Jason Stoddard
Sponsor: Schiit Audio
2024, Chapter 7
The Most Abused Audio Terms, Part 5: Gain
The final straw came when Evan and Stephan and Tyler and Zach and I were listening to a Skoll in the lab at Corpus Christi.
We started at 40dB of gain, typical of what you need for a moving-magnet cartridge. And, into a Freya+, run in tube gain mode, into a Vidar running Elac 6” bookshelves, we could pretty much turn the Freya all the way up—no significant noise.
At 70dB of gain, though, Skoll had audible hum with Freya at max.
“Oh hell, it’s noisy,” Tyler said.
“Yeah, at 3,300x gain,” I told him.
He looked at me blankly.
“70dB is about 3300x gain,” I said. “From the Skoll.”
“Wow, that’s a lot,” Tyler said.
“But actually, it’s more like 12,000x,” I said. “Because Freya has 4x gain in tube mode.” Except I realized this didn’t tell the whole story, either. “Wait a sec, there’s also the 20x gain of Vidar, so we’re looking at about 240,000x for the whole system. So yeah, it hums.”
“Whoa, I’m glad I didn’t drop the needle,” Stephan said.
“It’s like, picking up hum out of the air?” Evan asked.
“Out of the air, from the wires in the walls, anywhere,” I said. “240,000x gain is insane. That’s like 102dB total system gain. Which is more than the dynamic range of a CD.”
“A what?” Tyler ribbed, grinning.
Yeah bite me. I remember the time when “perfect sound forever” meant 16 bits, 96dB dynamic range, far far far ahead of 60 or so dB of a typical phono system.
“A typical streaming recording on any service you choose,” I said. Because most are still 16 bit. And if they were recorded in the pre-digital era, they might have significantly less resolution than that.
“Why do you need so much gain?” Zach asked.
“Because phono is basically amplifying a tiny signal stamped into vinyl, so it needs a ton of gain. Typical CDs will do 2V at 0dB, but phono was originally spec’d lower, like 400mV, and that’s after 40dB of gain at 1kHz, and our phono preamps have tons of headroom so you can run them at 2V or 4V if you want, but you still have to choose the right gain…”
I noticed the blank looks and trailed off.
And I remembered how many questions we get about “what is 2V RMS, and what does dBV mean, and I heard you don’t have to have a preamp at all, and why do you have to turn up a relay ladder attenuator more than a potentiometer to get the same volume…”
And it was at that point that I realized I needed to do this chapter on gain.
So let’s talk gain. Should be simple, right?
Riiiiiiighhht.
All Pain, No Gain
Let’s start with this: how much gain do you really need?
The answer is simple: you should have enough gain to run your amplifier to maximum output.
Sure, you’re gonna be running your amp at less than maximum output in most cases, because you value your hearing, and because your neighbors might come over and beat you up you if you blasted it all the time. But you should be able to hit maximum output when you want to.
And by maximum output, I mean, you should be able to run your amplifier to the rails, or, in engineering terms, to clipping. As in, to the Vmax shown here:
So how do you figure out how much gain you need?
If you know what voltage your DAC puts out, the gain of your preamp and amp, and where the amp clips, it’s easy to math it up. Like this:
Where:
Then, if:
you have enough gain in the system to hit maximum output.
“Ho hol hold on a sec,” you splutter. “I don’t know what voltage my DAC puts out. Preamp gain is in dB, the amp has a “sensitivity” rating, and there’s nothing about clipping, just maximum power.”
Yeah, that’s because we (the audio industry) screwed it all up. Instead of specifying everything in one unit that invites easy arithmetic, we fragmented it into a bunch of different specs, including:
But seriously, we’re making this way too hard.
Let’s look at things solely in terms of Vrms and arithmetic gain, using real numbers:
So, in this case:
And if you look at this in terms of Vidar, where:
It’s easy to see you have plenty of gain in the system.
“Wait a sec,” someone says. “Where is this +42V and -42V coming from? How does that relate to 30Vrms? Why in Vrms at all? How does this relate to watts?”
Got it. More math incoming:
But it can’t. Vidar’s rails aren’t that big, so you have tons of gain.
In fact, you have more than 5x the gain you need to drive the system to maximum output, at least due to this basic math. According to these numbers, you could easily run Freya or Kara in X1 or passive gain, and still have more than enough gain.
“Well, that’s why you use passive or buffer preamps,” some pedants say. “That’s clearly all you need. You just proved it!”
Yeah. Ah, weeeeeelllll…no.
Why?
Let’s talk about that.
The Myth of “Passive is All You Need”
The high output of modern digital sources (2 Vrms single-ended, 4 Vrms balanced are typical), coupled with the high gain of most modern amplifiers (20-50x), lead some to proclaim, “there’s no need for an active preamp! Passive or buffer is all you need!”
Yeah, well, until you listen a bit more.
Or have a turntable.
Or prefer the way an active preamp sounds.
Or have a bunch of older recordings.
Here’s the thing, passive preamp proselytizers: I was where you were. This was about 30 years ago. I ran the numbers. I went, “yeah, don’t need gain, this too much gain thing is dumb,” and hooked up a CD player to a power amp directly through a potentiometer, expecting nirvana…
…and it was…well, meh.
I found that, despite my calculations, I couldn’t run even a relatively modestly powered, high gain amp up to its limits with a passive approach. (For those interested, we’re talking a Sumo Nine Plus, which has 26dB gain and 60W output).
