Audio Basics 0 - Intro

[reggind]

I have been pondering this for a while. I thought I would write up a little series of blurbs on audio basics. Perhaps start with some info on cleaning up files and some of the tools that are available, with a focus on the free ones. None of this is cast in stone. I have been doing professional audio for many years now and thought something like this may be useful

I have spent a couple days walking around with parts of this in my head, so many thoughts, where to begin...

Perhaps the best place to begin is at the beginning. When we create a sound file...

Back in the old days sounds were recorded by an engineer, on tape. The engineer had an interesting job. If you were to record in a perfectly quiet room, you still got some noise. This came from many sources. Any electronic device has an inherent amount of self noise, this is noise that is not related to the sound you are recording. More often it is thermal noise and related to heat and the components. There is also induced noise. This is again not related to the audio you are recording. This time it may come from magnetic fields or radio frequency interference.

The self noise is not curable, but a lot of decent equipment today comes very close to having as little self noise as it theoretically possible. Better equipment will do better in this regard, but it is never totally curable.

Induced noise will always be around as well, but better cables are shielded from it better, as is better equipment.

So if there is a gem in that part, it is start with the best equipment you can.

The next source of noise is audio in nature. In a studio it can be a hiss or a rumble from the HVAC system. In our recordings it can be literally anything. A TV being on, cars going by, people talking, birds chirping. About the only things we can do here are get as close to the source of the sound as we practically can to make it proportionally louder than the extraneous sounds and we can also use a directional microphone to our advantage.

So what did our old sound engineer do back in the day, and what was the art? We have already spoken about the noises that take up the quiet end of the sonic spectrum. Obviously if a sound we want to record is fainter than the background noise, we are not going to be able to hear it. So one thing our engineer did was he kept is finger on the gain control, and made sure that the gain was turned up such that when was being recorded was louder than the ambient noise.

Now magnetic tape had another nasty little feature too. You could only magnetize it so much, once you reached that point it became saturated. Much like a sponge can only hold so much water before it becomes saturated. You can try to magnetize the tape more but once you have reached the saturation point that was it. In more practical words, you can only record so loud on the tape. If you try and record louder, it more or less just distorts.

So we have a lower limit, and an upper limit. There is a technical word for this, and it is called dynamic range. This is the range of sound levels that the device or medium is capable of recording or reproducing. Back in the tape days this may have ranged from perhaps 40 to 70 db or a bit more with electronic noise reduction and a good tape formula. With digital media it is not uncommon to see dynamic ranges in the 90-100 db range or greater.

So, our engineer had to sit there and turn up the gain on quiet passages so they would be louder than the hiss on the tape, and turn down the gain on the louder passages so they would not distort. The art was how they did it. And it was an art. A good engineer could let a very quiet passage start very very quiet and slowly ramp it up so you could comfortably hear it, and than quickly lower the gain for a loud part. The engineer had a brain and also knew the piece so he was also in a sense playing with the band.

Ok, a few notes here. If you look at dynamic range or signal to noise ratios on equipment make sure you are comparing apples to apples. Over the years many trade groups have come up with what they call "weighing's" . In some instances they even made sense. A weighing is filtering some frequencies out of the noise figure. For example if one manufacturer has put very large filter capacitors in their equipment to quell all of the 60 cycle hum, they may like an weighed reading, where another manufacturer would prefer a weighed reading on their that does not respond to the 60 cycle noise. If you compared a weighed and an unweighed reading you are not making an accurate comparison.

So, over the years on almost all tape machines and many digital recorders they have a thing called AGC or AVC. This is automatic gain control or automatic volume control This lets you make generally make pretty good recordings with out ever having to set any levels. It is a form of dynamic compression. That means in general it makes soft things louder and loud things softer. Compressors have time constants. How long something has to be amiss before they will "attack" the problem, and than they have another time constant for how long the problem has to be resolved for before they will "release". If you have ever turned an inexpensive recorder on in a quiet room you no doubt have heard this. At first you hear the click of it starting, than over the next few seconds you hear the hiss build up louder and louder. Than as soon as there is any sound being recorded the sound it over modulated for the first fraction of the second, and than it sounds good. What you heard with the ACG in action. With no sound it tried to make quiet things louder so it kept turning the gain up, until it was all the way up. Than as soon as there was a sound, bang, it turned the gain down to a level appropriate for recording that sound.

If you have a classical FM station that still plays classical LP records on occasion, listen sometime critically with headphones. When there is a quiet passage in the music and it is really quiet, the compressor at the station will crank the level up. If there happens to be a scratch in the record, the "pop" from the scratch will be very loud and it will cause the compressor to turn the gain down quickly, and it will "punch" a hole in the sound until the release time on the compressor allows it to turn the gain back up again. Once you hear it, it is easy to spot.

And that brings up a bit of a warning. Many audio tricks, once you understand how they work it makes them much less enjoyable for you as you will always be spotting artifacts of them. If you really just enjoy listening you may not want to peek under the veil.

And a note about listening. Often I prefer to listen to a wild recording. I think my brain makes a better filter most of the time than the electronics. But I will do a lot of manipulations. Often I can hear more what I want to hear but the recording is not as nice to listen to.

A good case in point is that glasgow recording here that I am so fond up. I love the wild recording. I also enjoy it in my winmap rotation with the EQ set and rocksteady, my compressor. It is still very listenable. I have versions processed through audacity that are very filtered and compressed. I can hear much more what was going on, but it is a much less enjoyable listening.

So, what tools do we have once we have a file. We have our brains. Often that is the nest tool of all for picking stuff out. Next we have about 3 categories of tools to play with. The first are filters, the second are dynamics, and the third are everything else.

Believe it or not our friends at Ma Bell have done tons and tons of research over the years and while we all love hi fi, the majority of the intelligence in audio is contained in one small little band, from 300Hz to 3000 Hz. Back in the old days you had to buy precision parts to build filters that had deep cuts, but now that the world has gone digital it is easy as pie. So, often times my very first thing to do to a file is snip odd everything outside of 300-3000Hz. In winamp I can use the EQ, ditto for VLC media player. You can get up to 15 DB of cut in winamp nativity, 30 DB of cut in VLC if you turn on the 2 stage, and I think 48 DB in Auditcy. Th later requires running it through once in HI PASS mode and putting the freq on 300 and setting it to the highest order, and than a second pass in low pass mode and setting it at 3000, and again the highest order.

This will not sound real good, but, and this is a big but, all the rumble will be gone as will most of the hiss, so you can crank the volume up and listen much more "closely" than you could before. But yes, it will sound like a telephone. Not hi fi at all.

Note, there is a physcoacoustic trick you can play on yourself if you want to. It is called perfuming. Another study, not bell this time as far as I know showed that adding some high frequency hiss to a file devoid of any high frequency information, made it sound "brighter". Apparently somehow in your brain the his gets modulated with the sound and you kind of synthesize the effect. So, now you can try cranking up the higher frequency faders on the graphic EQ and see for yourself. Sometimes just a little snort of noise actually makes a file sound a lot better, even if it makes it harder to "listen" to.

Cutting everything but 300-3Khz takes out all 60Hz remnants. If there is some other fixed sound, which is rare, but occasionally happens or if the sound "blares" at one frequency, you may be able to calm that with a parametric EQ application. A panoramic EQ is a cool tool. They generally only have a few bands, and the bands often times overlap. Unlike a graphic EQ where the center frequencies are fixed, and parametric EQ lets you move the center frequencies of the filters as well as the boost and cut and also the filters Q. The Q is the width of the filter. So you can boost or cut from a very narrow notch to a very wide band. For fun you c an crank the gain all the way up, crank the Q to the max, and sweep the frequency back and forth and listen to the wah wah effect on the audio, just like the guitar wah wah peddle.

The nice thing about doing all of this digitally is that the toys are free and you can try and tune it and try and tune it and try.

That's it for tonight. Next time I will revisit tweaking the dynamics, and some things that have helped me. Again, a lot of radical tweaks can often bring things out in a file, but there is a difference between bringing everything out in a file and having it sound good to you. Luckily you can make copies of both (smile)

[reggind]

Howdy! And welcome back!

Today we delve into dynamics, which is playing with how loud things are. In general when things are recorded they are compressed. Back in the days of LP's it was not unusual for an audiophile to have an expander. These little gems made loud things louder and soft things softer. With a bit of tuning you could make an LP sound much more alive, or less flat dynamically.

With the kinds of sounds that we are dealing with, and trying to pick little bits out around other things, most of the time we want to compress, and not expand. When audio is very compressed (to me) it is not pleasant to listen to. It is literally always at the same volume, be it the program material, hiss from the premap and mic, or dogs barking. But, it does let you hear much more activity. So again we get to that line in the sand where you have a nice sounding file on one side and a not so nice sounding file that you can make out a lot more info from on the other.

So, a compressor has a bunch of settings. It never helps that they call them different things on different ones, but it boils down to a few things, the lowest level that it will start working on. So if you have a file with people whispering in a tent that was recorded outside a few feet away, you want to set that level pretty low, perhaps -45db or so. You can always undo your changes and try again until you get the hang of it. Also start with a small file so the effects will be quick to run on the file. The next setting is the compression ratio. If you really want a "brick wall of sound" set it all the way to 10:1. Again, you can play with that later if you want to. The last things to play with are the attack and the release times. Generally the attack time should be pretty fast, on the order of 100ms, or 1/10th of a second. The release time is a bit more varied. If you set it too fast the file will sound very "pumpy" if it has a lot of noise in it. I generally start with about 300ms.

Now for a big note. If you are playing with your compressor and you hear a loud sound and than it gets quite for a few seconds and you look at your time sonstants and wonder what the hell is going on.. It is probably the AGC on the recorder. Don't forget most of the little recorders don't even have gain controls so they totally depend on the AGC to make decent recordings, and the time constants in their AGC circuits tends to be more in the seconds range than in the milisecond range. I have been fooled by this a few times myself.

One other thing worth mentioning here next to compressors and AGC is peak limiters. A peak limiter is a circuit that clips peaks off to prevent them from exceeding a threshold. This is typically near the loudest thing that can be recorded. In analog (tape) recording if you exceed that level it distorts and the output level can actually go down, and it just sounds bad. It sounds like you hit it too hard. But in the digital realm, even odder things happen. If the maximum level you can record is exceeded, the next level after that is zero. And then it starts over a 1 etc. For example if you have a 4 bit recorder (yuc, but some early phone systems used 4 bits for voicemail back in the day..) your lowest number is 0, or 0000 in 4 bit binary, and your largest number is 15, or 1111 in 4 bit binary. If you add one to 15 and make it 16, you wind up with 10000, but you only see the lower 4 bits, so the 1 is lost and your audio is all messed up. So, occasionally there will be a peak limiter that will as a last ditch effort clip anything that is just too big. The thought being the slipping sound bad but better than overflowing.

I use both the compressor in audicity and a plug in called rock steady in winamp for compression. They both work quite well, but the one in audicity is a more professional tool. Also because of the way winamp works with plugins, it takes a few seconds for any changes to take effect, so tweak your settings slowly, hit apply and wait until you hear the changes. It is easy to get ahead of yourself.

And last, we have the "everything else" section. There are only a few other tools that I use.

Sometimes I will use the noise removal tool. This is not the "best thing since sliced bread". If you have a consistent noise that you can not get rid of and have a good long sample of it, you can try this tool to get rid of the noise. Overuse of this tool will result in a file that sounds like it has been played through a guitar phase shifter effect pedal. Not good. This works by trying to cancel the sample from every part of the file, and has some parameters. It is kind of like the gadget in Photoshop that will let you tag a color and than tag anything "close" to it. Used sparingly and with a good quality input file, it can make it sound better. Used grossly it makes the file so bad you can't make out any words, even stuff you could hear before. If you are just listening for grunts, it might be OK, but I am in it for more than that (smile).

Another filter I use on occasion is the pop and tick removal filter. What? Am I dubbing records? Nope, but the pop and tick removal filter used sparingly does a good job of getting rid of spikey little impulse noise. You get a lot of that from those USB recorders when they are picked up and moved or people rollover on them or whatever. It is not a cure all, but it makes the sounds of the devices being touched with the gain cranked wide open sound a whole lot less offensive. Again, this is with a very sparing application. Also, if you use it to dub old albums, I have found that many mild passes works better than one harsh pass, but your mileage may vary.

And the last thing that I use very very rarely is... The time streatcher. That is the filter that will let you retain the proper pitch but expand or contract the playback time. Occasionally if you have a really hard time catching something, slowing it down but retaining the pitch is helpful. It is one of those things that makes the file IMHO, sound worse, but if you are just duying to hear what those muffled words were... This may be the ticket. I recall our phone system being swapped out years back and the old voicemail system had a gadget to allow you to slow the playback down, and the new one lacked that. The new system actually sounded a lot better, but a lot of people really missed the old feature saying it was hard to understand that people were saying at full speed in their messages.

That is about it for now, unless I think of anything else. I am always happy to help answer Q's and chat about hardware and software.

[mjj9994]

Great info and history of audiology, thanks!

If you get a chance could you maybe mention or add a tab to some of the tools you mention/use and what they do? I usually leave the software tweaking to the experts but in case someone may be interested. Thanks again!

[reggind]

I use two things for playing files, in my bedroom on that laptop I use winmap. I use one plug in for winamp, and that is rocksteady, which is a compressor. For casual listening on an older notebook that combo works well. The built in EQ in winamp is OK as is rocksteady. Warning if you play with this combo it takes a few seconds or longer for any changes to take effect. Play with the EQ first and crank one of the center (midrange) sliders all the way up with the EQ on and see how long it takes before you can hear it take effect.

On my better notebook I use the VLC media player, and it's built in EQ. I do not have a compressor. If I need to do more than I can do with the EQ I use Audacity.

Audacity is a great tool. It can do a lot of things. It takes some getting used to. It has a undo feature and if you have a fast processor that really encouraged experimentation. If you really hose something you can undo it and try again.

There is not too much magic to cleaning up audio files. If you depend on other people to do it for you, they may be going for something that you are not going for. Some people cancel tons of noise and leave the file sounding all phase shifted and funky. I don't like that sound, I would rather use less noise reduction, and have more noise than the artifacts of removing it. Some people might like the sounds of people sliding on the sheets more than the little deep breaths and moans and coos. I really don't like the physical noises all that much, so I would not be happy with the file they "cleaned" up. Some files I am really into hearing ever breath and sigh, and often times that takes a lot of compression and some playing with the time constants of the compressor. The end product is not "easy" nor real nice to listen to, but I can pick out what I am after.

The one thing that you will find when you start playing is that all of the tools have artifacts and after a while you will become very sensitive to them, even in non covert audio. The veil has been lifted...