MaSzyna International > EU07 Simulator English forum
Own developments, suggestions, ideas
TZs:
Hello, Witam!
They'll be here my developments, corrections, questions, ideas, suggestions in theme of Polish railways.
Dodano: 23 Marca 2015, 22:46:22 First theme.
Editing sounds.
I observed, in the Maszyna's recurrected sound samples take place in one direction (example: converters, electric motor noises, runningnoises, etc.).
I have a trick: copy the full "mid" sample, and paste behind. Make a reversal on the second sample, then merge the two sample. Many sample sound the same backwards as forward (except for example: a Hasler sample).
This technic's second benefit, a really short sample too will double in length.
I've created an example.
Type 303E-s "przetw3" mid sample was a deeper tone noise at the joint/replay point.
I remerged the start/mid samples, from mid delete around the fist 2 seconds (where was the deep tone error), reversed it, and start-mid samples I cut, that the splice point should be on 0 dB (so there is no popping noise).
Tried filtering a white noise/crepitation from the end sample, unavailingly... So, just use a short fade out, and after deleted the end crepitation.
Here are the re-edited example samples, try it, enjoy!
Accidentally, sample of "start" exported on 44kHz. Audacity add a terrible white noise to any unsigned 8bit exported sample, therefore should resampling from 16bit to 8 bit in Goldwave. And a little bit that stuff strange for me.
I'm so sorry! Here is the 22kHz reduced sample.
joey:
Hello.
As for editing sounds your looping trick may fail if the audio was cut in a wrong place (zero crossing can be wrong place too) before copying and reversing. There may still be clicks or pops although not as severe as when looping an edit done entirely wrong.
There's another looping method, more clean in my opinion. The longer the source sound the better as this method halves the original length. In Audacity (I see you mentioned Audacity) import the source audio and split it precisely1 in half ('split1.png' attachment). Cut the latter half and paste it as a new track ('split2.png' attachment). When you have two tracks apply the 'Cros Fade In' effect on the first track (first half) and 'Cross Fade Out' effect on the second track (latter half) ('crossfade.png' attachment). Play it looped (Shift+Space, or press Shift and click on the play button) to check if the loop sounds good. If it does, merge the tracks into single one (select all (Ctrl+A) then Tracks > Mix and Render)2. It's good idea to save a full quality backup copy at this point to avoid doing things over again in case the sound proves unsatisfactory (too loud/quiet etc). If the sampling rate of the source file was more than 22050Hz, change the project rate in the bottom-left corner of the window to 22050Hz, then resample (Tracks > Resample) to 22050Hz. When exporting, choose 'Other uncompressed files' and in options choose 'WAV (Microsoft)' header and 'Unsigned 8-bit PCM' encoding.
As for the noise you experience when exporting from Audacity, my guess it that you have not disabled dither. 8-bit resolution is so low, that you might want to disable dither (Edit > Preferences > Quality > High-quality Conversion > Dither, set to 'None'). Make sure Sample Rate Converter is set to 'Best Quality (Slowest) for best quality. ;)
It's best to avoid messing with sounds in 22050Hz/8-bit format as further modifications (especially loudness changes) can reduce quality of the sound.
1 Splitting audio precisely in half might be tricky. Choose 'hh:mm:ss + samples' display format from a menu in 'Selection Start' box at the bottom of the window. Then Select all the audio and check its length. Treat the samples count as a remainder. The samples count should be divisible by 2 for precise splitting, if it's not, remove final sample of the file.
2 It's good to have default sample format set to 32-bit float in quality settings. Mixing of crossfaded tracks can lead to overdrive (samples exceed the maximum value) if the source was too loud and in consequence lead to audible distortion. If that's the case use 'Amplify' effect and apply negative amplification so that the peak value in no larger than 0dBFS. This only works with 32-bit float sample format.
TZs:
Hello!
Many-many thanks, that was my fault, a few days ago I switched through the dither to enabled! Now correct, thank You!
You hear click, when playing mid sample as continued? Or in start-mid transition?
Unfortunately, from start to mid transition correct usually from 7 to 2 times in simulator. In sound editor, transition was faultless, countless test occasion.
My theory, maybe the problem, simulator's sound engine "memorize", where "stopped" was a sample. And in recall, the sample continued from this memorized point, never from the beginning of the sample. Of course, this is just a theory, be my fault too.
I try to work better and better.
This technique very interesting, and a good one! Already made some attempt.
However, many case frequent problem, sample is too short to repeat, like too long. Principally than cog wheel gear/traction motor noise. Often the usable session less than 3-4 sec.
Thanks again for your help in Audacity, and this new method!
joey:
As for the clicks I mentioned in my post I meant that sometimes a loop done with your method (copy and reverse) might not work well, I did not mean your files.
In Maszyna there's some kind of bug that sometimes causes a short pause between the sounds that were meant to be played in sequence (horns, engines etc).
I think that the most important transitions are start-to-mid, mid looped (these two are kinda obvious) and start-to-end (in case the user releases the key immediately) and these have to be seamless. Mid-to-end doesn't matter that much because we cannot control when the user will release the key or turn the engine or something off as the mid sound will be cut off immediately and the end sound will play. This may lead to clicks but I guess it can't be helped.
TZs:
Yes, yes, but now listen on high volume the sample, and make click every 12 sec. And it's funny, because, the reversed sample's _half length_ longer, than 23sec.
And in editor it's inaudible.
I don't understand.
Try make this something.
Nawigacja
[#] Następna strona
Idź do wersji pełnej