Why?
Simple: because assuming that a digital recording will deliver a 0dB peak output at maximum is not always a great idea. Not all music is compressed, sausaged-out, gain-normalized pop. Older recordings were frequently mastered at 3-6dB lower than peak. And better recordings today frequently use less compression as well, for a lower average level.
Or, in short, because music isn't perfect sine waves. It looks more like this:
“3-6dB doesn’t sound like a lot,” you say, over crossed arms.
Yeah.
Except 3dB down is only 70% of full output.
6dB down is half.
Oh, and since power is squared—as in, you get 4x the power for 2x the voltage—3dB down is half output, and 6dB is ¼.
Like this:
Starting to see why you might want some extra gain here?
The need for extra gain isn’t a surprise to classical music listeners. Many classical recordings have average levels 10-20dB lower than peak, to accommodate musical dynamics.
And then there’s phono. Many phono preamps are designed for a peak output of 400mV RMS, not 2V. That’s 12dB below “modern” devices, or a miniscule 1/16th the power out of your amp, unless you have more gain in the system. So you may want more gain than you expect there, too. Suddenly a Kara in high gain makes a ton of sense.
Oh yeah, and then there are portable devices. Many of these do 1V, not 2V, so that’s 6dB down. More gain is nice there as well.
And then there is the widely varying spec of digital output devices themselves. They should all be 2Vrms for single-ended, 4Vrms for balanced, but that doesn’t keep companies from doing 1.5Vrms products, for, say, voltage-limited applications (we did in early Modis, for example), and it doesn’t stop other companies from doing “hot” outputs of 3.5V in SE in 7V in balanced.
Oh yeah and you may want to convert from single-ended to balanced or vice-versa, so an active preamp may be necessary. Or maybe (gasp) you just like the way an active preamp with gain sounds.
So things can’t be over-generalized. Maybe a passive preamp is for you…and maybe it isn’t. Have a listen. Maybe you have enough gain. And maybe you don’t.
The main thing to remember is that neither answer is “right” or “wrong.”
Have fun with gain in your preamp…or without.
Life Below Clipping
“Okay, so how about when I’m not running my system at max,” someone asks. “You already said that if I did that I’d be deaf and my neighbors would come and beat me up.”
Well, maybe I was being a bit hyperbolic, but yeah. The reality is that most systems are going to be loafing along at 1/10 to 1/100th of its maximum output most of the time. Or in dB, -20dB to -40dB relative to its maximum output.
So how do we “lose” this gain? With attenuation.
Pick your poison:
A relay ladder is the best, offering performance that is beyond the capability of an APx555 to measure, but it’s relatively expensive to implement, complex, large, and requires a microprocessor and firmware. Potentiometers can be very good and are familiar, but they all will suffer from some channel matching problems at the very lowest attenuation levels. We use both depending on size and cost. Other companies offer solutions based on switched resistors, volume control chips, digital attenuation or other technologies.
Attenuators all have the same basic goals, though:
Great question. Time for some tables. Let’s look at a very good Alps audio-taper potentiometer:
“Whoa!” you say. “What’s with the multiple curves? And what’s this about audio taper? What does it all mean?”
Well, thankfully Alps’ diagram is plotted in percent output per rotation, which makes it easier to figure out. “50” on the diagram is noon, or straight up.
For a 15A taper, the most common audio taper, it gives 18%, or 0.18x total output. That’s an attenuation of only about -15dB.
For a 10A taper, it’s about 9%, or 0.09x, or -21dB.
Oh, and “why audio taper?” Look at this B1B (linear) taper.
For B taper, it’s 50% at 50%, or only -6dB.
Which translates to instant blast of sound as soon as you turn the knob. You’ll notice the audio tapers push most of the attenuation to the front end of the knob, allowing finer control at lower volumes.
This is why you don’t use linear potentiometers. Even with a “law-faking resistor,” they are pretty rough to use at lower levels.
But back to the numbers. Let’s compare to a typical relay attenuator at half, or noon, or “pointing straight up” for a typical front panel control:
Summarizing:
So yes, you have to turn it up more.
“Oh why oh why you do this to us?” someone says. “It’s weird and strange! Doesn’t it hurt the preamp to turn it up more?”
We do this simply because a relay ladder is the best attenuation method we know of. It’s literally unmeasurable on an APx555, and it offers precise 1dB or 0.625dB steps that are perfectly channel matched all the way down.
And “hurting it by turning it up more?”
LOL! You guys have some weird ideas about how amps work.
When you turn “up” a potentiometer, you’re merely controlling the input to the amplifier. The amp doesn’t have variable gain—it runs the same gain, full out, all the time. It doesn’t care about attenuation. Run a relay ladder all the way up—it’s fine. It doesn’t change the stress on the amplifier in any way.
“But what if I’m only a couple of clicks from the top?” someone asks.
Cool, fine. Note the Saga 2 table below. And note that 3 clicks down is only about 70% of total output, and 6 clicks is half. Relax. It’s OK. You’re not going to hurt anything.
“Well, couldn’t you make the relay ladder work more like a potentiometer?” some ask. “I like the way a potentiometer works.”
Short answer: Yes, but….
Longer answer: Yes, but let’s discuss your perception here. We’ve established that:
Given that, yes, there are ways to make it act more like a pot, but it’s all going to be in fakery. As in, you could simply drop in an audio-taper pot for the volume position sensor, and it would behave like an audio-taper pot. However, this will result in massive clicking when first turning the knob, and infrequent clicks once you get past noon. You could also fake it by skipping steps at first (basically an algorithmic cheat).
