Emax

Julian,

You are right in your description about how compression works in the Emax.
And EMXP emulates this compression/decompression with a similar (but not identical) algorithm as the one used in the Emax.
So EMXP doesn't write PCM to the disks, it writes the actual compressed 8 bit sound data.
This algorithm is sometimes called delta sigma but to me these two words sound more like a student society :-)

Anyway, it's a proprietary algorithm which took me quite some time to discover. And I think it's even slightly different from the one used in the Emulator II (but the principle is certainly the same). Well at least I had to change it a little bit when adding Emulator II support to EMXP.
The fact that there were recently some discussions in this group about distortion in a bass drum WAV which was converted by EMXP to the compressed emax format shows that it's not the most straight forward algorithm... (I will finetune the code to make it even more ressemble the orginal Emax compression logic) 

///E-Synthesist 

--- In emax@yahoogroups.com, "Julian" <jujulilianan@...> wrote:
>
> hmmm.... but then again.. if the compressed format is NOT PCM.. then how
> does EMXP work with writing wave files to disk? 
> 
> unless data format on disk is different to data format in RAM......
> 
> but then a disk only uses 512k of space, too.
> :-)
> 
> 
> interesting can of worms we've opened here.
> 
> 
> On Thu, 23 Jul 2009 13:36 +1000, "Julian" <jujulilianan@...>
> wrote:
> > actually - we're all wrong, according to my emax service manual...
> > :-)
> > 
> > it talks about the compression algorithm, and this algorithm DEFINITELY
> > is not just companding (apparently that was E1?) I always thought it was
> > digital companding in the emax.... would love to find where I read that
> > now!
> > 
> > it is also definitely not packing 12 bit sample words that come from the
> > A/D and to the D/A into bytes or 16 bit words. apparently the e-chip
> > actually compresses and decompresses PCM data into something else. from
> > the wording of the manual, it sounds like some kind of sample-sample
> > difference function.
> > 
> > they claim that that with this compression, they can represent 16 bits
> > of data in approximately 8 bits of memory. And that although it's not
> > 100% constant for every given sample, they claim they get a compression
> > ratio with the e-chip of around 2:1.
> > 
> > 
> -- 
> http://bleepin.com
> 
> -- 
> http://www.fastmail.fm - A fast, anti-spam email service.
>