Newcomer Help Required

There are many features of Har-Bal we still haven't discussed in this forum. Below we will start sharing a few items. Please feel free to add yours.
Post Reply
musicbox
Posts: 5
Joined: Tue May 03, 2005 4:51 pm
Location: Northampton England

Newcomer Help Required

Post by musicbox »

Hi
I purchased Har-Bal a couple of days ago and I am enjoying every minute of it so far! I have a problem though that I can not get to grips with. I have created a reference file for a track that I am working on. My track shows a couple of dips(Valleys or what have you) in the lower frequencies. I have pulled these up in line with the referance and everything looks and sounds good.
The problem is that when I record as (mytrack_eq) and then open up the new file in Har-Bal the valleys are back!! I have tried several times but I cannot save the file that I see.
Am I missing something?
Any pointers would be gratefully received.
By the way the product rocks! :?:
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Post by HarBal »

Have you toggled the EQ on? If the EQ button is up (off) and you write your file out the only change you'll have written is any gain change you make. For EQ changes to be written out you need to have the EQ button depressed when you do a record as operation.

If none of the above helps it would be helpful if you give us a step by step listing of what operations you are performing so that we can figure out where things are going wrong.

Regards,


Paavo.
musicbox
Posts: 5
Joined: Tue May 03, 2005 4:51 pm
Location: Northampton England

Post by musicbox »

Thanks for the reply. I did have the EQ button in (on) when I recorded the file. This is what I did:
Opened 'my track' using File, Open.
Opened a reference track using File, Reference, Open.
Pushed EQ button in.
Confirmed that EQ in was checked in drop down menu.
Pushed up the valleys on green trace until they near enough matched the reference.
Chose 'Record as'.
Accepted file name and pushed Save in the window showing the appropriate folder.
The box appeared showing the record operation and scrolled to 100%.
I then:
Closed the main window.
When asked if I wanted to save the filter chose' yes'.
Accepted the name of the file and clicked save.
Then:
I opened the new file (mytrack_eq) and the valleys were back exactly as before I edited them.
I then repeated the whole process after a reboot but I still can not save these changes on this file.
I have not had a problem with other tracks that I have tried. Could the file be corrupted?
As I said previously the track sounds fine ( no glitches ) so there is no sonic evidence of corruption.
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Post by HarBal »

One question:

When you re-recorded the track and opened it up a second time did it re-analyse the track? It may be that the date stamp on the my_track_eq.anl file is newer than the file itself (if you've been playing around with the clock recently), in which case it might not be re-analysing the file when it should. My suggestion is to delete the my_track_eq.anl file and re-open my_track_eq.wav. If that doesn't fix it then I'm at a loss to explain what is happening.

Regards,


Paavo.
musicbox
Posts: 5
Joined: Tue May 03, 2005 4:51 pm
Location: Northampton England

Post by musicbox »

Hi Paavo

That seems a plausable theory. I am not in the studio today but I will check it out when I get back later and let you know the results. I think that the analysis scroll bar appeared but I can not be sure. I did notice, though that there were a number of 'mytrack'.anl files with various _eq status so maybe it is opening a previous version. I will remove everything, i.e. the mytrack file, the reference file and the mytrack -eq files from the directory and from Sound Forge and Wavelab and import it from scratch. I will post my findings later.
P.S. I am very grateful for your help.
musicbox
Posts: 5
Joined: Tue May 03, 2005 4:51 pm
Location: Northampton England

Post by musicbox »

Hello Paavo

I hope that I am not being a pain but I guess that it is important for you guys to be aware of any quirks. O.K. I have tried everything that I can think of. I removed all traces of 'mytack' from my system and then relaoded it, opened it in Har-Bal, opened the reference file, adjusted the curve and recorded as 'mytrack_eq. No luck I am afraid! All changes in the higher frequencies were saved as per the adustments that I had made but the two valeys in the low frequencies re-appear when I open the file.
I understand that you don't have an answer to this issue right now but perhaps if anything occurs to you in the future you would let me know. I hate loose ends! I am sure that you will agree that there must be a reason for the anomally. As I said before, I can only think that maybe the original file is corrupted in some way? I definately do not get problems like it with other tracks.
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Post by HarBal »

No, You're not being a pain, but perhaps if you'd mentioned your findings in a little more detail to start off with I could have given you the correct answer quicker.

From your latest comments it sounds as if HarBal is behaving as it should and you've simply come up against a filter resolution issue. Basically, the frequency selectivity of an FIR filter is limited by how long its impulse response is. Essentially, an FIR can't be expected to have a selectivity of better than 1/T Hz where T is the length of the impulse response in Hz. It's actually worse than that but this is a simple upper bound.

For HarBal and a sampling rate of 48kHz, this limit is 48kHz / 8192 = 6Hz approximately. That means if you try and make a narrow band cut with a bandwidth of around 6Hz or less then the filter realisation HarBal comes up with won't be able to do that. Mathematically, what actually happens is something aking to taking the true frequency response that you what and passing a sliding averaging window, whose width is 6Hz, over the frequency response curve. The net result is that for fine detail the frequency response get's smeared so you end up losing selectivity. This only shows up at the LF end because at the HF end the bandwidths are much greater than 6Hz for a resonable Q filter.

Note that if you are using a higher sampling rate the frequency selectivity issue becomes proportionately worse:

48Khz/8192 = 5.9Hz
96khz/8192 = 11.7Hz
192kHz/8192 = 23.4Hz

I don't know what sampling rate you are using but if it's 192kHz I can imagine you'd have dificulty trying to kill indivual peaks below 100Hz. With progressively higher sampling rates all you can do at the bottom end is bring the peaks down with a lower Q filter response. If you try a high Q one the cut will be lost is the frequency response smearing and it will be more like not applying a cut at all.

Maybe in a future release I should actually put a Q constraint on the parametric tool so that it can't realize a response that HarBal can't realise.

Hope this helps explain what you've seeing.

Paavo.
musicbox
Posts: 5
Joined: Tue May 03, 2005 4:51 pm
Location: Northampton England

Post by musicbox »

Hello Paavo

Thank you for the response. It does explain what is happening perfectly. Sorry that I missed some critical information early on but I have only had the program for a week. I guess that I was a little unlucky in encountering the problem on one of the very first files that I worked on. I am really enjoying using it and I thank you again for your help.
Post Reply