But the way a relay ladder works is inherently consistent, as in, each click is a fixed amount of attenuation (1dB for our 64-step products, 0.625dB for the 128-step products). Ironically, some people like the 64-step products better, thinking the 128-step products are too “fine” in the degree of control. I wonder if it’s possible to do a 32-step attenuator in 2dB steps—after all, many stepped attenuators are even more coarse than this—but I’d rather have more control, myself.
In short, yes, we could make it more like a potentiometer. But it wouldn’t improve it. At least not in our experience.
So, relax. Keep turning up that relay attenuator until you smile. You’re not hurting anything at all.
Going Negative
“Well, what about negative gain?” someone asks. “When you pick negative gain, doesn’t the amp, like, negatize the input in an active way? Where does that fit?”
Sigh.
While there are negative gain active stages, the reality is the best way to get negative gain is simply to divide the input down, or, even better, divide it interstage if you don’t have overall feedback to worry about, because then you don’t have to worry about input impedance or additional amp stages—but that’s getting a bit too deep for a chapter on gain.
So negative gain = fixed divider, in about 99.999999% of cases.
Yes, I understand that we use this terminology in Magni and other amps that have an attenuation setting, but to be clear: this is attenuation, not gain. A magic active stage is not needed to shrink a signal; you can simply divide it down with two resistors.
And yes, I understand that “negative gain,” or “fixed attenuation,” is an important feature for some headphones, in order to get enough volume range on small potentiometers. It’s less important if you’re talking relay ladder attenuators with perfect tracking.
“So why use potentiometers at all? Why have negative gain? Why not just use relay attenuators everywhere?” the questions are like a barrage.
Simple: cost.
Relay attenuators cost more to implement, plus they need a microprocessor. They’re going to cost 15-40x more than a small potentiometer, depending on number of steps, resistors used, and sensing arrangement.
Sounds expensive? Oh no. You can buy special audiophile potentiometers that cost from 2-20x more than a relay ladder!
So yes, for entry-level components, if you’re looking for a good volume control range that includes good tracking at low volumes, you’re probably looking at a negative gain (fixed attenuation) option. And there’s nothing wrong with that.
Bringing the Noise
“Gain, attenuation…I don’t get it,” someone says. “Why worry about any of this? Put on the music, crank the knob, if it sounds good, who cares?
Simple: because of the relationship between gain and noise.
The more gain, the more noise.
Period.
This isn’t us being dicks, this is a physics thing. Remember the Skoll example that started this chapter? It was quiet at 40dB and noisy at 70dB. Which isn’t super surprising when you realize you’re going from 100x gain to 3,333x gain.
You can see it in Skoll’s specs: signal to noise ratio drops with each increase in gain, usually fairly proportional to gain increase.
So you really don’t want to use any more gain than you really need, in order to keep the noise floor down.
Now, in the case of a modern digital system, it’s going to be much less dire than a phono preamp. It’s entirely possible to have a speaker system with 20-30dB excess gain and have it be essentially noise-free in the room.
But it’s also possible that, with other transducers, like super-high-efficiency IEMs, gain matters. If you’re looking at a 105dB/mW or 115dB/mw product, you may need low gain, or negative gain, for zero noise floor.
“Well, can’t you turn down the volume and turn down the noise?” someone asks.
Welllllllll…maybe.
Here’s the thing: if the noise is coming from your source, yes, turning down the volume control will attenuate the noise. You can hear this on old recordings, etc, that have inherently high noise floor, or from a phono system, which has an, er, less than state-of-the-art noise floor.
But if the noise is coming from the amplifier, you’ll need to switch to low gain to reduce the noise floor…
…maybe!
Because this is assuming that the amplifier is reducing gain by increasing feedback. Increased feedback decreases gain…and the noise floor.
Aaaaaaand, even if the amp is using feedback to reduce the noise floor, there’s a limit to how low its noise floor can go.
“Wait, what?”
Look into equivalent noise resistance. Thermal noise has limits. The lower the equivalent noise resistance is, the lower the potential noise floor. This is why phono preamps frequently use heroic methods like paralleling a bunch of input devices to lower the equivalent noise resistance. It’s also why the impedance of the RIAA network in a passive phono preamp matters. It’s also why any feedback used in a phono preamp is usually very low impedance. It’s also why potentiometers have maximum noise at about their midpoint—the source impedance is highest in that case.
The relationship between gain and noise also explains a lot of the design decisions in audio that can seem, well, a bit crazy.
As an example, someone asked why all of our amps don’t have the same gain, and why gain varies between our headphone amp products. Couldn’t we make them all the same?
Short answer: yes, probably (there may be stability constraints—wayyyyyy too much to get into for this chapter).
But they have different gains, mainly due to the need to drive the amp to max output, and to balance that with noise. Our lower-output amps typically have lower gain, because they don’t need as much gain to reach maximum output…and also because the lower the gain, the lower the noise. For a nearfield amp like Gjallarhorn, a lower gain than Vidar makes sense, because you may be sitting very near the speakers. You don’t want any chance of hearing hiss or hum, so lower gain makes sense. Rekkr is lower still.
