Needed advice to train my ear for subtle changes.

[5argon]

I have a big issue here. 80% of people says Glue Compressor is good, but I can never detect difference between Glue and normal Compressor! (asides from GUI stuff like you cannot set knee in Glue etc.)

Another example is when I was told to add a bit of Dynamic Tube to my kick drum. Yeah when solo-ing I can hear the difference, but not in the whole mix. Do I supposed to believe "The kick sounds better on its own, so now it's probably doing better in the mix although I cannot hear any difference!" like this?

The same thing happened on mastering stuff. You have to made subtle adjustment. But each individual adjustment is barely intelligible.. only as a whole then you can hear the difference. But then, I cannot make any judgment if each things I adjust barely change anything and in the end the result I got is often unsatisfactory.

I'm making music for about 1 year and probably haven't improved at all in this regard compared to day one. (I can say I certainly get better at using Live and knowing plugins stuff better.) Now I think I have missed some important point and/or specific ear training required to understand subtle stuff. Any idea to help me improve myself in this? Thanks.

[re:dream]

*bookmark*

Thanks for raising this issue. I'd be interested in hearing about this too...

[snakedogman]

A year is not very long to train your ear to hear stuff like this (although it depends on how much time you spent producing in that year of course). Don't worry too much about it, and it's probably best to don't think about "mastering" yet at all. Just try to make your mixes sound as good as possible first without any "mastering". Depending on your skill level and time invested it can take a couple of years before you get the hang of it. Especially with compressors it often not so obvious what is happening to a sound and what the difference is between compressors. It can be helpful to first experiment with the different compressors you have usign rather extreme settings. That will make it easier to hear the effect and the difference. Then you can slowly dial down the effect and see what it does to the sound and whether you like what it does. Also with compressors try to compare the signal with and without compression at the same level (use the gain control in the compressor to match up the level so it sounds the same when you switch the compressor on or off). This way you can more easily judge what the compressor is really doing to the sound.

And in the end, some people will simply have better ears for the small details than others, and of course a lot depends on your monitors and your room as well in order to be able to actually correctly hear the stuff you're doing in the first place!

But again my advice (and keep in mind I'm not a professional ) would be not to worry too much about when making music, but do allocate some time to experiment with effects, starting with really extreme settings then dialling back the intensity and see where you can still hear the effect. You will get better at it in time. Also check out some of the mastering/mixing guides by forum member Tarekith (http://tarekith.com/). He also provides a per-track consulting service where he gives you feedback on the mix of your track. (it's on his studio's site http://innerportalstudio.com/). I haven't tried this myself yet and I'm not affiliated with him But it's something I've wanted to try myself at some point because it seems to me one could definitely learn some interesting pointers from this.

[re:dream]

Wise words.

If I may add my own (totally newbie) 2 cents: I am sorry if this is totally obvious, but I found it very helpful:

A lot of the time it is about changes you don't hear, but that make a big difference to the final render.

Example: when I print my final mix (e.g. for Soundcloud) I will normalize it to 0 db. This means the average loudness will be pretty much determined by my single loudest transient (s) - 3 or 4 peaks that are a few dB louder than the others.

I have found that applying some very light compression (soft knee of 3 db; ratio of about 1:2, threshold a few dB or so lower than my absolute peak) will cut those few transients. It is quite impossible to hear the effect: we are talking about a transient of a few milliseconds being attenuated by three dB. But this allows a gain of about 3 or 4 dB in my average loudness. That makes a huge difference when I render out the track.

Obvious, I know - but it took me a long time to realize I did not have to hear the effect of the compression at all: I just had to calculate by how much I want to tame the peaks.

(In fact, what I trained is not my ear but my eye: by eyeballing the shape of the waveform in my wave editor, I can see that the percussion transients are too loud in relation to the synths.

Somewhere on the interwebs, I recall someone putting up pictures of three waveforms: a 'squashed' overcompressed one, a feathery undercompressed one, and a gently compressed track with a fair amount of dynamics retained, which was somewhere in the middle. I found that very helpful.)

