VOC files (Creative Labs) when saved become noisy

I use Audacity to fix clicking and removing noise from VOC files.

When I finish my work on a file and export it in the same format as the original (unsigned 8-bit) the result is a very noisy file.
When playing this saved file, you can clearly hear, how noisy it is become. In some cases the result was worst than the original, it become noisier!
Audacity literally add noise to the already cleaned file, and it can be comfirmed by reloading the new cleaned file.
You can even see the noise that added to the file by just a glance to the waveform.

I attaced an example, one of the worst cases, that you can compare how the end result become noisier than the original.
noisyresult.zip (10.9 KB)

When using 8 bit (which is low quality at best), it is essential that you use as much dynamic range as possible - that is, amplify / normalize to 0 dB.
The less bits that are available, the more digital noise.

The number of bits refers to the number of distinct amplitude levels that are available. For unsigned 8 bit audio, the amplitude of each sample is represented as a binary digit between 0 and 256. If you only use half the available track height, then you are only using 7 out of the available 8 bits. If you only use 1/4 of the track height, then you are only using 6 out of 8 bits. For each halving of the peak amplitude, you use one less bit.

In the example files that you posted, the peak level of the original is -16.569 dB, which is a little over 1/8th of the available track height. In effect that means that the audio is using little over 5 bits per sample. Out of a maximum 256 possible levels, you are using 256/6.736743 = 38 distinct amplitude levels (6.736743 is -16.569 dB converted to linear scale). With only 38 possible values per sample, it is unsurprising that the audio quality is poor.

By way of comparison, normal 16-bit WAV files have 65536 distinct amplitude levels per sample.

Thanks, for the deep technical explanation.

I new to Audacity. I have a little experience in audio editing.
Back in the 90 we have a band and I do some digital afterwork on the recorded audio, recording samples, make rhythm loops, effects in those oldschool PC trackers and mix them with the recorded live singing, syth and drums. Or somtimes whole tracks made in trackers. In the beginning of 90’s rave music is just a very new thing almost nobody knows about it. We experimented with mixing oldscool disco music and digital, techno-like rhythms.But even with those old hardware we used to record sometimes as low as 22Kh and 8 bit (samples used to make effects, sounds, noises for the tracker to use in our music), noise never was a problem!
But I’m not really a musican, I always do my work by hearing and by looking the waveforms on the screen.

These files is from the old The Elder Scrolls Arena game, I didn’t satisfied with the existing sound patch, that fixes the audio (missing lot of ambient effects and not CD version, no voice acting in cutscenes) so I started to do it myself.
These files are the game’s effects that become worst then the original.
I didn’t know what the game engine can handle in term of highest KHz that I can use, so I kept everything as the original.
The files use various Hz range between 6427 and other similar weird stuff. I amplfied the sound level as loud as the original, because some are quiet ambient sounds. I think guys who made the game played with the sound level to get “sounds that come from a distant” 3D sound feeling or whatever.
So I just saved them as the originals.
I used amplified / normalized to -1 dB. I think that is the default in the program. I watched a guy in youtube too, and as I see (and hear) what he can do I followed how they do things.I dont know why, but he always used -1 too.I very satisfied with his work so I copied it.

The CD live audio tracks (cutscenes voice, 22KHz) are so noisy they become very good after cleaning them and saving, they become better than the original and I was happy with it.
Only these effects become worst.

If I save these files as 22KHz they will be better and less noisy?
According to the game’s cutscenes, 22KHz can be played by the game-engine.

A little help will be very good, how I can make them better in Audacity.
I not knowing technical therms, but I very happy if I can read them too. Maybe I learn a bit more technically what audio editing means.
But, like I said I work and edit by hearing and by visually.