And in some cases, we do multiple gains, from Magni to Ragnarok, so you can select the right gain for the system you’re using. In the case of Ragnarok, you can go from a gain of 2 to a gain of 30—a vast range—simply because Ragnarok can run both headphones and speakers from the same amp stage. A gain of 30 is a bit bonkers for headphones, but perfect for speakers, while a gain of 2 is silly for speakers, but right for sensitive headphones.
So yes, gain matters. You can have too much, just as you can have not enough.
How can you tell? It’s easy:
Too much gain =
In this case, maybe not. Bottom line, if you’re happy with your system—if it has enough volume for your needs, you’re fine.
Conversely, someone else will say, “My system is noisy, what do I do?”
Well, if you have adjustable gain, set the amplifier or preamp (or both) at lower gain. Did that help? If not, you may have a different problem. A high level of hum can indicate a ground loop.
Also, if you’re using phono, remember that phono isn’t, well, exactly at state of the art levels of performance. Some hiss and hum (and pops) are part of the package. That said, choosing a lower gain setting on the phono preamp might help.
“Sounds like you need to do a chapter simply on noise,” someone says. “Because there are sure a lot of questions on that.”
And yeah, you’re probably right.
Until then, I hope you, er, gained something from this latest chapter!
The Most Abused Audio Terms, Part 5: Gain
The final straw came when Evan and Stephan and Tyler and Zach and I were listening to a Skoll in the lab at Corpus Christi.
We started at 40dB of gain, typical of what you need for a moving-magnet cartridge. And, into a Freya+, run in tube gain mode, into a Vidar running Elac 6” bookshelves, we could pretty much turn the Freya all the way up—no significant noise.
At 70dB of gain, though, Skoll had audible hum with Freya at max.
“Oh hell, it’s noisy,” Tyler said.
“Yeah, at 3,300x gain,” I told him.
He looked at me blankly.
“70dB is about 3300x gain,” I said. “From the Skoll.”
“Wow, that’s a lot,” Tyler said.
“But actually, it’s more like 12,000x,” I said. “Because Freya has 4x gain in tube mode.” Except I realized this didn’t tell the whole story, either. “Wait a sec, there’s also the 20x gain of Vidar, so we’re looking at about 240,000x for the whole system. So yeah, it hums.”
“Whoa, I’m glad I didn’t drop the needle,” Stephan said.
“It’s like, picking up hum out of the air?” Evan asked.
“Out of the air, from the wires in the walls, anywhere,” I said. “240,000x gain is insane. That’s like 102dB total system gain. Which is more than the dynamic range of a CD.”
“A what?” Tyler ribbed, grinning.
Yeah bite me. I remember the time when “perfect sound forever” meant 16 bits, 96dB dynamic range, far far far ahead of 60 or so dB of a typical phono system.
“A typical streaming recording on any service you choose,” I said. Because most are still 16 bit. And if they were recorded in the pre-digital era, they might have significantly less resolution than that.
“Why do you need so much gain?” Zach asked.
“Because phono is basically amplifying a tiny signal stamped into vinyl, so it needs a ton of gain. Typical CDs will do 2V at 0dB, but phono was originally spec’d lower, like 400mV, and that’s after 40dB of gain at 1kHz, and our phono preamps have tons of headroom so you can run them at 2V or 4V if you want, but you still have to choose the right gain…”
I noticed the blank looks and trailed off.
And I remembered how many questions we get about “what is 2V RMS, and what does dBV mean, and I heard you don’t have to have a preamp at all, and why do you have to turn up a relay ladder attenuator more than a potentiometer to get the same volume…”
And it was at that point that I realized I needed to do this chapter on gain.
So let’s talk gain. Should be simple, right?
Riiiiiiighhht.
All Pain, No Gain
Let’s start with this: how much gain do you really need?
The answer is simple: you should have enough gain to run your amplifier to maximum output.
Sure, you’re gonna be running your amp at less than maximum output in most cases, because you value your hearing, and because your neighbors might come over and beat you up you if you blasted it all the time. But you should be able to hit maximum output when you want to.
And by maximum output, I mean, you should be able to run your amplifier to the rails, or, in engineering terms, to clipping. As in, to the Vmax shown here:
So how do you figure out how much gain you need?
If you know what voltage your DAC puts out, the gain of your preamp and amp, and where the amp clips, it’s easy to math it up. Like this:
Where:
Vdac = DAC output at 0dB digital signal
Gp = preamp gain
Ga = amp gain
Vmax = maximum amp output
Then, if:
Vdac * Gp * Ga > Vmax
you have enough gain in the system to hit maximum output.
“Ho hol hold on a sec,” you splutter. “I don’t know what voltage my DAC puts out. Preamp gain is in dB, the amp has a “sensitivity” rating, and there’s nothing about clipping, just maximum power.”
Yeah, that’s because we (the audio industry) screwed it all up. Instead of specifying everything in one unit that invites easy arithmetic, we fragmented it into a bunch of different specs, including:
- Volts RMS, or Vrms, for source components like DACs
- Decibels, or dB for preamps and amps
- Sensitivity, which is Vrms for max output, for amps
- Watts, for amplifiers
Aside: and that’s not even talking about some companies that don’t clearly specify output levels of their source gear, or rate things in terms of peak watts or total watts, or don’t specify the gain of their preamps at all, or use professional references like dBu. It’s a mess.
Another-side: professionals, I know why you use dBV and dBu, but outside of production, it’s totally confusing and insane. Nobody knows what a +4dBu signal is. I get it, it came about during the tape era, because you needed to know when you were oversaturating the tape, and continued into the digital era, because it was way way way more important to know if you were overdriving something, but it’s a real pain in the arse when it comes to consumer audio.