[Daniel_S]

I have found that applying some very light compression (soft knee of 3 db; ratio of about 1:2, threshold a few dB or so lower than my absolute peak) will cut those few transients. It is quite impossible to hear the effect: we are talking about a transient of a few milliseconds being attenuated by three dB. But this allows a gain of about 3 or 4 dB in my average loudness. That makes a huge difference when I render out the track.

Obvious, I know - but it took me a long time to realize I did not have to hear the effect of the compression at all: I just had to calculate by how much I want to tame the peaks.

(In fact, what I trained is not my ear but my eye: by eyeballing the shape of the waveform in my wave editor, I can see that the percussion transients are too loud in relation to the synths.)

This will definitely help - every track I render out, I'll always throw it into Audacity to normalise to 0db too. Althought the overall track wave structure is fairly consistent, there's always the odd couple of peaks that stop the normalisation from doing it's best. The thing, is that when I then go back into my track to the time where those peaks were SEEN, I could not HEAR anything to suggest a spike in volume.

The trick then is to find out which track is the culprit and compress that a little.

I've yet to master this at all, but the understanding is there anyway haha

[Stromkraft]

Example: when I print my final mix (e.g. for Soundcloud) I will normalize it to 0 db. This means the average loudness will be pretty much determined by my single loudest transient (s) - 3 or 4 peaks that are a few dB louder than the others.

I have found that applying some very light compression (soft knee of 3 db; ratio of about 1:2, threshold a few dB or so lower than my absolute peak) will cut those few transients. It is quite impossible to hear the effect: we are talking about a transient of a few milliseconds being attenuated by three dB.

I think it's very unwise to normalize at all. The waveform at 0FBfs is uninteresting per se as an OK looking waveform doesn't guarantee that listeners won't get inter-sample clipping.

Instead I use a Limiter that compresses the sound safely to -0.3dBfs as well as raises the overall volume. You should realize that the MP3 encoders were constructed in a world where releases were not absurdly loud.

Striving for loudness in absurdum will only end up in tracks sounding distorted in the real world. I listened to the Football Championship mix CD that came with DJ Mag last summer. What a sordid affair. It really sounded like vomit smells!

Anyone promoting shit sound, including SHM, Guetta, DJ Mag or whoever can go **** themselves.

There's something wrong with the loudness war because as a DJ I match levels anyway. If a track distorts I throw it out and I won't buy a track with that artist for a very long time.

I can recommend this article over at hometracked if you want more informed opinion on the matter. There are many more out there too.
"10 Myths About Normalization" (Hometracked.com)

Quoting from Myth #6 in the article:

"Louder tracks leave less headroom before clipping occurs. Tracks that peak near 0dBfs are more likely to clip when processed with EQ and effects.
Normalizing to near 0dbfs can introduce inter sample peaks."

[re:dream]

(1) I don't think adding a bit of light compression to a few milliseconds of my track means I am striving for loudness 'in absurdum' I don't make dance music, rather chillout and downbeat electronica; I am not interested in the loudness wars.

(2) I read the article with interest. I don't really think I believe in any of the myths about normalization it purports to debunk. But thanks for posting it anyway; it's useful and common sense. I should say that if I were to send the tracks to a mastering engineer, I would of course not normalize. I would print the track at -6 dBFS and -18 RMS like everyone says I should. But since I am just making music for fun, and don't have money to send it for mastering, I don't worry about leaving headroom for someone else: I just try to make sure my music is more or less equally loud to all my other tracks. Which is not very loud at all.

(3) My point is not about normalization. It is that I find that I find that I can make my tracks sound better on the whole by using compression to tame the loudest transients. What I find interesting is that although the difference to the peaks is almost inaudible, it allows for an improvement in the balance of the elements in the track as a whole. That is so whether or not I normalize.

[fishmonkey]

I think it's very unwise to normalize at all. The waveform at 0FBfs is uninteresting per se as an OK looking waveform doesn't guarantee that listeners won't get inter-sample clipping.

