Hi,

I wanted to use pvsceps to try some pattern recognition on voices. Including pvsceps in my code produces segmentation faults both when running in CsoundQT 0.9.2.1 or when running  Csound 6.07 on the terminal.

I verified that the same problem exists when using the pvsceps.csd example (example 706) from the manual.

what kind of values do I expect for keps[] from the line

keps[] pvsceps fsig ?

when trying to print them with printk2 I only get NANs.

How are the coefficients in keps[] related to 1/f of the frequencies you can retrieve via 

kflag	pvsftw fsig,ifftAmps,ifftFreq?

cheers,

Karin

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Cannot reproduce from pvsceps.csd on Linux


On Mon, 25 Jul 2016, Karin Daum wrote:

> Hi,
>
> I wanted to use pvsceps to try some pattern recognition on voices. Including pvsceps in my code produces segmentation faults both when running in CsoundQT 0.9.2.1 or when running  Csound 6.07 on the terminal.
>
> I verified that the same problem exists when using the pvsceps.csd example (example 706) from the manual.
>
> what kind of values do I expect for keps[] from the line
>
> keps[] pvsceps fsig ?
>
> when trying to print them with printk2 I only get NANs.
>
> How are the coefficients in keps[] related to 1/f of the frequencies you can retrieve via
>
> kflag	pvsftw fsig,ifftAmps,ifftFreq?
>
> cheers,
>
> Karin
>
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
>

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

on the oter hand using valgrind

==10758== Invalid read of size 8
==10758==    at 0x4F34C1A: perf_iceps (arrays.c:1864)
==10758==    by 0x4FA090D: kperf_nodebug (csound.c:1558)
==10758==    by 0x4FA1EFD: csoundPerform (csound.c:2051)
==10758==    by 0x40158B: main (csound_main.c:318)
==10758==  Address 0x19152b40 is 0 bytes after a block of size 4,112 
alloc'd
==10758==    at 0x4C2B200: calloc (in 
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==10758==    by 0x4E8DAD7: mcalloc (memalloc.c:113)
==10758==    by 0x4E75D32: csoundAuxAlloc (auxfd.c:53)
==10758==    by 0x4F38579: init_iceps (arrays.c:1851)
==10758==    by 0x4E8AF28: insert (insert.c:300)
==10758==    by 0x4E8F815: process_score_event (musmon.c:768)
==10758==    by 0x4E91E3C: sensevents (musmon.c:962)
==10758==    by 0x4FA1F0D: csoundPerform (csound.c:2042)
==10758==    by 0x40158B: main (csound_main.c:318)
==10758==

migt relae


On Mon, 25 Jul 2016, Karin Daum wrote:

> Hi,
>
> I wanted to use pvsceps to try some pattern recognition on voices. Including pvsceps in my code produces segmentation faults both when running in CsoundQT 0.9.2.1 or when running  Csound 6.07 on the terminal.
>
> I verified that the same problem exists when using the pvsceps.csd example (example 706) from the manual.
>
> what kind of values do I expect for keps[] from the line
>
> keps[] pvsceps fsig ?
>
> when trying to print them with printk2 I only get NANs.
>
> How are the coefficients in keps[] related to 1/f of the frequencies you can retrieve via
>
> kflag	pvsftw fsig,ifftAmps,ifftFreq?
>
> cheers,
>
> Karin
>
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
>

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

I’m on mac OSX 10.10.5. using Csound 6.07 64 bit 

copy & paste from the terminal:

=mbp-kd-2:/x01/usr/daum/Csound 15:13> Csound -d  pvsceps.csd 
virtual_keyboard real time MIDI plugin for Csound
WARNING: STK opcodes not available: define environment variable RAWWAVE_PATH
(points to rawwaves directory) to use STK opcodes.
0dBFS level = 32768.0
Csound version 6.07 (double samples) May 16 2016
libsndfile-1.0.21
WARNING: could not open library '/Library/Frameworks/CsoundLib64.framework/Versions/6.0/Resources/Opcodes64/libjacko.dylib' (-1)
WARNING: could not open library '/Library/Frameworks/CsoundLib64.framework/Versions/6.0/Resources/Opcodes64/libjackTransport.dylib' (-1)
WARNING: could not open library '/Library/Frameworks/CsoundLib64.framework/Versions/6.0/Resources/Opcodes64/librtjack.dylib' (-1)
UnifiedCSD:  pvsceps.csd
STARTING FILE
Creating options
Creating orchestra
Creating score
WARNING: STK opcodes not available: define environment variable RAWWAVE_PATH
(points to rawwaves directory) to use STK opcodes.
rtaudio: PortAudio module enabled ... 
using callback interface
rtmidi: PortMIDI module enabled
Elapsed time at end of orchestra compile: real: 0.003s, CPU: 0.003s
sorting score ...
        ... done
Elapsed time at end of score sort: real: 0.003s, CPU: 0.003s
--Csound version 6.07 (double samples) May 16 2016
displays suppressed
0dBFS level = 1.0
orch now loaded
audio buffered in 1024 sample-frame blocks
writing 4096-byte blks of shorts to test.aif (AIFF)
SECTION 1:
new alloc for instr 1:
diskin2: opened 'fox.wav':
         44100 Hz, 1 channel(s), 121569 sample frames
B  0.000 .. 60.000 T 60.000 TT 60.000 M:  0.00000  0.00000
Score finished in csoundPerform().

csound command: Segmentation fault
=mbp-kd-2:/x01/usr/daum/Csound 15:14> 

> On 25 Jul 2016, at 13:54, jpff  wrote:
> 
> Cannot reproduce from pvsceps.csd on Linux
> 
> 
> On Mon, 25 Jul 2016, Karin Daum wrote:
> 
>> Hi,
>> 
>> I wanted to use pvsceps to try some pattern recognition on voices. Including pvsceps in my code produces segmentation faults both when running in CsoundQT 0.9.2.1 or when running  Csound 6.07 on the terminal.
>> 
>> I verified that the same problem exists when using the pvsceps.csd example (example 706) from the manual.
>> 
>> what kind of values do I expect for keps[] from the line
>> 
>> keps[] pvsceps fsig ?
>> 
>> when trying to print them with printk2 I only get NANs.
>> 
>> How are the coefficients in keps[] related to 1/f of the frequencies you can retrieve via
>> 
>> kflag	pvsftw fsig,ifftAmps,ifftFreq?
>> 
>> cheers,
>> 
>> Karin
>> 
>> Csound mailing list
>> Csound@listserv.heanet.ie
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>> Send bugs reports to
>>       https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> 
> 
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Hi Karin,

>From what I understand, pvsceps should give you the real cepstrum
derived from the fsig.  cepsinv should then be used with the results
of pvsceps to get the spectral envelope.

The code for pvsceps is found at:

https://github.com/csound/csound/blob/develop/Opcodes/arrays.c#L1760-L1803

and is similar to:

ceps = fft(log(abs(fft(sig))

which is then liftered if the option is given to remove coefficients.

cepsinv would then do something like:

spectrum = exp(ifft(ceps))

to get the smoothed spectrum. (The above as pseudo-matlab code.)

I used a printks with lenarray and it shows the length of the array as
513 for the pvsceps.csd.  The window size for the fsig is 1024, which
would give 512 positive and 512 negative frequencies.  The cesptrum is
analysed from the magnitudes of the positive frequencies (512).  The
additional value is... what I'm forgetting, but I think it's the value
for the Nyquist frequency...

Well, that's what I understand at least.  (I think this is the cue
where Victor comes in and corrects me. ;) )

As a side note, I'm not getting crashes here on Windows 64-bit, but
I'm using the latest from the develop branch.

