Using Har-Bal to analyze dynamic properties

This is an area where you can share your most unusual or creative application(s) using Har-Bal. We receive emails from folks who are using Har-Bal in ways we never would of imagined!
Post Reply
Michael Hansen

Using Har-Bal to analyze dynamic properties

Post by Michael Hansen »

In a quest to find out more about how some of my favourite albums were compressed, I got the idea that the difference btwn the average power plot and the peak power plot could tell something about the compression - not really sure what exactly - so I loaded a track both as reference and source.

First approach: I moved the whole average plot (green curve) roughly so it overlayed the reference peak plot (yellow), then I finematched the overlay by hand. The result was not positive in regards to the overall sound of the song, but somehow it all became more even distributed. It sounded like all the hard compressed areas were reduced in volume.

Second approach: Instead of overlaying the average onto the peak power plot, I did the opposite (somewhat problematic due to the default cursorsnapping to the green curve, but not impossible), so the yellow curve was matched to the green. The result of this transformation again was not positive, and resulted in a more boomy sound, like all the hard compressed areas got higher in volume.

The interesting part comes when you shift view to the frequency response after overlaying one curve upon the other. Now you see the difference between the power and the peak plot. The closer the curve is zero, the less difference you have between the power and peak plot. What does this tell us? That these areas "close to zero" in the frequency response are more compressed or what?

I'm interested comments about comparison of the curves and their difference. If anything at all what exactly does this tell us in regards to compression? Is it useful for anything at all (not only in regards to compression)?
Last edited by Michael Hansen on Tue Nov 21, 2006 2:39 am, edited 1 time in total.
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Post by HarBal »

I don't quite follow what you are doing or trying to do here. In terms of dynamic range interpretation, the space between the green and yellow lines gives some indication of how compressed it is. The more compressed the closer they will be together (as a general rule).

The difference in shape between peak and average traces arises out of your program material having a specific structure in time. By that I mean things like thematic changes, instrument solos and the like. For tracks that have those characteristics the shape of the peak and average trace can and should look different. If it is a full on rock song then they'll more than likely look very similar in shape.

Fundamentally Har-Bal is an EQ and not a dynamics processor so using its EQ features in an attempt to radically change the dynamics is likely to fail.

Regards,


Paavo.
Michael Hansen

Post by Michael Hansen »

HarBal wrote:I don't quite follow what you are doing or trying to do here. In terms of dynamic range interpretation, the space between the green and yellow lines gives some indication of how compressed it is. The more compressed the closer they will be together (as a general rule).


Thanks for your quick reply :)
That "general rule" exactly was my initial guess before I did the experiments.
What I'm doing (or trying to) is not to use Har-Bal as a dynamics processor, rather in this "unusual application" I am using Har-Bal to analyze general dynamic properties of the program material.

In addition to my explanation above, let me clarify:

I am loading the same track as both source and reference.
When I then match (overlay) the green curve of the source to the yellow curve of the reference (the first approach above), what I get is equal to subtracting the difference curve of the two from the green curve: Gs + (Yf - Gs). Switching to the frequency response view after doing this, will show the difference curve (Yf - Gs).

In the second approach I did it the other way around, matching the Yellow curve of the source to the green of the reference: Ys + (Gf - Ys). Switching to the frequency response view after doing this, will show the difference curve (Gs - Yf), and this time the freq. curve is inverted in comparison to the freq. curve of the first approach.

The audible results are really not interesting, but confirms the application. In the first approach everything sounds flat or lifeless, in the second it sounds boomy or overcompressed. This makes sense since one approach reduces the dynamic differences and the other increases them.
But, it is not really the resulting EQ that I'm after, its the frequency response (as a source of information), because this tells me how the material were compressed in the different frequency areas. Information which could be useful for configuring multiband compressors, or learning how compression was done with given material (at least that was the idea..).

See the idea?
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Post by HarBal »

Yes, I see your idea but I think it can only be truly informative (regarding the compression settings they may have used) if you have access to the source prior to compression and post compression and then compare pre and post spectrums. Only with both can you tell what effect dynamics processors have had on the spectrum.

On the other hand, the difference between peak and average traces for a given track can be used as a guide for setting up a multiband compressor as you've pointed out.

Cheers,


Paavo.
uncajesse
Posts: 27
Joined: Tue Apr 17, 2007 10:53 am

Post by uncajesse »

I just thought I would add, for the benefit of your sanity Michael, that you can switch which of the currently enabled EQ filter plot lines (peak/mean/average) your shelving/parametric cursor is "focused" to when adjusting by... holding down the left mouse button to make an adjustment, then while still holding the left mouse button - pressing and releasing the right mouse button. it will also make new adjustments on that plot line untill you switch it again, and/or disable that plot.
Post Reply