Some thoughts on sounds.

JCitron

Trainzing since 12-2003
Since the discussion of sounds has come up, I took a look at some of the sound-producing assets such as the older sounds such as birds, seagulls, etc., for example. There are many issues which need to be addressed, here and like many Trainz issues, the problems falls in both the laps of N3V developers and the content creators with the latter folks being lead astray by lack of clear documentation, and general content-creator created issues.

There's one setting in there that caught my eye to reduce the sound radius, which I think for starters is something the content creator should have looked at in more detail. The problem I see is the content creators had cloned a sound asset and substituted another sound for it and kept the parameters of the previous assets. This isn't an unusual process for creating assets, heck we all do it. Copy an existing one and change the data and names to protect the innocent, and viola, we've got a new asset.

This works most of the time for those mundane things such as buildings, and train cars, however, with sound-producing assets this can be a problem. If there's an explosion sound, or some other loud sound-producer cloned to produce crickets, we now have crickets that can be heard two kilometers around their placement.

The other issue is of course the sounds themselves. The requirement for a low bit-rate .wav file is quite obsolete today. This was okay in the days of the Sound Blaster 16 and other older RealTek sound cars, but we've come a long way since then. The now highly compressed low-quality sounds, really sound low-quality and really compressed.

Then there's the issue of the quality of the recording. I noted way back in the TRS2004 days that the train sounds sounded as if someone went on a field trip with a cassette recorder to capture the sounds. The overly loud, and way too saturated sounds, causes awful distortion due to being clipped out at the maximum. The lack of harmonics and distorted sounds makes some engines sound like vacuum cleaners instead of diesel electrics for example. Other files were recorded similarly by various people so that bells, car horns, birds, crickets, and so on are out of balance.

A more recent issue occurred around the time that N3V introduced the initial content repair initiative. Some repairers increased the gain rather than the amplitude to make some sounds louder. This in turn offset the starting point of the loop from zero sound to a higher level. This higher level causes the pops and clicks we hear now in some assets such as crossing gates, birds, and others.

So what do we do?

I think part of the problem can be resolved within the CRG and N3V where the actual sound-producing assets such as seagulls, birds, crickets, etc. are examined and adjusted for the obvious things such as gain anomalies, amplitude (loudness), and maybe the distance parameter in the sound script.

http://online.ts2009.com/mediaWiki/index.php/"Soundscript"_container

Code:
soundscript
 {
   morning
   {
     ambient 1
     value-range 1, 0.1
     volume 0.3
     sound
     {
       0 ctry_day_1.wav
       1 ctry_day_2.wav
     }
   }
   night
   {
     ambient 1
     value-range 0, 0.9
     volume 0.3
     sound
     {
       0 night_loop.wav
     }
   }
 }

With this part settled, and putting all the sound-producing assets on a level playing field, N3V can update the sound-engine to modernize, and perhaps add in defaults such as the ambient tag where required which affects how a sounds is played. This isn't a one-shot and quick process, however, and like anything else will require a concerted effort on both N3V's part and on that of the community.
 
I have noticed some real issues with some steam engine sounds that create very strange noises that somehow manage to create repetitive "vocalising " . I can't put it more precisely than that but it resembles the old chestnut of the " I think I can " speech that steam engine is supposed to be repeating in children's books. In other words the sound resembles garbled repetitive human speech with one phrase repeated over and over again.
 
A good post John.

You mentioned, amongst other things, the issue of the quality of the original recordings. I would add to that the issue of "non authentic" sounds being used - where the sound of one locomotive has been substituted for another. Now I can (usually) tell the difference between the sound of a steam loco and that of a diesel loco but distinguishing between the sounds of different individual locos is beyond me, but it seems to get some posters hot under the collar.

My suggested solution to both issues: Those who are concerned about the quality and authenticity of the original sound recordings should contribute to a library of recordings that meets their standards that all content creators can freely access.
 
A good post John.

You mentioned, amongst other things, the issue of the quality of the original recordings. I would add to that the issue of "non authentic" sounds being used - where the sound of one locomotive has been substituted for another. Now I can (usually) tell the difference between the sound of a steam loco and that of a diesel loco but distinguishing between the sounds of different individual locos is beyond me, but it seems to get some posters hot under the collar.

My suggested solution to both issues: Those who are concerned about the quality and authenticity of the original sound recordings should contribute to a library of recordings that meets their standards that all content creators can freely access.

Excellent suggestions.

