NGWave Audio/Sound/MP3 Editor: Test 3: Quality
NGWave uses 32-Bit floating-point numbers to store audio data. We use an unbounded single precision floating-point format. This has several advantages over standard integer formats:- Higher dynamic range, especially at very low or very high levels
- No clipping
- High precision editing
- Less data conversions; many functions (any FFT functions for example) work natively with floating-point data
See, raw audio sources that haven't yet been mixed-down and mastered will often have much lower levels, and will require a lot of processing. Each processing step reduces the resolution, since everything must be rounded to the nearest integer. Any excess is simply dropped. The accumulation of rounding errors is referred to as Quantization Errors.
Adding a few bits helps, but using a floating-point format, instead of integers, helps even further, by spreading the dynamic range over a larger area, and reducing the amount of data actually lost due to rounding. It also spreads the error relatively linearly over the entire range of data. Integer formats are more prone to error in the lower ranges, while they tend to top out easily if you push the levels too high.
Following Along
You may not be able to mimic these tests in some other editors. Many simply don't offer the kind of resolution to be able to do some of the drastic things here. However, this test should show you first-hand why a floating-point audio format is important for editing.The Tests
Action: With a sound file loaded, highlight a few seconds worth of sound. Click on Process, then on Equalizer. Crank the 60-Hz band to +24 db, and leave the others at 0 db (no change).The result should be heavily distorted. This is what you would expect.
Now, click the Volume icon, and reduce the volume by -21 db or so. Notice that the resulting sound is no longer clipped. That clipped data -- that is normally lost once clipped -- is back!
That's because we never actually clipped your data; clipping was only applied for display and audio playback, but the data itself remained in tact at some +18 db or so. This is what we mean by unbounded.
Wait a minute -- Headroom in Digital?
You've probably heard that digital audio has no headroom. In reality, digital audio devices and existing applications usually have no headroom. The computer can easily handle much "louder" data, it only needs to clip when converting to an integer format, or when playing back the sound to a limited real-world device like your sound card.There is nothing inherent about digital audio that precludes having some headroom, and in fact a lot more room is available than with any analog source. However, most implementations offer no headroom, and instead maximize the available resolution.
Considering that NGWave never clips internally, you can think of it as having a few hundred decibels of headroom, as well as a few hundred decibels of resolution! While this seems impossible, remember, this only applies to the internal data. The source data was still (most likely) limited to 16 bits (96 db), as is the output format. It is only during the editing session that these boundaries are removed. You can think of it as Unlimited Virtual Headroom.
There are many situations where you may accidentally clip your audio during an edit. With NGWave, all you have to do is lower the level a bit, and no actual hard-clipping has occurred.
Action: Undo the volume change, and then click Normalize. Notice that it actually recommends that you reduce the level! It does this because many samples are actually above 0 db. Normalizing the volume simply adjusts the level so that the highest peak is 0 db (full volume). In this case, because we went nuts with the EQ, it happens that it needed to reduce the level, instead of the typical increase in volume.
Hint: It's a good idea to hit Normalize before saving your work to an integer-based format -- this way you never have to worry about clipping. NGWave will remind you of this upon saving the file if you have a lot of clipping (configurable).
Action: Reduce the volume by -48 db, or to approximately 0.4%. Some editors may not allow you to do this; in this case, choose 1%, and do it two or three times. The audio is, for all practical purposes, muted.
Now do it again. That's right, reduce the level by 96 decibels. Now it's not just effectively muted; in 16-bit format (such as when playing back the audio), it is muted!
Next, increase the volume by 48 db (increase 25600%), twice. Notice that your audio has been brought back from the dead!
Only editors with an extremely high resolution retain any quality at all when doing this (or any audio at all). NGWave's storage format allows most of the detail to remain present, even at such extremely low audio levels. 16-Bit data -- as used by many "professional" editors -- retains none of the audio, or worse, may have resulted in heavy distortion.
IEEE float data on the other hand loses very little resolution, and will most likely sound just the same once you crank the volume back up. Only the most picky ears will tell the difference, and in practice -48 db signals will be extremely rare, and -96 db signals -- well, unless you have a very high-quality low-noise sound card, you simply won't have to go that far. But, it's nice to know that your software would...
Summary
You should be able to see by now that NGWave was designed from the ground-up with new ideas in mind, new ways of thinking. NGWave truly is the Next Generation in Audio Editing. As development progresses, you will see more and more new ideas.Also consider that so-called "professional" editors -- some that cost 20 times what NGWave costs -- don't offer many of these features. Try any of the above in such an editor, and you'll see that most do not offer true real-time preview (some let you preview a couple of seconds, but NGWave is truly real-time); most do not offer any sort of recovery from a crash or improper shutdown; most do not take advantage of the available resolution in your PC, and limit everything to 16-bits of integer data... the list goes on.
