In this document we will discuss the various aspects of a sound-set as used to program the CT SL51-4 decoder (SL75/GE75 are identical for programming). Although I have not experimented with every area of it yet, I hope this will provide a useful insight into how it fits together and how the various sound files relate to what actually gets played during operation of the loco.

Basic Concepts

The decoder uses digital recordings of sounds stored in flash memory and plays them back through an integrated amplifier and attached speaker. Most sounds are played back 'as is' without any modification, such as whistles, brakes etc. Other sounds are modified on-the-fly by the decoder to adjust what the loco sounds like - for example, the pitch of a chuff is modified depending upon the loco's current speed, the load, and whether it is accelerating, braking or moving at a constant speed.

It might seem complex, but actually there isn't much going on!

The SL51-4 has the ability to play up to 3 different sounds at the same time (this is what we mean when we say 3 channel playback). Up to 3 sounds effectively have their waveforms merged before sending the signal to the amplifier/speaker. 3 is just enough for a loco sound decoder - any less and you'd have to cut out 1 sound while the whistle blew, for example, and it would give a stuttered, disjoint effect. LokSound 3.5 decoders and Zimo sounds decoders provide 4 channels, and LokSound V4 decoders provide a whopping 8-channels. More channels would be nice, but in practise 3 is adequate.

As mentioned, the decoder possesses some flash memory, which means that you can download your own sound recordings onto it, and it will remain stored there until you replace it with something else (or you swing a big magnet near it!!). Computers use this type of technology extensively nowadays to store data which wont be lost when you turn it off.

The flash memory of the decoder also stores the CV values, but that's a topic for a separate discussion!

The sound recordings are stored individually into things called slots. You can think of the slot essentially as equivalent to a file on a computer's file system. Rather than having names though, they are referred to by their slot position. Each slot is used for a different purpose, which we'll talk about next.

What Sample Rate Should I Use?

The sample rate of a digital sound indicates the number of chunks of information recorded per second of the sound. The higher the sample rate, the finer the sound quality... at least in general.

CT sound decoders support 22khz and 11khz sample rates and if there is enough storage room remaining on the decoder, you should always go for the higher rate - at least keep the 'master' copies of your sound files at the higher rate.

However, in reality there is little perceiveable difference between 22khz and 11khz for many sounds - particularly those of low-to-mid range frequency. High-frequency sounds, with little metalic clinks and clicks will suffer most or high pitched hissy noises. Experiment with the effects of reducing the sample rate on the computer to hear the loss, if any.

Good candicates for a lower sample rate might include long platform announcement sounds. Whistles and horns are often largely unaffected too.

What Sample Size Should I Use? 16bit or 8bit

The sample size of a digital sound defines the accuracy of each sound sample.

Let's imagine you had a 2bit sound, this would mean that the individual sound sample values could only be ON or OFF - nothing in between. The more 'bits' you have, the finer the description of the sound can be. 16bit sounds more accurately describe the sound than an 8bit sound.

The reality is that actual sound is totally analogue - it is an infinitely fine curve of 'values', and any digital respresentation of the sound is at a loss with the original real-world sound.

Of course the ear can be easily fooled, in the same way that the eye can be fooled by a TV picture - which displays at 25, 50 or 100 frames (pictures) per second, rather than a real-world continuous image. The 'sample rate' is a bit like this, and the 'sample size' defines the number of digital values that a sound can be represented by.

In a high-quality home hi-fi system you can easily tell the difference between 8bit and 16bit sound (CDs use 16bit), but in a tiny model train speaker the difference becomes blurred. This is why some manufacturers choose to use 8bit (Zimo), and others choose to use 16bit (CT). The reality is that there is no discernable difference in a model train!

What Sound Slots Are Available?

So, a slot is an individual sound recording, but how does the decoder use each slot? Well, each slot is numbered, so let's simply list the available slots and put a description by them...

* Sounds marked with an asterisk above indicate that they are part of the main running sounds, rather than additional effects.

As you can see, many of the slots are grouped together, and the decoder will play them in sequence within that group, often providing the ability to loop the sound for a number of repetitions during the middle slot of the group. For example, Effect 1 could be a long whistle sound, which comprises of the following component sounds:

• Slot 37 = start of whistle sound
• Slot 38 = continuation of whistle
• Slot 39 = middle of whistle sound (which loops until the function is switched off again)
• Slot 40 = start of the end of the whistle (you know what I mean!?!)
• Slot 41 = end of the whistle sound

When you activate this sound, say from a function key on your controller, it would first play Slots 37 and 38 sequentially, followed by Slot 39 which is would continually loop until you switched the function off, at which time it would then continue through Slots 40 and 41 until the sound is finished. The decoder decides if the middle slot is a looping sound or not based on CV 145 - a different bit of CV145 is used to specify if looping is activated for each of the first 8 Effects.

Effects 1 to 8 may also be sounded by the decoder's random play feature (effects 9-16 are not possible to include in random play). Any sound activated because of random play may also be looped, but because it is not user-controlled directly, you need to program the decoder to specify how many times the middle slot is looped before it moves onto the final slots. This definition is done using CVs 122 to 128, bits 2-4 and bits 5-7. Please refer to my CV Commentary guide for the SL51-4 for examples and a fuller explanation. The choice of which of the first 8 Effects are included in random play are decided by the values in CVs 112 and 113 (CV 112 for those at standstill, and CV 113 for those during movement).

