you do?”

“I tried mixed-mode feedback, kind of a derpy motion feedback thing. Maybe. I mean, if the headphone drivers even flop around that much. Because it’s dependent on back EMF. I mean, we don’t have access to a separate voice coil for real motion feedback, and it might only matter, like, with a 15” woofer in a ported box, but for that woofer you may even be able to get some better measurements, due to damping of the resonance frequency—”

Tyler held up his hands. “Wait a sec. Are you saying this might get you better measurements—”

“Not likely with headphones,” I interrupted. “Maybe not likely at all. If it worked at all, it would work best for floppy things, like ported speakers—”

“But if it worked…and it made the acoustic performance better…”

Tyler just trailed off, and we sat staring at each other for a while, not saying anything.

Because if we could create better performance at the transducer, holy crap that was the holy grail. Everything blown up. Because a -110dB and a -120dB amp into a -50dB transducer both give you -50dB for the system.

But if one amp bumped that -50dB to -55 or -60…

Holy schiit!

That would literally be a game-changer.


Like Questing for the Grail

But it really couldn’t be that easy, could it? I mean, this wasn’t real motion feedback, like they do with subwoofers, this was dependent on resonance and floppy drivers and (most likely) ports and stuff like that, all of which is alien to headphones, and especially alien to planar headphones, which, although they may have resonance peaks, they don’t have impedance changes at those peaks, and who knows what kind of back EMF they have, because that’s something that isn’t really studied.

Aside: and why isn’t that studied? Same reason as for most of audio: no money in it. Things get grants and R&D investment because they’re hoping to make billions (or maybe even trillions). A solid audio result moves millions at best. And that’s at best, because there are still going to be endless arguments on each side about the validity of the data, the design of the experiment, etc. Sigh. Audio research is so derpified that there are tons of companies willing to cite another company’s standard, just because they did some measurements, not thinking about the reality of, “Hey, we’re citing another company’s research, not an independent entity, not something that’s been extensively confirmed.” Argh. But that’s the reality. It’s not going to change anytime soon.​

To cut to the chase: no, it isn’t really that easy.

In fact, getting reproducible results from an acoustic transducer, even with isolation chambers, even working around known resonance points, is super difficult. Maybe impossible.

Why “maybe impossible,” and not “impossible?”

Simple: we don’t have infinite time. Our early results (done by a third party) showed a potential improvement in THD in the low frequencies when run through our mixed-mode feedback topology. Later results were mixed or null. There’s not enough data for us to claim improvement, but there’s also not enough experimentation to say there’s no ‘there’ there.

So, here’s an idea: anyone who wants to try proving a difference, and who has an acoustic measurement rig for testing headphones, contact us. I’m sure we’ll be happy to get you a Midgard to play with. And yes, we promise to report all data, including nulls.

Which, after all, is what I expect: nulls. Because we’re only measuring a tiny corner of the audio universe—steady-state, typically single-sine or low-amount-of-sines for test signals, etc. Transient response, complex signals…we need to look a lot deeper if we expect to find any correlation with subjective results.

Aaaaaannnd…there still may be no correlation with subjective perception. Or unpredictable results. Because everyone has different preferences, and there are certainly differences in auditory capability.

I feel like I hammer on this a lot, but I’m not in the mood to codify it into the Three Laws of Audio Measurements. Let’s just leave it at this: there is no One Number to Rule Them. And there never will be.

Anywhoo, back to Magnius 2/Midgard:

• Most people who listened to the prototypes noted a significant difference between the balanced and single-ended outputs, with all favoring the balanced outputs. The only difference between these outputs was the mixed-mode feedback schema, which makes the balanced outputs meaningful, in the context of a single-ended amp stage.
• We performed experiments to see if we could show a measured difference in the transducer output between the single-ended and balanced outputs, but results were mixed or null, so we’re not claiming any measurable benefit.

“Wait a sec, what is this ‘Midgard’ thing,” you ask. “I thought this was Magnius 2.”

