2023 Chapter 5
The Most Abused Audio Terms, Part 2: Balanced
“Balanced,” like “Class A,” suffers from nominative determinism—as in, the term itself primes people to prefer it over everything else.
Skeptical?
Nah. Think about it.
Class A? Anyone who can get an “A” doesn’t want a “B,” “C,” or “D.” Like duh. We’ve all been programmed by too many years in school.
Balanced? Again, who picks “unbalanced” when they can avoid it? Well, except for a few avant-garde designers who probably still think mixing OSB and stainless with recycled banana leaves and hand-hewn pine on painted concrete floors is a bright idea.
So, yeah, “balanced.” You hear it. You want it. Because of course you want balanced. It just sounds better!
And, of course, just like “Class A,” the marketing pukes love to slather “balanced” on everything, because you’re biased to think it’s better.
Here’s the catch:
what looks balanced may not be what you expect.
It may be balanced technically, but not in a way that you get any benefit from it being balanced.
Or it might implement balanced in a way that requires more circuitry and conversion back and forth internally, leading you to wonder if that’s really what was intended all along.
Or it may not be balanced…
at all.
Lesson 1: Seeing Isn’t Believing
Here’s the thing. Most people see XLR connectors on a product and assume:
- It’s balanced
- It’s inherently better
- It’s the best way to connect other components
None of these assumptions are necessarily correct.
In fact, when you see XLR connectors on a product, you can pretty much assume only one thing:
that it has balanced connectors on it. It’s entirely possible that those connectors aren’t actually connected to anything balanced.
“Ah come on!” someone says. “No credible manufacturer would just put balanced connectors on a product just for the sake of checking a box, right? I mean, that’s not even possible, is it?”
Well, it is possible, and it does happen. It’s probably not the most common way of dealing with balanced, but this approach—simply hooking up the pin 2 connection from the XLR and ignoring pin 3—is out there.
“Wait a sec, what’s this pin 2/pin 3 nonsense?” someone asks. “What does that mean, and why is it bad…and what other ways are there of dealing with balanced inputs and outputs?”
Glad you asked. It’s probably best to break down some of the typical ways you would deal with balanced inputs and outputs, complete with diagrams.
Balanced Type 1: Balanced Differential. This maintains the balanced differential signal all the way through the product, from input to output. This is the approach that promises the greatest benefits to going balanced—noise rejection, ground independence, and topological advantages that can increase overall performance. And this kind of balanced is what the marketers want you to think about when you see XLR inputs.
A caveat on the diagram: as titled, this is a general approach. Balanced in via XLR, processing done in a balanced stage (in this example an IC, like an LME49724, but there's no reason it can't be discrete.
So why doesn’t everyone do Balanced Differential? It’s simple:
- This kind of balanced means more parts (but not necessarily double, as some of the anti-balanced mafia would like you to think)
- It means more specialized parts, like 4-gang potentiometers and 4-pole switches, for stereo applications
- It also may mean you want to sum to single-ended for best single-ended performance, which means more parts and higher complexity
In short, it’s more difficult to do and more expensive, so not everyone does it. Simple as that. But it’s not the only way to do balanced. Let’s move on.
Balanced Type 2: Balanced Separate. Balanced products don’t have to be differential at all. It’s perfectly fine to simply use two separate gain stages per channel, one for each phase. This kind of balanced has exactly double the parts of an equivalent single-ended circuit, because it’s exactly the same as two single-ended circuits. And it still requires the same specialized parts as a “differential all the way thru” balanced.
As above, this is shown in the diagram using two IC gain stages; there's no reason you can't do this discrete, though.
So what’s the downside here? Well, if you have a balanced source, and if you’re feeding a balanced amp, this might be a perfectly fine approach. But if you have single-ended input, then things may get a bit weird. Here’s why:
- A single-ended source won’t be converted to balanced, so you won’t have true balanced differential output from a single-ended source
- Without a true balanced output, you may not be able to drive some downstream components as well as you’d like
- You also won’t be benefiting inherently from common-mode noise rejection as with a differential amp
In short, it’s more difficult to do, and more expensive, and has some gotchas. We don’t do balanced like this. Others do. And hey, it does have at least one huge advantage over the next kind of balanced: no conversion.
