[StimpyWan]

@Jason Stoddard,

Were both the Sumo "Ten" and "The Nine" amps designed by James Bongiorno...? Thanks.

 

[ScubaMan2017]

Aww yeah!
Good year ahead.

Please in black!

New developments!? Interesting times, no?

 

[fm540i]

2023, Chapter 1
The Most Abused Audio Terms: “Class A”


It’s time to call out some of the most misused and abused audio terms. Marketers looooovve to abuse terminology, including "multibit" for delta-sigma DACs, and, lately, "discrete" for designs that are 100% op-amp based. No kidding. Look it up. There are lots of marketing idjits out there.

But let’s start with “Class A.”

Because it is—literally—the worst. As a phrase, it sounds great! Everyone likes to use it! Anything else seems like you’re compromising! And so marketing pukes put it on everything, including things where it doesn’t matter, things where it’s trivial, and things that aren’t Class A at all, not even with the most broad definition.

And yeah, I know, I’ve blathered about Class A before, but now it’s time to really break it down, with diagrams and stuff, so you really know what Class A is, what kinds of Class A there are, how to know when Class A matters, and how to know when Class A ain’t Class A at all.

“Wait a sec,” someone is already protesting. “What you do you mean ‘what kinds of Class A there are?’ Isn’t Class A, like, Class A?”

Ahhhahahhhahhahhahaahahahahahaa!

You guys are funny.

Oh no, the Class A confusion is sooooooo much worse than that.


What is Class A (for Real)

The real, old-skool, most-stringent definition of “Class A” is “an analog circuit where the transistors are never cut off.” As in, take a single output device, and run it at the maximum current it can supply. Connect to a load. The current is then shared between the device and its load. Done.

“Well, that sounds easy,” you say. “I don’t see what’s so difficult about that.”

Well, what’s so difficult is…that it’s difficult.

As in:

• iI’s super inefficient
• It’s blazing hot
• And it’s physically impossible when you start getting into serious power

I know you guys hate math, but let’s do a simple calculation. Let’s say you want to make a Class A amp that does 100 watts into 8 ohms. To do that, you need to swing 80 volts across those 8 ohms.

And 80V into 8 ohms means you need to swing 10 amps.

Aside: I = V/R is the equation you’re looking at.​

“So?” you ask, all innocent and non-engineeringy.

Sooooooo…your Class A amp, in its most pure form, needs to run 10 amps across 80V all the time. Which runs us into another equation: P = VI

Or, power = voltage x current.

Or, to put a really sharp point on it, your 100W Class A into 8 ohm amplifier has to dissipate 800 watts at idle.

Read that again: 800 watts. Per channel.

Eight hundred watts constantly. All the time. Two channels? That’s a room heater. That is completely and utterly bonkers. How bonkers? The best audio output transistors are rated for about 100-150W dissipation each. But that’s also given pretty much perfect heatsinking, so figure you’re gonna need at least twice that many—16 per side—to get a viable amp. And that may even be too much. Asking an output device to dissipate 50W all the time is a pretty hard job.

Heatsink? Oh yeah, get a few feet of Aegir heatsink to try to get rid of that heat. Even a more efficient heatsink profile, like the one we used on the old Sumo Ten, with 2.5” deep fins and 27” depth, needed 14” to keep it at 60-65 degrees C.

Transformer? Yeah that’ll be huge too. Oh yeah and it has its own dissipation. So you’ll be dissipating more than 800W for that 100W Class A amp.

“But…but…that’s completely nuts!” You cry. “How can anyone do something like that?”

It’s easy: because they don’t.


Types of Class A

Here’s the thing. That super-simple hypothetical 800W room heater described above is a setup so we can discuss the types of Class A, look at the benefits and drawbacks of each approach, and discuss where they might be used.

Aside: and, in the end, make fun of some stuff that just can’t be Class A, no way, no how, nevergonnahappen.com.​

Type 1: Single Ended Class A. That super-simple room heater? That’s a single-ended Class A amp. It’s arguably the most “pure” because it can be a single transistor, and it can never be driven out of Class A. But it’s also the least efficient.