But seriously, we’re making this way too hard.
Let’s look at things solely in terms of Vrms and arithmetic gain, using real numbers:
Vdac = 2 Vrms (typical of SE output products)
Gp = 4x (like Kara/Freya)
Ga = 20x (like most amps)
So, in this case:
Vmax = 2 Vrms * 4 * 20 = 160Vrms total potential system output
And if you look at this in terms of Vidar, where:
Vmax = 30Vrms (typical of Vidar into 8 ohms, stereo)
It’s easy to see you have plenty of gain in the system.
“Wait a sec,” someone says. “Where is this +42V and -42V coming from? How does that relate to 30Vrms? Why in Vrms at all? How does this relate to watts?”
Got it. More math incoming:
- Why Vrms, or Volts RMS? Because it relates directly to Wrms, or Watts RMS, which is what all good amplifiers are rated in. 100 Watts RMS equates to about 28 Volts RMS into 8 ohms, thanks to the power equation of P = V^2/R. Vidar does at bit more, so let’s use 30V RMS here.
- What’s this 42V stuff? Volts RMS translates to Volts peak-peak by multiplying by 2.83, so Vp-p = 2.83*Vrms. So, for 30V RMS, you’re looking at about 84V p-p, or, for a typical bipolar power supply, about +/-42V or so.
But it can’t. Vidar’s rails aren’t that big, so you have tons of gain.
In fact, you have more than 5x the gain you need to drive the system to maximum output, at least due to this basic math. According to these numbers, you could easily run Freya or Kara in X1 or passive gain, and still have more than enough gain.
“Well, that’s why you use passive or buffer preamps,” some pedants say. “That’s clearly all you need. You just proved it!”
Yeah. Ah, weeeeeelllll…no.
Why?
Let’s talk about that.
The Myth of “Passive is All You Need”
The high output of modern digital sources (2 Vrms single-ended, 4 Vrms balanced are typical), coupled with the high gain of most modern amplifiers (20-50x), lead some to proclaim, “there’s no need for an active preamp! Passive or buffer is all you need!”
Yeah, well, until you listen a bit more.
Or have a turntable.
Or prefer the way an active preamp sounds.
Or have a bunch of older recordings.
Here’s the thing, passive preamp proselytizers: I was where you were. This was about 30 years ago. I ran the numbers. I went, “yeah, don’t need gain, this too much gain thing is dumb,” and hooked up a CD player to a power amp directly through a potentiometer, expecting nirvana…
…and it was…well, meh.
I found that, despite my calculations, I couldn’t run even a relatively modestly powered, high gain amp up to its limits with a passive approach. (For those interested, we’re talking a Sumo Nine Plus, which has 26dB gain and 60W output).
Why?
Simple: because assuming that a digital recording will deliver a 0dB peak output at maximum is not always a great idea. Not all music is compressed, sausaged-out, gain-normalized pop. Older recordings were frequently mastered at 3-6dB lower than peak. And better recordings today frequently use less compression as well, for a lower average level.
Or, in short, because music isn't perfect sine waves. It looks more like this:
“3-6dB doesn’t sound like a lot,” you say, over crossed arms.
Yeah.
Except 3dB down is only 70% of full output.
6dB down is half.
Oh, and since power is squared—as in, you get 4x the power for 2x the voltage—3dB down is half output, and 6dB is ¼.
Like this:
| Level | Vout, percent of full scale | Power Output, 100W rated amp |
| 0dB | 100% | 100W |
| -3dB | 71% | 50W |
| -6dB | 50% | 25W |
| -10dB | 32% | 10W |
| -20dB | 10% | 1W |
Starting to see why you might want some extra gain here?
The need for extra gain isn’t a surprise to classical music listeners. Many classical recordings have average levels 10-20dB lower than peak, to accommodate musical dynamics.
And then there’s phono. Many phono preamps are designed for a peak output of 400mV RMS, not 2V. That’s 12dB below “modern” devices, or a miniscule 1/16th the power out of your amp, unless you have more gain in the system. So you may want more gain than you expect there, too. Suddenly a Kara in high gain makes a ton of sense.
Aside: Run a Bifrost into a Kara on high gain into a Gjallarhorn—a 10 watt amplifier—and you hit Vmax at almost 3:00 on the Kara’s knob. Yes. High Gain. Yes. 10W. Yes. Relay ladders are crazy—Kara has almost 20dB more, or 10x more output, at 3:00 on the knob. We’ll get into the details of that a bit later.
Oh yeah, and then there are portable devices. Many of these do 1V, not 2V, so that’s 6dB down. More gain is nice there as well.
And then there is the widely varying spec of digital output devices themselves. They should all be 2Vrms for single-ended, 4Vrms for balanced, but that doesn’t keep companies from doing 1.5Vrms products, for, say, voltage-limited applications (we did in early Modis, for example), and it doesn’t stop other companies from doing “hot” outputs of 3.5V in SE in 7V in balanced.
Oh yeah and you may want to convert from single-ended to balanced or vice-versa, so an active preamp may be necessary. Or maybe (gasp) you just like the way an active preamp with gain sounds.
So things can’t be over-generalized. Maybe a passive preamp is for you…and maybe it isn’t. Have a listen. Maybe you have enough gain. And maybe you don’t.