Instead I use a Limiter that compresses the sound safely to -0.3dBfs as well as raises the overall volume. You should realize that the MP3 encoders were constructed in a world where releases were not absurdly loud.

the distinction between normalising and limiting that you are making is a bit confused. normalizing and limiting both increase the loudness of a track. for example, a limiter set with a high threshold and makeup gain to take you close to 0 dBFS is effectively working as a normalizer.

if you are already happy with the dynamic range and loudness of your mix, then there is nothing wrong with normalizing, although it is probably a good idea to leave a little headroom for lossy compression that might create intersample peaks.

although i see plenty of intersample peaks in lossy versions of official releases these days...

[re:dream]

if you are already happy with the dynamic range and loudness of your mix, then there is nothing wrong with normalizing, although it is probably a good idea to leave a little headroom for lossy compression that might create intersample peaks.

That's a useful observation. So if uploading to Soundcloud, I will normalize to -3 dB or so.

[fishmonkey]

if you are already happy with the dynamic range and loudness of your mix, then there is nothing wrong with normalizing, although it is probably a good idea to leave a little headroom for lossy compression that might create intersample peaks.

That's a useful observation. So if uploading to Soundcloud, I will normalize to -3 dB or so.

it doesn't have to anywhere near that drastic. usually a fraction of a dB is necessary. experimenting is the best way to learn. convert some files to MP3, and play them using a meter that shows intersample peaks...

[Tarekith]

A few things:

1. Learning to hear the effects of something like compression can take a long time to learn, especially when you're talking about different characteristics of multiple compressors. Dont be discouraged, just keep trying and eventually you'll get there. Also, a lot of it can depend on your monitoring environment, and you didn't mention what you're using.

2. Usually I find that compression is more noticeable in how it feels versus how it sounds. Turning up your monitors while you play with the attack and release helps, see if you can feel how the punch of the sound hits you in the chest more or less as you tweak. Unfortunately I find that having an amp powerful enough to react to these transients quickly helps A LOT, and sadly a lot of the cheaper self-powered monitors just don't react fast enough to make this easy to notice. Still, that doesn't mean it can't be done, you just need to keep practicing and referencing what you do on other monitors and listening environments.

3. There's nothing wrong with normalizing per se, especially if you have a pretty dynamic mix. You're less likely to get intersample-peaks with normalizing versus limiting a lot of the times in fact, even with the limiter output set to something like -0.3dBFS. Oversamples happen when you get a lot of samples peaking in a row, it gets reconstructed and can be interpretted as a peak over 0dBFS, and thus causes tiny, tiny amounts of "distortion". When you limit something, you're more likely to make lots of contiguous samples all hit the same max value, and even with an output as low as -1dBFS you can get intersample peaks at times.

All that said, it's important to keep in mind that these intersample peaks are very often totally inaudible, especially with today's DAC's. This was more an issue early on in digital audio than today, though it's good to be aware of it. FWIW, in my mastering business I pretty much always leave the limiter output at -0.3dBFS (unless I'm prepping MFiT files) and I know I get intersample peaks, but their inauduble to me, so no biggy. Almost all songs limited for volume these days (eg, the vast majority) have intersample peaks, so just something to keep in context.

4. The production guides I wrote that were references earlier are now hosted here, not on my Tarekith.com site:

http://innerportalstudio.com/guides/

Hope that helps, feel free to let me know if you need any clarification. I'm away from internet a lot at the moment with my move, but I'll do my best to reply to any questions in this thread as soon as I can.

[re:dream]

This is very useful, thanks very much.

Any recommendations for a good intersample peak meter? I see one by Solid State, but it's not 64 bit compatible...

I notice that my wave editor (ocenaudio) warns me about 'possible clipped samples'; I suppose that would refer ISPs?

It's a pity that Live's export audio function doesn't allow you to set the level you want to normalize at; it assumes 0dBFS

[fishmonkey]

i use K-Meter:

http://www.meterplugs.com/kmeter

[crumhorn]

Let your ears rest frequently while working. Otherwise you loose perspective. I often waste hours tweaking to get the sound 'just right' then when I listen to it the following day it sounds crap.

[re:dream]

Yup. So doing final mix goes well with domestic life. Take out the trash; take the 11 year old to her sleepover, do a bout of shopping ... in between little snatches of time at the controls