Some notes on this diagram, and all diagrams following:

• You ain’t gonna build an amp from this. It’s wayyyyyyy oversimplified. Front end gain isn’t shown at all. I mean, the “bias” is just a voltage source here. Voltages aren’t specified. Resistor values aren’t specified. Transistor types aren’t specified. This is just a conceptual rendering.
• However, it shows the basic idea: in the example above, a single transistor sits on top of a current source. The current source sets the quiescent current. Quiescent current is the current flowing through the stage all the time
• To calculate quiescent current in this example, it’s (Vbias-0.6V)/R. 0.6V due to transistor diode drop. MOSFETs and JFETs will be different. Look it up yourself if you want to know details. Sorry, this is not a class.

Cool? Nobody’s gonna try to build the circuit above, and blame me when they use 2N2222 transistors, +/-80V rails, a 1.8V LED as bias, and a 1 ohm current source resistor…and it goes up in flames, right?

Benefits of single-ended Class A:

• Can be very simple
• Can be easy to implement, with very few parts
• If same devices are used top and bottom, helps cancel distortion
• Output transistor cannot turn off, by definition
• Absolutely, 100%, totally, unarguably Class A

Drawbacks of single-ended Class A:

• Oh my god it’s a power hog
• Cannot ever supply more current than what’s flowing through it (low impedances can be challenging)
• Not good for more than fleapower without heroic heat management
• Will need to use more than one transistor for more than fleapower due to power dissipation per device
• Is still dependent on the linearity of the parts being used (Early voltage, nonlinear beta/transconductance still apply)

Anyone who’s owned an Asgard or Asgard 2 knows how inefficient a single-ended Class A amp can be—because that’s exactly what they were. They used current sources and a single output transistor (in Asgard and Asgard 2’s case, a MOSFET) per channel to deliver 1 watt of power into 32 ohms. In doing so, these amps dissipated 30-35W to create those two 1W channels. They ran quite hot, despite using the entire 9” x 6” x 2” aluminum case as a heatsink. And there was no getting more power out of them—when they were out of current, they were out of current. 1W into 32 ohms became 1W into 16 ohms—that’s all they had current for, so the maximum voltage output just halved.

But those early Asgards also show where single-ended Class A can be used: in small, low-power applications like headphone amplifiers. Not 100W, but 1W. Not 8 ohms, but 32 ohms. At these power levels, single-ended Class A might be viable. But even then, it will never run cool or be able to push extra current into demanding loads. So it’s still a tough choice.

How tough? We’re planning to do a Mjolnir 3 this year that can run in pure, unadulterated, single-ended Class A. It uses 8 10A-rated output devices per channel. It has heatsinks that are as big as my head. It runs very hot…and it does 2 watts per channel into 32 ohms.



Yes, this is a 2W/32 ohm amp. Yes, that’s all.

Sounds wimpy? Well, that’s why MJ3 will also have a front panel switch so you can run push-pull Class A. And suddenly it’s an 8W/32 ohm amp.

“Wait, what?” you ask. “How did you suddenly get so much more power?”

Through the magic of push-pull Class A.


Type 2: Push-Pull Class A. What if you could dissipate only, say, 200-400W instead of 800W for that 100W Class A into 8 ohm amp? Although that’s still horrible efficiency, it sure sounds a lot better, right? To do this, you need to change topology to push-pull Class A.

Is it still Class A? Sure. Or it can be ish. It can be gray. Read on. Let’s start with two diagrams, two different ways to get push-pull Class A.



Here’s a classic way to get push-pull Class A: the White Cathode Follower (or, in this case, the White Emitter Follower, since we’re using bipolar transistors). In this case, you sample the current running through the top transistor and use it to modulate the bias of the bottom transistor. Your constant current source is no longer constant. You can turn it off.

And there’s the magic. Now, each transistor can drive the load separately. So your current needed to remain in Class A is reduced by half. Boom! Way higher efficiency.

As with previous diagrams, it’s oversimplified. I’ve left out all details about how you modulate your current source, for example. But it’s detailed enough that the sharp readers will see a problem with this stage: the loss through the sampling resistor. If the sampling R is large (like in a tube stage), the power loss can be significant.)



And here’s another classic way to do push-pull Class A: the totem pole output stage. No sampling resistor to muck stuff up. But what’s with the dual bias? And where’s the bias connected to? And what’s with the IN (BIG) and IN (SMALL)?