I am also more on the macro level when it comes to this stuff too. I can tell the difference between an EMD diesel engine versus an ALCo. The ALCo sounds as though it has loose rings and lifters while the EMD is smooth sounding, but getting into the subtle differences within the families is beyond me. If these people know the difference and have sounds they can share, by all means do so, but they have to be recorded well and not overly saturated an distorted because the recording volume is up too high to begin with.
 
I have noticed some real issues with some steam engine sounds that create very strange noises that somehow manage to create repetitive "vocalising " . I can't put it more precisely than that but it resembles the old chestnut of the " I think I can " speech that steam engine is supposed to be repeating in children's books. In other words the sound resembles garbled repetitive human speech with one phrase repeated over and over again.

I've noticed that as well. This has to do with the short length of the recorded sound loops and the repeat rate at which they are applied. There's a similar phenomenon with older digital keyboards. The sound sample is so short that when a note is held for any length of time, that the sound loop when repeated, as it should be, but has an obvious wow-wow-wow like affect as the sound starts and stops its sequence.

Unfortunately repairing the sound files themselves is something that can only be accomplished by rerecording the audio because the final results are still only as good as the original source not unlike a texture image.
 
Warning, long post incoming...
The other issue is of course the sounds themselves. The requirement for a low bit-rate .wav file is quite obsolete today. This was okay in the days of the Sound Blaster 16 and other older RealTek sound cars, but we've come a long way since then. The now highly compressed low-quality sounds, really sound low-quality and really compressed.

This right here is my main issue with sounds in this game! The gross misuse of the .wav filetype almost hurts to see. I'll jot down some notes, findings, and thoughts about what is currently in Trainz, what could be in Trainz, and stuff about audio files in general.

Many default sounds I can find in my copy of Trainz have stats similar to the following:
-Maxing at 5kHz
-Sample rate at 11025 Hz or less (8000 is another common number)
-Single low bit depth audio channel (some have two, but too many have only 1)
-Highly varying bitrates - I've found some as low as 88 kbps, some as high as 353)
Example: Built in rain sound file (5kHz, 11025 Hz, 353 kbps 2 16-bit channels)
A select few sounds in my game also crash my spectrogram analyzer program called Spek, a free audio visualizer for Mac, Windows, and Linux. This problem was most common in the car horn files I noted.

Now, a few notes on audio files in general.

Trainz is making use of the .wav file type - what is unique about .wav? Wav, along with other filetypes like FLAC and AIFF a lossless file. There are two main types of audio files we use - lossy and lossless. Lossy is the most common, being used with MP3, AAC, WMA, M4A, and many more, while lossless is less common. Lossless simply means that the file does not loss any data if compressed, meaning that it retains full sound quality after the fact. Lossy does see heavy compression taking away from the audio within - this is most notable when listening to a song at a high volume, a lossy, compressed MP3 will have more static as certain elements of a track are eliminated to preserve space.

Wav is a uncompressed lossless file, meaning larger file sizes but at no risk at all of losing that audio data. FLAC, for comparison, is a compressed lossless file that allows for the same data to take up less space but also not lose any data (this is why FLAC is often regarded as the best type of audio file type, though it took until 2017 for all major operating systems and devices to be able to make use of it). What N3V and Auran before them are doing is using a completely worthless wav file. The data being stored on the builtin wavs are more suited towards an MP3 if they insist on using such poor quality.

Ideal, highest quality audio files, will have some of the following stats:
-Maxing at 20 - 22kHz (20 is the max the human ear can generally perceive (this is highly debated however), but 22 is the maximum that can be stored on most file types without going into absurd file sizes. It is not necessary to go above 20, but 22 is the real maximum most people aim for)
-Sample rate of 44100Hz (48000 is common but like 20 vs 22kHz, the differences are minimal and cannot easily be heard by most human ears)
-2 16-bit audio channels (24 bit is indeed higher but results in file sizes significantly larger than 16 bit, and there is much debate as to whether the difference between the two is indeed that notable. My personal experience says there is little difference)
-Bitrate of 320 kbps (1411 is another common number, but again, above 320 the differences you can hear are usually quite minimal)
Example: A 2017 remake of an unreleased 2007/8 Led Zeppelin remix (22kHz, 44100 Hz, 1411 kbps, 2 16-bit channels)

The built in .wav files in the game don't even come up to half of that of an ideal high quality sound file. The vast majority have single channels and bitrates under 200, the rain example is one of the better sound files I've found... which isn't a good sign truth be told. These files sound just plain bad. I'm not even joking when I say the data in the car horns in my copy of Trainz is unreadable in Spek, which is something I've never even seen before! I managed to capture this screenshot before the program crashed when trying to open a car horn: https://i.imgur.com/E9lvm7q.png

