I have spent the last month enjoying working with CSound5 (Windows/varga).  It has been a real pleasure to see everything work so well, and I had spent a lot of time struggling with various aspects of CSound 4.  

However I have a problem with lpc.  I started out trying to duplicate the cross synthesis orchestras found in the CSound book , CH 27, but I couldn’t get past the lpc part of it.  So I tried a simple LPC orchestra that works for me in CSound 4.20.  That sco/orc are part of the .doc file attached to this report.  The file seems to be ok on analysis.  The four parameters, krmsr, krmso, kerr, kcps are all in the right place when I run the lpanal utility in csound5  with –v2.  However on resynthesis, the file seems to blow up, as you can see below.  I note the WARNING  lpfile srate !=orch sr.  That is not true.  I note that the bin directory of csound5 includes an lpanal32.exe file but I can’t get the program to find that file when I include it in my command line (see the .doc attachment).  The file blows up  whether I use csound or csound 32 as my .exe….  
I note that the documentation indicates that the –a option is the only way in which lpc files are now formatted, but including or not including the –a option does change the program information regarding what type of file storage is used.  However when the lporc/sco perf is run the lpread/reson doesn’t seem to notice that.

Any suggestions?
Thanks,
Bruce McKinney

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
>csound32   c:\csound5\orcsco6\pvocs\blp04a.orc  c:\csound5\orcsco6\pvocs\blp04a.sco


0dBFS level = 32768.0
PortMIDI: selected input device 0: 'In-A USB MidiSport 4x4' (MMSystem)
orch now loaded
audio buffered in 128 sample-frame blocks
winmm: opening output device 0 (DIGI96/8 PAD Out (1+2)(1))
writing 512-byte blks of shorts to dac
SECTION 1:
ftable 1:
B  0.000 ..  1.000 T  1.000 TT  1.000 M:      0.0      0.0
new alloc for instr 1:
file lp.124 (63400 bytes) loaded into memory
Using filter coefficient type of file.
WARNING: lpfile srate != orch sr
B  1.000 ..  3.000 T  3.000 TT  3.000 M:      1.$      1.$
         number of samples out of range:    88198    88198
inactive allocs returned to freespace
end of score.              overall amps:      0.0      0.0
           overall samples out of range:    88198    88198
0 errors in performance
Elapsed time at end of performance: real: 3.195s, CPU: 3.188s
1034 512-byte soundblks of shorts written to dac

Bruce McKinney wrote:

 > So I tried a simple LPC orchestra that works for me in
 > CSound 4.20. That sco/orc are part of the .doc file attached
 > to this report.  The file seems to be ok on analysis.
 > The four parameters, krmsr, krmso, kerr, kcps are all in the
 > right place when I run the lpanal utility in csound5 with –v2.
 > However on resynthesis, the file seems to blow up, as you can
 > see below. I note the WARNING lpfile srate !=orch sr.
 > That is not true. I note that the bin directory of csound5
 > includes an lpanal32.exe file but I can’t get the program to

Did you try doing the analysis with csound32.exe -U lpanal, and
then reading the .lpc file with csound32.exe ? Reading a 32 bit
.lpc file with a 64 bit build of Csound (or vice versa) will
surely not work, and it may be possible that LPC does not work
with doubles at all. Of course, the OPCODEDIR and OPCODEDIR64
environment variables should be set correctly as well.

 > find that file when I include it in my command line (see the
 > .doc attachment).

Would it be possible to send reports as plain text ?
Not everyone can read .doc files, and they are not really
a preferred format on the Csound lists.

I have found that LPC seems to be problematic and I was planning to look
at it. It used to work for me in the past, but at the moment, both csound 4.23
and 5 seem to be problematic. The analysis seem to work only with filter
coefficient analysis, the pole method seems to be broken.  In resynthesis,
lpreson seems to work OK, but the pitch and rms estimation of the excitation
don't seem to work, producing a very messy output.

Victor


At 16:04 08/11/2005, you wrote:
>I have spent the last month enjoying working with CSound5 
>(Windows/varga).  It has been a real pleasure to see everything work so 
>well, and I had spent a lot of time struggling with various aspects of 
>CSound 4.
>
>However I have a problem with lpc.  I started out trying to duplicate the 
>cross synthesis orchestras found in the CSound book , CH 27, but I 
>couldn’t get past the lpc part of it.  So I tried a simple LPC orchestra 
>that works for me in CSound 4.20.  That sco/orc are part of the .doc file 
>attached to this report.  The file seems to be ok on analysis.  The four 
>parameters, krmsr, krmso, kerr, kcps are all in the right place when I run 
>the lpanal utility in csound5  with ­v2.  However on resynthesis, the file 
>seems to blow up, as you can see below.  I note the WARNING  lpfile srate 
>!=orch sr.  That is not true.  I note that the bin directory of csound5 
>includes an lpanal32.exe file but I can’t get the program to find that 
>file when I include it in my command line (see the .doc attachment).  The 
>file blows up  whether I use csound or csound 32 as my .exe….
>I note that the documentation indicates that the ­a option is the only way 
>in which lpc files are now formatted, but including or not including the 
>­a option does change the program information regarding what type of file 
>storage is used.  However when the lporc/sco perf is run the lpread/reson 
>doesn’t seem to notice that.
>
>Any suggestions?
>Thanks,
>Bruce McKinney
>
>;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
> >csound32   c:\csound5\orcsco6\pvocs\blp04a.orc 
> c:\csound5\orcsco6\pvocs\blp04a.sco
>
>
>0dBFS level = 32768.0
>PortMIDI: selected input device 0: 'In-A USB MidiSport 4x4' (MMSystem)
>orch now loaded
>audio buffered in 128 sample-frame blocks
>winmm: opening output device 0 (DIGI96/8 PAD Out (1+2)(1))
>writing 512-byte blks of shorts to dac
>SECTION 1:
>ftable 1:
>B  0.000 ..  1.000 T  1.000 TT  1.000 M:      0.0      0.0
>new alloc for instr 1:
>file lp.124 (63400 bytes) loaded into memory
>Using filter coefficient type of file.
>WARNING: lpfile srate != orch sr
>B  1.000 ..  3.000 T  3.000 TT  3.000 M:      1.$      1.$
>          number of samples out of range:    88198    88198
>inactive allocs returned to freespace
>end of score.              overall amps:      0.0      0.0
>            overall samples out of range:    88198    88198
>0 errors in performance
>Elapsed time at end of performance: real: 3.195s, CPU: 3.188s
>1034 512-byte soundblks of shorts written to dac

Victor Lazzarini
Music Technology Laboratory
Music Department
National University of Ireland, Maynooth 

I had a similar problem with LPC the one time that I tried to use it with
4.23.  IIRC, my analysis file ended up with one infinite value in it.  I
converted it to text with lpcexport, fixed the errant value, and
lpcimport-ed it, then everything worked fine.

I did not investigate why the infinite value was occurring.

Anthony Kozar
anthonykozar AT sbcglobal DOT net
http://akozar.spymac.com/


Victor Lazzarini wrote on 11/8/05 1:03 PM:

> I have found that LPC seems to be problematic and I was planning to look
> at it. It used to work for me in the past, but at the moment, both csound 4.23
> and 5 seem to be problematic.

> At 16:04 08/11/2005, you wrote:

>> However on resynthesis, the file
>> seems to blow up, as you can see below.