Hope that's useful!
steven







On Mon, Jul 25, 2016 at 5:33 AM, Karin Daum  wrote:
> Hi,
>
> I wanted to use pvsceps to try some pattern recognition on voices. Including pvsceps in my code produces segmentation faults both when running in CsoundQT 0.9.2.1 or when running  Csound 6.07 on the terminal.
>
> I verified that the same problem exists when using the pvsceps.csd example (example 706) from the manual.
>
> what kind of values do I expect for keps[] from the line
>
> keps[] pvsceps fsig ?
>
> when trying to print them with printk2 I only get NANs.
>
> How are the coefficients in keps[] related to 1/f of the frequencies you can retrieve via
>
> kflag   pvsftw fsig,ifftAmps,ifftFreq?
>
> cheers,
>
> Karin
>
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>         https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Hi Steven,

thank you for your explanation and the reference to the source. This seems to be what I was missing for relating the cepstrum coefficients to 1/freq from what I can get via pvsftw.

apart from the segmentation fault (for which John found the reason) I have the problems when printing the coefficients with printks2 I only get nans. 

cheers,

Karin


  
> On 27 Jul 2016, at 05:14, Steven Yi  wrote:
> 
> Hi Karin,
> 
> From what I understand, pvsceps should give you the real cepstrum
> derived from the fsig.  cepsinv should then be used with the results
> of pvsceps to get the spectral envelope.
> 
> The code for pvsceps is found at:
> 
> https://github.com/csound/csound/blob/develop/Opcodes/arrays.c#L1760-L1803
> 
> and is similar to:
> 
> ceps = fft(log(abs(fft(sig))
> 
> which is then liftered if the option is given to remove coefficients.
> 
> cepsinv would then do something like:
> 
> spectrum = exp(ifft(ceps))
> 
> to get the smoothed spectrum. (The above as pseudo-matlab code.)
> 
> I used a printks with lenarray and it shows the length of the array as
> 513 for the pvsceps.csd.  The window size for the fsig is 1024, which
> would give 512 positive and 512 negative frequencies.  The cesptrum is
> analysed from the magnitudes of the positive frequencies (512).  The
> additional value is... what I'm forgetting, but I think it's the value
> for the Nyquist frequency...
> 
> Well, that's what I understand at least.  (I think this is the cue
> where Victor comes in and corrects me. ;) )
> 
> As a side note, I'm not getting crashes here on Windows 64-bit, but
> I'm using the latest from the develop branch.
> 
> Hope that's useful!
> steven
> 
> 
> 
> 
> 
> 
> 
> On Mon, Jul 25, 2016 at 5:33 AM, Karin Daum  wrote:
>> Hi,
>> 
>> I wanted to use pvsceps to try some pattern recognition on voices. Including pvsceps in my code produces segmentation faults both when running in CsoundQT 0.9.2.1 or when running  Csound 6.07 on the terminal.
>> 
>> I verified that the same problem exists when using the pvsceps.csd example (example 706) from the manual.
>> 
>> what kind of values do I expect for keps[] from the line
>> 
>> keps[] pvsceps fsig ?
>> 
>> when trying to print them with printk2 I only get NANs.
>> 
>> How are the coefficients in keps[] related to 1/f of the frequencies you can retrieve via
>> 
>> kflag   pvsftw fsig,ifftAmps,ifftFreq?
>> 
>> cheers,
>> 
>> Karin
>> 
>> Csound mailing list
>> Csound@listserv.heanet.ie
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>> Send bugs reports to
>>        https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
> 
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

On Wed, 27 Jul 2016, Karin Daum wrote:

> Hi Steven,
>
> thank you for your explanation and the reference to the source. This 
> seems to be what I was missing for relating the cepstrum coefficients to 
> 1/freq from what I can get via pvsftw.
>
> apart from the segmentation fault (for which John found the reason) I 
> have the problems when printing the coefficients with printks2 I only 
> get nans.

Do you have a simple example where you get NaNs?  I can look at it then
==John

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

here it is:





sr = 44100
ksmps = 128
nchnls = 2
0dbfs = 1.0

instr 1
kcnt	init 0
printks2" next kontrol cycle %d \n",kcnt
kcnt +=1
a1 diskin "fox.wav",1,0,1
fsig pvsanal a1,1024,256,1024,1
keps[] pvsceps fsig
kind=0
until kind==513 do
printk2 keps[kind]
kind +=1
od
endin




i1 0 0.1



You will realise also another bug. this time related to printks2 which prints in addition to what is expected from the statement additional characters in the next line. I simply didn’t report it yet. A similar problem was with the prints opcode on the 6.07 beta version. which I’ve reported at that time and which was fixed then. ( I will report the printks2 bug on git)