I would consider myself a bit of an audiophile, and that's why I take the topic of audio in this game so seriously. Compressed audio is just unpleasant to listen to in general. For those who play with sound off or low it isn't a big deal, but sound is a very critically undervalued aspect of immersion - if the sounds are poor, then it can take one out of the experience entirely. Some audio in this game is indeed attributed to content creators, and they often do a better job at adding in decent files themselves. I had a look at some K&L whistles and see they're reaching 22kHz, 44100 Hz, and even 706 kbps! The only negative I can find with these is they are single channel, but with Trainz rather poor sound engine, there is little to fault with that (Don't get me started on how poorly that doppler effect is in this game! I had to edit that out when I made Trainz videos on YouTube...).

I suppose the TL;DR of this rather long post is that: the default sounds in Trainz are of definitive low quality and are in dire need of an update to match with modern day audio standards.

If I had to propose the most ideal utilization of files, I would say that new default sounds should have these minimum settings:
-20kHz max (22 optional)
-Sample rate of 44100 Hz
-2 16-bit audio channels (this may involve better working the sound engine in the game itself, especially the doppler effect. I could be wrong on this note however)
-Bitrate of 192kbps (Lower filesize and still sounds quite acceptable. 320 is the minimum for lossless files, but strictly lossless isn't necessarily a requirement)

There are many matters of sound this game needs to improve on, but the one I think is the most important (in my own personal opinion) is the actual files themselves. So many default sounds - weather, cars, junctions, track, wheels, etc - sound horrendous and could be improved to really make the game more simulator-esque. Graphics are all well and good, but sound is equally important in my opinion.

Cheers,
SM
 
Last edited:
A good post John.

You mentioned, amongst other things, the issue of the quality of the original recordings. I would add to that the issue of "non authentic" sounds being used - where the sound of one locomotive has been substituted for another. Now I can (usually) tell the difference between the sound of a steam loco and that of a diesel loco but distinguishing between the sounds of different individual locos is beyond me, but it seems to get some posters hot under the collar.

My suggested solution to both issues: Those who are concerned about the quality and authenticity of the original sound recordings should contribute to a library of recordings that meets their standards that all content creators can freely access.

Is it out of the realm of possibility that one should expect more realistic sounds from a game trying to market itself as a "realistic train simulation"? Sounds are just as important as visuals, especially when it comes to selling the game as "realistic". Some of us don't have the resources, time, or ability to "contribute to a library of recordings that meets their standards that all content creators can freely access". Wanting better sounds doesn't translate to asking to much, i think.

How would you feel if every asset in your Trainz installation used a single-note square wave for sound? Oh wait, you'd just turn the sound off. That's great for you, but i for one prefer to hear what a train actually sounds like. Besides, is it not right of the customer to ask more of the developers when we feel they're slipping?

You essentially told John and the others that have mentioned this issue that that asking for better sounds is not allowed. That's quite rude, don't you think? Supposedly you think it's wrong of the consumer to ask the company for a better product.

EDIT: Leo pretty much hit the nail on the head, in terms of the technical side of things.
 
Last edited:
You essentially told John and the others that have mentioned this issue that that asking for better sounds is not allowed. That's quite rude, don't you think? Supposedly you think it's wrong of the consumer to ask the company for a better product.

No, you have misunderstood my meaning. I am not against better quality anything in Trainz.

Since the sound files used on the DLS were put there in assets created by the users themselves using sources best known only to themselves, it should be the responsibility of those users, or those who have access to better sources and who want to see improvements in this area, to contribute to the solution of this problem.

N3V provide the platform, it is the users who provide the content.

There are thousands of locos on the DLS and while many of them are the same make and model, there must be hundreds of different varieties some of which would no longer exist. How do you suggest is the best, quickest and least disruptive way of solving this issue?
 
Fair enough, i'll admit i misunderstood you in the sense that i'm referring to Trainz's default sounds. I suppose i should've worded my post differently.

However, that does not excuse how poor Trainz's default sounds are. As Leo pointed out, they are of pretty low quality, despite being on a file type that allows for "lossless audio". Take the rain sound for example: it's horrible, no way around it. Sound is just as important as visuals in most cases, and i believe N3V should take steps to provide better audio for a game being made in 2018-2019.

There are high quality enginesounds, hornsounds, and ambient sounds for Trainz. But stuff like the rain sound in the game is something most users don't or cannot change. Therefore, N3V should improve the sound quality of their default sounds.
 
I agree. Sound technology seems to have left the Trainz franchise behind and is one area that will have to be looked at, hopefully before too long.
 