Yes. It started as Magnius 2, and then evolved to Midgard. Why? Mainly because we are falling out of love with “ius” names, as we feel it minimizes the significance of products bearing those names. Is Modius just a big Modi? No, it’s significantly different, right down to the DAC chip. Is Magnius just a big Magni? No, it’s totally different.

There was also the discussion about this product and Asgard. Asgard 3 is its own thing, and produces its own sound—a sweeter, mellower take on absolute accuracy. It’ll continue on as it is. The prototype Magnius 2, however, s a significantly different overall topology even if it was still single-ended, a focus on excellent measurements, and a new take with mixed-mode feedback.

All in all, the prototype Magnius 2, we realized, was perhaps our most universal amplifier—delivering high power without balanced connectors, delivering full performance thru a standard TRS jack, and offering a unique feedback arrangement that included the driver in the loop.

“What about Midgard?” Tyler said, finally.

“As a name for Magnius 2?” I asked.

“Yep.”

“As in, Earth. Our realm.” I said.

“Yeah, it makes sense—” Tyler began. But I got it. Earth was here, earth was us, earth was grounded. Midgard made sense for the first amp with this new toplogy. It was, in effect, grounded.

So you get Midgard.

And yes, I know. So now we have Midgard and Asgard and which one will sell better aren’t we worried and all of that?

But again, we’re getting ahead of the plot. We’ll break down Midgard vs Asgard 3 a bit later on. For the moment, let’s talk a bit more about dev and buzzword bingo.

“Buzzword bingo?” you ask.

Yep. Because, as you know, we have a habit of naming things we consider important: Unison USB™, Nexus™, things like that. This new mixed-mode semi-motion-feedback topology really deserved a name, but what? I mean, any number of “completing the circle” or “including the transducer” single-word things come to mind, but many of them crash into our existing names.

And, to be frank, the new topology promised great things—literally holy grail things—but we weren’t able to reproduce good enough results, so it might not deliver anything.

After agonizing for a while, something struck: Halo.

Why “Halo?” Because it’s a continuous ring, as in, completing the circle with the transducer. Also because, if anyone proves actual transducer-level measurement improvement, it is a holy grail. And also because, if nobody proves better measurements at the transducer, it’s still a shiny object, gold leaf adornment on a fancy painting.

Aside: and yeah, I know lots of people think these names are silly, but we try to keep them kinda-sorta in balance. As in, there are companies that don’t name anything, and then spend 50 or 500 or 5000 words trying to explain why they have some unique juju, and there are other companies that name nearly every damn thing, minimizing the importance of each individual name. We’re trying to name only the really important stuff. And we’re not above simply not using a lame name, like Equipoise. I mean, I use it here, but we don’t make a big marketing deal about it.​

So yeah. Now you have Midgard. The first Halo topology headphone amp.




Practical Considerations

Okay, okay. So now you have a ton of questions. Let me guess a couple of them.

• I’m sure you’re wondering why, if this Halo thing is such a great idea, why stick it on a cheap amp rather than a super-expensive one?
• And you’re wondering if it really works, why did we go with incomplete and inconclusive data, instead of being really truly sure?

Actually, these two questions are connected. So let’s talk a bit about how companies develop products, and the different ways those products can reach the market.

Many companies like to develop and test and measure and test again and benchmark and test and redevelop and test and measure and redevelop and test and maybe their endless development and test is symbolic of internal doubt, so maybe better to kill the product through nervosa rather than admitting you’re scared of your own idea.

Aside: this is frequently known by the fantastic phrase “paralysis by analysis.”​

Now, this is a great approach in some cases. Want to get something on Mars and only have one shot? Yeah, best to be certain.

Aside: and even then it doesn’t always work out, does it?​

In other cases, all the testing and redevelopment may mean the market has moved on by the time you get there, which ain’t so great. Or maybe, if the idea is truly bad, best not to put the effort into production ever, so maybe it also works in that case.

You also have companies that decide to try to charge you for their R&D time, because they spent a ton of person-years developing it, and hey you know that’s worth something, right? Well, yeah, maybe, unless someone else has also been developing the same thing and they know R&D is best seen as sunk cost and it’s best to win market share with lower prices.

