Bugs in spectrum filter

GoldWave general discussions and community help
Post Reply
Moonmist
Posts: 30
Joined: Sat May 16, 2009 3:24 pm
Location: The Milky Way Galaxy

Bugs in spectrum filter

Post by Moonmist »

Hello, wonderful software, it's the best in terms of ease to use imho! And the expression evaluator is the most genius idea ever brought to audio editing! I use it all the time! However there has been a bug that has been bothering me in the spectrum filter that has been present ever since you implemented it (and is still there in 5.52 even). The bug is this... at FFT 16 (65,536 samples) frequencies are displayed only up to 65,536 hz and not up to 100,000hz! (On a file I've recorded at 200,000 khz sample rate using a different program to set the sample rate and then opened in goldwave and recorded, wish you would allow us to choose up to 200,000 as my sound card supports it!).

But that's not the big problem... the big problem is when you apply a filter at any FFT above 14... parts of the frequency spectrum get totally removed or shifted down by a certain db offset. I don't remember exactly what frequencies were cut off the top of my head but I'll go test it out and get back to you. I seem to remember frequencies between 20,000 and 65,536 hz being totally removed.

Also there is a similar bug in the lower end of the spectrum that affects all FFT sizes... Basically GoldWave thinks that 10hz is 20hz... if I do a high pass at 11 hz using the filter points it ends up high passing at 20hz.. and if I high pass at 20hz same thing... so basically it's not recognizing frequencies 10 to 20 hz it seems. That is a big problem for me especially because I do a lot of stuff in low octaves below 5 string bass guitar even, including pipe organs, I've converted earthquake seismographs to sound waves before, I have an 8 string bass guitar in which I tune to subcontra notes like C#0 (17.32hz) etc.

I'm very experimental with stuff, I get bored of standards really easily. (I'd personally love it if you didn't impose a lower limit on frequency, but 10 is nice, my headphones output as low as 5 hz however, lol).

I think maybe nobody noticed because no one works with frequencies that high, but that's where I come in, lol. And that's mainly why I signed up for this forum, it was driving me crazy for the last year or two, I thought you would have fixed it with the release of 5.50 but you didn't so I had to come out of the woods, lol.

Anyway if you could fix that bug I would be so very greatful! And yes I'll test it out to give you firm figures on those bugs, I might even make some screenshots of the spectrum filter window for ya, and I'll get back to you later, thanks for making and continuing to develop and improve such a great application! Oh and btw, I kinda like the new gradient feature... looks good with light blue-silverish-blue colors ;)

Moonmist
GoldWave Inc.
Site Admin
Posts: 4372
Joined: Wed Mar 10, 2004 6:43 pm
Location: St. John's, NL
Contact:

Re: Bugs in spectrum filter

Post by GoldWave Inc. »

Thanks for the details. I'll look into that before the next update.

Chris
Moonmist
Posts: 30
Joined: Sat May 16, 2009 3:24 pm
Location: The Milky Way Galaxy

Ok here is the info I got, screenshots included :)

Post by Moonmist »

Ok here are some screenshots of what I'm talking about... first screenshot is of the FFT 16 not showing the full spectrum. As you can see frequencies from 65,536 hz to 100,000 hz are not being displayed when they in fact are present in the audio. (This is a spectrum of a baritone guitar note boosted 60 db over 0 and then applied the default dynamics preset to chop off everything at 0 db so as to create overtones for display purposes. Was recorded at 200,000 hz sample rate.)

Image

And now this is the same guitar note with an 88,200 hz Lowpass set up in the spectrum filter... The top is before processing what everything looked like with an FFT of 16 (we can see frequencies from 65,536 to 100,000 hz missing still of course, but they are there). Then in the middle is after processing now what the spectrum looks like at FFT 16... All frequencies between 20,000 hz and 32,768 hz have been totally removed. But we can't see what happened up higher above 65,536 hz because the FFT 16 won't show it ... so that leads to the bottom illustration... FFT 15 shows what happened up higher... Upon enlarging the spectrum window and using points I was able to measure the second gap being from about 86,250 (I suppose this is suppsed to be the 88,200 hz we wanted?) to about 98,250 hz!

Image

And in this final image I used a blank file at a sample rate of 1,000 hz so that I could have a better resolution in the low end to illustrate the problem better (that's why you see it cut off above 500 hz but that's normal). I inserted white noise and the top half shows the spectrum at FFT 16 before processing... now in the bottom half it shows after I applied the 10 hz highpass but you see that it actually highpassed at 20 hz and cut out frequencies from 10 to 20 hz!

Image

Hope this helps you Chris :)
Post Reply