The Song Fight Bulletin Board

I've been reading 'mastering audio' and am intrigued by 'upwards expansion' .. anybody here have any thoughts on that? I'm a bit annoyed that none of my fancy dynamics processors can do it.

-craig

I haven't had such exposure, either. But it's the domain of high end equipment and I don't live in that domain either. This synopsis may help (from soundonsound.com):

"The next two chapters discuss automatic dynamic control in great detail, starting with the downward processors — compression and limiting. Single and multi-band devices are considered, along with the effects and uses of each of the generic controls, and some sage advice about how to avoid making 'hypercompressed' recordings. Some rather less common dynamics processing techniques are the subject of the next chapter, with upward (or parallel) compression and the even more rare upwards expansion techniques (found in the Dbx Quantum processor and the Waves C4 plug-in, for example). To close the chapter, Katz brings us back to our starting point by describing ways to change microdynamics manually."

Are you talking about cascading a series of compressors (with minimal compression setting) one after the other, to achieve a high output, but transparent signal

If so that has been a staple technique used in pro mastering houses for years. However these mastering studios have many thousands of dollars of top draw outboard, fantastic treated rooms, magnificent monitors and superb ears

All of the things I want to have, but seriously lack


Just wondering if thats what you mean. Ive never heard that term, but Roy mentioned parallel compresion, and the waves device, which I've been told does have this feature....

dubbly post

Here is a FREE plug in that purports to do upward expansion.

This page has the download.

http://www.kvraudio.com/forum/viewtopic.php?p=934428

This page explains the thing.

http://www.platinumears.com/

Steve

oooh, thanks sdurand.
i will try it out.

me$$iah, with upwards expansion, the dynamic range is expanded by increasing the volume of the louder bits (as opposed to the more common downwards expander that lowers the volume on quiet parts).

if you are with a typical dynamics processor's transfer graph, no compression or expansion happens when you have a straight line from the lower left to the upper right corner. upwards expansion corresponds to a more vertical line above a threshold.

-craig

Awsome.

Thanks cats,
Ive downloaded the one that Steve Durand has linked to and Im gonna play with it. Sounds interesting.

Learn, learn,learn

starfinger wrote:oooh, thanks sdurand.
i will try it out.

me$$iah, with upwards expansion, the dynamic range is expanded by increasing the volume of the louder bits (as opposed to the more common downwards expander that lowers the volume on quiet parts).

if you are with a typical dynamics processor's transfer graph, no compression or expansion happens when you have a straight line from the lower left to the upper right corner. upwards expansion corresponds to a more vertical line above a threshold.

-craig

Never really heard about this, but couldn't you do the same thing by compressing the lower end of the dynamic range and leaving the highs intact? (On the transfer graph, draw a concave curve below the diagnonal) I can see how it's not really the same thing since you're necessarily gating the lowest dynamics, but is that such a big deal?

Mostess wrote:I can see how it's not really the same thing since you're necessarily gating the lowest dynamics, but is that such a big deal?

I think its' all about what is emphasized. The motivation in the katz book is that upwards expansion increases the dynamic range in a way that flows with the music. He particularly likes it on percussive parts, because the actual hits of the rhythm jump out more. (Unlike compression which tends to bring up the space between the beats)

I encourage you to check out the plugin that sdurand linked. It definitely sounds different than the more typical downwards expander.

that plugin also does upwards compression which is cool too.

-craig

Just throwing this out there to see if anyone has tried it:

When I very first started recording and hadn't discovered compression, I would basically compress everything by hand - that is to say I'd use a volume envelope and actually set volume curves on every part of the track. Acid has a very intuitive envelope system (that I wish every program had) that made the task a reasonable feat for songs with like 4 tracks (those were the days).

How different is that from a really good compressor? What exactly is it that makes a shitty compressor sound like shit?

Sorta related to TSI's question: in the pre-digital days, a compressor was / is a very time-based processor - it can't pull the volume down until it "hears" how loud the input volume is, and there is some amount of "reaction time" before it actually starts working (and then maybe a real drastic - audible - volume reduction).

Given that, with digital sound, everything is sitting there in .wav (or whatever) files, it seems to me like a "look ahead compressor" would make a lot of sense - the compression at time X would be based on what the levels are going to do in the next Y seconds - if there's a huge peak half a second from now, the compressor would already be pulling the volume down. Particularly for mastering stuff, this just seems like it makes sense.

Anybody know if anybody makes something like this? Or maybe it's been tried and there is some fundamental flaw in the idea?

Charles

roymond wrote: "The next two chapters discuss automatic dynamic control in great detail, starting with the downward processors — compression and limiting. Single and multi-band devices are considered, along with the effects and uses of each of the generic controls, and some sage advice about how to avoid making 'hypercompressed' recordings. Some rather less common dynamics processing techniques are the subject of the next chapter, with upward (or parallel) compression and the even more rare upwards expansion techniques (found in the Dbx Quantum processor and the Waves C4 plug-in, for example). To close the chapter, Katz brings us back to our starting point by describing ways to change microdynamics manually."