-2 16-bit audio channels

One question - why 2 audio channels? I was thinking the audio comes from a point source, and so should be a mono file that would be mixed to the two output channels depending on the orientation of the camera to the audio source?

I myself have always struggled with audio in my assets. I don't have a good baseline that I can use to verify my loudness against; for example.

Thanks,
Curtis
 
One question - why 2 audio channels? I was thinking the audio comes from a point source, and so should be a mono file that would be mixed to the two output channels depending on the orientation of the camera to the audio source?

Its been many, many years since I did an acoustics course in my Science degree but I do seem to recall that having two channels increases the "3D sound depth". Even though there may be just a single "point source" for the sound by using two spaced microphones means that the two channels will not be identical and this can give the ears (or more accurately, the brain) 3D spacial information about the source of the sound.
 
The simple fact is that many of the sounds used in TANE still emanate from the UTC era - the default GM and Alco engine sound and horns. Given the talent which has produced some of the excellent sound sets for MSTS and the DTG product where is the equivalent for this game? Obviously there are exceptions - the Railsim UK guys did a superb job with the sounds for the S&C pack including their freeware after market additions. The built in default sounds such as track clatter sit firmly in N3V's purview to sort out, you would have thought after 18 years of development they would have someone they could send out or commission or even buy in from the indies doing stuff for DTG (such as Armstrong Powerhouse) to get better sounds in there.
 
Vern, there is nothing you have said that I would disagree with.

I suspect that sound has a much lower priority than graphics or "game-play" since it is the latter that mostly sells the product. If you look through the forums, it is graphical and operational issues (e.g. "AI") that have generated the most complaints and demands for improvements. You only occasionally see the issue of sound quality raised. Perhaps it is long overdue.

We may, or may not, see in the upcoming release.
 
Railroads do not hold the romance of "yester-year". Many N3V customers have probably not heard a steam engine or paid attention to the sounds of the rail. They accept that a product such as TANE contains legitimate quality sounds thus Tony needs to do a cost benefit analysis of publishing some improved sounds. His competitor brags about recording actual train sounds and shows staff with recording equipment in an engine cab. Sound either annoys, is ignored, or enjoyed. Darn, I just tossed the CDs for RailWorks.
 
What's also unclear is just what kind of sound producing device the Trainz audio engine is expecting. There is an extremely wide variance in Trainz users' experience in this area. Some users find the sounds to be wonderful, others have total audio dropouts every few seconds. Some use onboard audio, some use sound cards, but there seems to be no correlation between the hardware and the quality of the resulting sound. My own PC contains what was a high end Soundblaster at the time (2012), and Trainz sounds perfectly awful with it.

I can't help but think that a lot of the problem has to do with how many audio channels have to be simultaneously mixed to the final output. When every creator feels they have to add a sound component to every little asset (I'm half-expecting telephone poles that whistle in the wind one of these days) the unnecessary proliferation of sound sources has to be a potential source of problems. Whether this is handled by the Trainz sound routines or is referred to the PC's output software/hardware I have no idea, but perhaps some clarification on this from N3V would be helpful in determining what an optimum hardware setup would be. This would also be to N3V's advantage, as the current confusion and widespread experience with poor sound reflects (quite possibly unfairly) on them.

--Lamont
 
Memory may not serve me well here, but I seem to recall one significant change to sound and that was fade in of general track noise (rail joint sound) at 30mph or thereabouts in TRS2006. This effect, I note, continues to this day.

I thought that was a good response to feedback from users at that time.

Now, we have continuously welded rail, and so, the disappearance of general track-rail-joint noise in the real world except for the occasion track spread (rail spreading out of gauge) and the associated yawp-yaw rail noise when rolling stock goes hunting on the rail head.

So there is a tension between modern railroading practice (less of "road noise" and the periods we are modelling (more of 40 foot rail and its associated road noise).

Perhaps, one day, we may be able to set the track sound as an event in surveyor?
 
This thread got me interested in what real engines sound like. There are many websites and youtube clips. One thing quickly became very clear, the sound of an engine at high speed is not just the sound of the engine at low speed just played faster. See this series of clips https://www.youtube.com/watch?v=whcHZntoduE

I don't think the purists will ever be truly satisfied. But I do agree there is much lacking with the current system.




But I guess we could carry that discussion to smoke as well...
v5KyyGO.jpg
[/IMG]
s15tloT.jpg
[/IMG]
The first pic is the loco at low speed, the second at high speed.

In N3V smoke is smoke regardless of speed.
 