Now, we’ve never tried to capitalize on the insane amount of person-years we’ve put into True Multibit, or Unison USB, or Autonomy, or Nexus, or any other of our innovations. We don’t do that. In addition, our focus is on affordable gear, so we’re not going to try to pull what-the-market-will-bear pricing.

And, at the same time, we’ve been willing to experiment and see what people really want, rather than tweak and redevelop forever internally—see our product Thunderdomes, for instance.

So it really makes sense we decided to start with Halo on an inexpensive product, and without full data: it fits our ethos. It’s a choice as valid as spending 5 years getting data, applying for a patent, crowing to the moon about how this is the greatestest invention everer, and putting it in a $12,000 machined-from-a-solid-block-of-titanium coffin.

Aside: because inexpensive products are super-important for getting people into this crazy audio thing, and because, even if we had a stack of data, it still might not be the greatestestest thing evereverever—or not for everyone!​

Now let’s go back to the question of Asgard 3 and Midgard—how are they different, which is better, and what works for you?

Let’s break it down:

	Asgard 3	Midgard
Output Power	3.5W/32 ohms	4.8W/32 ohms
THD	-103dB at 2V RMS/300 ohms	-118db at 4V RMS/300 ohms
SNR	115dB ref 2V A weighted	128dB ref 4V A weighted
Inputs	RCA SE, plus optional DAC cards	XLR balanced and RCA SE
Outputs	¼” TRS headphone RCA preamp outs	4-pin XLR, Halo headphone ¼” TRS headphone RCA preamp outs XLR preamp outs
Modularity	Yes, via optional DAC, phono, or RCA input cards	No
Power supply	Internal transformer with integrated stacked, dual-regulated power supply, +/-24V and +/-18V rails, plus +/-15V for servos	External wall-wart with internal stacked, dual-regulated power supply , +/-30V and +/-20V, plus +/-15V for servos
Topology	Fully discrete, current feedback, with Continuity™ output stage	Fully discrete, current feedback
Price	$279	$219

“Wait a sec,” you might be saying. “Seems like Midgard offers a lot more, and better measurements, for a lower price. How does that work?

How it works is simple: Asgard 3 is in a heavier metal chassis, with an internal transformer. No wall-wart. Plug the AC cord right into the chassis. And that chassis and transformer cost a lot more than Midgard. So Midgard can pack in additional features at a lower price.

Midgard also gives you the ability to compare Halo and non-Halo topologies. Halo requires a 4-pin connector with separate negative returns in order to include the transducer in its feedback. It can’t be done on a 3-pin TRS jack.

So try it: connect via 4-pin, then try the TRS. See what you think. You might be surprised.

With Midgard, you can also easily integrate into a balanced system. Have a balanced-out DAC, or need balanced outs for powered monitors? No problem. And yes, the balanced inputs are summed to SE, and the balanced outs are derived from SE, because Midgard is an inherently single-ended amp, but those inputs and outputs are there—and they will work flawlessly with balanced I/O.

And remember—if you have an acoustic measurement kit and want to play with Midgard, contact us. Maybe there’s something to this after all. Or not.


The Future?

With this hot new topology, you’re probably wondering what our plans are with respect to other products.

Weeeeeeeeelllllll…it’s murky.

Here’s the thing: implementing Halo on actual balanced products is, ah, more interesting. And not a ton of fun. So we’re really talking “do we plan to use Halo on other single-ended products?”

In short: yeah, but…

Yeah, as in it’s interesting enough to try in another format. So another product will be launching eventually.

But, as in, it ain’t proven, and it doesn’t need to be on everything until it’s more than a shiny maybe-something.

So, I hope you enjoy Midgard. And the other Halo product.

As to the rest…we’ll see!

@Jason Stoddard
Congrats on Midgard!
Alex

 

[Hyde00]

Aside: during development, the “Manius” moniker fell away, to be replaced by “Skoll,” which was suggested by Tyler. I agreed it was time to get away from the “ius” names, which kinda minimize the products they’re attached to, and so Manius became Skoll, the wolf that chases Sol in Norse mythology.

