Voxengo Premium Membership - All Voxengo Plugins For a Fixed One-Time Fee
Forums     Plugins     CurveEQ HarBal

This topic was created before release of the latest product version, and it may contain details irrelevant to this version.  Replying is disabled for this topic.

"it's just using an A-weighted analysis to ensure the output file has a similar perceptual loudness to the input file.  It's only purpose is to make pre- and post-EQ comparisons easier." Andrew Milne.

It applies the compensation in virtual real time as you are making the adjustment.  That is what is so truly useful about it.

If you think making an Eq edit as step one and then doing the loudness compensation afterwards separately as step two can ever beat having it all done together in real time, then I can only suggest you're living in cloud cuckoo land.  The freedom that Harbal gives the engineer is remarkable and if you can't see that then I suggest you continue to do it the hard way and screw up your mixes.  It is really remarkable that people make comments on products where they have no hard user experience of using them - only a perception based on opinion and feedback.

I'm a user of Voxengo products - absolutely wonderful, but you won't see me, (or I suspect many other users of Harbal), rushing to stop using the Harbal product any time soon.

They say ignorance is bliss and I'm telling Andrew Milne that blissfully lounging around for too much of the time has its dangers!


wakey, could you give some guidelines on how CurveEQ could get the similar functionality?  Will equal loudness compensation of the spectrum be enough?

Wakey Wakey you are very rude.  I understand perfectly well how HarBal works, thank you very much.  By using the terms input and output file, I did not mean to suggest that this was not a real-time process, so apologies if I gave a misleading impression on that aspect.

As a software developer and a newcomer to audio DSP I wanted to better understand the different approaches CurveEq and HarBal take.  I have read quite a lot e.g. http://www.nonoise.org/library/sndbasic/Sound.pdf.

Let me quote from this document :

"The human ear can not detect or "hear" lower frequencies as well as higher frequencies.  The A-weighting network is used to duplicate the sensitivity of the human ear."

Great.  It begins to make sense.  HarBal analyses the spectral content of the whole of the input source and displays the results after applying an A-Weighting compensation.  It then makes it very easy for the user to modify the spectral curve and to A/B the results all in real time.  This approach is based on the observation that good mixes for different music genres exhibit similar spectral patterns.  The fact that this results in user appreciation does rather support the view that this approach can work.

Aleksey Vaneev asked :

"wakey, could you give some guidelines on how CurveEQ could get the similar functionality?  Will equal loudness compensation of the spectrum be enough?"

My impression is that A-Weighted compensation is the correct approach.  What I am saying is that if I were to implement such a system, I should use A-Weighted compensation based on the many documents and papers I have read over the past few weeks.

I note that there are many requests for Harbal to be made into a plugin amongst many others smaller improvements.  Clearly this would make it much more useful.  If CurveEq could feature similar functionality, I am sure it would receive even more appreciative plaudits than it already does.  This brings me to make a point regarding the development of these tools.  As a developer, I am amazed at how close people like Aleksey and others get to their users and how responsive they are to change requests.  This is great and only a good coder would be comfortable with this.

I have bought many Voxengo plugins and I have purchased a licence for HarBal.  I am fascinated by what is coming next from you guys...


As far as 'A' weighted (or any weight for that matter) compensation goes there needs to be a switch (off/on) if it were implemented in CurveEQ.

Curve EQ has both single track mixing and stereo buss mastering uses.

For example during mixing with CurveEQ inserted on a single track sometimes you want a resonant bump and you adjust the EQ to give you exactly that without making you reach for the track fader at the same time.

I like the idea of a weighted compensation and would like to play with it in CurveEQ2, hehe.  I would think that keeping the track at the same average or rms level would work fine also (I guess).  This could even be an arbitrary setting made by the user that CurveEQ would always keep the average or rms at -23dB for example.  Otherwise wouldn't CurveEQ have to read the entire track first to determine it's relative levels ?  That's what Har-Bal does.

So I guess my vote would be that CurveEQ simply have an on/off switch and a knob or silder that lets you set a value for a compensated constant average dB or rms level - the compensation could be 'A' weighted or 'C' weighted or any reasonable weight.

ED: Come to think of it for CurveEQ to 're-calibrate' or compensate to any level it would have to read the entire wave first to get the level in the first place - in other words get the level context.  Maybe this isn't that hard since it seems to be what a portion of the curve matching feature already does.  Once a control point is moved then it would have to re-read the entire wave again - I guess.  That might be the tricky part to do in real time.  Har-Bal does it somehow though in real time maybe by sampling.


"A" weighting is a pretty unsophisticated and inaccurate way to estimate "loudness".  There are better psycho-acoustic models available (e.g.  Zwicker).  Of course the best method is called the human ear and a volume slider.  I apologise in advance for being facetious.

I just read this thread again - hehe repeated myself a few times - did I mention I wanted a switch... :)

wakey - you're overreacting - nobody here told you not to use Harbal - get Andews' ankle out of your mouth !  I take it you've been beat up on other forums for using this - hehe ME's didn't react too kindly to the 'replace your mastering engineer' early promos did they ?

HarBal is the only EQ that I know of right now that does any kind of 'loudness compensation' so it's a tool that has it's uses like anything else.


There is so much to learn in so little time...

Anyway, can I recommend a couple of books that are helping me understand the complex subjects of mixing and mastering.

The first is called "The Art of Mixing" by David Gibson.  This uses a 3D visualization technique which is fascinating.  ISBN is 1-91837-117-1.

The second book is called "Mastering Audio" by Bob Katz.  ISBN 0-240-80545-3.

These books are helpng me enormously, along with the posts in these forums.  However, apologies if this is the wrong place to post this.


By the way, anybody heard about the SPL compensating curve?  I.e. the one expressed in dBA and and dBC?  Seems like it is exactly the curve which can be used for general fine-tuning.  Equal loudness curve may offer fine tweaking while sound pressure curve may help to get a better overall balance.

Yes I remember - we were talking about SPL dBA/dBC earlier last summer.  I have a radio shack SPL meter and a DEQ2496 that has a meter weighted that way.  An Eq curve could be built using this link:

http://www.measure.demon.co.uk/Acoustics_Software/a_weight.html

Ian - I've got the Bob Katz book, thanks for the other link.  I figure I'll get all the stuff together this lifetime then learn how to use it properly in the next one - hehe :)

Concerning visual spectral balancing techniques and various tools tonight I'm going to repair a bad mix (ugly resonances at 100, 160, 320, 650, 950, 1.2K, 4K) using a variety of tools.  I used Har-Bal last nght.  Tonight it will be:

1.  Eqium - set a tight parametric filter (.06 octave) at each resonance to kind of let the air out of each one by about 3-6dB depending on what is needed.

2.  Soniformer2 - Set control points at each resonance (hehe I actually need 2 more control points to do this correctly !) dynamically 'limit' each resonance only after crossing a certain well placed threshold another 3-6 dB or whatever sounds transparent.

3.  GlissEQ - I might be able to get Gliss to dynamically control left over resonances if I can figure it out.

4.  Elephant - Pushing into this guy about 3dB sounds rather good I think !

The combination of controlling resonances using both EQ and dynamics is probably the best bet in this case.  This mix should have never left the mixing stage but what can I say, I think it's too extreme to balance just using an EQ.

ED: I forgot to mention I'll be watching the spectrums using either of CurveEQ, Ozone3 or Firium real time spectrum analzers.  I use Ozone3 because I can set a smooth 1 second decay along with peak decay and CurveEQ if I want to zoom in.

This topic was created before release of the latest product version, and it may contain details irrelevant to this version.  Replying is disabled for this topic.