Clipping and Compressor Usage

Hello,

This is my first post, and I appreciate any help that is offered!

I am recording an accordion, which has a very wide dynamic range. I am confused about the best way to go about setting the volume for the mic. From what I’ve read on this forum, I should try to maximize the volume into the red range, but avoid actual clipping. I find this rather iffy and am sure there must be a better way.

I thought maybe I could use the compressor to compress the highs somewhat, but if I should, when do I use it? That is, should I record without using the compressor, and then apply the compressor later as an “editing” tool?

At one point, I thought the “normalize” function would do the trick, by simply setting the maximum to some level and then allowing the program to recalculate the wave file. However, the more I’ve tried to “study” this, the more confused I have become.

Perhaps someone could recommend some good articles that would help me to sort this out. I am wide open to suggestion, and I appreciate any help that is offered!

Thanks!

Tom.

If you’re recording at 16bit you should try to get your levels high without clipping. For a dynamic instrument, try to get them mostly in the yellow, only occasionally jumping briefly into the red. It will take a few tries to get find teh right input level. If you’re recording at 24bit (i.e., you bought a special 24bit audio interface and paid extra fro the 24bit version of n-Track), then you can record with a “safer” (lower) level where clipping is not a possibility.

Stay away from “normalize.” It almost never helps anything.

Record without the compressor or any other effets/plugins. After you’ve recorded the track, use the track EQ. If your mic has added any coloration or tone you don’t like, try and EQ this color out and make the track sound as natural as you reasonably can.

Now add the compressor. The way a compressor works is to reduce the volume of the track by a ratio amount whenever the track exceeds the threshold level. This allows you to bring the quieter parts up without causing the louder parts to clip; you are compressing the dynamic range. One of the main reasons you’d want to do this is to help a dynamic instrument (or voice) sit more evenly in the mix. In a full mix a single dynamic track often sounds alternately too quiet and too loud, never just right.

Unfortunately, because the behavior of compressors is dependent on how often the track crosses the threshold - which will be different for every track - it’s very hard to give blanket recommendations about settings (many compressor plugins I’ve seen don’t even have presets programmed in). But as a place to start, try a ratio of 4:1 and a threshold of -9dB (I’m guessing you recorded mostly in the yellow which is centered at -9dB). Play with lower and higher thresholds and greater and lesser ratios. Compression can be made to sound like an effect or it can be subtle and virtually transparent. You have to play with it and get a feel for it.

Hope this helps!
•KªRL•

Compressors are used for lots of things, including what Capt. D recommends (although I’ll bet you’ll find that a ratio lower than 4/1 will be the ticket). They can also be used as limiters at the time of recording, which can really help get hot signals - and I’m certain that I am wrong about this, but even with 24/96 or whatever, hotter signals always sound better to me. So if you have the option, you might try a (hardware) limiter going into the soundcard, just to make sure you catch any overs that might creep in, and try to get as much as you can.

For myself, I like to hear the dynamic range on the instrument, especially if it’s solo. Putting accordion in a mix might require a lot more compression to get it to sit. But I’ve always loved accordion; put it in a nice room with a nice mic or two not too close to the instrument and let it rip. :)

Hey folks, TomM was the 2006th registered forum member, doesn’t that numerological coincidence entitle him to some free plugins or something? :D

Thank you, Karl!

So, a few questions come to mind…it sounds like I should upgrade to the 24bit version of n-Track. I am using an M-Audio USB Duo, which I believe is 24bit capable. I don’t really understand how this would reduce/eliminate the clipping, but it sounds like a good way to go.

I should add that what I’m doing is pretty simple. I’m recording myself playing the accordion into a single mic (AT3031), so I end up with one single track (I suppose I should have bought 1-Track!!). Clipping has been a problem for me (I assume because of the wide dynamic range). I want to make these recordings for my Mom, and I don’t want any artificial harshness in there. She likes what I’ve done for her, but that’s because she’s my mother! (I hear little bursts of harshness that I don’t like, and which I assume originated in the clipped wave file.)

So, Karl, are you saying that if I use the 24bit version of n-Track , I won’t have a clipping problem to worry about? That sure sounds like the way to go.

I am totally new to the compressor, so I will take your advice and play around with it. The key thing you’ve told me is to record without any effects. By the way, the compressors I have available are the one that came with n-Track Ver. 4.2.1, and the Kjaerhus Audio Classic Compressor (which was recommended in one of the posts I read in this forum). Which would you recommend?

