Re: a little SP0250 data

From: Jess Askey <jess_at_magenta.com>
Date: Wed Jun 18 1997 - 18:06:29 EDT

Hi all!
  Well I knew that my dump of the Three Stooges speech data probably
wouldn't work
in a TMS5220 (and it didn't) but what the hell, I tried anyway.
  As I was doing it I realized that even the TMS5220 doesn't use
straight LPC-10. TI has
some special bits (repeat, unvoiced) in their coding.
  Now, since Al found this small amount of data about the SPO250 I will
dig out my 5220 stuff
again and do some comparing. They seem really similar. The SPO250 seems
to be using actual
floating point values x 32,??? for each reflection coefficient where the
5220 uses a ROM lookup table
to narrow each parameter down to the "most useful" coefficient values.
So maybe the SPO250 is not
to far from true LPC10 except for the coding of the repeat and
voiced/unvoiced flags??? Also, since the repeat
flag is mostly used to keep the data stream as small as possible, it is
not necessarily manditory to use it
in testing for compatibility.
  Clay... You know the 5220 from talking to Larry, any thoughts?? I will
write Michel to see what he thinks it all.
  later!
    jess

Al Kossow wrote:
>
> here is something that I just got from a request on comp.speech about
> the part. someone had a 4 page preliminary data sheet on it.
>
> There is just one page. The rest is electrical stuff.
>
> There are 15 parameters each of 1 byte, which are input to pins 3 2 28
> 27 26 25 24 23 (3 is Most Significant Bit). The bytes are labelled (in
> order) C21, C11, Amplitude, C22, C12, Pitch, C23, C13, Repeat, C24, C14,
> C25, C15, C26, C16.
>
> The C's are the twelve (reflection?) coefficients, where the MSB is the
> sign, 1 = pos.
> The amplitude in direct data mode has the 3 MSBs as exponent.
> The pitch does not have a sign bit.
> Repeat has MSB set to 0, MSB-1 set to V(u), remaining bits the repeat
> value. It does not say what V(u) is (there is a bar over the u). It may
> be the voiced/unvoiced flag.
>
> Finally there is a description of a byte which may be an alternative to
> the amplitude. The MSB is sign 1 = neg, the rest is amplitude. A note
> says "exponent from normal mode remains until changed".

-- 
Unofficial Atari Game Page:http://magenta.com/havoc
Lots of other General Arcade Game info at
      http://arcadegames.miningco.com
Received on Wed Jun 18 15:09:09 1997

This archive was generated by hypermail 2.1.8 : Fri Aug 01 2003 - 00:31:37 EDT