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.

Aleksey - I have HarBal, and, frankly, I think it is very overrated.  It is certainly not a "considerable improvement" over CurveEQ, especially if you compare the feature sets of both.

To me, using pink noise spectrum matching with CurveEQ produces results that are just as or more useful for radical frequency balancing as what can be done with HarBal.  It may be a few extra steps, but, if one is essentially trying to remix in the mastering, you've already got a rough row to hoe anyway, so who cares.

That being said, for an amateur mastering engineer like myself, compensating curves are helpful, and I would think an SPL curve would be very useful.

John

PS - Hey Kyle, did you get my email?


John, thank you for your support.  I'm currently seeking for some useful weighting curves.  I have found that C-weighted curve mixed with A-weighted curve yields useful results.  By the way, the 'Parabolic' curve I've implemeted in Soniformer 2 is very close to the generally accepted C-weighting curve.  Possibly, adding a Fletcher-Munson compensation can give some more precision.  Anybody knows where I can find the numeric chart for the Fletcher-Munson equal loudness curve?  I can only find graphs on the Internet which are not very useful.

Hey John - I didn't get your email yet - I'll check at home later sometimes the web mail goofs up !

Aleksey - all these talks of curves are reminding me of grey noise (inverted 'A' compensated applied to white noise).  I always thought the 'B' compensation would sound good.

I think I tried compiling a fletcher munson eq a while back but since it's kinda the way the inner ear reacts to sound pressure of varying frequencies it sounds very bad to apply that as a compensation to anything - as does applying the full strength of the 'C' weighted curve.  Am I wrong there ?  I'll have to see if I can find my notes (mostly packed up since I'm moving).

I like the parabolic so far in Soniformer2 - it does seem to smooth out the edges and it worked well last night.  I had Har-Bal in the equation though, hehe and I really clobbered a couple of peaks I shouldn't have...


I think it's found in ISO226, but you'll have to buy it from ISO website.  Here are a couple of approximations that might be useful:

http://mambo.ucsc.edu/psl/audfaq.txt

http://groups.google.com/groups?selm=B9C08AD3.7CB8%25robert%40wavemechanics.com


Some thoughts...

First of all, ISO 226 weighting is WRONG to use with audio recordings (and I've heard it is mostly used for noise level estimation).  I've tried it many times with audio and never succeeded.  Averaged C- and B-weighting curves can be used to help in bass and high freq regions, only.  Now, I don't really understand how equal loudness controur can help at all, especially after understanding such things like masking psychoacoustic effects.  ELC simply won't help here.

Secondly, on the HarBal topic.  The only difference of HarBal to any other equalizer in existance is that it is an absolute EQ curve equalizer and thus it requires one to analyze the whole track.  Other equalizers are simply relative.  I have to admit, HarBal's hype have simply hidden this simple fact.


Aleksey Vaneev wrote:

Some thoughts...
First of all, ISO 226 weighting is WRONG to use with audio
recordings (and I've heard it is mostly used for noise level
estimation).
I had thought that ISO 226 was the Fletcher-Munson data, not the A-weighting curve.

I've tried it many times with audio and never

succeeded.  Averaged C- and B-weighting curves can be used to
help in bass and high freq regions, only.  Now, I don't really
understand how equal loudness controur can help at all,
especially after understanding such things like masking
psychoacoustic effects.  ELC simply won't help here.
That's pretty much what I was trying to say 9 posts back.  All of these weighting systems (A, B, C) or even a more accurate representation of the inverse of the Fletcher-Munson curve are not terribly accurate representations of our (human) perception of volume (and the measurements have a wide statistical margin of error -- i.e. your personal ELT probably won't be the same as the Fletcher-Munson curve).  I have read that the Zwicker method is the most accurate (because it takes into account frequency and temporal masking), but the "loudness" given by Zwicker (or obviously any accurate loudness estimate) will be contingent on the frequency content of the material on a moment by moment basis, so it cannot give a single volume level that will match pre and post eq at all times.  In other words there is no way of matching pre and post-eq levels over time with a single volume change -- the best we can do is approximate (and a B or C weighted analysis will give a better approximation than nothing), BUT honestly how hard is it to use the volume control to match the levcels yourself for your own ears?


ISO 226 is pretty close to the A-weighting curve.  I had an idea that the A-weighting curve is the simplified Fletcher-Munson curve of the threshold of hearing.  So, pretty the same 'unusefulness' problem applies to the A-weighting.

On your question about the volume control, I'm pretty sure I don't have such problem, personally.  Automatic volume matching will make things even harder to me I suppose.

After getting to these conclusions I don't understand HarBal's positioning.  They could simply state that their approach can be described as an absolute equalizing which, when coupled with the full-length track analysis, allows one to see the resonances (peaks and notches) and to correct them 'in-place'.  This can be convenient at times.

You know, I have to understand the real flaws of the competing products, so please forgive me if I'm sounding too repetitive. :)


Aleksey Vaneev wrote:

On your question about the volume control, I'm pretty sure I
don't have such problem, personally.  Automatic volume matching
will make things even harder to me I suppose.
I agree.  By the way that question wasn't directed at you, more at the heavens :-)
After getting to these conclusions I don't understand HarBal's
positioning.  They could simply state that their approach can be
described as an absolute equalizing which, when coupled with
the full-length track analysis, allows one to see the
resonances (peaks and notches) and to correct them 'in-place'.
This can be convenient at times.
Doesn't sound so exciting though, does it ;-)
You know, I have to understand the real flaws of the competing
products, so please forgive me if I'm sounding too repetitive.
:)
This weekend I have been correcting the spectral imbalances of a number of real-space impulse responses I collected last week.  I did this by convolving white noise with them (individually) and then matching the convolved audio's spectrum to the white noise spectrum using Curve EQ.  I assume that I could have simply tried to match the spectrum of the impulse response themselves to white noise, but I wasn't sure how long CEQ needs the audio to be to do a proper match.  Anyway it worked very nicely but was a very long-winded process.

Anyway I thought that this might be one advantage that a stand-alone EQ could have, and that is if it could batch-process a whole bunch of audio files so that they are matched to a predefined spectrum.  AFAIK HarBal doesn't allow for batch processing, depite being stand-alone.


Your approach of convolving the white noise with the IR and the matching it to the plain white noise is OK and of course will give better results than the direct matching of the IR to white noise spectrum.

OK, thanks for confirming that.
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.