Here’s the deal: totem pole output stages are a mess. They can’t really be viewed without their associated voltage and drive, like in this classic JLH 10W Class A amplifier:



Totem pole output stages suffer from their own problems, perhaps the largest of which is hinted at by the IN (BIG) and IN (SMALL) on my diagram. The top transistor is an emitter follower. It has no voltage gain. The bottom transistor isn’t. It has huge voltage gain. So you need to carefully balance the drive between the top and bottom transistor, and/or linearize it with feedback. It doesn’t take a ton of parts to do so, as the JLH amp shows, but it is a limitation.

Benefits of push-pull Class A:

• Much higher efficiency = more power for the same heat
• Still pretty easy to implement
• If same devices are used top and bottom, helps cancel distortion

Drawbacks of single-ended Class A:

• Still inefficient
• Still can’t supply more than 2x quiescent current in examples shown
• Starts getting gray—are you turning off the bottom transistor or not when you’re modulating the current source?

Anyone who has a Valhalla or Valhalla 2 has a White Cathode Follower push-pull Class A amp. I went with the White Cathode Follower arrangement because you can’t run much quiescent current through tubes. While Asgard ran about 250mA, Valhalla ran about 20mA. You can’t get much Class A power with 20mA single-ended. But you can get a lot more with push-pull!

“Wait a sec,” says someone who’s seen a lot of Class A amp schematics. “What I’m looking at doesn’t look anything like what you show above!”

Righto. That’s because most Class A amps, these days, are complementary Class A amps. Which have the HUGE advantage of being able to source more than their quiescent current, but also a HUGE identity problem—which is, are they a “real” Class A amp, or are they just high bias Class AB?

Well, you know Class AB is also Class A below a certain output level, right?

“What!” you cry. “Now I’m totally confused!”

Yeah. And there you go. Let’s talk about complementary Class A.


Type 3: Complementary Class A. The vast majority of large, hot-running speaker amps billed as Class A are complementary Class A amplifiers. The Sumo Ten was a complementary Class A amplifier. Complementary Class A is fantastic for high power speaker amps, because it’s as efficient as push-pull, and it can source much more current than its Class A bias—an order of magnitude or more.

But that’s also where it gets gray, because if it’s sourcing more current than its Class A bias, one transistor or the other is turned off, so is it Class A any longer? Or is it just a high bias Class AB amp?

To illustrate, let’s look at a typical Class AB output stage:



And a typical complementary Class A output stage:



No, your eyes aren’t deceiving you. They are exactly the same. One just has more bias.

“Oh, cool!” someone says. “I’ll just turn up the bias on my Class AB amp and I’ll have a Class A amp!”

No. Actually what you’ll have is something like a smoking crater, because your Class AB amp wasn’t designed to dissipate the kind of heat a Class A amp does. And its bias was probably carefully chosen to minimize the crossover distortion that happens when the transistors turn off.

Aside: why is it bad when transistors turn off? Because it produces distortion. Transistors don’t like to be run down to zero and then cut off, because their gain characteristics aren’t the same at different voltages, especially low voltages, and they aren’t the same at different currents, especially low or no current, so things go kinda whomper-jawed as everything crosses zero. Despite this, Class AB amps can be very linear and sound very good, especially if they have enough bias to run Class A for a watt or so of output.​

So don’t just go turning up your bias. A complementary Class A amp is still different than a Class AB amp, because it had to be designed to dissipate more heat. It’ll have more output devices and larger heatsinks.

Benefits of complementary Class A:

• Similar to a Class AB amp in implementation; familiar to amp designers
• Can source much more current than its Class A bias
• Much better for high power designs; scales easily
• If used within its Class A bias, the output transistors don’t turn off, so it is technically Class A

Drawbacks of complementary Class A:

• Still inefficient relative to Class AB, G, H, D
• Output transistors can turn off when driven beyond Class A bias
• Gray area: is this Class A or high bias Class AB?
• Complementary transistors often aren’t really complementary

So is a complementary Class A amp really Class A? Sure it is, as long as you don’t exceed its quiescent bias. The output transistors won’t turn off, so it’s Class A. Same as a Class AB amp run at low volume. Some Class AB amps have a decent amount of bias on them, and might run Class A for a watt or so.