Yes. It started as Magnius 2, and then evolved to Midgard. Why? Mainly because we are falling out of love with “ius” names, as we feel it minimizes the significance of products bearing those names. Is Modius just a big Modi? No, it’s significantly different, right down to the DAC chip. Is Magnius just a big Magni? No, it’s totally different.

Oh no! My dream of having Lyrius and Asgardius shattered LOLLLL. (Though I guess Asgardius would just be Jot 2..... maybe......????)

I guess one can still dream about Asgard+.

Jokes aside, looking forward to try the Midgard!

 

[Snowpuppy77]

2023 Chapter 11
Fish in a Barrel, and All That


So what do you do when you’re faced with such an obvious decision that it seems like a trap?

I mean, like this:

You have a top-selling product, and it’s the only one in its category that you make…

…aaand…

…the logical step-up from that product is the #1 most searched on in Google?

Yeah, I’m talking about Mani 2. The successor to Mani. Both “the little phono preamps that could.” Both top sellers. Both very high in the rankings of value per dollar in the phono preamp realm.

But also both entry level.

For a long time, Tyler and Alex have been bothering me to do a “higher end” phono preamp, something above the Mani 2. Mainly because, based on the strength of Mani 2, they thought it would sell. And sometimes business needs adults in the room to say, hey, you know sometimes it’s the bottom line that counts and sometimes maybe we shouldn’t just go off and tilt at Atoms or make crazy desktop grille/amps for crazy people.

But that’s where Tyler and Alex’s design direction stopped. Not “higher end balanced” or “higher end discrete,” or “higher end tube,” or “higher end just for the sake of being more expensive and fancy higher end.” Just “something more than Mani 2.”

And I, for better or worse, waved them off. I had other things to do. And Mani 2 was sooooooooooooo good, why bother, right?

Aside: and I do have a healthy aversion to doing products just for the sake of doing another product. If we can’t find a way to make it significantly better, different, less expensive, easier to use, with more relevant features—as in, if we can’t find a way to “enhance the category,” I don’t want to do it. That’s one of the reasons we haven’t really done a portable or a dongle or a streamer.​

So I managed to sidestep the issue of doing a better phono preamp for quite a while. But then, Rina came to me and said, “You know, a ton of people are searching for a balanced phono preamp, do we make one?”

I sighed and said no.

She looked at me askance and asked why, and I told her about “enhancing the category,” and she gave me that ah yeah you’re lazy and sometimes dumb look, and said, “I can’t believe you can’t do better than the other stuff out there.”

Sigh.

And so, with Tyler and Alex’s ask, with Rina’s browbeating, and the irksome knowledge that some other manufacturers love to blang on about the limitations of Mani 2—namely, op-amp based, limited rail voltage, blah blah (I know you guys are reading this)—I put three and three together, and I decided:

Hey, let’s do a new phono preamp.

But what?

Gooooood question.



First, the Research

At first, I figured: someone’s gotta be doing a logical upgrade to Mani 2. I’ll find it and that’ll shut everyone up and I won’t have to do anything.

Because, hey, at this point in the dev cycle, it’s still 2022, and things are still super busy, and I got other things to do like Syn and Mjolnir 3 and Kara and stuff I don’t know if it’ll be released by the time this chapter hits.

And what did I find?

(Insert AI-generated video of a desolate wasteland with tumbleweeds blowing across the field of view.)

I mean, sigh.

I mean, come on guys, please try with this stuff. It’s not hard, and we make Schiit in the USA.

So here’s what I found:

• Insanely priced stuff with some interesting tech.
• Insanely priced stuff with some crap tech.
• Moderately priced stuff claiming stuff that wasn’t technically correct (I mean, balanced is balanced, right, and discrete is discrete, right?)
• Tumbleweeds.

Sigh mark II.

I’ll spare you the pain, but let’s review the log line for Skoll: the least expensive truly balanced really 100% absolutely discrete, 60V rail, zero feedback, microprocessor managed, remote control phono preamp.