The main thing to remember is that neither answer is “right” or “wrong.”
Have fun with gain in your preamp…or without.
Life Below Clipping
“Okay, so how about when I’m not running my system at max,” someone asks. “You already said that if I did that I’d be deaf and my neighbors would come and beat me up.”
Well, maybe I was being a bit hyperbolic, but yeah. The reality is that most systems are going to be loafing along at 1/10 to 1/100th of its maximum output most of the time. Or in dB, -20dB to -40dB relative to its maximum output.
Aside: This translates very well to the subjective experience of a relay-ladder preamp like Kara or Freya. Both are -40dB at noon, -20dB around 2:30 on the knob.
So how do we “lose” this gain? With attenuation.
Pick your poison:
- Relay attenuator. The best overall solution. Uses only resistors and switches (relays) in the signal path. Extremely precise control and pretty much perfect volume matching all the way up and down the range. The price for this perfection is hearing the clatter of relays when you change volume, and momentary drop-outs as the relays change (but no drop-outs when you’ve settled on a volume, of course).
- Switched resistor attenuator. Perhaps the most technically perfect solution. Uses only resistors and switches in the signal path. However, typically has less steps than a relay attenuator, so volume adjustment can be “too loud” or “too soft” between adjacent positions. Also no easy way to remote control.
- Potentiometer. Inexpensive and familiar. Uses a tapered carbon or plastic element to divide the input signal down. Easy to use and simple to implement, but channel matching can suffer at low volume levels. Larger (27mm, for example) potentiometers are usually better than smaller (9mm, for example) ones, but are much more expensive.
- Volume control chip. Some IC manufacturers offer integrated “volume controls,” which use electronic switches on a single chip to replicate the function of a relay attenuator. None offer the performance of a relay attenuator.
- Digital attenuation. Loses resolution at higher attenuation levels. You can also simply scale the input in the digital domain, but even the most advanced algorithms don’t evade resolution loss at lower levels.
- Others. Companies have also used technologies such as LDRs and autoformers for attenuation. LDRs, in our experience, have high measured distortion, and autoformers will be limited by the number of taps, as with a switched resistor attenuator.
A relay ladder is the best, offering performance that is beyond the capability of an APx555 to measure, but it’s relatively expensive to implement, complex, large, and requires a microprocessor and firmware. Potentiometers can be very good and are familiar, but they all will suffer from some channel matching problems at the very lowest attenuation levels. We use both depending on size and cost. Other companies offer solutions based on switched resistors, volume control chips, digital attenuation or other technologies.
Attenuators all have the same basic goals, though:
- Preserve signal quality (don’t add distortion or noise)
- Reduce the size of the input signal
- Provide a wide range of control
- Match the two stereo channel levels
Great question. Time for some tables. Let’s look at a very good Alps audio-taper potentiometer:
“Whoa!” you say. “What’s with the multiple curves? And what’s this about audio taper? What does it all mean?”
Well, thankfully Alps’ diagram is plotted in percent output per rotation, which makes it easier to figure out. “50” on the diagram is noon, or straight up.
For a 15A taper, the most common audio taper, it gives 18%, or 0.18x total output. That’s an attenuation of only about -15dB.
For a 10A taper, it’s about 9%, or 0.09x, or -21dB.
Oh, and “why audio taper?” Look at this B1B (linear) taper.
For B taper, it’s 50% at 50%, or only -6dB.
Which translates to instant blast of sound as soon as you turn the knob. You’ll notice the audio tapers push most of the attenuation to the front end of the knob, allowing finer control at lower volumes.
This is why you don’t use linear potentiometers. Even with a “law-faking resistor,” they are pretty rough to use at lower levels.
But back to the numbers. Let’s compare to a typical relay attenuator at half, or noon, or “pointing straight up” for a typical front panel control:
| Type | Attenuation at Half | Attenuation at Half, dB |
| Alps B1 (linear taper) | 0.5x | -6dB |
| Alps 15A | 0.18x | -15dB |
| Alps 10A | 0.09x | -21dB |
| Lyr, Saga 64-step relay attenuator | 0.025x | -32dB |
| Freya, Kara 128-step relay attenuator | 0.01x | -40dB |
Summarizing:
A Kara or Freya has about 1/10 the output of a potentiometer-based product when both are set with the volume control pointed straight up.
So yes, you have to turn it up more.
“Oh why oh why you do this to us?” someone says. “It’s weird and strange! Doesn’t it hurt the preamp to turn it up more?”
We do this simply because a relay ladder is the best attenuation method we know of. It’s literally unmeasurable on an APx555, and it offers precise 1dB or 0.625dB steps that are perfectly channel matched all the way down.
And “hurting it by turning it up more?”
LOL! You guys have some weird ideas about how amps work.
When you turn “up” a potentiometer, you’re merely controlling the input to the amplifier. The amp doesn’t have variable gain—it runs the same gain, full out, all the time. It doesn’t care about attenuation. Run a relay ladder all the way up—it’s fine. It doesn’t change the stress on the amplifier in any way.
“But what if I’m only a couple of clicks from the top?” someone asks.