But if you, say, connect a 2 ohm load and crank it up, that complementary Class A amp may come out of Class A. It may be able to source enough current to run that load, but only at the cost of cutting off the output transistors. That may be a fine compromise for efficiency.

“Wait a sec,” another sharp-eyed reader says. “What’s this ‘complementary transistors aren’t really complementary’ thing?”

Aha. Ho-ho. Someone noticed that we went from all-NPN designs to complementary NPN/PNP designs. Yep. Now we’re using both NPN and PNP transistors. And, while it’s popular to think of a PNP transistor as an “inverted” or “mirror” NPN, it’s actually a completely separate device with different construction. So, even if a manufacturer says, “hey, these devices are the same and work together—they are complementary,” a quick look at the datasheet will have you wondering…


Do these transistors look complementary to you? Yeah. They aren’t exact. They never will be. So if you’re handing off from NPN to PNP, even running in Class A, it’s never a perfect handoff. They aren’t a mirror image. Some will argue that the emitter resistors will swamp any differences. Others will argue that paralleling transistors will improve matters. Or degrade matters. Or won’t matter.

Yeah, I know. This is my world. Wanna join the fun?


Type 4: “Class A” Nonswitching/Sliding Bias/Other Exotics. If you’ve been around this audiophile world a while, you’ve probably heard claims of “Class A without the burn,” AKA various flavors of:

• Nonswitching: keeping a trickle of current flowing thru the transistors so they don’t technically cut off, even if they aren’t really doing anything. While this flavor of “Class A” may be technically true, it’s, well, a bit disingenuous. If one side is trying to source 10A of current, and you have 10mA, or 1/1000 of that amount, running thru the other side, yeah, it’s technically not turned off, but, is it really Class A?
• Sliding bias: just keep adjusting that bias so it matches what you need, and you get something that doesn’t switch. At least in theory. Managing that sliding bias so it doesn’t short you when you need it into a complex and reactive load is a heckuva trick.
• Other Exotics: I’m talking everything from Class AC, mode switching, current dumping, modulated rails, and every other idea under the sun. If it doesn’t technically shut off the transistors, someone has probably tried to call it Class A. And that’s fine.

Sorry, no schematics here, because I never really tried any of these schemas out, and the best of them are probably patented.

Are they Class A? Maybe technically.

Do they sound good? No clue. Depends on which one. Have a listen. If they sound good to you, they’re good. Period. Does anything else matter?


Continuity™. And if you’ve been around this Schiity thread, and around our Schiity website, and around my Schiity talks, you probably have heard of our own weirdo constant-transconductance output stage that tries to extend the benefits of Class A beyond its Class A bias region.

“So why didn’t you lump it in with other ‘exotica?’” someone asks.

Simple: because we’re not claiming that Continuity is Class A.

We’re clearly stating this is a high bias Class AB amp that has a constant-transconductance schema to combat the transconductance droop outside the Class A operation region, and furthermore also uses both NPN and PNP devices on either side of the output, helping to compensate for “complementary” devices that aren’t truly complementary, and finally, the output stage does not have gain, as in it is not a CFP arrangement.

Continuity is our answer to the problem of transconductance droop as posited by Bob Cordell and John Broskie. As far as we know, it’s unique, most likely because it’s still a relatively inefficient, hot-running stage, and one that we’re not claiming the “Class A” moniker for.

You’ll find Continuity in a whole lot of Schiit gear:

• Asgard 3
• Lyr 3
• Lyr+
• Jotunheim
• Aegir
• Tyr

And you’ll also find it in the Nitsch Magni Piety, but in a mutated form that we never really named. In development it was called the “Programmable Output Stage,” because it could be tweaked to simulate a square law device. It’s complex and inefficient—even less efficient than Continuity.

“Well, if Continuity is so great, why don’t you use it in preamps?” someone asks.

LOL, that’s funny.

Here’s why: because doing Class A in preamps is trivial.


Class A Elsewhere

I always laugh when someone asks why we don’t use Continuity in preamps, just as I laugh when a discrete or tube preamp proudly proclaims “Class A” as a marketing feature.

Why?

Simple: because unless the designer is utterly incompetent, every discrete or tube preamp will of course be Class A.

Let’s go back to some math. Yeah I know you hate it. So did I. I barely made it thru differential equations. (And never used them, by the way, so neener.)