Period. From any company, anywhere.

Yeah. Fish in a barrel. And all that.



So how did we get here? Another good question.


Internal Thunderdomes Ahead

One of the benefits to being an (ahem) more mature audio company (13, we still act like we’re 13, that’s fine, isn’t it?) is that we have a lot of interesting tech to play with.

I mean, heck, this “Manius” (internal dev name) could end up being something based on TI’s excellent LME49724 op-amp (we did a version of this phono preamp back in 2019), or it could be based on our Nexus stage, which would also allow that wonderful candy-like badge of “Transimpedance” to be added to the mix, or it could be a super-simple, no-feedback stage based on the differential buffer of the Freya+. All would allow the valued badge of “balanced” to be attached to the phono stage.

But op-amps were out. We’d tried doing a higher-end balanced op-amp phono pre back in 2019, and it was OK, but in the end op-amps are best suited for a $149 phono stage like Mani 2. At least in our opinion. Yes, I know there are more expensive op-amp phono pres out there, but yeah, not our bag.

Which left:

• Based on Nexus
• Based on Equipoise, a horrible name for the topology introduced in the Freya differential buffer, and is now writ large in Mjolnir 3

Now, the sharp-eyed are going “Wait a sec, a differential buffer implies no gain, so how is that gonna work in a phono preamp that needs a ton of gain?”

Right. Gotcha. Because the buffer is only a buffer due to degeneration, and it can actually have a decent amount of gain. Especially when cascaded in a passive RIAA phono preamp, where you have something like:

Gain-->RIAA-->Gain-->Out

But the question was: would we have enough gain? Equipoise is a no-feedback, single-stage topology. Could it get to the 60dB of gain we’d need for moving coil? Which really means 80dB of gain with 20dB thrown away for the passive RIAA. Which is a lot of gain.

And we also had a competing topology, one I already knew worked at high gain, but also used a ton of parts: Nexus. Nexus also worked well as a transimpedance input amplifier, a fact I ascertained while working on a different future product. So that opened up the whole Grand Badge of Transimpedance, which would be cool.

Aside: what is all this “transimpedance” stuff, you ask? It’s the latest hot word in phono preamps. In some ways it’s super cool, and in some ways it has significant limitations. The super cool stuff starts with the fact that it can kinda auto-set gain for moving coil cartridges. And, it can kinda-sorta eliminate a lot of the crazy loading stuff for MC cartridges. The catch is that the auto-set gain isn’t perfect; it will have higher gain for lower output impedance MC cartridges, which is usually about the way it should work—lower impedance, lower output, so higher gain is better—but it may not be exactly what you need. The other catch is this is an MC cartridge thing only, and it’s really a balanced thing only, at least if you want to do it right, in our opinion. So if you have an MC cartridge and if you are running balanced, and if you have a topology that is suited to it, it’s super cool. However, that doesn’t mean it’s a panacea.​

So yeah, Equipoise vs Nexus? No feedback and simple vs feedback and complex? Both, of course, would have passive RIAA in-between the gain stages, so no feedback there, but the Nexus stages do use current feedback, while Equipoise does not.

It was really a conundrum, compounded by lots of questions:

• Was it worth having literally 2x as many parts to do Nexus? That’s a lot of parts, and a big difference in cost.
• What kind of hidden gotchas were in each implementation? Enough gain in Equipoise, vs unknowns with Nexus?
• Did the whole idea work, period, with our idea to do a relatively complex, relay-switched control system?

“Wait, what?” someone is asking. “What’s with the relays? What’s with ‘control system?’”

Okay, let’s back up. Here was the idea for the “step up from Mani 2” phono preamp:

• Balanced or SE input.
• Significantly easier to use, with front panel controls and, ideally, remote control.
• Significantly better performance in at least one metric, and the metric could be subjective.

Front panel controls means:

• Icky stuff like running all the signals to the front panel and back. Not a great idea in a design that has gains of 1000x or more.
• Horrible mechanical stuff like long-shaft switches and pots to bring the controls to the back of the chassis. We’d never done this before, and the idea didn’t excite me.
• Untested stuff like relays, which there was some debate as to whether they’d be able to handle the very low-level signals of a phono input.