Cool, fine. Note the Saga 2 table below. And note that 3 clicks down is only about 70% of total output, and 6 clicks is half. Relax. It’s OK. You’re not going to hurt anything.
| Vol Position | Relay Input, Binary | Attenuation, A | Note | Decimal |
| 0 | 111111 | 1 | Max vol | 63 |
| 1 | 111110 | 0.89 | 62 | |
| 2 | 111101 | 0.79 | 61 | |
| 3 | 111100 | 0.707 | 60 | |
| 4 | 111011 | 0.63 | 59 | |
| 5 | 111010 | 0.562 | 58 | |
| 6 | 111001 | 0.5 | 57 | |
| 7 | 111000 | 0.446 | 56 | |
| 8 | 110111 | 0.398 | 55 | |
| 9 | 110110 | 0.354 | 54 | |
| 10 | 110101 | 0.316 | 53 | |
| 11 | 110100 | 0.281 | 52 | |
| 12 | 110011 | 0.251 | 51 | |
| 13 | 110010 | 0.224 | 50 | |
| 14 | 110001 | 0.199 | 49 | |
| 15 | 110000 | 0.177 | 48 | |
| 16 | 101111 | 0.158 | 47 | |
| 17 | 101110 | 0.141 | 46 | |
| 18 | 101101 | 0.125 | 45 | |
| 19 | 101100 | 0.112 | 44 | |
| 20 | 101011 | 0.1 | 43 | |
| 21 | 101010 | 0.089 | 42 | |
| 22 | 101001 | 0.0794 | 41 | |
| 23 | 101000 | 0.0707 | 40 | |
| 24 | 100111 | 0.063 | 39 | |
| 25 | 100110 | 0.0562 | 38 | |
| 26 | 100101 | 0.05 | 37 | |
| 27 | 100100 | 0.0446 | 36 | |
| 28 | 100011 | 0.0398 | 35 | |
| 29 | 100010 | 0.0354 | 34 | |
| 30 | 100001 | 0.0316 | 33 | |
| 31 | 100000 | 0.0281 | 32 | |
| 32 | 011111 | 0.0251 | Noon | 31 |
| 33 | 011110 | 0.0224 | 30 | |
| 34 | 011101 | 0.0199 | 29 | |
| 35 | 011100 | 0.0177 | 28 | |
| 36 | 011011 | 0.0158 | 27 | |
| 37 | 011010 | 0.0141 | 26 | |
| 38 | 011001 | 0.0125 | 25 | |
| 39 | 011000 | 0.0112 | 24 | |
| 40 | 010111 | 0.01 | 23 | |
| 41 | 010110 | 0.0089 | 22 | |
| 42 | 010101 | 0.00794 | 21 | |
| 43 | 010100 | 0.00707 | 20 | |
| 44 | 010011 | 0.0063 | 19 | |
| 45 | 010010 | 0.00562 | 18 | |
| 46 | 010001 | 0.005 | 17 | |
| 47 | 010000 | 0.00446 | 16 | |
| 48 | 001111 | 0.00398 | 15 | |
| 49 | 001110 | 0.00354 | 14 | |
| 50 | 001101 | 0.00316 | 13 | |
| 51 | 001100 | 0.00281 | 12 | |
| 52 | 001011 | 0.00251 | 11 | |
| 53 | 001010 | 0.00224 | 10 | |
| 54 | 001001 | 0.00199 | 9 | |
| 55 | 001000 | 0.00177 | 8 | |
| 56 | 000111 | 0.00158 | 7 | |
| 57 | 000110 | 0.00141 | 6 | |
| 58 | 000101 | 0.00125 | 5 | |
| 59 | 000100 | 0.00112 | 4 | |
| 60 | 000011 | 0.001 | 3 | |
| 61 | 000010 | 0.00089 | 2 | |
| 62 | 000001 | 0.000794 | 1 | |
| 63 | 000000 | 0.000707 | Min vol | 0 |
“Well, couldn’t you make the relay ladder work more like a potentiometer?” some ask. “I like the way a potentiometer works.”
Short answer: Yes, but….
Longer answer: Yes, but let’s discuss your perception here. We’ve established that:
- Turning up a relay ladder more doesn’t stress the amp more
- A relay ladder provides a perfect, completely predictable output
- A relay ladder provides perfect channel matching all the way up and down the volume range
Given that, yes, there are ways to make it act more like a pot, but it’s all going to be in fakery. As in, you could simply drop in an audio-taper pot for the volume position sensor, and it would behave like an audio-taper pot. However, this will result in massive clicking when first turning the knob, and infrequent clicks once you get past noon. You could also fake it by skipping steps at first (basically an algorithmic cheat).
But the way a relay ladder works is inherently consistent, as in, each click is a fixed amount of attenuation (1dB for our 64-step products, 0.625dB for the 128-step products). Ironically, some people like the 64-step products better, thinking the 128-step products are too “fine” in the degree of control. I wonder if it’s possible to do a 32-step attenuator in 2dB steps—after all, many stepped attenuators are even more coarse than this—but I’d rather have more control, myself.
In short, yes, we could make it more like a potentiometer. But it wouldn’t improve it. At least not in our experience.
So, relax. Keep turning up that relay attenuator until you smile. You’re not hurting anything at all.
Going Negative
“Well, what about negative gain?” someone asks. “When you pick negative gain, doesn’t the amp, like, negatize the input in an active way? Where does that fit?”
Sigh.
While there are negative gain active stages, the reality is the best way to get negative gain is simply to divide the input down, or, even better, divide it interstage if you don’t have overall feedback to worry about, because then you don’t have to worry about input impedance or additional amp stages—but that’s getting a bit too deep for a chapter on gain.
So negative gain = fixed divider, in about 99.999999% of cases.