Cheers

Karin

> On 27 Jul 2016, at 15:43, jpff  wrote:
> 
> On Wed, 27 Jul 2016, Karin Daum wrote:
> 
>> Hi Steven,
>> 
>> thank you for your explanation and the reference to the source. This seems to be what I was missing for relating the cepstrum coefficients to 1/freq from what I can get via pvsftw.
>> 
>> apart from the segmentation fault (for which John found the reason) I have the problems when printing the coefficients with printks2 I only get nans.
> 
> Do you have a simple example where you get NaNs?  I can look at it then
> ==John
> 
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

I see no NaNs -- just some junk afte " next kontrol cycle 1 "
#is that  the other bug?


On Wed, 27 Jul 2016, Karin Daum wrote:

> here it is:
>
> 
> 
> 
> 
> sr = 44100
> ksmps = 128
> nchnls = 2
> 0dbfs = 1.0
>
> instr 1
> kcnt	init 0
> printks2" next kontrol cycle %d \n",kcnt
> kcnt +=1
> a1 diskin "fox.wav",1,0,1
> fsig pvsanal a1,1024,256,1024,1
> keps[] pvsceps fsig
> kind=0
> until kind==513 do
> printk2 keps[kind]
> kind +=1
> od
> endin
>
>
> 
> 
> i1 0 0.1
> 
> 
>
> You will realise also another bug. this time related to printks2 which prints in addition to what is expected from the statement additional characters in the next line. I simply didn’t report it yet. A similar problem was with the prints opcode on the 6.07 beta version. which I’ve reported at that time and which was fixed then. ( I will report the printks2 bug on git)
>
> Cheers
>
> Karin
>
>> On 27 Jul 2016, at 15:43, jpff  wrote:
>>
>> On Wed, 27 Jul 2016, Karin Daum wrote:
>>
>>> Hi Steven,
>>>
>>> thank you for your explanation and the reference to the source. This seems to be what I was missing for relating the cepstrum coefficients to 1/freq from what I can get via pvsftw.
>>>
>>> apart from the segmentation fault (for which John found the reason) I have the problems when printing the coefficients with printks2 I only get nans.
>>
>> Do you have a simple example where you get NaNs?  I can look at it then
>> ==John
>>
>> Csound mailing list
>> Csound@listserv.heanet.ie
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>> Send bugs reports to
>>       https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
>

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

> On 27 Jul 2016, at 19:03, jpff  wrote:
> 
> I see no NaNs -- just some junk afte " next kontrol cycle 1 "
> #is that  the other bug?
> 
yes, this is the other bug.

I’m using a MAC Csound 6.07 64 bit official release.

here is a short part from the printout showing fans and infs:

--Csound version 6.07 (double samples) May 16 2016
displays suppressed
0dBFS level = 1.0
orch now loaded
audio buffered in 1024 sample-frame blocks
writing 4096-byte blks of shorts to test.aif (AIFF)
SECTION 1:
new alloc for instr 1:
diskin2: opened 'fox.wav':
         44100 Hz, 1 channel(s), 121569 sample frames
 next kontrol cycle 0 
/-èigÿ^? i1 161188980270881738219061248.00000
 i1 -9533147304789584751989096448.00000
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1 -757314508004701827133196347606983177014548462211128625122041845579193470998882550126599485003247975588181797255126468590474486889077873312365414859510685628470110604058051997346623741228072774947084583987625473258196174764098073850156596665021694040796468705184574693441536.00000
 i1 -766207275976928179907905945799800306236616435509236269990430500670821982078649495699057551648878021775894226303200279833575865669738667208436754528208564895635840391781953176874625185116310328180772503927588985599191003532631958936134471215593374650337804816813454031585280.00000
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         nan
 i1         inf
 i1         nan
 i1         nan
 i1         nan
 i1         inf
 i1         nan
 i1         nan
 i1         inf
> 
> On Wed, 27 Jul 2016, Karin Daum wrote:
> 
>> here it is:
>> 
>> 
>> 
>> 
>> 
>> sr = 44100
>> ksmps = 128
>> nchnls = 2
>> 0dbfs = 1.0
>> 
>> instr 1
>> kcnt	init 0
>> printks2" next kontrol cycle %d \n",kcnt
>> kcnt +=1
>> a1 diskin "fox.wav",1,0,1
>> fsig pvsanal a1,1024,256,1024,1
>> keps[] pvsceps fsig
>> kind=0
>> until kind==513 do
>> printk2 keps[kind]
>> kind +=1
>> od
>> endin
>> 
>> 
>> 
>> 
>> i1 0 0.1
>> 
>> 
>> 
>> You will realise also another bug. this time related to printks2 which prints in addition to what is expected from the statement additional characters in the next line. I simply didn’t report it yet. A similar problem was with the prints opcode on the 6.07 beta version. which I’ve reported at that time and which was fixed then. ( I will report the printks2 bug on git)
>> 
>> Cheers
>> 
>> Karin
>> 
>>> On 27 Jul 2016, at 15:43, jpff  wrote:
>>> 
>>> On Wed, 27 Jul 2016, Karin Daum wrote:
>>> 
>>>> Hi Steven,
>>>> 
>>>> thank you for your explanation and the reference to the source. This seems to be what I was missing for relating the cepstrum coefficients to 1/freq from what I can get via pvsftw.
>>>> 
>>>> apart from the segmentation fault (for which John found the reason) I have the problems when printing the coefficients with printks2 I only get nans.
>>> 
>>> Do you have a simple example where you get NaNs?  I can look at it then
>>> ==John
>>> 
>>> Csound mailing list
>>> Csound@listserv.heanet.ie
>>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>>> Send bugs reports to
>>>      https://github.com/csound/csound/issues
>>> Discussions of bugs and features can be posted here
>> 
>> Csound mailing list
>> Csound@listserv.heanet.ie
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>> Send bugs reports to
>>       https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>> 
> 
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>       https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here

Kixed the printks2 problem -- just checking it in

On Wed, 27 Jul 2016, Karin Daum wrote:

> here it is:
>
> 
> 
> 
> 
> sr = 44100
> ksmps = 128
> nchnls = 2
> 0dbfs = 1.0
>
> instr 1
> kcnt	init 0
> printks2" next kontrol cycle %d \n",kcnt
> kcnt +=1
> a1 diskin "fox.wav",1,0,1
> fsig pvsanal a1,1024,256,1024,1
> keps[] pvsceps fsig
> kind=0
> until kind==513 do
> printk2 keps[kind]
> kind +=1
> od
> endin
>
>
> 
> 
> i1 0 0.1
> 
> 
>
> You will realise also another bug. this time related to printks2 which prints in addition to what is expected from the statement additional characters in the next line. I simply didn’t report it yet. A similar problem was with the prints opcode on the 6.07 beta version. which I’ve reported at that time and which was fixed then. ( I will report the printks2 bug on git)
>
> Cheers
>
> Karin
>
>> On 27 Jul 2016, at 15:43, jpff  wrote:
>>
>> On Wed, 27 Jul 2016, Karin Daum wrote:
>>
>>> Hi Steven,
>>>
>>> thank you for your explanation and the reference to the source. This seems to be what I was missing for relating the cepstrum coefficients to 1/freq from what I can get via pvsftw.
>>>
>>> apart from the segmentation fault (for which John found the reason) I have the problems when printing the coefficients with printks2 I only get nans.
>>
>> Do you have a simple example where you get NaNs?  I can look at it then
>> ==John
>>
>> Csound mailing list
>> Csound@listserv.heanet.ie
>> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
>> Send bugs reports to
>>       https://github.com/csound/csound/issues
>> Discussions of bugs and features can be posted here
>
> Csound mailing list
> Csound@listserv.heanet.ie
> https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
> Send bugs reports to
>        https://github.com/csound/csound/issues
> Discussions of bugs and features can be posted here
>

Csound mailing list
Csound@listserv.heanet.ie
https://listserv.heanet.ie/cgi-bin/wa?A0=CSOUND
Send bugs reports to
        https://github.com/csound/csound/issues
Discussions of bugs and features can be posted here