Thanks again for your help!

Tom.

TomS said:
<!–QuoteBegin>

Quote
Compressors are used for lots of things, including what Capt. D recommends (although I’ll bet you’ll find that a ratio lower than 4/1 will be the ticket). They can also be used as limiters at the time of recording, which can really help get hot signals - and I’m certain that I am wrong about this, but even with 24/96 or whatever, hotter signals always sound better to me. So if you have the option, you might try a (hardware) limiter going into the soundcard, just to make sure you catch any overs that might creep in, and try to get as much as you can.


Hi, TomS! Thanks for your input. As I said in my previous post, I have very little experience with using a compressor, so I’ll have to play around with them. Sorry, due to my inexperience, I don’t even know what a “hot signal” is, and what it would mean in my situation. Perhaps you could say more about this. I don’t have a hardware limiter, so I can’t try that idea. My setup is very basic…just a mic going into the Duo, and the Duo feeding into my computer.

TomS then said:
<!–QuoteBegin>
Quote
For myself, I like to hear the dynamic range on the instrument, especially if it’s solo. Putting accordion in a mix might require a lot more compression to get it to sit. But I’ve always loved accordion; put it in a nice room with a nice mic or two not too close to the instrument and let it rip.


As I mentioned in my other post, I’m playing solo and making simple recordings for my Mom. I actually am very pleased with how it’s going so far, but this clipping/distortion issue is something I’d like to learn to control. And, like you, I really enjoy hearing the complete dynamic range, from ppp to fff. Being a neophyte at this recording business, it seems to me it will be quite a challenge to capture all the dynamics.

Hey, I like your idea about the free plugins! Only one problem…at this point in my recording career I probably wouldn’t know what to do with them! :D

Thanks again for your input, Tom!

Tom.

<!–QuoteBegin>

Quote
So, Karl, are you saying that if I use the 24bit version of n-Track , I won’t have a clipping problem to worry about? That sure sounds like the way to go.


Recording at 24bit will allow you to retain high resolution (audio quality) at a lower input level. So you can record at a safer level, staying out of the red alltogether without being worried about how close you’re coming to clipping. At 16bit you generally want to record as hot as you can without clipping to get the maximum resolution. To get a little more technical, the lower the input level, the fewer bits you’re actually using to encode the sound. So while a low level at 24bit might still be using 16bits, the same level at 16bit would only be using 12bits.

I’d work at 16bit and get used to recording correctly before springing for 24bit - just my $0.02

Well, a “limiter” is a compressor with a high ratio setting - 10 to 1 or greater is the standard definition, I believe. They are often put in line between the preamp and recorder to make sure that signals that are too “high” (or “hot”) don’t cause clipping. Theya re also used in other contexts - e.g. to protect speakers in high volume settings like concerts.

The concept you need to think about in this regard is “gain structure.” Think of all of the separate steps in recording - typically: the sound comes out of the instrument, goes into the microphone, if it is a condensor mic it will have some electronics in it that will amplify the signal a bit, then it goes into your preamp, then into whatever external or “outboard” processing you want to apply to the signal before it is recorded (e.g. limiting or compression or EQ or whatever) and then into the recorder. At each step the level in must be correct, and the level out as well (so that the level into the next step is correct). So you can get clipping at the mic, in the mic’s electronics, at the preamp, at any outboard processing, at the recorder (e.g., the digital converter, which is where you are having problems) and (if one is recording to tape) at the recording head on the tape (“tape saturation”). Not all of this overloading is necessarily bad - most rock records are recorded pushing the preamps, and they used to be recorded with tape saturation often enough, and still are when tape is used. But that’s not what you want, not right now - you want a clean representation of your playing. So…

Your mic is a really nice mic, and the m-audio duo I’ve heard is also great, I know that it has a 20 db pad on it and a clipping indicator, so gain structure should not be a problem at all. Personally, I’d go register the 24 bit version and do exactly waht Captian D. said in his first post, but he has a really good point about learning using 16 bits first. But I’d still go for the 24 bits. It’s not like it costs very much, and you will get all the extra headroom from 24 bits, plus better sound from the 96 sampling rate.