Yes, I understand that we use this terminology in Magni and other amps that have an attenuation setting, but to be clear: this is attenuation, not gain. A magic active stage is not needed to shrink a signal; you can simply divide it down with two resistors.
And yes, I understand that “negative gain,” or “fixed attenuation,” is an important feature for some headphones, in order to get enough volume range on small potentiometers. It’s less important if you’re talking relay ladder attenuators with perfect tracking.
“So why use potentiometers at all? Why have negative gain? Why not just use relay attenuators everywhere?” the questions are like a barrage.
Simple: cost.
Relay attenuators cost more to implement, plus they need a microprocessor. They’re going to cost 15-40x more than a small potentiometer, depending on number of steps, resistors used, and sensing arrangement.
Sounds expensive? Oh no. You can buy special audiophile potentiometers that cost from 2-20x more than a relay ladder!
Aside: why oh why?
So yes, for entry-level components, if you’re looking for a good volume control range that includes good tracking at low volumes, you’re probably looking at a negative gain (fixed attenuation) option. And there’s nothing wrong with that.
Bringing the Noise
“Gain, attenuation…I don’t get it,” someone says. “Why worry about any of this? Put on the music, crank the knob, if it sounds good, who cares?
Simple: because of the relationship between gain and noise.
The more gain, the more noise.
Period.
This isn’t us being dicks, this is a physics thing. Remember the Skoll example that started this chapter? It was quiet at 40dB and noisy at 70dB. Which isn’t super surprising when you realize you’re going from 100x gain to 3,333x gain.
You can see it in Skoll’s specs: signal to noise ratio drops with each increase in gain, usually fairly proportional to gain increase.
So you really don’t want to use any more gain than you really need, in order to keep the noise floor down.
Now, in the case of a modern digital system, it’s going to be much less dire than a phono preamp. It’s entirely possible to have a speaker system with 20-30dB excess gain and have it be essentially noise-free in the room.
But it’s also possible that, with other transducers, like super-high-efficiency IEMs, gain matters. If you’re looking at a 105dB/mW or 115dB/mw product, you may need low gain, or negative gain, for zero noise floor.
“Well, can’t you turn down the volume and turn down the noise?” someone asks.
Welllllllll…maybe.
Here’s the thing: if the noise is coming from your source, yes, turning down the volume control will attenuate the noise. You can hear this on old recordings, etc, that have inherently high noise floor, or from a phono system, which has an, er, less than state-of-the-art noise floor.
But if the noise is coming from the amplifier, you’ll need to switch to low gain to reduce the noise floor…
…maybe!
Because this is assuming that the amplifier is reducing gain by increasing feedback. Increased feedback decreases gain…and the noise floor.
Aside: don’t say feedback never did anything for ya!
Aaaaaaand, even if the amp is using feedback to reduce the noise floor, there’s a limit to how low its noise floor can go.
“Wait, what?”
Look into equivalent noise resistance. Thermal noise has limits. The lower the equivalent noise resistance is, the lower the potential noise floor. This is why phono preamps frequently use heroic methods like paralleling a bunch of input devices to lower the equivalent noise resistance. It’s also why the impedance of the RIAA network in a passive phono preamp matters. It’s also why any feedback used in a phono preamp is usually very low impedance. It’s also why potentiometers have maximum noise at about their midpoint—the source impedance is highest in that case.
The relationship between gain and noise also explains a lot of the design decisions in audio that can seem, well, a bit crazy.
As an example, someone asked why all of our amps don’t have the same gain, and why gain varies between our headphone amp products. Couldn’t we make them all the same?
Short answer: yes, probably (there may be stability constraints—wayyyyyy too much to get into for this chapter).
But they have different gains, mainly due to the need to drive the amp to max output, and to balance that with noise. Our lower-output amps typically have lower gain, because they don’t need as much gain to reach maximum output…and also because the lower the gain, the lower the noise. For a nearfield amp like Gjallarhorn, a lower gain than Vidar makes sense, because you may be sitting very near the speakers. You don’t want any chance of hearing hiss or hum, so lower gain makes sense. Rekkr is lower still.
And in some cases, we do multiple gains, from Magni to Ragnarok, so you can select the right gain for the system you’re using. In the case of Ragnarok, you can go from a gain of 2 to a gain of 30—a vast range—simply because Ragnarok can run both headphones and speakers from the same amp stage. A gain of 30 is a bit bonkers for headphones, but perfect for speakers, while a gain of 2 is silly for speakers, but right for sensitive headphones.
So yes, gain matters. You can have too much, just as you can have not enough.
How can you tell? It’s easy:
Too much gain =
- Audible noise
- Poor volume control range
- Can’t drive system to maximum output
In this case, maybe not. Bottom line, if you’re happy with your system—if it has enough volume for your needs, you’re fine.
Conversely, someone else will say, “My system is noisy, what do I do?”
Well, if you have adjustable gain, set the amplifier or preamp (or both) at lower gain. Did that help? If not, you may have a different problem. A high level of hum can indicate a ground loop.
Also, if you’re using phono, remember that phono isn’t, well, exactly at state of the art levels of performance. Some hiss and hum (and pops) are part of the package. That said, choosing a lower gain setting on the phono preamp might help.
“Sounds like you need to do a chapter simply on noise,” someone says. “Because there are sure a lot of questions on that.”
And yeah, you’re probably right.
Until then, I hope you, er, gained something from this latest chapter!
Last edited:
