The Digital Age

The 192kHz Lie (Part 3)

Continued from The 192kHz Lie (Part 2)

Resource Cost

Resource cost will have three parts for the purpose of our discussion.

  • Increased processing time in the ADC and DAC
  • Increased processing load on the processor of the computer or digital recorder we are using
  • Increased storage requirements (we need to store all those ones and zeros)

ADC and DAC Cost

As it turns out, processing time in the ADC and DAC are not a very big factor. Your sound card will either support the higher rates or not. A sound card may cost more if it supports the higher rates but this is not guaranteed. Today’s home studio hardware is cheaper than ever to manufacture and you can find products at all price points supporting very high bit depth and sample rates. The capabilities of the ADC and DAC are built into the hardware so it really produces no extra load on those components to operate at their highest possible rates.

Processor Cost

Processors do not get off as easily. Doubling the bit depth doubles the needed processing time and doubling the sample rate doubles the needed processing time. Doubling both will quadruple the amount of processing time needed to handle all the math of mixing and manipulating those ones and zeros. If we only had one sound file to deal with then it is a no brainer and you can just use the highest possible rates you can. We are focused on home recording though and that means multi-tracking. A few tracks of vocals, guitars, bass, drums, keyboards, etc. and the computational horsepower requirements can really start to rise. Add to the equation our desire need to use EQ, compression, reverb, and other effects to sweeten our tracks and realize our musical vision. Imagine this as a balancing scale with processing our audio files on one side and processing all our effects on the other. If you add more processing to the audio files you have to take some away from the effects. This is what I mean by the processor resource cost.

Storage Cost

All these audio files need to be recorded somewhere. Our modern equivalent to the tape reel is almost universally the computer hard drive. Hard drives are getting bigger all the time but they are still of finite size. They can only hold so much information before you start deleting your songs or backing them up. I don’t know about you but I don’t like the idea of casually deleting songs that I’ve worked hard on. I want to keep my master tracks around for years of remixing pleasure or just the comfort of knowing I have them. That leaves data backups as our only other solution. The home recording solution to backing up data will usually be to copy master tracks out to CDR or DVDR. To give you some rough ideas, I recently completed a session where we recorded 10 tracks of live instruments for four songs. The master tracks for each song took up nearly a gigabyte. That is two CDs to back up each song or one DVD to back up the whole session. If we increase the size of our master files we increase the physical number of discs needed to back them up. You then have to store them all somewhere. As someone who has stacks of a few hundred discs of backups sitting around, I can say the less discs you use, the better!

Hard Numbers

The increase in processing cost and increase in storage cost are closely related to the amount of information needed to represent the audio. Here are some actual numbers to illustrate the increase in these areas. Through our previous experiments we have seen that our greatest gains can be had from increasing bit depth or increasing bit depth and sample rate together. We will not discuss the cost of increasing only sample rate because of these findings. How much information is needed to represent audio data at some common bit depths and sample rates?

  • Audio at 16-bit / 44.1kHz takes up 5.3 megabytes per minute of mono audio
  • Audio at 24-bit / 44.1kHz takes up 7.9 megabytes per minute of mono audio
  • Audio at 24-bit / 96kHz takes up 17.3 megabytes per minute of mono audio
  • Audio at 24-bit / 192kHz takes up 27.6 megabytes per minute of mono audio

The compact disc is at 16/44.1 as are typical mp3 files and other encoded internet audio formats. We will consider this to be the baseline size as your song is most likely destined to end up at this resolution. Use this information to calculate the increased storage and processing cost of each rate.

  • Audio at 16-bit / 44.1kHz is the baseline
  • Audio at 24-bit / 44.1kHz has 1.5 times the base cost
  • Audio at 24-bit / 96kHz has 3.2 times the base cost
  • Audio at 24-bit / 192kHz has 5.2 times the base cost

Those are very staggering numbers. Increasing the bit depth alone is a marginal increase in resource cost but increasing both bit depth and sample rate is three to five times more expensive! What does this mean in terms of storage requirements? Lets explore the amount of hard drive space and back up space needed for a typical project.

Johnny Rock and his friends have a kickin’ song to record. They are getting awesome reviews around town and everyone loves their three and a half minute pop-punk ditty, “Love Hurts And Now So Do I.” Johnny wants to hire you to record the band playing this song. They want to lay down two guitar tracks, bass, keyboards, vocals, two backup vocals, and three tracks of drums (kick with stereo overheads because they want that “seventies” sound). There intro, outro and guitar solo don’t have vocals on them so the vocal track will only take about 2.5 minutes of actual audio and Johnny wants to overdub them after the band lays down their tracks. The backup vocals are only on the last chorus so they only take 30 seconds each, also overdubbed. The rest of the instruments are going full bore for the whole song, that is seven tracks of audio. Here is how that works out in terms of total minutes of audio needed.

  • Lead vocals for 2.5 minutes per take
  • Background vocals for 30 seconds x 2 = 1 minute per take
  • Instruments are 3.5 minutes x 7 = 24.5 minutes per take
  • Grand total is 28 minutes of audio per take

Now how much storage does that translate into for our hard drive and back ups?

  • 16/44.1 is 5.3 megs x 28 mins = 150 megs per take
  • 24/44.1 is 7.9 megs x 28 mins = 220 megs per take
  • 24/96 is 17.3 megs x 28 mins = 485 megs per take
  • 24/192 is 27.6 megs x 28 mins = 775 megs per take

We don’t know ahead of time how many takes they will need to really nail it on “Love Hurts And Now So Do I” but at the highest rates we are eating up 3/4 of a gigabyte of audio for each take! You could hand Johnny and his pals a single backup CD of their tracks for up to four takes at 16/44.1 or three takes at 24/44.1. Beyond those rates and you are handing out more expensive DVD backups (but you’re making them pay for it anyway, right).

Beyond the storage requirements, that is a lot of extra processor horsepower going to handling larger files when it could be used to do real time mixes with plenty of EQ, compressor, and reverb plugins going. Remember, increase the size of your audio files and decrease the amount of effects your computer can reasonable process.

Where It Matters Most

The base sample rate of a CD is 44.1kHz as we have already discussed. That is 44 thousand and one hundred vertical lines per second. That is a very high resolution already. Increasing the sample rate doesn’t have that much good effect on the accuracy of our audio (Figure 8 and Figure 12).

The base bit depth of a CD is 16 bits. Solution 2 showed us that increasing bit depth alone has an enormous effect on increasing the accuracy of our audio (Figure 10). 16 bits is also quite high resolution already. As the bit depth corresponds to the vertical scale, this means it also corresponds to the loudness of the audio. Though we are not as concerned with the effect on loudness as we are concerned with the effect on the quietness of the audio. Representing loudness is not a problem. The quieter an audio signal gets, the less and less bits we have to represent it, the more and more it starts looking like a square wave (Figure 2). Johnny’s keyboard player might like the occasional square wave but I’m sure the rest of the band and their fans would no appreciate it on the entire mix.

Higher bit depth translates into more definition in reverb trails and room ambience. This is a very tangible benefit. You also get a much higher dynamic range meaning Johnny’s most pained screams of agony can be quite a bit louder than his softest loving plea for forgiveness. Bit depth increases are a clear gain.


Congratulations for sticking with me until the end. I hope this exploration of digital audio has given you some insights into bit depth and sample rate and the affect they have on the accuracy of your audio. Recording at a higher bit depth can have very audible results in your mixes. I tend to resist the urge to increase sample rates because the resources required to do so are better spent on a more detailed mix. Be your own judge, but either way you’ll be better equipped to make an informed buying decision when considering your next piece of digital gear!

5 replies on “The 192kHz Lie (Part 3)”

Yes. The final decision should be based on the intended format. This article is primarily targeted toward people who have been told “higher is better” no matter what they’re doing.

Nowhere is the Nyquist theorem mentioned here while the graphic example gives no specification as to the frequency of the wave being sampled. There is a direct correlation to the accuracy of higher frequencies at higher sample rates. Waveforms are very complex. Especially from instruments like cymbals, drums, guitars, pianos and anything that consists of a wide spectrum and combination of frequencies, including harmonics (which you won’t see in a graphical representation of a simple sine wave). When you play an open string on a guitar you are not just hearing the note played, but also the harmonic frequencies of that note as well as the physical properties of the instrument. Much of that information can exist at the very top of the perceptible frequency spectrum. Lower and middle frequencies will be represented better digitally at lower sample rates but a 20 kHz wave will be sampled just over twice at 44.1 kHz looking more like the square wave graphic where the same frequency will be more accurate at 88.2 kHz, 96 kHz, 172.6 kHz and of course 192 kHz. Bit depth corresponds to amplitude (dynamic range) and the bigger myth is the 32 bit depth available from some DAWs. 24 bit/ 192 kHz is the closest digital representation we have to old fashioned 2 inch 24 track tape which has a bias of around 100 kHz. When applying the Nyquist theorem the sample rate would be around 200 kHz. Bigger factors are what device is being used to convert the analog signal to digital data and back again. An Apogee A-D/D-A @ 16/44.1 may sound better than say an M-Audio A-D/D-A @ 24/96.That is where higher may not be better.

Mr. Jim Diaz presented here nice THEORY which has a little flaw – most mature people cannot hear anything past 16-17Khz, also in the age of mp3’s where lossless compression like FLAC is used by enthusiasts mostly those 192Khz doesn’t really matter. Still – the better the sources, the better the result will be, but again – I’m with the author that 44,1Khz 24bit is a good compromise and since most of the people wont hear the difference – it’s pointless to go audiophile’s way. Unless…

Unless you want the recorded material to be heavily post-processed and resampled (modern electronic musiccians do this a lot). But yet again – most of electronic music producers base their work on samples and VSTi’s so it’s not about audio interface possibilities, but about good quality sample sources and high quality internal audio engine of your DAW.

Comments are closed.