The other thing i’d think about is getting a second 3031, 'cause if you have a nice sounding room (living rooms often work pretty well!) then you can get very natural stereo recordings (but note that there are placement issues when you are working with two mics, b/c they can cause comb filtering due to phase cancellation - which means exactly what it sounds like it means, some frequencies can be cut and others enhanced and the result is often unpleasant - so if you decide to do this, you might want to look at some of the on line lit on mic placement - DPA microphones has some great general articles on this, featuring their mics of course, but still useful). I’ll find the link, but it used to be under dpa mics “microphone university” - which had an article about placement for recording accordion, I do believe. :D

It’s really very simple. If you’re getting clipping, reduce the recording level a bit and try again. Repeat until you can record without any clipping. Bingo, done.

When recording at 16 bits, you generally want the resulting recording to peak at -12dB or higher, because every 6dB below 0 essentially wastes a bit. So, peaking at -12dB, you really have a 14-bit recording (which, despite all the arguments, is still a very good recording). But 13 or 12 bits? You’re starting to suffer.

With a 24-bit recording, if you waste a couple bits, you still have plenty left. Enough so that you no longer need to think in terms of wasting bits, and instead worry about the analog noise floor of your mic preamp and A/D converter, and then only for signals with rather low levels. For example, recording a piano solo that ranges from ppp to FFF, the ppp section will be very quiet. My suspicion is that an accordion part doesn’t generally vary in volume quite THAT much, and you should be quite OK.

PS: Captain, your idea is right but your math is wrong. If you’re peaking at -48 dB, you’d get 16 bits on a 24-bit recorder (if it really recorded 24 bits, but they don’t) or 8 bits on a 16-bit recorder, not 12 bits.

Cheers
Jeff

Karl, Tom, and Jeff:

Thank you all for your contributions. I’m printing this off and I will refer to your guidance quite often.

Karl, putting all the comments together, and trying to understand the difference between the 16 and 24 bit recording (which I do not yet understand), I’m thinking I’ll upgrade to the 24 bit. I get the sense that it "can’t hurt."

I’d like to understand the 16 bit/24 bit thing better. Can you guys recommend any reading for me?

Tom, about the 2nd mic, and recording in stereo…I have done some experimenting with this, and have (for the present time) decided I like the sound I get using only one mic. That’s not to say it’s “better”…just a personal preference. There has been some debate about this within the accordion community.

What I’ve been doing is placing the mic about 3 to 4 ft. in front of the accordion, but aimed at the treble side. The bass side is still captured with this arrangement. Then, as recommended to me by an accordion friend, I have cloned the track, panned the two tracks a little right and left, and combined into a stereo track. Not very sophisticated, I admit, but it seems to make a nice sounding recording. My Mom (90 years old, bless her soul!) thinks I’m right in the next room playing for her.

Of course, I would like to learn how to make my recordings better and better as I go along. I’ve been reading the n-Track User Guide, but I think I find more helpful answers right here on this forum! (Not to say the User Guide isn’t a good thing!)

Thanks again for the help! You guys are great. :)

Tom.

wow—yes good stuff! i never realized that you actually used up bits till i read this. so by having 24 bits as opposed to 16, one is pretty much in la la land without having to worry about sound degradation. right? ???

Right, jwgeetar, although you still have to sweat the details in your analog chain. In general, the analog chain becomes the “weak link”, rather than the number of bits per sample, and it becomes far less important to crowd the signal near 0dB; you can leave much more headroom. Note that the nubmer of bits also affects calibration of your analog gear to your digital gear. With 16-bit recording, I’d have 0dBVU on my mixer much closer to 0dBFS on my DAW – say, -9 or -12dB. With 24-bit recording, since the “floor” is much lower, I can set it in the -18 to -24 dBFS range. Most home-rec folks don’t even bother “calibrating”, they just guess & by gosh the levels in the chain and frankly, when done with sense (keeping signals in the green along the chain), it works just fine for all but the most demanding cases.

TomM, a little primer on how digital recording works.

44,100 times a second, your soundcard measures the voltage of the input line and records it as a number. For 16-bit cards, that number is between -32768 and +32768 – a number that fits in 16 bits (2 bytes) in a computer. For 24-bit cards, the number is between +/- 8.4 million (give or take), a number that fits in 24 bits (3 bytes) in a computer. The DAW software saves this data to disk and remembers the file name for playback.

These numbers we’re talking about are called samples. And they’re integers – whole numbers, no fractions. So, clearly, the 24-bit card has lots finer gradation between values.

In both cases, though, they take the biggest number that can be represented and set it to represent the same level, “full scale”, and everything is relative to that. This makes it what we call “fixed point”. Let me translate that to the decimal system we use for normal arithmetic as an example to show the difference.