Class A math looks a lot different for preamps. Instead of trying to run 80V into 8 ohms and needing 10 amps and 800W to do it, you’re looking at running 6V into 10-100kohms, which requires 0.6-0.06mA, or, with say 30V rails, P=VI, 1.8mW worst case.

Yeah. To summarize the results:

• 100W Class A SE into 8 ohms: 800W
• 2VRMS Class A preamp stage into 10k ohms: 1.8mW

The top one is a room heater. The bottom one you wouldn’t notice that it's on.

Now, of course the preamp example is a cheat. That’s just the dissipation thru the output stage. To be Class A all the way, every stage needs to run enough current to stay in Class A. And there are multiple stages. And you want a buffer of maybe 5-10X just in case, so you might be running 3-6mA, not 0.6mA. And there are two channels. And power supply dissipation. And transformer losses. So in the end your solid state preamp will dissipate a handful of watts. Maybe 5-15W.

But that’s not a room heater. That’s not an implementation where you need 16 150W output transistors and a 20 lb transformer to maybe kinda-sorta work.

In short, there’s no barrier to running preamps—phono or line level—at Class A. Tubes, yeah, they dissipate more power because they run higher voltages, but they’re designed for that. Still not a huge issue. Still gonna be Class A.

In short, preamps: of course they’re Class A. Why wouldn’t they be?

Aside: Well, there’s one reason—op-amps. Some op-amps run so lean that they may, in effect, have Class AB output stages. But those are fairly crappy parts. Newer stuff that’s aimed at audio is much better. So even an op-amp based preamp may be technically Class A.​

Class A Bullschiit

“Okay, fine,” someone says. “I get it, there’s a lot of different ways to do Class A, and some of them are kinda sketchy, and some of them may not be Class A at all. How do I tell?”

Let’s start with the most basic logic:

• Big and hot? Makes other people go, “you’re crazy, dude!” That’s Class A.
• Small and cool? Makes other people ask, “where’s your amp?” Not Class A.

Yes, I know, it’s oversimplified. But go back to that Mjolnir 3 photo and look at the size of the heatsink. That’s for 2 watts of Class A SE. 2 watts. That’s it. Bottom line, if something is small and runs cool and is claiming to be Class A, it really should peg the doubt-o-meter.

Let’s drill down with some more logic:

• Preamp: probably Class A. Kinda weird if it isn’t.
• Headphone amp: might be Class A, could be any flavor from SE to complementary. When in doubt, ask the manufacturer for details. No claim probably means Class AB or op-amp based. Claimed Class A should be subjected to the basic logic above. If someone is telling you a Magni-sized amp is doing 5W Class A SE, you should cross your arms and ask for proof.
• Speaker amp: if more than 50W per channel and claiming Class A, you can safely assume they’re talking about complementary Class A. Even then it may bigger than a Tyr. There may be some smaller Class A speaker amps running SE or push-pull topologies; when in doubt, ask the manufacturer.

Bottom line: pretty much everyone wants to use “Class A.” I mean, nobody wants anything less than an A. But if you want to know what kind of Class A you’re getting—or if it’s Class A at all—it’s worth learning a little bit of the whys and wherefores, and a little bit of math.

I hope this helps a bit!

This is great Schiit, really well written article. Thanks. Very enjoyable article as usual.

 

[US Blues]

Bake-offs???

Bake-offs? I'm still trying to keep my salmon lit.

 

[yonson]

@turtlerod: Out of curiosity, when you tri-amped your Martin Logans, how did you do the split?
.

I'm pretty sure he meant he was running the LCR with a Vidar on each, (I ran my LCR like that for a while as well).

 

[corndog71]

I used 3 vidars to run my Martin Logan front speakers and then switched to two Tyrs. The improvement was as close to a religious awakening as I will ever get. (I usually do not post from my phone because I find it difficult to review my typing. Based on my last post I will never again post from my phone.)

You can always edit your post. I do it often thanks to autoincorrect.

 

[theeclone]

You can always edit your post. I do it often thanks to autoincorrect.

Pobody's nerfect!

 

[JamminVMI]

Same warranty as the old Asgard: 5 years.