Balanced Type 3: Balanced Converting. This is where a balanced input comes in, and a balanced input goes out, but all processing inside the device is done single-ended. Yes, that’s right—
balanced is converted to single-ended, run through amplification, filtering, etc, and then
converted back to balanced again at the output.
Two diagrams are show above, one converting via an instrumentation amplifier on the finput and a differential amplifier on the output, and one converting via transformers. There are lots of other ways to do this, including line transceivers, discrete instrumentation amplifier, an instrumentation amplifier made of separate ICs, a differential amplifier using a single IC, a discrete stage with differential input, and so on. I just picked a couple typical ways to do it.
“Wait a sec,” you say. “No matter how you do it, that sounds like a whole lotta conversion going on! How can that be good and why the heck would you do that?”
Well, yes, that is a lot of conversion. It’s also the way quite a bit of pro gear operates, and also quite a bit of high-end gear as well. As to why you’d do it, let’s look at the benefits:
- Cuts wayyyyy down on the amount of circuitry inside the device, since it’s only processing in single-ended
- Standard 2-gang pots and 2-gang switches are easier and less expensive to get
- Provides for good SE output; no need for summing, etc as with some differential topologies
- Many options for conversion: transformers, instrumentation amplifiers, line receivers, differential summers, etc.
See the last point. One of the indicators of how popular this approach is, is the sheer number of pre-packaged line receivers, instrumentation amplifiers, balanced-to-SE and SE-to-balanced converters, and, yes, even line transformers out there. There are a ton of options. Converting balanced to SE and SE to balanced with good measurements is just a couple of $1-2 chips away. Or you can get fancier and do your own discrete approach, or you can pick an exotic-core transformer.
Each conversion method has its upsides and downsides, though:
- Canned solutions mean you should look carefully at their in-built feedback network and gain limitations
- Instrumentation amps involve a lot of parts
- Transformers rarely measure at the limit of today’s tech
- Simple solutions (Broskie cathode follower, etc) have their own limitations, middling measurements, and can be noisy if you’re not paying attention
So why would someone choose to do this, except out of sheer cussedness or cheapfulness? Well, if you’re looking at an insanely complex device, that suddenly would need 2x the parts to do balanced, this may be the best solution. Loki Max converts, as does Lokius. When we do this, we let you know, and we let you know why. In the case of Loki Max, it’s the difference between 70 and 140 relays, and $1500 and $2700 or so.
Again, this is not an inherently bad approach. But you should be aware of it—especially when your balanced in and out may be running through a lot more circuitry on the balanced side than you expect.
The next approach? There’s a strong case for “inherently bad” here.
Balanced Type 4: Not Balanced. This is where someone (a cynical marketer, usually) simply hangs XLR connectors on a product, but simply doesn’t hook up pin 3 (the inverted pin). What does this do?
- It fools unsuspecting customers into thinking the product is balanced.
- It saves a ton of money, because you don’t need additional circuitry, fancy/weird parts, etc.
- It usually works (ish), but…
- …may cause compatibility or performance issues with some components.
(3) and (4) are why this approach is inherently bad—it doesn’t necessarily offer the noise rejection of a true balanced circuit, and it may have performance problems if the circuit coming before or after the connector expects real, true, no-kidding balanced.
What kind of performance problems?
Well, consider the example of one of our power amps that uses a balanced input to run mono. If you don’t have an actual balanced output from the product, then you won’t have half of the amp operating. So, it might “work,” but there won’t be any additional power on tap in mono.
Or consider a transformer input circuit that expects both pin 2 and 3 to be active. If pin 3 is simply disconnected, you won’t have a functioning input at all. And tying Pin 3 to ground in that case is dangerous. If you then connect a real balanced source, you could fry half the circuit—or at least put the product that came before it into protection mode. Or it could limp along at high distortion. Or lots of other bad things.
One special case where you can get away with fake balanced is with a 4-pin XLR headphone output (or TRRRS). In those cases, sure, tie the negative outputs to ground, and you can happily connect a 4-pin headphone safely. The catch, of course, is there are no balanced benefits—no higher power, no noise rejection, etc. But it will work.