Let’s say soundcards worked in decimal rather than binary. The same principles apply. Instead of “16 bit” and “24 bit” converters, though, we’d have something like “4-digit” and “7-digit” soundcards. The cheapies would record values from 0.0000 to 0.9999, and the expensive ones would record from 0.0000000 to 0.9999999. The good ones would have 3 more digits of precision for every sample they record.

When an actual voltage of 0.123456789 is presented to the two soundcards, the 4-digit one records 0.1234; the 7 digit one records 0.1234568. (Yes, it rounds off.)

As you can see, the 7-digit card gets more precise values. But what does that really mean?

Well, let’s say that we’re recording a sound that has one fairly loud part to it, and a softer, more subtle thing in it too – like a big man shouting while a small child is whispering. We have to set the levels so the big man doesn’t clip. If we recorded the child alone at that same setting, let’s say the sample values would fall in the range of +/- 0.0000500. Well, the 4-digit soundcard would hardly register it at all! But the 7-digit soundcard would record three digits of it – enough for it to be clearly recognizable, even if not as good as you’d like if recording it alone.

Of course, anything we record has multiple components, some of which are loud and some of which are quiet. Now, whether we can actually HEAR anything that’s so quiet the 16-bit card wouldn’t even have registered it (buried in a signal that’s full scale on that soundcard, like the big man shouting) – that’s a subject of debate. But note that earlier we said we’re going to give up some of the high end of the range to avoid clipping – a “safety headroom”. That makes everything get recorded more quietly (lower signal values). Later we adjust the fader to put the level where it belongs in the song, and if it’s a boost, we’ll be glad that we have a few extra digits on the right hand side to retain the clarity of the original.

As an example, go back to the 4-digit converter. Let’s say we reserve the top digit just in case that big guy shouts louder when the party starts than when we asked him to set levels. (Which is almost ALWAYS the case!) Well, now we really only have a 3-digit converter! That top digit might not even get used, and if it does, it’s just to catch a whoop here or there that was louder than expected. Well, I’d sure feel better recording it with the 7-digit converter, so that we have 6 perfectly good digits.

Taking this example a bit further: maybe the guy went over the top when setting levels, and in the actual party, he’s much quieter. Let’s say it’s enough so that now we hardly use the top two digits. With the 4 digit converter, we’re down to only 2 digits – a range from -100 to 100. Um, not really good enough for anything but cheap games. With that 7-digit converter, we’re down to 5, still better than the 4-digit converter under ideal conditions.

HTH
Jeff

I’m in awe, Jeff.
I didn’t get it before I read your post.
Now I get it.
Thanks for the extraordinarily thoughtful and articulate explanation!
PR

Jeff,
Now I understand WHY 24-bit is much better.
Thank you, now I can justify buying a 24-bit S/PDIF device to my woman. :D
Any help with that?

Excellent clarification Jeff! I appologize if my “math” confused anybody. I was just trying to describe the principles. YOu shouldn’t expect a lot of accurate math out of musicians before 9 AM! :D

Quote (Captain Damage @ Aug. 24 2006,06:03)
Excellent clarification Jeff! I appologize if my “math” confused anybody. I was just trying to describe the principles. YOu shouldn’t expect a lot of accurate math out of musicians before 9 AM! :D

i thought 24bits was hard, now—whats 9 am? :D

<!–QuoteBegin>

Quote
i thought 24bits was hard, now—whats 9 am?


9AM is: 00111001 01000001 01001101 :p :D
Quote (Gizmo @ Aug. 24 2006,14:13)
i thought 24bits was hard, now---whats 9 am?


9AM is: 00111001 01000001 01001101 :p :D
OMG!! 3 chords, a six pack, distortion pedals and have gone digital! :D

do any of you guys write songs just for the helluvit? :p :D :D

Hey, Jeff!

Thanks very much for the excellent primer on digital recording. I read thru it and decided to do a search for “24-bit” or “24bit” or “24 bit” or “24” and I found 6 pages of stuff. Lots of reading to do. Your primer has prompted me to want to understand this better, and I thank you for that!

Looks like it will be a few days before I’ll be able to get to all these postings. My daughter came to visit, along with her 3 boys, for a few days. So that kinda puts a temporary halt to Grandad’s fun! :(

Just wanted to tell you I appreciate your comments very much! :)

Tom.

You’re welcome, Tom. Cheers! :) And enjoy the kids.