Fun fact: I have an old Asgard that has been running probably about 80% of the time for the last 12 years. It's still going. They run hot, but they're plenty reliable. And we can always still fix them.

You tastefully omitted something (which I’m happy to include..) - the “old” thing (in my case, a used Asgard 2 I picked up for a song) is also a brilliant-sounding amp. No wonder it’s been on so much!

 

[JamminVMI]

Dualit 4 slicer, made in the U.K., with timer - to match your Schiit gear -

I have had a 3-slot Dualit Combi (3rd slot is wide with a carrier for sandwich or bagel) since ‘88 or so… Best toasters ever.

 

[dstrimbu]

Has anyone with Vidar/Aegir auditioned Tyr? I've got a set of bi-amped Magnepan 3.6Rs (Aegir/Vidar) and was contemplating Tyr or perhaps going dual Aegir.

I've owned a pair of Vidar monos (now in my bro-in-law's family room), and currently have a pair of Aegir monos (driving LS50s) upstairs and a pair o' Tyr (driving B&W 804 D3s) downstairs.

Have never owned Maggies. But to me Aegir/Tyr would be the perfect bi-amp pair... and I just may try it myself when I get some downtime. Biggest issues are a) available mains power and b) if it works, I need to buy two more Aegir... <g>

Urd first, pls.

p.s. Listening to Sarah McLachlan's Angel via the Tidal PC app right now, through the TrueMultibit DAC in my Jotunheim; balanced out to 2x Aegir, to the LS50s. Simple system. The clarity and imaging, coupled with the black background of this setup - in near-field - is mind-blowing. I have actually been shunning my headphones over the last few days, and I am over-the-freaking-moon satisfied (maybe blown away?!) by what I'm hearing. Aegir is a big wow IMHO.

p.p.s. Another possible experiment would be to bring the Tyr pair upstairs to drive the LS50s... like having huge headphones, indeed.

p.p.p.s. It the Tyrs were upstairs, I wonder if the Aegir monos would (could?) do a good job of driving the B&Ws to moderate volumes. With the stereo RELs augmenting, just maybe. (Dammit. Another weekend shot to schiit!)

So many possible setups, so little time. Also, more and more each day, I favor listening over tweeking. I must be getting old(er). /ds

 

[Timster]

We are weird. We're considering a solar house with rainwater capture (for all house water, no well) in Texas. I'd love to blow up some California heads with that one.

100% rainwater for us here in Oz. 50k litres for house + a 25 k litre overflow to capture excess for the garden. Closest reticulated main is way too far away to contemplate. Have 6.6KW PV solar too, but not off-grid until batteries have a reasonable ROI in single digit years. All hot water is heated via the excess from the solar during the day.

 

[senorx12562]

Yes that is a 1966 Corvette 427/450 roadster. Which I still have.

We are weird. We're considering a solar house with rainwater capture (for all house water, no well) in Texas. I'd love to blow up some California heads with that one.

I always forget about the pre-68 cars. It was the "eyebrows" that threw me. Lol.

Edit: nvm, it's a plastic bag.

 

[OldRoadToad]

For me, it comes down to, can anyone really tell if they're listening to class A, AB, D, etc in blind testing? My guess is, no.

Yup. I don't care about "class" except when it comes to how the company conducts bidness and by that I mean they have "class", not they tell me anything about how to live, etc. or that I will experience an epiphany when I put different colored M&Ms on the tops of my speakers and depending upon where I place them and the colors I choose...or that I use a green felt marker to darken the edges of my CDs to prevent homeless (i.e, "transient") light from escaping and losing "bits" or if I chant the title of Grand Funk Railroad's song title "T.N.U.C." backwards into a mirror then musicalifragelisticexpealidociousness will be achieved and all will be revealed...

If an audio product reproduces and amplifies my music to where I hear my music but do not hear anything else, including my imagination and...the product is pleasing to my eye then I am happy and if I can afford it within my personal "buyers remorse zone o' comfort" then I'm singin' in the rain.

'n' Schiit.

Because all preferences of a personal nature aside regarding how their stuff looks, Schiit does NOT make crap.

ORT

 

[OldRoadToad]

Bake-offs? I'm still trying to keep my salmon lit.

Better than bakelite.

ORT

 

[ArmchairPhilosopher]

I just stumbled upon this and thought some of you might get a kick out of it.