Yeah baby! That sort of talk is giving me an upward expansion!

charles, there are a lot of look-ahead compressors out there. if you want to play with a demo of a great plugin that can do this, check out Trackplug.


-craig

most plugins are still trying to act like their real-world counterparts. twiddling knobs.

The Sober Irishman wrote: that is to say I'd use a volume envelope and actually set volume curves on every part of the track
...
How different is that from a really good compressor? What exactly is it that makes a shitty compressor sound like shit?

are you talking about larger-scale volume changes over the course of the track (like between different parts of the song). because compressors are definitely better for manipulating the dynamics of drum hits or problem vocals (bob katz calls them "microdynamics").

As far as what makes a compressor "good", it could be because it's very transparent, or because it colors the sound in some pleasing way (analog circuitry and all that), especially when compressing aggressively.

-craig

From reading their page, seems like Trackplug has a "lookahead" function that lets you specify how far in advance it scans the sound. Doesn't sound to me like what Charles wants, and I'm with him. Why doesn't the plugin scan the whole sound file and make decisions based on the entire thing rather than taking it as it comes? Seems like a no-brainer.

It should be able to scan the wav, stroke its chin and go "hmmm, it's really quiet here in the beginning, but gets way loud later on. I'll adjust the envelope of my compression automatically to adjust for that."

The problem with doing that as a plug-in is, the plugin isn't necessarily going to be passed a .wav filename as input. And before you say "Well it should be!", imagine what life would be like if plugins were reading the .wav files directly rather than the audio that the host app sent to it. For a start, they could only ever be used as insert effects. For another, they could only be the very first effect in the insert effect chain.

None of this explains why you can't get a stand-alone app to do what you want tho', and I'd be surprised if there wasn't one.

well, charles's wishlist is very much like trackplug's lookahead, but the time scale is way different (typical lookaheads are more like 1-5ms)

if there's a huge peak half a second from now, the compressor would already be pulling the volume down.

I think this is to keep attacks from distorting when the compressor kicks in. it wouldn't really sound normal to fade out slightly half a second before a drum hit.

-=craig

obscurity wrote:The problem with doing that as a plug-in is, the plugin isn't necessarily going to be passed a .wav filename as input. And before you say "Well it should be!", imagine what life would be like if plugins were reading the .wav files directly rather than the audio that the host app sent to it. For a start, they could only ever be used as insert effects. For another, they could only be the very first effect in the insert effect chain.

None of this explains why you can't get a stand-alone app to do what you want tho', and I'd be surprised if there wasn't one.

In cubase there's a whole menu for this sort of thing Audio -> Process -> whatever. Normalization is under there. It runs a thing on the wav file, but not in real time. That's where the sort of analysis I'm talking about should be at. I've always thought this is just another problem with audio apps using the physical mixer metaphor. It's data analysis, but it's not presented as such in ways that improve things over, say, one of those audio analyzer things we used in acoustics class in college.

Not that plugin designers should abandon what people are used to, but why does that seem to be the only path they go down? The UI of these things leaves so much to be desired, I want to scream.

jb wrote:In cubase there's a whole menu for this sort of thing Audio -> Process -> whatever. Normalization is under there. It runs a thing on the wav file, but not in real time. That's where the sort of analysis I'm talking about should be at.

That's where it is in Logic, as well. I haven't explored other plug-ins for audio processing yet, but now I'll look them up.

I've always thought this is just another problem with audio apps using the physical mixer metaphor. It's data analysis, but it's not presented as such in ways that improve things over, say, one of those audio analyzer things we used in acoustics class in college.

Exactly. But the problem is often with user acceptance of new ways to approach old problems. people know about compressors and such, and that they have knobs and you monitor a live audio stream as it plays. Full-stream profiling is certainly a different level of challenge for the UI developers.

Seems easy to me:

<center><b>[ AWESOMEIZE IT ]</b></center>

jb wrote: In cubase there's a whole menu for this sort of thing Audio -> Process -> whatever. Normalization is under there. It runs a thing on the wav file, but not in real time. That's where the sort of analysis I'm talking about should be at.

Yeah, I'd forgotten about the batch processing. I still wouldn't expect it to be do-able there, tho', because I would expect those plugins to also work on audio passed in from the host rather than at file level, as they have to work on audio parts (which don't equate to audio files). They'd need two passes to do what you need, and I bet cubase just isn't designed to handle multipass plugins. Wavelab would seem a more likely bet, I wonder if that's got anything that could do it?

jb wrote: I've always thought this is just another problem with audio apps using the physical mixer metaphor. It's data analysis, but it's not presented as such in ways that improve things over, say, one of those audio analyzer things we used in acoustics class in college.

Not that plugin designers should abandon what people are used to, but why does that seem to be the only path they go down? The UI of these things leaves so much to be desired, I want to scream.

Total agreement here. The whole way audio channels get routed, with an input, a sequential step of modifications then an output, totally bugs me. Why can't we have a sequencer with the routing flexibility & power of Reaktor? I'd totally pay good money for that.