Any news about next version

This area of the Forum is where you can post your suggestions and ideas for future update releases of Har-Bal, as well as any gripes you may have on the software, its useability and its interface.
DuncJF
Posts: 31
Joined: Mon Oct 16, 2006 4:26 am

Any news about next version

Post by DuncJF »

Hi,

Just wondering if you are looking to get a Beta version of the re-write out this year, or, as I suspect with doing such a drastic code re-write, is it looking to be next year before something is available ?

No rush, rather have something good than quick... :D

I guess I am just curious as to how you are progressing with the new analysis engine etc


Dunc 8)
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

Hi Dunc,

I'd say more the later than the former (next year that is). I've been thinking about talking about my progress to date but didn't get around to it. You've given me the opportunity to now.

At present, the analysis part is essentially complete but I have a major bit of design work figuring out the details of how the filtering design will work. I can tell you that the analysis engine does time based analysis on a nominally 1/20th of a second basis and also provides for calculating averages as per current Har-Bal (it does time based analysis first and can then calculate averages from that, typically less than a second to do the average of a typical song).

By doing time based analysis it allows for the ability to arbitrarily partition a track into segments and have average analyses for each plus individual filters for each segment. Not only does the analysis engine do time based analysis but it also does analysis for mid,side,left and right channels. The engine is written in a general sense so it can (theoretically) also handle 5.1 and 7.1 setups. It is the time based filtering aspects of the filter design process that still needs sorting out as far as the back end design is concerned.

Time based analysis also offers other possibilities, possibly most notably is an integrated compressor. The time based analysis gives you access to a smooth volume envelope that can then be used to drive compression. That envelope will be altered through filtering but there is sufficient information to predict the change in envelope from the filter characteristic and the spectrum by time. All in all, it should make for a clean sounding compression.

Another possibility that I've catered for in the design proposal is controlling errant peaks in a track. The spectrum analysis has information in it that will allow you to quickly locate the time segments that are responsible for particular peaks in the peak spectrum trace. With that time selection centred on a particular errant peak you will be able to build a filter to just control that, yielding a process which you might refer to as my take on multi-band compression.

One thing I'm also working on doing is separating the filter design process from the filtering process. This will allow for a separate standalone and free Har-Bal player, allowing people with Har-bal licenses to design filters and make them available for people to hear without the need for them to purchase the software.

Finally, the way I do the filtering will be generalised into a cross coupled filter block which allows support for independent filtering of mid, side, left and right channels as well as Haas zone ambiance processing and integrated room EQ (compensating for room acoustics). That is all achievable with a filter structure that has only twice the computational complexity of what Har-Bal has at present.

Har-Bal Air as it is now, will be replaced with a more powerful and natural sounding Haas zone ambiance process. That way you use it will be the same but it will sound better and have a greater range of possibilities, one of which is stereo'ising mono sources.

On the GUI side of things the geometric mean trace will cease to be and is being replaced by a realtime trace which displays the spectrum at a particular 1/20th second time slice. I've actually prototyped this behaviour and it works quite well. It needs some optimisation though.

That's a fair summary of where I'm at an where I'd like to go. Hopefully it should give you some idea as to why I'm taking so long doing this re-write.

Cheers,


Paavo.
DuncJF
Posts: 31
Joined: Mon Oct 16, 2006 4:26 am

Re: Any news about next version

Post by DuncJF »

Hi Paavo,

After reading the above (several times, because on first reading, most of it went over my head higher than a Boeing 747 :lol: ) I can understand why it is taking a while. I have a few comments and questions to follow.
By doing time based analysis it allows for the ability to arbitrarily partition a track into segments and have average analyses for each plus individual filters for each segment. Not only does the analysis engine do time based analysis but it also does analysis for mid,side,left and right channels.
This sounds like exactly the feature a few of us have eagerly been awaiting. Is the change between segments going to be instant or can it be "morphed" over a selectable time frame ?

