Please can you explain what the "8192-point" refers to.
In Waves' LinEQ (Broadband) there are 6 bands (control points). Does Harbal in effect have 8192?
Thanks,
ubuntu
What does 8192 point linear-phase filter mean?
What does 8192 point linear-phase filter mean?
It means that Har-Bal specifies and implements the filter as a finite length impulse response whose length is 8192 points. It is almost like having 4096 sample points (linearly spaced) in the frequency domain but not quite owing to the lobe spreading caused by the windowing function used.
At the lowest frequency (20Hz) this corresponds to around 1/3 octave selectivity (at 48kHz sampling) whereas at 1kHz it is around 1/100th octave selectivity. For the most part it means that there is more than enough resolution to cope with almost any EQ problem you can through at it.
Linear phase refers to the fact that the filter phase response has a constant slope with respect to frequency which in turn means a constant group delay over all frequencies. Some find the pre-ringing of linear phase filters objectionable so version 1.5 also supports minimum phase filter realisations.
Unfortunately the release of 1.5 is delayed owing to some ASIO teething problems. Hoepfully in the next coming weeks all will be resolve but I'm sure you all heard that before many a time (I'd rather release it late as a robust product rather than having to cope with excessive tech support issues).
Regards,
Paavo.
At the lowest frequency (20Hz) this corresponds to around 1/3 octave selectivity (at 48kHz sampling) whereas at 1kHz it is around 1/100th octave selectivity. For the most part it means that there is more than enough resolution to cope with almost any EQ problem you can through at it.
Linear phase refers to the fact that the filter phase response has a constant slope with respect to frequency which in turn means a constant group delay over all frequencies. Some find the pre-ringing of linear phase filters objectionable so version 1.5 also supports minimum phase filter realisations.
Unfortunately the release of 1.5 is delayed owing to some ASIO teething problems. Hoepfully in the next coming weeks all will be resolve but I'm sure you all heard that before many a time (I'd rather release it late as a robust product rather than having to cope with excessive tech support issues).
Regards,
Paavo.
-
ubuntu
Linear Phase v Minimum Phase
Paavo,
Many thanks for your detailed reply.
(1) Does this mean that at 20Hz Harbal's resolution is just as good as Freefilter's (it is a 1/3 octave linear phase filter) and at any frequency higher than 20Hz Harbal's resolution is signifinantly and increasingly better
than Freefilter's?
(2) You mentioned that:
"Some find the pre-ringing of linear phase filters objectionable"
Is pre-ringing something that one can hear? If so what should one be listening for? How does one decide whether to use "linear phase" as opposed to "minimum phase"?
Regards,
ubuntu
Many thanks for your detailed reply.
(1) Does this mean that at 20Hz Harbal's resolution is just as good as Freefilter's (it is a 1/3 octave linear phase filter) and at any frequency higher than 20Hz Harbal's resolution is signifinantly and increasingly better
than Freefilter's?
(2) You mentioned that:
"Some find the pre-ringing of linear phase filters objectionable"
Is pre-ringing something that one can hear? If so what should one be listening for? How does one decide whether to use "linear phase" as opposed to "minimum phase"?
Regards,
ubuntu
Linear Phase v Minimum Phase
Whether you will hear it or not will depend a lot on the nature of the EQ filter. If the filter isn't doing much high Q filtering chances are you won't be able to notice it. I'd argue that it is difficult to tell the difference in normal useage unless you have a controlled A/B testing arrangement. The difference in sound is quite subtle.
As to which you should be using you could argue that minimum phase is the best choice as it is the most natural sounding (linear phae systems don't generally occur in nature though minimum phase systems do). The default is minimum phase in version 1.5 and I generally leave it in that mode for the work I do with HarBal.
On resolution issues I'd imaging the point you make about HarBal compared with Freefilter is correct though I cannot conclusively say as I don't know how they have implemented their filter. As far as PC based EQ FIR filters go, there doesn't seem to be many about with 8192 points or better. I believe CurveEQ uses a 3072 point FIR.
I chose 8192 points on the basis of about as much resolution as you'd ever want for EQ of 48kHz sampled music. Higher resolutions would be nice for higher sampling rates though but that adds a level of complication that I wasn't prepared to deal with now. Maybe sometime in the future.
Regards,
Paavo.
As to which you should be using you could argue that minimum phase is the best choice as it is the most natural sounding (linear phae systems don't generally occur in nature though minimum phase systems do). The default is minimum phase in version 1.5 and I generally leave it in that mode for the work I do with HarBal.
On resolution issues I'd imaging the point you make about HarBal compared with Freefilter is correct though I cannot conclusively say as I don't know how they have implemented their filter. As far as PC based EQ FIR filters go, there doesn't seem to be many about with 8192 points or better. I believe CurveEQ uses a 3072 point FIR.
I chose 8192 points on the basis of about as much resolution as you'd ever want for EQ of 48kHz sampled music. Higher resolutions would be nice for higher sampling rates though but that adds a level of complication that I wasn't prepared to deal with now. Maybe sometime in the future.
Regards,
Paavo.
-
ubuntu
8192 v over 8000 bands in Ozone
Paavo,
Thanks you very much for the excellent clarification.
I also use Ozone. It has a "matching eq" funtion and according to the Ozone manual:
"The matching EQ (in Ozone 3) is a digital linear phase EQ, with the ability to use over 8,000 bands of frequencies for very precise matching. "
Are these 8000 "bands" similar to what HarBal calls 8192 "points" ?
Thanks,
ubuntu
Thanks you very much for the excellent clarification.
I also use Ozone. It has a "matching eq" funtion and according to the Ozone manual:
"The matching EQ (in Ozone 3) is a digital linear phase EQ, with the ability to use over 8,000 bands of frequencies for very precise matching. "
Are these 8000 "bands" similar to what HarBal calls 8192 "points" ?
Thanks,
ubuntu