Bottom line, balanced is fundamentally different. It should be treated correctly—whether that’s with a differential, separate, or converting approach. All are valid. All have their applications. Fake balanced doesn’t. Unfortunately, it’s out there. When in doubt, ask.
Lesson 2: Balanced Isn’t Unquestionably Better (or Worse)
Some audiophiles think that “going balanced” is a natural progression from single-ended, that balanced is always better, and that they should shoot for having everything in their system balanced. Sounds like a logical goal, right?
Wrong.
In reality, things are much squidiger. As in, you can make a state-of-the-art, end-game system from all-balanced components, all-single-ended components, or a mix of both. Sometimes SE sounds better. Sometimes balanced does. Sometimes going balanced means adding a whole lot more conversion steps to your system. Sometimes going SE means performance or noise penalties. Unfortunately, there’s no universal answer.
Want more confusion? Some people have also put forth the idea that balanced is
always worse. They cite the increased complexity, increased thermal noise, higher output impedance, and higher cost as issues. But, in many cases, these arguments are inaccurate or oversimplifications. Balanced doesn’t have to mean complex. There are very simple balanced circuits out there. Noise? Usually 6dB more noise is accompanied by 6dB more gain, so it’s a wash. Higher output impedance? Yes, sure, but this matters most for open-loop or low-feedback all-tube topologies. Higher cost? Well, there are plenty of car-priced single-ended components out there too. Balanced can be done at reasonable cost.
So what are the actual advantages and disadvantages of balanced, and where are the gotchas? Let’s break it down:
Advantages:
- Noise rejection: balanced differential approaches have high common mode rejection, so noise picked up on long lines to and from the product will be rejected. As will internal noise, such as power supply noise, so the product can have lower noise than an equivalent single-ended design. Need to run 25 feet (or more) of cables? Balanced is your best bet. Real balanced, that is.
- Ground independence: balanced differential doesn’t rely on having a super-clean ground, nor on the overall ground performance, as much as SE. This can also help with noise performance, and, perhaps, in transient response (though it’s unlikely this is measurable).
- Power supply rejection: if you’re staying balanced, or summing back to single-ended with an inherently high-PSRR circuit, you can frequently go with a much simpler power supply. Heck, the original Mjolnir and Mjolnir 2 have exactly zero regulation on the output stage, which means they have significant ripple on each of the balanced output phases. But because the ripple is the same, it cancels. And it cancels very well. Without this inherent balanced advantage, Mjolnir would have a signal-to-noise ratio of only about 20-30dB!
- Topological advantages: if you’re talking about a real differential balanced topology, there are other advantages that can increase overall performance. And, in terms of “real differential balanced,” there are many ways to implement this, from simple discrete differential amps to instrumentation amplifier topologies (op-amp or discrete) to specific naturally balanced architectures like supersymmetry, cross-shunt push-pull, and Nexus™. Supersymmetry in particular offers distortion cancellation via cross-coupling complementary stages.
- Woo-woo handwaving: and yeah, let’s face it, balanced is still a bit fancy or weird in the audiophile realm (not so in pro sound), so there’s plenty of woo and bullschiitery about how it “sounds better” by “driving both sides actively” and other really not-supported-by-reality claims. In my experience, yes, balanced can sound better. And it can sound worse. It’s certainly not the only way to get great sound.
Disadvantages:
- Higher complexity: For any given topology, balanced will use more parts than single-ended. This isn’t necessarily bad, though, even for those audiophiles afraid of “complex topologies.” The reality is that you can get really amazing performance from simple balanced circuits, especially if you couple balanced with distortion cancellation, such as in CFP stages or in current-sourced Class A outputs. This is the basis of the Freya+ differential buffer, which offers exceptional performance from a circuit that only has 4 active devices per phase.
- Higher noise: In this case, we’re talking about internal thermal noise, or equivalent input noise, not external noise picked up by long interconnects. And, while this may be technically true, this “additional noise” is usually accompanied by higher gain—which means it’s not really higher. Higher gain means higher noise, because physics. It’s also counterbalanced by the external noise rejection and internal power supply rejection advantages mentioned above. Bottom line, balanced can end up being quieter than single-ended.