Last edited:
One question - why 2 audio channels? I was thinking the audio comes from a point source, and so should be a mono file that would be mixed to the two output channels depending on the orientation of the camera to the audio source?

I myself have always struggled with audio in my assets. I don't have a good baseline that I can use to verify my loudness against; for example.

Thanks,
Curtis

Its been many, many years since I did an acoustics course in my Science degree but I do seem to recall that having two channels increases the "3D sound depth". Even though there may be just a single "point source" for the sound by using two spaced microphones means that the two channels will not be identical and this can give the ears (or more accurately, the brain) 3D spacial information about the source of the sound.

Pretty much this. Single-channel audio is strictly mono, 2-channel audio is stereo. Stereo is much more realistic for simulators like Trainz, in which many objects and the camera have to move around and thus the audio will change depending on where they're moving, how fast, etc. In music, where most of my work with audio lies, 2-channel audio files are responsible for receiving a certain instrument more in one ear/speaker - such as how it's very common for the left headphone/speaker to have more guitar emphasis on some slightly older tracks, I've noticed. With Trainz, multiple channels makes the sound not appear from a fixed point, which is crucial in a game like this one.

Cheers,
SM
 
I am happy that sound is finally having a proper discussion in this community. I was very disappointed when Tony made no mention of improved sound quality or effects to Trainz. Trainz has been around for a long time now. As some have pointed out, most of the sounds we hear in TANE dates back to the UTC days of Trainz! As someone who has played with the freeware BVE train simulator many years ago, I always admired how well it simulated train physics and sound effects. BVE allowed a lot of flexibility to content creators in making trains sound quite realistic.

With that said, I will propose some ideas on how N3V could tweak the sound engine without the need of special scripts or coding by the content creator. These ideas I base off how BVE handles sound effects.

Locomotive sounds...

The current way Diesel locomotives handles Notch up - Idle - Notch down is currently set where if you go from Notch 0 to notch 8, you will hear all the notch up's and idle samples until you get to idle 8.WAV. The reverse happens when you notch back down to notch 0. It would be ideal that the new sound engine skips the idle samples until it reaches the correct notch of the throttle.

The addition of adding air compressor sounds into the enginesound KUID's would be ideal. this could look like Compressor-start.WAV compressor-idle.wav compressor-end.wav

Also adding to the enginesound KUID would be the air brake sound effects. This would completely replace the default "yet annoying" air break sound in Trainz where it sounds like someone is releasing air from a tire. These files could be labeled as Air-apply.wav Air-release.wav and Air-dump.wav

Dynamic breaking should also be added. Dynamic-break.wav This file should be pitched based on the force amount of breaking selected by the driver.


Tracksounds...

Tracksounds should no loner be allowed to be tied to bogey KUID's. Tracksounds should strictly be tied to track KUID's. A track over wooden ties sounds different from tracks over a bridge or in a tunnel. They may also sound different based on various methods tracks are constructed these days. Also, locomotives, passenger cars and freight cars sound different when going over these tracks. With that said, I propose the tracksound KUID has multiple tracks based on what is on them. Here is an example of what I am saying.

L for locomotive
LRUN1.WAV
LRUN2.Wav
LRUN3.wav
etc
P for Passenger
PRUN1.WAV
PRUN2.Wav
PRUN3.wav
etc
F for Freight
FRUN1.WAV
FRUN2.Wav
FRUN3.wav
etc

Following these steps would mean we could finally retire the default clickety clack sound. Squealing flanges effects should also be added particularly in curves. Trainz does this a bit, but it can be hugely improved! It should also produce such sounds at higher speeds.

Bogeysound...

A bogey sound should be a new KUID that is introduced. In this file should be where the traction motor sounds of a locomotive or a multiple unit lives. Currently, this is one of the biggest weaknesses with Trainz. It is horrible with traction motor sounds which is awful for electric trains. traction motors base their pitch to speed. The faster the train goes, the higher the pitch of the traction motor. Also some trains have multiple sounds to their traction motors when starting from a stop position. No traction motor sound should be herd if the throttle is in the neutral position. This is exactly how Trainz should replicate traction motor sounds.

https://www.youtube.com/watch?v=DJOnjng9poc
https://www.youtube.com/watch?v=G1qyivrX6eM

Junction sounds should be included in the Bogeysound KUID's. Trainz badly needs to remove the default junction sound. Ether N3V get a professional to record trains going over a junction or let the content creators make much better sound effects.

These are some of the ideas that N3V can do to vastly improve the train sound effects in Trainz. All this is so far just off the top of my head and not so hard to implement.
 
Last edited:
Back
Top