Another possibility that I've catered for in the design proposal is controlling errant peaks in a track. The spectrum analysis has information in it that will allow you to quickly locate the time segments that are responsible for particular peaks in the peak spectrum trace. With that time selection centred on a particular errant peak you will be able to build a filter to just control that, yielding a process which you might refer to as my take on multi-band compression.
Never thought of this as a feature requirement but, now that you mention it...YES, bravo :D
One thing I'm also working on doing is separating the filter design process from the filtering process. This will allow for a separate standalone and free Har-Bal player, allowing people with Har-bal licenses to design filters and make them available for people to hear without the need for them to purchase the software.
Another good idea that will help to keep the nay sayers at bay when they hear what Harbal can do.
Har-Bal Air as it is now, will be replaced with a more powerful and natural sounding Haas zone ambiance process. That way you use it will be the same but it will sound better and have a greater range of possibilities, one of which is stereo'ising mono sources.
I'm sure this would also be very useful to a lot of us and I am chomping at the bit to hear what this can do.
On the GUI side of things the geometric mean trace will cease to be and is being replaced by a realtime trace which displays the spectrum at a particular 1/20th second time slice. I've actually prototyped this behaviour and it works quite well. It needs some optimisation though.
Would there be a way to Freeze the trace ( I just have a vision in my head that the trace would be jumping all over the place and be slightly distracting and difficult to interpret) or are you saying that you will be able to select a particular time slice and see the trace for just that slice ?

Would you be able to implement a VST Effect slot (as a last process in the chain) into the new Harbal or would that still be something that is not an option ?

It is not particularly important and probably difficult to tell at this stage, but what kind of increase in computing cycles is it looking like taking ?


This all sounds brilliant to me. If you need somebody for Beta Testing (especially for longer files of 20 minutes and more with numerous segments) sign me up.


Keep up the good work.

Dunc 8)
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

Filter changes will be (morphed) over a segment. The segment resolution is nominally 1/20th of a second, which should be fine enough for what I'm trying to do. Selecting time segments for filtering will be achievable explicitely, through selecting areas on a time line, or indirectly through the peak taming mechanism I referred to.

On the real time trace, when not playing the trace is frozen to the time segment that the track time is cued to. Change the cue position and the time trace will change. In that sense you can easily get a sense of where something happens without loosing it when you blink!

Putting in a VST slot at the end of chain should be perfectly possible though I probably won't be implementing it first up as this new feature set has put enough on my plate as it is. That sounds like a good feature update after the first release though.

On the CPU usage front, with real time display off the increased CPU usage should be about 2 to 2 1/2 times what it is now. The filtering is strictly 2 times the CPU but how much extra CPU compression might take I'm not sure though I don't anticipate it taking too much. With real time display it is significantly more on account of the screen refreshing going on, however, it shouldn't impact on the playback quality much if at all because of the way I'm implementing it. The GUI will just fall behind in updating if CPU gets used up.

Cheers,


Paavo.
analogic
Posts: 19
Joined: Thu Apr 27, 2006 1:08 am

Re: Any news about next version

Post by analogic »

HarBal wrote:Another possibility that I've catered for in the design proposal is controlling errant peaks in a track. The spectrum analysis has information in it that will allow you to quickly locate the time segments that are responsible for particular peaks in the peak spectrum trace. With that time selection centred on a particular errant peak you will be able to build a filter to just control that, yielding a process which you might refer to as my take on multi-band compression.
That sounds really fantastic. 8) With the new approach I guess it will be possible to use HarBal as a more dynamic EQ correction tool instead of the static filtering way. I think applying the necessary filtering to the particular time segments only, without affecting the rest of the track, would possibly result in more natural sonic results.

Hope to see the new version soon. 8)

Thanks Paavo, and keep up your great work!

Kursat
djaq
Posts: 2
Joined: Fri Oct 24, 2008 5:49 am

Re: Any news about next version

Post by djaq »

Oh, I feel a bit dumb for my last post. I had not read this post, but my offer of helping still stands.

-djaq
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

To follow up on things a bit, the re-write is still progressing at a steady rate but as mentioned in the above thread, there is still much to do. Maybe soon I'll set up a page on the website showing the progress to date but I'm reticent to at the moment for fear of being inundated with questions about the next version that may drag me away from what I should be doing.

regards,


Paavo.
janpetter
Posts: 45
Joined: Sat Feb 05, 2005 7:00 pm

Re: Any news about next version

Post by janpetter »

Wow.
It seems to be a major upgrade with new features.
Can't wait to get my hands on this one :)
All the best
Jan
DuncJF
Posts: 31
Joined: Mon Oct 16, 2006 4:26 am

Re: Any news about next version

Post by DuncJF »

Hi Paavo,

Just a curious user wanting to know how the new Analysis Engine and Gui is shaping up.