So, for us, that meant relays.

But would they work? Or would they cause noise or connectivity issues? And, if we threw a microprocessor in there to run the show, would that be too noisy? Or could we have it go to sleep? And would that be enough? There were tons of debates online, mostly between people who had never really built a commercial phono preamp—

Aside: isn’t that always the case? Bench racers galore.​

—or who were repeating anecdotes from other designers.

Ah, heck with it, I shrugged. Let’s do both, let’s do both with relays, and let’s see how they do. An internal Thunderdome, so to speak.

So I put together two “best guess” designs, each based on a similar power supply, and each designed around a wall-wart, which made it a Corpus Christi product, and sent them out to become boards.

When the boards came back, I put them together, thinking, Ah hell, this is gonna be a whole lot of crazy. Probably neither will work.

So what happened?

Both of them worked.

At least kinda.


Two Flavors of Gotcha

“Kinda working” is the analog designer’s hell, especially if you’re doing discrete. A lot of stuff is “kinda working.” But “kinda working,” doesn’t mean “this is a safe, sellable, compelling product.” And “kinda working” can look pretty robust…usually until you’re just about ready to go to market.

Luckily, both of the Skoll prototypes had pretty obvious gotchas, even if they were “kinda working.”

Aside: during development, the “Manius” moniker fell away, to be replaced by “Skoll,” which was suggested by Tyler. I agreed it was time to get away from the “ius” names, which kinda minimize the products they’re attached to, and so Manius became Skoll, the wolf that chases Sol in Norse mythology.​

Both prototypes had one same gotcha, which was my bonehead mistake: wrong microprocessor layout. Which meant we couldn’t fit the microprocessor on the board.

Which usually means it’s time to do a new board. But when you’ve already stuffed 400 or so components when you realize the microprocessor won’t fit, sometimes it’s better to fire it up and switch the relays by hand, just to see if you had a product.

And that’s what I did with the Nexus version. I already had tons of parts in place, so it made sense to see how it worked.

And it did, kinda. As in, it started up, the voltages were right, there was no magic smoke, and Nexus was Nexusing fairly happily (Nexus being in-balance is critical, dependent on matched parts and layout, so this is no small thing.)

So I pulled the relays to give me a calculated 40dB of gain, turned on the input and output, and gave it to the guys in Corpus to hook it up. They have a test setup right there with Mani, speakers, a couple of turntables, and one of them is very much into vinyl.

We dropped the needle. Two seconds in, Evan said, “Well, that kills Mani pretty much stone dead.”

And it did sound pretty damn good. It sounded so good I suspected something was wrong, something euphonic, some screwup in the RIAA, something like that. But the measurements weren’t too wonky. RIAA wasn’t as good as Mani in terms of accuracy, but it was very close. So there was something nice going on here.

But it didn’t take long to find a pretty serious issue: on some records, the Nexus prototype would go out of Nexus, which led to horrible distortion on one or more channels. Left to sit by itself, it would happy stay Nexused, but some records—usually the more dynamic and bombastic ones—could cause problems.

Maybe the rumble? I wondered? After all, Nexus’ operating point servos ran at similar frequencies to record warps, and there was Nexus into Nexus interaction, so maybe that was it. I staggered the servo frequencies, and things got a lot better.

And when it was working, holy moly, this this was nuts. It hardly even looked like a phono preamp on the analyzer, delivering performance that I’d never seen from a phono preamp before.

But still, it wasn’t perfect. It still would go whomper-jawed from time to time. Nexus, when run cascaded with a passive RIAA, would probably need a differential servo or two-stage oversight to make sure everything ran fine all the time. That didn’t give me a warm and fuzzy feeling, because that added complexity, and could end up with some really interesting problems that we really didn’t need in a still-inexpensive phono preamp.

Aside: target price for Skoll was “under $500.” Happy to see we significantly undershot that price. ​

And, in any case, we still had the question of whether or not we’d be able to get away with relays for switching, and a microprocessor running the whole thing.