- Higher output impedance: Absolutely. Real balanced will have 2x the output impedance of an equivalent SE circuit. This absolutely matters if you’re using a Class A tube OTL circuit with an output impedance of 14-140 ohms. This is really a canard if you’re talking a circuit with an output impedance of 0.01 ohms—that’ll end up being swamped by cabling.
- Summing for SE: Single-ended output from balanced usually requires summing for highest performance. Summing is where you combine the two balanced phases into a single single-ended output. Summing requires additional circuitry, so now you’re looking at higher complexity, no matter how you slice it. Summing can be done in any number of ways, including very simple topologies such as single transformers and two-transistor Broskie Cathode Followers, to super-complex multistage amplifiers (discrete or IC).
- Gain issues: Lots of balanced topologies don’t offer a 0dB gain option without dividers—many of them are x2, or +6dB. Something to consider when going balanced.
With both advantages and disadvantages to balanced (as with everything in life), there are gonna be gotchas. Gotchas like:
- System matching problems: This is really the biggest one. This can be as gross as simply not having the right connectors—as in, for example, a headphone amp that only has balanced output, when all your headphones are 1/4” TRS SE, or your favorite source that only has single-ended output coupled with a preamp where all the other SE inputs are taken. It can be more subtle, too, like if you have a fake balanced product that you’re trying to run into a transformer-input converting balanced product, you may end up with something that doesn’t work very well—or even at all. Using one phase from a differential stage that really needs summed may not result in best performance. But, bottom line, the moment you venture away from the tried-and-true consumer audio connectors (that is, single-ended output on RCAs), you may end up with system matching issues.
- Balanced nervosa: System matching can also be subtle—as in, “well, if I insist on everything being balanced—source, preamp, amps—then everything is best!” Except this isn’t really the case in a world of balanced differential, balanced separate, balanced converting, and fake balanced products. Insisting on balanced may mean you end up with a system that has significant additional conversion steps in its signal path—which may or may not end up sounding better—which may or may not be what you expect. So, yeah, even insisting on balanced isn’t a panacea.
- Bigger, more expensive products: Like it or not, going balanced means bigger, more expensive products, at least until there’s something other than the XLR standard for interconnection. Yes, I know, there are 4.4mm proprietary TRRRS connectors out there for headphones, but those aren’t used for system interconnection. And yeah, I know, there are mini-XLR connectors, but those never really took off other than on some headphones, and there are custom and proprietary DIN cables and whatnot, but locking someone into your own kind of cables is a super dick move, like making a car that only runs on your brand of gasoline. So yeah, when you want a preamp that handles a bunch of inputs and is also balanced, you’re looking at a big box. Bigger is more expensive, no matter how you slice it. More metal is more expensive. More finishing is more expensive. PCBs are costed by surface area, the bigger the pricier. Etc, etc. Seems like a small thing? Not in an age when people don’t want giant racks of gear.
Lesson 3: Overcoming Balanced Nervosa
“With both good and bad aspects to balanced, how do you choose?” someone asks. “You just said that simply insisting on balanced isn’t a panacea. So now what?”
Cool. Let’s cover how to overcome balanced nervosa in one step:
1. Don’t worry about it.
Sounds too simple? It isn’t. Bottom line, most balanced gear will interconnect and work with other balanced gear, and other single-ended gear, just fine. If you like the product, go ahead and go with that. Mix and match. It won’t offend any audio gods, no bolt of lightning will come down from the sky to smite your heresy, and nobody else besides you will notice or care.
“But you just said that some things won’t work together, or won’t work as well as they could together, what about that?” someone continues.
Yes, and turning the volume pot to midpoint on most gear introduces enough thermal noise that precious super-low THD+N measurements go out the window. Do you worry about that?
Or how about those long RCA cables you have running right on top of your power cords? Guess what, those aren’t perfectly shielded, and it may be causing a much higher noise floor. Of course, “Much higher,” is probably still “Quite inaudible.”
Or how about that balanced DAC you just had to have, because it’s balanced. But is it balanced based on hardware for each phase, or is it based on a balanced-output DAC, or is it deriving the balanced output from single-ended? Unless you know the topology, or if the manufacturer specifies clearly what’s going on, you may not know.