The CVs mentioned above are interesting and useful, and you can tweak these in a loco that already has a CT SL51-4 decoder installed. However, in order to download your own sounds onto the decoder you will need the right equipment for the job - please read my Commentary on the CT Programmer for more information.

Where Can I Obtain Sound Recordings?

Not everyone can get to a place where the real engine is running, and even if you do, there is no guarantee that the conditions will be suitable for recording, or that the engine driver will kindly blow the whistle for you when you ask! Of course, sometimes this can be arranged - and I would suggest that if you are serious about getting your own original recordings, that you ask permission before you set off for your visit.

For most however, you will need to source your recordings elsewhere, and the obvious place is the Internet - YouTube in particular is full of useful videos. It does of course involve spending A LOT of time sitting in front of the PC watching immensely boring stuff, and if you have a spouse-in-the-house you should probably warn them that you are going to be 'a while'! No kidding though, you can spend many hours just watching clips to find the right ones, with suitable sounds that you can make use of.

Always check that you are allowed to download/record the clip. If you are using it for yourself then this is usually no problem at all, but if you intend to resell the loco then you need to be quite wary about copyright. I would never dream, for example, of reselling a Thomas The Tank Engine sound-set!

Pay particular attention to background and ambient noises when searching for good recordings. It is hard to find completely crystal clear recordings, but the louder the target sound, and the quieter the ambient noise, the better. Of special interest is often the whistle, which, although relatively easy to come across, is rare to find on its own without overriding steam hiss and chuff sounds. It is worth continuing the search in the hope of something better.

Sounds can be cleaned up quite substantially later on, but like many end-products, the better the original recording, the better the eventual results.

As well as the Internet, there are numerous libraries becoming available specifically of steam train recordings, such as those on www.steamsounds.org.uk. Take your pick, and use whatever suits you best.

Of course there is also a plethora of sound-sets out there already for specific decoders. LokSound in particular have a huge library of sounds for many steam engines which you can simply request from them when you purchase the decoder. Zimo is similar, providing a reasonable selection of sound projects, albeit mostly European. CT is very limited on this front, so you'll most likely want to create your own sound-sets. Beware though that many of these sound projects are copyright to that company, or the individual who compiled it. As we can see, it takes a lot of effort and skill to put together a good set, and the authors generally like to be rewarded for their efforts!

How Should I Edit Sound Files?

This is where I could get very technical, and this discussion could get very opinionated. There are so many ways to edit your sounds that it really is down to personal preference.

I would recommend using editing software on the PC to do the job, although there are other, more traditional methods (don't go there!). My experiences with different software packages is pretty limited, as there is so much choice. Lots of freeware of shareware is out there which can do an adequate job, or if your budget stretches, then you can go right up to professional-grade studio packages which have every bell and whistle (no pun intended) you could dream of.

I'll tell you what I currently use, and leave it at that, without the heavy opinion: GoldWave. This is a relatively low-cost editing package which has a lot of good features. It misses a few features I would like, but in general it is packed with effects and filters, and most importantly is quite easy to learn and use. It costs around £30 to buy online. If you are interested, take a look on www.goldwave.com.

How you actually go about editing the sounds is again largely a matter of preference. There are a wide variety of filters and effects that you can play with until you improve the sound enough to be happy:

• Cropping - once you find some sounds, you can grab them from source (GoldWave for example can record any sound playing through the PC), but then you need to chop the start and end off accurately of the bits you don't want. This is cropping.
• Stitching - sometimes it is necessary to stitch 2 or more sounds together to form a longer sound, or to achieve a seemless loop. For example, something like a looping whistle which you switch on, then loops in the middle until you switch it off, requires a central sound which loops neatly. This is actually one of the hardest tricks to achieve without getting clicking noises on the join. You'll need to zoom right into the sound to chop/stitch at exactly the right cross-over points in the wave form. An alternative to stitching involves a combination of stitching and cross-fading i.e. you overlap the end of the first sound with the start of the next sound, and cross-fade them so that they blend into each other. Sometimes this works better, and sometimes you end up with an even bigger mess! Experiment to get experience!
• Filtering - Many live recordings will have lots of background noise, often referred to as white noise. This is essentially a hissing sound from the microphone that took the recording, perhaps of the wind blowing across it. It is often possible to use filtering features in your editing software to reduce this sound, but beware - do not reduce it too much or you lose the crispness of the main sound that you are trying to keep.
• Change pitch - often it is useful to adjust the pitch of a sound to make it better suit the model. I often tweak whistle pitches so that they translate better into the smaller scale, but even more often I play around with the pitch of the chuff and engine sounds according to the loco.
• Change volume - again, a very useful feature is to be able to adjust the volume of a sound. Although many sound decoders have CVs to allow you to adjust the volume of each sound, it is best to make sure that your original sound-set is good before you start. Try to begin with the loudest possible recording, and turn the volume down on each sound individually as you test it in the loco. I often go through a repeating cycle of loading onto the decoder, testing the loco in action, then changing the volume back on the PC for anything which sounds wrong or distorts a little. It is worth going through a number of cycles like this until you have a good balance. It is common to find that each speaker has different tolerrances, and what sounds clear on one speaker may distort on another, so you just need to turn down the volume a little for that one. It is important to avoid clipping of sounds by increasing the volume too high - it will just sound distorted if you go too far.

As you can imagine, it is a complex business, and a very subjective one, and you'll really only get good at it over time with lots of practise. This document is just meant as a starting point giving you some of the basics to kick you off.