Hopefully, you are now on the top side of the re-write and have the light at the end of the tunnel showing you the way to the end (of the Beta version). :lol:


All the best...

Dunc
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

Yes, I can see the light at the end of the tunnel and its getting closer but probably not as close as you'd want.

The Analysis engine is still evolving, not majorly, but it has needed some restructuring to cope with complex time selections. The filter model I had before couldn't calculate the updated spectrum predictions (based on filter changes) correctly when including complex time selections (that is ones are non-continuous and select only a subset of the track based on some selection criterion). I had to re-think the design somewhat but have come up with a far better approach that I would have needed in any case, to translate the filter plan into a realisation for the player and renderer.

I'm at the tail end of the restructure and should have that fully resolved in the next few weeks. At that point I'll be in a position to work on getting the player and renderer to translate that filter design into a real filter and apply it to a track. That is the moment I'm waiting for! To be able to hear the fruits of my labour.

Whatever the case, I'm confident of having a release before the year is out and I'm feeling confident that it will be a great step forward from what Har-Bal is now. Just being able to see how the spectrums of problem tracks stack up when split in appropriate places gives me confidence in the approach. I'm sure the end result will be a revelation, albeit a late one.

Cheers,


Paavo.
SynthRiot
Posts: 10
Joined: Fri Jul 29, 2005 3:11 am
Location: Germany - Bavaria - Munich

Re: Any news about next version

Post by SynthRiot »

I'm curious about the new version!

Good luck!

Nitro from Germany...
It's the question that drives us ...
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

The state of play is that the analysis and filter design code is complete, and as far as I can currently tell, functioning as it should.

I am half way though writing the filtering code to support the filter design. When that's complete, I'll spend some time verifying the correct behaviour for the filter design and rendering processes. No doubt, I may spend some time fooling around with it when I hear it for the first time.

After that, most of the core back end functionality will be in place and I can concentrate on finishing off the GUI front end, which itself, will take some time as there is much left to implement and refine.

That's pretty much the state of play. When I have the filtering engine up and running I'll probably post an update on how fruitful it has been.

regards,


Paavo.
DuncJF
Posts: 31
Joined: Mon Oct 16, 2006 4:26 am

Re: Any news about next version

Post by DuncJF »

Hi Paavo,

Just a short call to see how the rewrite is going ?

I presume by now you are tackling the GUI and refining the filter engine. I'm hoping to be able to use some kind of Beta by year end, but if it's not to be, that's fine. Better to be well written/tested and take a while to get out.

if you get the chance, an update on how it's going would be nice to see.


Thanks :)
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

H'mmm, well, this new version has some mind bogglingly complex detail to it.

I'm on the verge of having my filtering code up and running though I keep on coming across subtle details that I missed in my initial implementation that I have to go back and fix up. It's been a real mental challenge but I'm almost there with the playback side of things.

As I mentioned earlier, I did intend having the back end side of things finished at this stage, but due to the unforeseen complexities involved, I've slipped a bit. I need to re-jig some file formats into a more optimal form before I get the playback side of things going. I guess when that is done and all is fired up, I'll find out what else I've overlooked.

Apart from the delays, so far so good. My apology to those of you waiting patiently, though I really, really, really want to get this right as I can see a mountain of future headaches if I don't. So far so good though.

cheers,


Paavo.
HarBal
Site Admin
Posts: 761
Joined: Mon Apr 19, 2004 8:18 pm
Contact:

Re: Any news about next version

Post by HarBal »

Thought I post a follow up to bring you up to date on progress.

As alluded to, the processes required to implement this new technology is quite complex and that has slowed me up a bit, but there is good news. I have managed to implement playback with new complex filter realisations and have demo'ed it on a number of difficult tracks to check the validity of the concept.

As far as I can ascertain, the idea is spot on. Track partitioning and independent partition filters makes for transparent sounding processing easily superior to current Har-Bal and I'd argue, superior to multiband compression. There is very little artifact resulting from this type of processing, In fact, for the examples I tried I could not hear any filter transitioning at all.

On the down side, the complexity of the filter design process makes for a big computational workload to calculate the filter realisations. As such, I've got some serious optimisation issues to work through. I need some time to think those issues through so at this moment I've switched to the more "brain dead" GUI coding whilst I get my head around the optimisation issues.

That's it for now.

Cheers,


Paavo.
Post Reply