Header

Objective Reviews & Commentary - An Engineer's Perspective

September 19, 2011

All About Gain

nwavguy gain diagramINTRO: For those interested in headphone amps, gain is an important topic. Here are the essentials about gain with lots of links to more information. In addition to this introduction explaining gain, there are later sections on:

RELATED ARTICLES: The following may be useful understanding how gain fits into the bigger picture:

WHAT IS GAIN? Put simply, gain is the maximum amount an amplifier can increase a signal. In the world of headphones, it's usually voltage gain. The diagram above shows a source with a weak output made 3 times stronger by an amplifier. Applying gain to a signal doesn’t normally change the dynamic range, it just makes everything louder. Other ways to specify gain, such as power and current, are rarely used for headphone gear.

YACA (Yet Another Car Analogy): The relationship between gain, volume, and maximum power can be confusing to some. Using car analogies might help:

  • Gain Is Like The Gear Selection - When a car is in its lowest gear its top speed is limited to a fairly low value but it can climb steep hills. It's somewhat the same with a headphone amp set to low gain. Think of a gain switch as 1st, 2nd, and 3rd gear on a car's transmission. Different gears are used for different driving conditions.
  • Volume is like the Throttle (Accelerator) - The volume control adjusts the overall power to the headphones much like the throttle in a car adjusts the power to the wheels.
  • Maximum Power is like Top Speed - Most cars are capable of going faster than most people need. While a car might have a top speed of 100 MPH, in first gear, it might only do 40 MPH. So gain can be used as a way to limit maximum output and better match the output to a given pair of headphones. The difference between the top speed and the highest speed limit is somewhat analogous to excess gain.

WHY DOES GAIN MATTER? If you don't have enough gain, your headphones probably won’t get loud enough. If you have too much gain, you will be forced to use only a small portion of the volume control's range, there may be increased channel balance problems, more noise, more distortion, and you could even damage your headphones more easily. Most any amp will perform worse at higher gain settings so you want to use the least amount of gain that gets the job done.

GAIN AND VOLUME ARE IMPLEMENTED DIFFERENTLY: Turning down the volume is not the same as lowering the maximum gain (with a few rare exceptions). A race car doesn't turn into a family car if you never use more than half throttle. Certain compromises were made in the race car that make it less suitable for family car duty even if you try to drive it slowly. The same is true with a high gain amp. It's important to not have much more gain than you really need.

GAIN CAN BE USED TO LIMIT MAXIMUM POWER: If you have some fairly efficient headphones that are as loud as you would ever want with only 0.7 volts of audio. But you have a high-end desktop headphone amp that can put out 10 times as much (7 volts). By reducing the gain, even if you accidentally turn the volume all the way up, you can limit the maximum power to your headphones. This can help prevent hearing damage and even headphone damage. It also also allows using much more of the volume control's range making volume adjustments easier and more accurate.

GAIN CAN CHANGE WITH DIFFERENT HEADPHONE LOADS: If an amp doesn’t have an output impedance below 2 ohms the gain will change with different loads. The higher the output impedance, the more dramatic the change. For example, an amp with a 120 ohm output impedance with 5X gain at no load, will have only 3.5X gain with 300 ohm headphones and only 0.6X gain with 16 ohm headphones. This is another reason it’s important to know the output impedance.

GAIN CANNOT CORRECT FOR A LACK OF MAXIMUM POWER: If your headphones need 5 volts but your amp is only good for 2 volts cranking up the gain won't help any and will just make the amp clip sooner. Clipping is what happens when an amp runs out of power. The rest of this article assumes your amp is capable of enough power for your headphones. For more on this topic see: More Power?

 


GAIN DETAILS

HOW IS GAIN EXPRESSED? Gain is usually expressed as a factor (ratio), such as 4X, or in decibels (dB) such as 12 dB. In this case, 4X is the same as 12 dB and simply means with the volume control set to maximum, the output voltage of the amplifier will be four times higher than the input voltage. If you put 1 volt in you will get 4 volts out (assuming you don't exceed the maximum output capability). I show how to convert from ratios to dB and back in the Gain Calculations section below.

WHAT ARE TYPICAL GAINS? Headphone amps that have just a single fixed gain typically range from 2X (6 dB) to 5X (14 dB). For amps with two gain settings, low gain might be 2X to 3X and high gain 5X to 8X. Some amps have three or more gain options while others allow changing internal jumpers or resistors to set the gain.

A ROUGH GUIDELINE: If you want to skip the more technical details, the table below offers some idea of what gain will be usable with various headphones and sources:

Headphones Portable USB DAC HOME < Source
IEM 100 - 115 dB/mW 16 - 32 ohms 1X/0 dB 1X/0 dB 1X/0 dB  
Sensitive Low Impedance Full Size Grado, Denon, Etc. 2X/6 dB 1X/0 dB 1X/0 dB  
Average Full Size HD5xx/600/650 100 - 300 ohm 5X/14 dB 3X/10 dB 2X/6 dB  
High Impedance Full size 300 - 600 ohm 10X/20 dB 5X/14 dB 3X/10 dB  

DO I NEED AN AMP AT ALL? If your headphones already get plenty loud enough, and you're happy with the sound, odds are you don't need an amp. But if they don't get loud enough, an amp may be required. The table above shows several combinations where the gain is listed as 1X which implies no amp is required. But you might still want to use an amp to lower the output impedance of your source. For example, the iPod Touch 3G has a 7 ohm output impedance. With balanced armature IEM headphones this causes audible problems. See: Output Impedance. So while the IEMs don't need any more voltage they can still benefit from a lower output impedance. Some amps, like the O2, can be easily configured for 1X gain and improve the sound of higher impedance sources. For more on headphone amps in general, see: Headphone Amps Explained

 


EXCESS GAIN

EXCESS GAIN: If your amp has just enough gain for your headphones to hit the desired level at full volume with a 0 dBFS signal it's sort of like having a car with a top speed equal to the highest speed limit in your area. There might be times when you want to go even faster. But how much extra do you need? The amount of volume control range above where a 0 dBFS signal clips is considered “excess gain”. Here are some points to consider:

  • Excess Gain Is A Tradeoff - As described above in Why Does Gain Matter, too much gain has negative side effects. So it's a tradeoff between those side effects and how much excess gain you want.
  • Excess Gain Is Only Useful For Quiet Tracks - Most properly recorded digital music is designed to hit, or get within 1 dB of, 0 dBFS. If your amp has excess gain, using full volume with normal tracks will cause the amp to clip (badly distort) and may even damage your headphones. But, without excess gain, rare quiet tracks may not be loud enough even with the volume at 100%. One solution is having a reasonable amount of excess gain. Another solution is to normalize quiet tracks to 0 dBFS using Audacity or other software. This simply raises the levels so the loudest portion of the track is at 0 dBFS. Yet another solution is to use ReplayGain which helps equalize all tracks to roughly the same perceived volume. 
  • ReplayGain (volume leveling) - If you use something like ReplayGain nearly all of your music should play at a consistent subjective level. ReplayGain is essentially like having someone automatically adjust the volume for each track (or CD). The default target loudness of 89 dB leaves 14 dB of headroom between the average volume and 0 dBFS. For the most compressed pop music this means the peak levels will be as much as 6 dB below 0 dBFS, while dynamic music will get much closer to, or even hit, 0 dBFS. If you use ReplayGain, or something similar, for volume leveling, you may want up to 6 dB of excess gain to allow for the overall volume reduction ReplayGain applies to some tracks. For more on average volume see: More Power?
  • Dynamic Range – It’s important to note that neither the gain of an amplifier, nor ReplayGain, normally changes the dynamic range of music. Gain changes normally apply equally to the softest and loudest portions of the signal (music). To change the dynamic range you have to apply compression, limiting, (or dynamic range expansion).
  • Clipping vs Maximum SPL – Some amps have plenty of power and your headphones, or ears, will give up before the amp does when dialing up excess gain on a recording that already hits 0 dBFS. But, in some circumstances (especially with low sensitivity headphones), the amp might run out of power first. If you want to be assured your amp will never clip, just work backwards from the amp’s maximum output into your headphone impedance to determine the gain. If your amp maxes out at 7 volts, and your source is 2 volts, set the gain to 3.5X and the amp will never clip even at full volume with a worst case recording.  See: More Power?
  • 9 dB Excess Gain Is A Reasonable Maximum - I would aim for at least 3 dB of extra gain but somewhere around 9 dB the negative side effects already mentioned start to outweigh the advantages. A few might want to go as high as 12 dB but only if they've used something like ReplayGain to normalize their music library to a lower than normal average volume or have other unusual requirements.
  • The Math – 3 dB of extra gain means multiplying the minimum gain by 1.4 and for 6 dB multiply by 2, and for 9 dB multiply by 2.8.

THE CHANNEL BALANCE PROBLEM: Devices with conventional volume controls may have audible channel imbalance at very low volumes. It's extremely difficult to manufacture volume control potentiometers that maintain tight channel balance below about -40 dB (referenced to full volume). Having too much excess gain forces using only the lower portion of the volume control's range with normal recordings. If your amp has 10 dB of excess gain, for example, -40 dB below 0 dBFS on the volume control is -50 dB below full volume. The channel balance error will likely be much greater and more audible. Some products get around this problem by using digitally controlled electronic volume controls that can maintain better channel balance at low settings but these can add distortion and often controlled by up/down buttons rather than a simple knob. Stepped precision attenuators are another solution but are expensive and sometimes their step size is too great.

 


GAIN CALCULATIONS

HOW DO YOU CONVERT FROM Gain Ratio X TO dB? The math requires using a function known as base 10 logarithm (LOG on a calculator). But you can also just use an online dB calculator:

  • Online dB Calculator
  • Gain in dB = 20 * LOG10 ( Gain Factor ) [i.e. 20 * LOG10 ( 4 ) = 12 dB]
  • Gain Factor = ANTILOG10 ( Gain in dB / 20 ) [i.e. ANTILOG10 ( 12 / 20 ) = 4X]

HOW MUCH GAIN DO I NEED? This comes down to only three things:

  • How much voltage your headphones need
  • How much output your source has
  • How much excess gain you want.

HEADPHONE REQUIREMENTS: For a given pair of headphones, the More Power? article helps determine what your headphones need. This is related to the sensitivity and impedance of your headphones which vary widely. The most efficient IEMs only need about 0.1 volts to play loudly, Sennheiser HD600s need about 2.3 volts and Beyer DT880-600s need about 6 volts or 60 times more than the IEMs.

SOURCE OUTPUT LEVEL: Sources have a maximum output level. It's the output voltage at 0 dBFS (the loudest digital music can get) and at full volume (if the source has a volume control). Sources can be roughly divided into the follow categories (all voltages are Vrms):

  • Portable Players - Most of these have a maximum output of 0.5 - 1.0 volts from either their headphone jacks or LOD (Line Output Dock) connectors. The iPod Touch is 0.5 volts from the LOD.
  • USB Powered DACs - Most of these have a maximum output of 1.5 volts or less.
  • Normal Home Audio Gear - The Redbook standard for any home equipment that plays digital audio is 2 volts. Some go slightly higher up to 2.5 volts. There are also a few USB powered DACs that can manage similar voltages (such as the HRT Music Streamer II).
  • Unusual (rare) Home Sources - A few companies, arguably unwisely, have decided to go well above the established standard and output more than 2.5 volts. But I'm not aware of any that go above 3.3 volts unless they have a volume control.
  • Balanced Outputs - These can be misleading because balanced outputs normally have twice the output of unbalanced outputs. A 4 volt balanced output, used with an unbalanced adapter or cable, is really only 2 volts.
  • Unknown Sources – See DIY Gain Measurements below.

CALCULATING THE REQUIRED GAIN: Once you've figured out the above three requirements, it's fairly easy to calculate the gain you need. Just follow these steps (examples are given for the HD600's driven from an iPod LOD with 3 dB excess gain):

  • Calculate Minimum Gain: Gain Factor = Vout / Vin [i.e. 2.3Vout with 0.5V Vin = 2.3/0.5 = 4.6X]
  • Convert To dB: Minimum Gain in dB = 20*LOG( Gain Factor ) [i.e. 20*LOG(4.6) = 13.2 dB]
  • Add Excess Gain: Final Gain in dB = Minimum Gain + Excess Gain [i.e. 13.2 + 3 = 16.2 dB]
  • The result is 16.2 dB (6.5X) of gain for the example case.
  • You can also multiply 4.6X by 1.4 to get 6.5X (1.4 is 3 dB excess gain as a ratio)
  • If you know the output impedance multiply by (Zload + Zout) / Zload

INPUT OVERLOAD: Also be aware that many devices, especially portable ones, have inputs that can be overloaded by certain sources. Because the Redbook standard is 2 volts, a lot of devices are designed to only handle about 2.1 volts such as the FiiO E9 desktop amp. And the portable FiiO E7 overloads at about 1.2 volts on the input because it was designed for use with portable players which are generally under 1 volt. Sometimes the input overload point may depend on the gain settings. So check the documentation for a given amp--especially if you plan to use home sources with a portable amp, or sources with more than 2 volts of output.

nwavguy gain resistors diagramRESISTOR VALUES: Some amplifiers let you set the gain by changing resistor values. The diagram to the right shows a typical non-inverting amplifier stage. The gain is given by 1 + 1000/500 = 3. For all the math, see Wikipedia Gain and Wikipedia Amplifiers. Some amplifiers may have multiple gain stages, in which case you have to multiply the gain of each stage together. So an amp with a 2X and 3X stage has a total gain of 6X.

DIY GAIN MEASUREMENTS: If you have a decent DMM (multimeter) that can measure x.xx volts AC (i.e. 0.01 volt resolution) you can play back a 60hz 0 dBFS sine wave and measure the output of your source in Vrms. You can create the file with Audacity. Ideally the voltage should be measured using a “Y” cable with your source connected to the amplifier. You can then also (carefully so not to cause short circuits) measure the output of your amplifier. Don’t connect your headphone as they could be damaged. And don’t try higher frequencies as most meters are not accurate much above 60 hz. Vout/Vin = Gain

BOTTOM LINE: Gain is important. There's a definite sweet spot for most headphones and sources. The closer an amp comes to having the ideal gain, the better overall performance you will get.

52 comments:

  1. Cool. Another useful article outlining the basics.

    I'm liking these "Headphones 101" articles. You're going to be a full-fledged educational site soon enough.

    You might want to put together a post that describes and links to all your explanatory articles and put a link to it on the sidebar called something like "Headphones 101", "Introduction to Headphones and Headphone Amps", or something similar so people can find it all in one place.

    ReplyDelete
  2. Love all your audio articles, refreshes alot of stuff i have learned but forgotten over the years.

    ReplyDelete
  3. I seriously think you should compile these articles into a book. This is the only site that I found that actually explains stuff in a lay-man acceptable way - kudos for the good work!

    ReplyDelete
  4. Thanks all. I'm trying to at least cover the basics and most common questions.

    @Maverick, That's a good idea once I have more articles and it can be reasonably complete.

    ReplyDelete
  5. This is fast becoming the most valuable resource for logic regarding headphone hi-fi on the internet, and I can say that without a hint of hyperbole. The basic electrical principles (and the not-so-basic) you relate here are very rarely relayed with such clarity, let alone in such a way to allow anyone to instantly apply them in decisions regarding their headphone rig.

    ReplyDelete
  6. Damn....I love the car analogy....that's the simplest explanation of gain I've ever heard. It's made my (limited) understanding much clearer in my head.

    ReplyDelete
  7. I was wondering, NwAvGuy, if you will be reviewing some of Fiio's upcoming products. I am especially interested in the E10 (dac/amp) and the D3. If you have any comments on them please let me know.

    Basically I am planning on upgrading my audio equipment in the following months, both dacs and amps are interesting. I am really looking forward to the desktop 02. I am selling my E9 and keeping my E7 for now. I currently own Allesandro MS-1, but more powerhungry cans will probably be bought in coming years and months.

    So I am thinking either:

    1) Get the fiio e10 if it gets good reviews.

    2) Get the O2 amp when the desktop version is ready and available. + HRT Music Streamer II or other suitable DAC. (looking forward to your DAC measurements).

    Do you have any recommendations here? This will be used at my desktop only rig. Is it important with a Dac with asynchronous usb? I am not in a hurry and can easily wait till you have som recommendations of Dacs.

    Thanks you for a great site!



    Best Regards,

    Delirious

    ReplyDelete
  8. The last comment above should really be in the E7 or E9 comments as it has nothing to do with gain but Blogger doesn't let me move comments around. Yes I plan to test the E10 even if it might share some of the E11's and Mini3's flaws.

    ReplyDelete
  9. Another great article :)
    I was wondering where it was :D

    On the book suggestion someone had - considering you want to keep the site non-commercial, you could offer a free "ebook" (plain pdf) and a paid for book via say amazon (that allows people to sell their own work)

    On that note - any update on the email I offered to translate for you? If you can't find my address I can send you an email again.

    ReplyDelete
  10. I'm trying to visualize gain. Is it more like elevating the entire soundfloor, or is it more like decrompressing an audio signal, increasing its dynamic range? I understand the elctrical implications of it, but in terms of visualizing it like a waveform, isn't it more like spreading the peaks and troughs?

    ReplyDelete
  11. That's a good question akgk and thanks as that's a point worth clarifying so I'll revise the article. If you ignore noise and clipping, applying gain just raises the level of everything. It doesn't change the dynamic range.

    ReplyDelete
  12. Apologies for off-topic, but that's the way these comments work...

    There was some talk about contacting Lake People (Violectric?) about their DAC module? Apparently Fried Reim is the one to contact. I checked around a while back out of curiosity and noticed there's a post on HF where someone reposted a PM he had with Fried Reim regarding Violectric products. Somebody else notes that Fried is approachable and doesn't speak audiophile BS. I presume the PM was in English, so translation seems not to be necessary.


    PS: as your blog's gotten more popular, I think I saw somebody else post as Mike (or mike; too common a name after all) so I'm going to revert to mikeaj for these purposes. ;)

    ReplyDelete
  13. Thanks Mikeaj. I sent an e-mail to Fried Reim. It is still a relatively expensive solution (especially outside the EU) but would allow for an elegant "one box" solution.

    The new FiiO E10 likely uses the same Tenor USB chip and is less than half the cost. And perhaps FiiO will do a USB DAC-only product (like the D5) that will be even less expensive and perform well.

    I receive immediate notification of comments posted to any article here not just the newest ones. Please try to keep things reasonably on topic. This really belongs in one of the O2 articles.

    ReplyDelete
  14. "...Yes I plan to test the E10 even if it might share some of the E11's and Mini3's flaws. "

    So the E10 has an active ground channel?

    ReplyDelete
  15. @Mikeaj, Violectric responded, please see my comment in the O2 Details article.

    ReplyDelete
  16. Thanks again for the great article.

    I'm still struggling with the part about dynamic range expansion/compression regarding amps and software solutions like ReplayGain.

    Wouldn't the Voltage-Wave look more like the middle one, rather than the one on the right, in this picture, if there is no dynamic range expansion by an amp?
    Picture: http://www.abload.de/image.php?img=wvbux6.png (sorry for my bad drawing)

    Also how can software like ReplayGain change the volume of a recording without compressing the range? Those are purely digital changes right? That means I could use the (digital) output, burn it on a CD, and if I compared it to the original CD it would play louder/quiter, right?

    ReplyDelete
  17. You're looking at a drawing of a sine wave. That's very different than looking at the volume of a music recording over time. Put simply, let's say the background noise between tracks is the quietest thing on a CD at -85 dBFS and the loudest peaks hit -5 dBFS for a total dynamic range of 80 dB. If we add 3 dB of gain, now the noise is at -82 dB and the peaks at -2 dB. The dynamic range is still 80 dB.

    ReplayGain doesn't change the digital bits in the music file at all. It's just a "tag" in the header information that tells the software playing the track to adjust the volume up or down. It's exactly like a human adjusting the volume.

    To change the dynamic range, ReplayGain would have to be constantly adjusting the volume during each track. That's how a compressor works. But it doesn't do that. It only adjusts the volume once at the start of the track or entire CD depending on how it's configured.

    ReplyDelete
  18. Hm.. ok, but the software playing back the audio file that was tagged by ReplayGain would be applying the compression then (in order to actually make it louder/quieter)?

    ReplyDelete
  19. I'm not sure how else I can say it Guilty Spark. Please see the links in the article to explanations of compression. It's not a one time volume adjustment. Nothing is doing compression in the signal path we're talking about.

    ReplyDelete
  20. I would mention, that after replaygain your music is never bit-perfect. Only if you have -6dBFS peak level and add 6dB. This is basicly doubles or "shifts" every sample, but in this case you only have 15 bit music, not 16 (on a CD). It works much better if you are using a 24 bit DAC, but it wont be bit-perfect.
    And -6dB is rare, i would say everithing today is 0dBFS (sadly).

    Also after the stupid EU volume regulation, having proper gain is really an issue ie. with an HD600, so thank you for the article.

    ReplyDelete
  21. I've wondered for ages how replay gain tags work and if they could audibly degrade the sound.

    ReplyDelete
  22. @tszaboo

    Unless I am mistaken the output power limitation in Europe only affects portable devices.
    This means a headphone amplifier wouldn't be affected at all.

    Further, your HD600 are 300 Ohm headphones, they wouldn't ever play very loudly on any portable device.

    If you want something loud you just need very efficient headphones with a low impendance - on the IEM side I haven't seen anything that's possibly louder than Sennheiser IE8.
    http://www.headphone.com/headphones/sennheiser-ie-8.php
    Sensitivity: 125 dB/V
    Impedance @ 1kHz: 16

    ReplyDelete
  23. Thanks DetlevCM. The O2 amp is portable and will drive the HD600's nicely :) Also, the Ultimate Ears TripleFi 10's are the most sensitive current headphones I know of at 117 dB/mW which is 132 dB/V! But the IE8s will certainly get plenty loud on portable gear.

    The sad thing about the EU volume regulations is a lot of companies want to make just one product to sell globally. Some products, when first powered up, let you choose a region or language and set the volume limit based on the selection. But other products have limited volume in all markets to allow selling them in the EU.

    ReplyDelete
  24. There are several different IEMs that are more efficient than the TripleFi, but not by a large margin. Only a few dB/mW. The highest I'm aware of are a few models that are rated at 120 dB/mW.

    ReplyDelete
  25. The highest IEM sensitivity spec I've seen was for Westone UM2s, which calculated to 135 dB SPL / 1 Vrms (119 dB / mW @ 27 ohms). That's pretty much an invitation to hiss and all kinds of spurious noises. You can tell that these BA drivers are primarily developed for the hearing aid market, where power consumption is extremely critical.

    Another problem associated with volume-limited players is some manufacturers (e.g. Sony) including 40+ ohm series resistors in the output. Great way of messing up the frequency response on multi-driver BA IEMs.

    Thankfully I have a Rockboxed Clip+ - which now even is about 3 dB louder than it used to be thanks to an experimental patch that increases DAC reference and analog supply voltage to OF level (patched build).

    I wouldn't worry too much about potential quality losses when using Replaygain. First of all it's a non-issue when using >16-bit samples and DACs, and even when going to 16 bit, dithering at the end of the processing chain should ensure the effect is limited to added random noise.
    If all of your material is gained down, advanced RG implementations have make-up gain (pre-gain) settings that can be used to minimize volume loss. (Or accomodate unusually dynamic recordings when adjusted the other way.) Besides, any recording mandating plenty of negative replay gain is likely to be effed up to the point of rendering the discussion pointless anyway...

    ReplyDelete
  26. Thanks Stephan. Do you have any good references on the audibility of losing bit resolution below 16 bits? I've read several arguments that talk about dithering, the ambient noise floor, etc. but I can't seem to find anything that's relatively definitive (i.e. referencing blind tests, etc.). I've tried searching the AES library but it's not a topic that lends itself well to searching.

    I agree ultra sensitive headphones present unique challenges (which is partly why I worked so hard to keep the O2's noise so low). My experience has been balanced armature IEMs nearly always sound better than less sensitive dynamic IEMs when driven from a source with a suitably low output impedance.

    ReplyDelete
  27. mmm....I have a TI6120 based headphone amp for my 600ohm headphones...with the gains (that are applicable to 600 ohm headphones, obviously no gain won't cut it), they are rated at +12dB and at +18dB. Which one should I use?

    ReplyDelete
  28. Karl, you need to figure out what your headphones need. Just knowing they're 600 ohm doesn't answer that question. Please see the More Power? article.

    ReplyDelete
  29. Re: gains for a TPA6120A2 amp, I'd say there is little use for more than about 12 dB when using a line-level source. The chip can run at +/-15 V max, where it'll do about 26 Vpp or ~9 Vrms of output. A 2 Vrms max source at 12 dB (4x) gain will do 8 Vrms max, close enough. A portable player would require 18 dB or more.
    I'd say try both - if 12 dB gets the job done just fine, stick with that.

    Re: audibility of losing bits, I know nothing in terms of papers. I do have a test CD from the '90s featuring the same sample at 16, 14, 12 bits, but quite honestly I've never really been able to tell these apart.

    From what I know, truncating to 16 bits undithered gives a dynamic range (SFDR) of about 72 dB. (Which is why 8-bit linear PCM undithered sounds so crappy @ 24 dB SFDR, hence telephone systems use log quantization.) Even then it takes multiple computing steps at 16 bits and very quiet passages for things to really become audible. There's a number of albums mixed with ye olde Sony PCM1610 system back in the early '80s, and those don't sound half-bad indeed (e.g. Peter Gabriel IV, pre-remaster). The processing may not even have been the biggest bottleneck, the historic ADC and massive analog anti-alias brickwall filter are at least as suspect.

    Losing one bit of resolution at the very end of the playback chain really is about the last thing I'd be losing sleep over. I mean, how much dynamic range do we really need (or get) in practice? Maybe 60 or 70 dB at normal volumes. (Say, ambient noise 20 dB SPL, average level 70 dB, peaks 90 dB. And that's for an unusually dynamic record and fairly quiet environment.) And let's be honest, any material with program dynamics exceeding approximately 50 dB gets somewhat hard on the ears. With the quietest spots about 20 dB above ambient noise, we're at 70 dB again.

    CDs produced with today's technology (24/96 recording, float32 processing, advanced limiting, proper resampling and dithering etc.) and mastered at about 1990 (or even mid-'80s) levels could sound stellar. An effective dynamic range of about 93 dB (>100 dB when using shaped dither) is nothing to sneeze at.

    ReplyDelete
  30. Thanks Stephan. I mostly agree but I would add about 20 dB to your numbers as 110 dB SPL seems to be closer to a real world peak listening level (see my More Power article).

    I thought there was an existing test on Hydrogenaudio that ABX compared different bit depths with Foobar but I can't seem to find it. It should be easy enough to do for anyone who has a bit accurate playback setup with an analog volume control.

    ReplyDelete
  31. Sorry, 600 ohm headphone, 96 dB/ mW SPL, 10 ohm output impedance.

    ReplyDelete
  32. Some of us crazy audiophiles love high GBW uber slew rate devices for gain, and end up having to use relative high gain to throw away some of that bandwidth just to get marginal stability without introducing external compensation.

    I've come to think it's usually not worth the trouble. Sometime playing with OPA637 followed by emitter follower leaves me feeling like I might be building a radio transmitter, and people dropping the newer high speed bipolars into existing circuits make me cringe just at the thought of it.

    ReplyDelete
  33. Thanks hoshiyomi. It makes me cringe too. See:

    Op Amp Myths & Facts

    ReplyDelete
  34. Also Stephan, I would argue that most humans won't be able to discern -75dB dynamic range. Ethan Winer has got a very good Audio myths workshop on Youtube that goes into dynamic range and here's a link to bit depth testing: http://www.ethanwiner.com/BitsTest.html

    His Audio Myths workshop is essential viewing as well. Get a drink beforehand as it's an hour long, but a very informative hour at that: http://www.youtube.com/watch?v=BYTlN6wjcvQ

    It's all about dynamic range and really doesn't matter about format. Redbook Audio with 16 bit, 44.1kHz sampling rate is more than enough dynamic range for music playback if that dynamic range is used. e.g. This track :) http://www.youtube.com/watch?v=Zi3i-HqDNFI

    ReplyDelete
  35. Karl, your source matters as well. But working backwards, your headphones need 3.9 Vrms to hit 110 dB. With 12 dB of gain that's 1 Vrms input with 18 dB it's 0.5 Vrms (an iPod LOD is 0.5Vrms).

    ReplyDelete
  36. DW, thanks to your article I figured it out myself factoring in excess gain, application of ReplayGain etc.. I need 5.5x which is inbetween the two software settings >.< I have the Asus ST, which should be extremely similar to the STX, which has a line output of 2.16v measured by Stereophile. 6v / 2.16v = ~2.78. 20x LOG(2.78) = ~8.88 dB. Since I use ReplayGain, add ~6dB of excess gain = 14.88dB = ~5.5x :/ Did I do it right? If I factor in a lot of loudness War'ed tracks that ReplayGain about -10dB, that would bring it up to the +18dB gain. Did calculations, still confused about which setting I should use lols.

    ReplyDelete
  37. Google is presenting comments for approval with substantial delay for some reason. So going back and addressing the ones I originally missed...

    Thanks Karl for the Ethan Winer link. That's exactly what I had in mind but had forgotten about his files. I haven't ABX'd them with Foobar myself but that should be enlightening. Although I'd probably pick 15, 14, and 13 rather than 13, 11 and 9.

    As for still being confused about 12 vs 18 dB try 12 dB and if it doesn't get loud enough try 18 dB. :) If you're using the software (driver) "gain" option on the Xonar ST that's not really gain at all as John Atkinson explained in his measurements. It just artificially limits the output of the soundcard at lower settings. I thought you were using an external amp.

    ReplyDelete
  38. @DetlevCM
    You are right. But this way, I am unable to listen to my headphone from a portable player, so I need an AMP for it, that is why gain setting is important.
    Also I'm not using my HD600 on the street that would be very bad :D I have a AKG K416P 126dB/mW (loud as hell)

    Gonzalo : it wouldn't change the sound, it is just a number. The playback software can just ignore it. It doesn't change the music so much until you have digital clipping, which is bad.

    ReplyDelete
  39. NwAvGuy, the material you post on your blog is amazing, it would be great if you would have a full-featured forum, there is a real need of science-based audiophile forum. Too many audio forums are filled with mumbo-jumbo.

    ReplyDelete
  40. I vote for NwAvGuy-Fi.

    ReplyDelete
  41. I don't see any mention of the "consumer" level of -10dBV (0.316V RMS) Is this obsolete now? \Greg.

    ReplyDelete
  42. -10 dBv is not a measurement unit, it's just a level sometimes used as a reference. The units most used with built in references are dBv and dBu. The Redbook standard for 0 dBFS consumer gear outputs is 2 Vrms which is about +8 dBu or +6 dBv.

    ReplyDelete
  43. My M-Audio Delta 66 audio interface (now quite old) has three level settings: +4dBu, "Professional", and "-10dBV". Also, the Line Level wiki page: http://en.wikipedia.org/wiki/Line_level suggests to me that "-10dBV" is in fact an absolute level that it used in consumer equipment. (am I misinterpreting it?). I.e - the reference is 0dBV (1V RMS), but the actual equipment has a level that is 10dB down from that reference. Maybe the Wiki is wrong, if I am not misinterpreting it. \Greg.

    ReplyDelete
  44. @SullivanG, perhaps you should message me privately with some of your questions. I think this is getting a bit off topic (or at least confusing for some) for the article comments. The Wikipedia article is correct, and manufactures are free to reference their measurements to any level they like just so they tell you what it is. More often than not they don't provide any meaningful reference and just use units of "dB". The beauty of using just dBu or dBv is they're self referencing without needing to specify any reference value.

    ReplyDelete
  45. Just some closure on this -10dBV issue - NwAvGuy very kindly helped me offline. I now agree with him that it's not very important or relevant in this context. (and there's actually more to it than I had realised). I am very grateful for his help! \Greg.

    ReplyDelete
  46. Hi NwAvGuy,
    what kind of dBSPL do you aim for when doing calculations? I have K701s and their sensitivity is 105dBSPL/V, so feeding 1V into it already seems excessive. By some quick calculations, 2.3V into 300ohm HD600s (98dB/mw) would be about 111 dB...and this is before the excess gain you put in. Am I underestimating how loud you need the highest volume to be?

    ReplyDelete
  47. @Anon, see my More Power article. It explains everything. The short answer is 110 dB PEAK SPL. That may seem loud to some, but that's typical of live performances--even with classical music in a concert hall. That can be adjusted down to 105 dB for some and up to 115 dB for others. The article explains it all.

    ReplyDelete
  48. Good info. Let's see if I understand the practical implications with an example that might help others as well.

    Setup: Foobar -> Benchmark DAC1 -> Adcom GFA-555 -> Dunlavy SC-IV. Cheap, non-true-RMS multimeter from Radioshack.

    Goal: set my DAC1's calibrated outputs so that I *cannot* clip the power amp.

