Here is the latest on Har-Bal 3 development. (from harbal forum)

Here is the latest on Har-Bal 3 development.  

Progress is steady but still quite a lot needs to be done though there is some good news as far as processing goes. I have to admit that I've been somewhat oblivious to the pace of change in the microprocessor world and in doing some processing speed tests, have found that the current generation of Intel processors are about as fast on floating point as fixed point. The fixed point implementation filtering code in Har-Bal was designed in the days of PII where that definitely wasn't the case so over that last few days I actually ditched all the fixed point DSP and updated it to 64 bit floating point DSP at not cost in performance (on my centrino duo development machine). In fact, performance was slightly better.  

The upshot of that change is that now DSP coding in Har-Bal is much simpler (fixed point implementations are complicated) and the signal path in Har-Bal is 64 bit floating point so added noise through processing is essentially non-existent other than the addition of the dither when rendering to a fixed point buffer. Another side effect is that there is no longer a filter boost limit in Har-Bal. You should theoretically be able to boost as much as you want to and have a realisable filter. With the fixed point implementation you were limited to around 24dB. 

I was thinking of not implementing the Har-Bal dynamics processing (compression) strategy yet out of implementation and performance concerns, though now I think I will at least have a stab at it as my concerns are probably unfounded.  

Other developments include the near completion of a Har-Bal player application (which will be freely available) that will allow someone to take a commercial track and a filter we've designed for that track to hear the effect that the next generation Har-Bal can have (hearing is believing). I guess it could become an important tool for people dealing with clients over the internet too as you'll be able to demonstrate your processing to the client without having to send them the entire processed file, just the filter.  

That's about it for now. Good news, good progress but still no release date I'm afraid. There's still a few bits in the back end that are unstable (specification wise) that need to be pinned down first. With those done I'll be able to be more clear about when a release might be possible though an alpha release is possible in a few months time. 

cheers, 


Paavo.