Re: complex output of pwelch

[Marius Schamschula]

Dmitri,

You are not seeing things.

I tried the same on two Macs, both running my builds of 2.9.9.

On the G5 I get

octave:1> a=rand(16,1)

a =

   0.624555

   0.412899

   0.191215

   0.969144

   0.543905

   0.015706

   0.556574

   0.740120

   0.931901

   0.187294

   0.234009

   0.835056

   0.086267

   0.802493

   0.554253

   0.477358

octave:2> b=pwelch(a)

b =

   1.31500 - 0.00000i

   0.44513 - 0.00000i

   0.03487 + 0.00000i

   0.09067 - 0.00000i

   0.15099 - 0.00000i

   0.04116 + 0.00000i

   0.04662 + 0.00000i

   0.04982 + 0.00000i

   0.00142 - 0.00000i

On the Mac Pro I get:

octave:1> a=rand(16,1)

a =

   0.229622

   0.913613

   0.095573

   0.894183

   0.973529

   0.315020

   0.147748

   0.597676

   0.767818

   0.045655

   0.715150

   0.064185

   0.940617

   0.891771

   0.015746

   0.182650

octave:2> b=pwelch(a)

b =

   1.2225856

   0.2785923

   0.0018920

   0.0435077

   0.0975156

   0.0717156

   0.0581892

   0.0836018

   0.0733558

Thus, this is a PPC issue.

On Mar 11, 2007, at 12:56 AM, Dmitri A. Sergatskov wrote:

I noticed that at some moment I started getting

a complex output while doing

b = pwelch(a)

(which must be real).

The imaginary part is order of eps, but it is still an annoyance.

Should we put

Pxx = real(Pxx)

or there a better fix?

Dmitri.

p.s. This is an example on PPC OSX with hpc binary of octave 2.9.9:

octave:85> a=rand(16,1)

a =

   0.23158279611096871

   0.13475686431041639

   0.11600916079881567

   0.66193574795860100

   0.28821200634388527

   0.33950150091274417

   0.03019499029666499

   0.53297717647050991

   0.00689129650662494

   0.40770131753552846

   0.76269004312663635

   0.64289433700013687

   0.48166615055996825

   0.61059075641665395

   0.09181145530069872

   0.96806357024737943

octave:86> b=pwelch(a)

b =

   0.694906450550431 - 0.000000000000000i

   0.168157410845850 + 0.000000000000000i

   0.044042332703282 + 0.000000000000000i

   0.013559565644394 + 0.000000000000000i

   0.020360604330361 + 0.000000000000000i

   0.043678042332714 + 0.000000000000000i

   0.030391967689350 + 0.000000000000000i

   0.043206552245520 - 0.000000000000000i

   0.069815447871313 - 0.000000000000000i

octave:87> max(b-real(b))

ans = 0.00000000000000e+00 + 2.82851361838824e-18i

_______________________________________________

Help-octave mailing list

address@hidden

https://www.cae.wisc.edu/mailman/listinfo/help-octave

Marius

--

Marius Schamschula,  Alabama A & M University, Department of Physics

    The Center for Hydrology Soil Climatology and Remote Sensing

   http://optics.physics.aamu.edu/ - http://www.physics.aamu.edu/

          http://wx.aamu.edu/ - http://www.aamu.edu/hscars/