Aaaaaannd…as long as it sounds good, does it really matter?
The reality is audio is imperfect, and there are no 100% safe and right and true answers. Not worrying about it so much makes total sense.
“But I just can’t let go,” someone says. “How do I choose balanced or single-ended? Come on, there has to be more than one step of ‘just say **** it!’”
Okay. Fine. Let’s go a bit deeper. But again, these are not rules. Feel free to experiment. You’re not going to be graded on this.
Go for balanced if:
- You need the noise rejection for long cable runs. Trying to run RCAs 25 or 50 or 100 feet? Yeah, maybe you really need to go XLR balanced. Real balanced.
- You’re going to be working with a lot of pro gear in the mix. Pro stuff is usually XLR balanced. Best to keep it there if that’s what you’re working with.
- If you know what kind of balanced it is, how it works together, and you like the way it sounds.
Stay away from balanced if:
- You want least cost for quality. SE does have an advantage in simplicity, so if you’re going for simple, and you want high quality, you have a better shot of getting it from SE.
- You want pretty much universal compatibility with consumer audio gear. Let’s face it, pretty much everything can accept RCA inputs. Balanced-only gear is scarce in the consumer audio space.
- You want it small. Connectors and circuitry can both be smaller on the SE side, so if you’re looking for small, you’re also probably looking at SE.
Now, none of the above is universal, of course. You’re gonna find people who can run 25 feet of RCA without issues, and tiny balanced dongle-DACs that pretty much do everything a stack of components can do.
The real question is, as always: do you like it?
If you do, go back to that single point above:
don’t worry about it.
Schiit On Balanced
This chapter wouldn’t be complete if we didn’t say a few words about why we went balanced in the first place, what our goals are with balanced, and what we mean by balanced when we talk about it today.
How’d we get into balanced?
Because I always liked circlotron-style topologies (aka cross-shunt push-pull). So we did the original Mjolnir, which was the first and only circlotron-style topology headphone amp at the time. It also didn’t have single-ended out, which was really kinda stupid, because it meant it was only compatible with a relatively small selection of headphones at the time.
But, as I said, I liked circlotrons. It’s a super-simple topology that offers very high performance, and I still like it to today. I just think we’ve surpassed it with Nexus, without the balanced-only problem, and without the super-complex power supply. That’s my opinion; I may be crazy.
With respect to goals, our first forays into balanced were always to push performance. First in terms of power output, with Mjolnir and the original Ragnarok, then overall, with Jotunheim. The original Jotunheim measured very, very, very well, especially for the size and price. Jotunheim 2 brought the relatively expensive Nexus™ topology down in price. From there, it went throughout the line—into preamps, into power amps, into inexpensive products like Magnius and Modius.
Nexus is a bit of an odd beast, and worth mentioning. Distilled down to an IC-like diagram, it looks something like this:
It looks kinda like balanced differential (general), but you see the critical distinction: input impedance. Since the feedback doesn't go back to the input, you have two inherently high-impedance nodes. This can be very convenient, especially if you want to put a volume pot in front of it and don't want to use buffers. Another way to get high impedance inputs is to go feedback-free, like we do in the Freya+ differential buffer.
But here's the thing: when you see balanced connectors on our products, they're usually connected to a balanced differential stage.
The only two cases where we do balanced converting (to save on a billion relays and circuitry), we specify what those products are, and what they are doing. I’m talking specifically of Lokius and Loki Max. Anything that has balanced connectors on it has proper interfacing with the balanced connectors, and a reason for them to be there; there’s no “fake balanced” anywhere. And that’s the way it’ll stay. That’s why Folkvangr—despite its high price—didn’t have any balanced connectors at all—not even a 4-pin headphone XLR for convenience. Because it’s not a balanced product, and there’s no reason for the connectors to be there.
Me, personally, what do I use?
Back to (1).
I don’t worry about it.
Last night was a Jotunheim 2. The night before was Syn. The night after might be Folkvangr or Ragnarok 2. Bottom line, if it sounds good, it sounds good.
Even if it is “unbalanced!”
Hence, no Raspberry PI etc.