    The SC-IVs are effectively a true 4-ohm load with 91dB/2.83V sensitivity. The Adcom clips a bit above 325W continuous into 4 ohms, so we'll shoot for 300W at 0dBFS to allow a little headroom & to keep things safe. This should give ~113dB (300W is about 22dB above 2.83V into 4 ohms) at 35V (square root of 300W * 4ohms).

    To calibrate:
    1) Disable all plug-ins, EQ, etc.
    2) Generate 60Hz tone with add location/tone.
    3) Set playback volume to -40dB. Voltage will be 1% of max (or 0.35V) & power only 0.01% (or 0.03W). Nice & safe!
    4) Measure speaker-terminal voltage at 60Hz.
    5) In calibrated mode, set DAC1's trimmers to 0.35V at the speakers.

    To verify this, we can measure the DAC1's output. The Adcom's 27dB gain implies a voltage ratio of 22.6. Therefore, 0.35V at the speakers should be about 15mV at the Adcom's input. (This is so low, we might want to bump it up a bit. The speakers should be fine with a steady input of a few volts. They're rated for 25W continuous pink noise, and 60Hz will stress only the woofers.)

    Questions:
    1) Have I missed anything?
    2) If I EQ in Foobar with auto-level, do I need to measure again?
    3) Any chance you'll discuss power amps soon? Because mine's getting old & I can't really replace the Dunlavies, I'd like to buy a modern, no-BS amp.

    Finally, many thanks for injecting some sanity into audio!

    ReplyDelete
    Replies
    1. The impedance of your speakers is a total unknown at 60 hz. That's close to system resonance of many speakers and the impedance may be far higher than the nominal value.

      Never run a sine wave at more than about 0.5 watts into a speaker for more than a second or two. While big beefy woofer voice coils can handle a few watts continuous, smaller drivers can't. Don't believe the MFG power ratings. Voice coils have little thermal mass and they're stuck in a very small enclosed space with little airflow. And, unless the driver has a metal cone, they have nowhere to "dump" their heat.

      So for your 4 ohm speakers, 1.4 Vrms is about the most you want to run into them. So I would use 1.4 V or 1.0 V rather than 0.35V.

      O dBFS is 0 dBFS regardless of EQ applied. You just need to use a 0 dBFS signal with no EQ.

      The best way to do what you want is to use a fast oscilloscope that can handle 50+ Vrms of input and monitor the amp playing a variety of music. The reason is the speakers are a complex reactive load and they may trigger the current limiting in the amp, and/or drag the power supply down (power amps nearly always have unregulated supplies) in ways that create clipping well below the theoretical point.

      Doesn't the GFA-555 have Adcom's "distortion indicators"? If so, I believe they're fairly accurate clip indicators in real world use. So you could gradually increase the level, trying a variety of music, until they start to flash and back off a bit from there.

      There's nothing wrong with the GFA-555. Despite what marketing people want you to believe, power amp technology (at least for linear class A/B amps) has not changed much since then. Your GFA-555 might benefit from new electrolytic caps (they dry out with age and lose their capacitance) but otherwise should serve you well.

      If you do want a new amp, I'd look hard at Emotiva.

      Delete
    2. Thanks for the reply.

      I'd not considered the impedance specifically at 60Hz, so that's a great observation. According to Stereophile (http://www.stereophile.com/content/dunlavy-audio-laboratories-sc-iv-loudspeaker-measurements), it is just over 4ohms.

      I've been VERY careful with the level of sines. BTW, Dunlavy's power ratings certainly seem more realistic/honest than most: 25W continuous pink noise & 250W for 10mS. Seems pretty conservative for 5 sizable drivers!

      It has become obvious that the multimeter isn't remotely fast enough for this. Perhaps a soundcard-based scope is worth a look, though the clip indicators should suffice. And I've never seen them flicker.

      Reasons for replacing the amp are age & lack of protection circuits. Your points about quality & technology match my experience, so I'll take a good look at Emotiva.

      Delete
  49. Quick question: For a high impedance headphone for the O2, would it be better to use the lowest gain and turn up the volume knob near 3/4 to max or use some gain and it at 40 to 50%?

    Awaits reply and eagerly awaits a new blog post! :D

    ReplyDelete
    Replies
    1. See the section above on Excess Gain (link at start of article). That's essentially what you're asking--how much excess gain should you have. It's a matter of personal preference. Some like to use the entire volume range at the expensive of having some extra gain to boost "soft" recordings. Some want more gain in reserve.

      Delete

PLEASE NOTE! If you see "LOAD MORE..." OR "LOADING..." just above this paragraph, you need to click that link to load the newest comments. It may take several seconds. When all the comments are displayed the Load More/Loading link disappears.

COMMENT APPROVAL: Comments may take anywhere from a few minutes to a few days to be approved. You might want to save a copy (highlight Ctrl-C) of your comments before submitting in case something goes wrong. The approval process is necessary to prevent obscenity, etc. Please keep comments on topic. Comments promoting commercial interests will not be approved and may be reported as SPAM.