So I moved on to the Equipoise layout, which I re-did for the correct microprocessor pinout so I could run the whole thing and see if we could do the switching arrangement I wanted.

This one was wayyyyyyy simpler—less than half the parts. Like Nexus, it’s a differential topology, but it’s a single-stage affair. And in some ways it allows us to claim even more audiophile woo. Because Equipoise is a no-overall-feedback stage. So you could add “no feedback” to “passive RIAA” and “Class A operation.”

Aside: of course it’s Class A. All preamps are. It’s not a big deal. Crowing about a “Class A preamp!” is like bragging about a “beef hamburger!”​

The big question with Equipoise was if it could give me the gain I needed with such a simple topology, and if that gain still had good enough distortion and noise performance. Because distortion would certainly be higher—after all, it had no feedback!

So I proceeded with this one pretty much like the Nexus stage: made sure it was basically working, set the gain manually (firmware would come later) and lobbed it at the guys in Corpus.

“Oh, this is niiiiicceee,” Evan said.

And it was. From the start, it was a very natural-sounding preamp. What was super-weird was that it also seemed quieter than Mani 2—quieter in terms of ticks and pops. Maybe the no-feedback thing? Maybe the higher rails?

But it definitely struggled a bit with distortion. What it needed was more volts, so we could run it more linearly, and so we could get larger output swing.

Now, tweaking power supplies and getting more volts was something I could do. So I tweaked it to an asymmetric supply with 60V total, and tweaked the operating points a bit, to get 2x the swing and 1/10 the distortion.

Now, the overall distortion numbers aren’t going to win awards in this age of op-amp everything, but come on: what’s the distortion like on the record? Definitely higher than this preamp. So neener. And what’s the distortion on your transducer? Also higher than this preamp. So double neener.

What surprised me is how quiet the Equipoise prototype was. I mean, this was a single-stage differential design with no overall feedback and passive RIAA—and I was seeing 100dB+ SNR, A-weighted, at 40+dB gain. Which was like 6dB better than Mani 2, which uses some very fancy op-amps in its design.

What’s more, it acted much more “like a product” than the Nexus version. As in, it played happily with every record, it didn’t go out of balance, it sounded good, and its RIAA was easily tuned to +/-0.1dB.

Now, if it just worked with the microprocessor.


Scarily, Into the Future

Why was I worried about the microprocessor and the relays? Simple: we’re talking about tiny, tiny signals here, and tons and tons of gain. Microprocessors mean there’s going to be a clock running, a high-frequency noise generator that might interact with those tiny signals.

Aside: that is, unless we could turn it off, and only have it wake up when necessary.​

I was also worried about the relays because I knew that we’d been able to measure noise from a ripply DC supply to the relays. Would our relay supply be quiet enough? I’d designed in additional filter stages, but I didn’t know if it would be enough.

So, armed with the measurements taken before the microprocessor went on the board, I dropped it on and told Stephan to proceed with the firmware.

The firmware development was relatively unremarkable, and the Equipoise prototype was soon up and running.

The first surprise was the LEDs. When first turned on, Skoll did a scrolling “Cylon scan,” of the long row of loading LEDs. Something instantly familiar to fans of the original Battlestar.

“I had to do it,” Stephan said, looking a bit sheepish.

“Cool. Leave it in,” I told him.

He looked surprised and happy, as if he thought I would tell him to do something a bit more, ah, boring.

I hooked it up to the analyzer, half-expecting to see intractable noise…

…and there was nothin.

Just the same as no microprocessor and shorted relays.

I ran it through different loading and gains, and it obediently did its thing. No noise, no pops, no drama, nothing to show the microprocessor was there at all, nothing to show there were relays in the signal path.

Woohoo!

But still, I am insane, so I asked Stephan if he could “sleep” the microprocessor and only have it wake when necessary. After a bit of grumbling about how it might be problematic with a remote, he had it working. And even if the noise floor didn’t drop any more, I feel a little better about having the microprocessor off. I hope you do to.


A No-Brainer?

So what did we end up with, after all of this internal thunderdoming? I know it sounds like hype and bull, but maybe we really do have a new reference in the phono preamp world.

I mean, it already takes the crown of “the least expensive fully discrete truly balanced phono preamp,” but it goes a lot farther than that, especially in an era where there’s a lot of fast-and-loose played with “discrete” and “balanced.”

It also throws in convenient front-panel control and remote control.

As in, you can change the input, gain, loading, and subsonic filter from your listening chair.

No more bottom-mounted DIP switches, no more carefully turning things off before making changes, just press and button and the microprocessor makes sure everything is safe.

Aside: with the usual proviso of “going from 40 to 50 to 60 to 70dB of gain makes things a lot louder, so maybe don’t start with the volume maxed, bro.”​

So what do you get for your $399, specifically?

• Two inputs (XLR balanced and RCA single-ended), switchable from the front panel or remote control.
• Two outputs (XLR balanced and RCA single-ended), both on all the time.
• A no-feedback, fully discrete, fully differential, Class A topology with fully passive RIAA, running on 64V rails with a fully linear, dual series-regulated, ultra-quiet power supply.
• 40/50/60/70dB of gain, selectable via the front panel or remote control.
• Selectable R loading of 47K or 10/50/100/150 ohms, via front panel or remote control.
• Selectable C loading of 47, 100, 150, or 200p, via front panel or remote control.

Criticisms will probably focus on two areas:

• The first is that we didn’t include 38 different R and C loads. Right. This product had to conserve somewhere, so we selected only the most common loads: 47K/47p for MM, and a small selection of low-resistance values for most MC cartridges.
• The other is that we didn’t include the weirdo balanced connector sometimes used on inexpensive balanced-out turntables. Also right. I’m sure there are adapters to convert that connector to the industry standard for balanced: dual XLR.

Measurements? Better than Mani 2 in terms of noise, worse than Mani 2 in terms of distortion. But again, THD is definitely better than necessary for the medium, especially in the context of your transducers.

RIAA accuracy? Better than Mani 2.

Intangibles? Mainly those related to a non-feedback, high-rail design. Subjectively it appears to minimize pops and ticks. Which, yes, I know, is super weird, but using phono in this day and age is super weird, so there’s that.

So yeah. Maybe a no-brainer to upgrade to this from Mani 2. While Mani 2 is really, really good, it is not the last word in ease of use (being very kind here). And Mani 2 maybe isn’t the first thing to come to mind for more expensive turntables and cartridges.

But don’t fear—Mani 2 will continue. It’s a great introduction to the possibilities of vinyl, and it’s still a very flexible, very high-performing product.

But now you have another choice: Skoll.

I hope you enjoy!

So glad you did this! My Pro-ject X8 TT has a true balanced wired tonearm and I am wanting to take advantage of it. Just need to get a moving coil cartridge and a Skoll. And the upgraded balanced cable that plugs into the tone arm. Need to butter up my wife first. Looks like there are several Pro-ject TT that have balanced wiring now. Including the latest version of this popular model:

https://pro-jectusa.com/product/debut-pro/

 

[Snowpuppy77]

Teac TN5-BB is what we ended up with. Pro-ject also makes one, but it has a nonstandard connector and will need an adapter cable to standard XLRs. Beyond that, I think it's gigadollar range.

But you don't need a balanced turntable (and an associated MC cartridge) to use Skoll. You don't even need to use the balanced outputs. It works fine with anything.

This is the cable I need to get for the Pro-ject X8 to use it with the balanced input to the Skoll:

https://www.audioadvice.com/pro-jec...MIwYWBzMPugQMVo1VHAR0lHwr5EAQYAiABEgKXLvD_BwE

Same cable would be used with all balanced Pro-jects. Pro-ject seems to have balanced models down to $799 now according to their USA website.

 

[XMarsX]

Order placed for Midgard. I need another amp like I need another herniated disc, but Halo has me super curious.

@Jason Stoddard Is the Halo sound on the XLR output affected by the input type? In other words, RCA vs XLR input in regards to Halo?