High Pass Filter....Oops

[kozikowski]

I may need somebody else with a Mac to try this.

One of the least damaging ways to eliminate DC level is apply a High Pass Filter 24dB/10Hz. Except in Audacity 1.3.7 you can't do that. The system will not allow me to type "10" in the frequency box. It defaults to "1". It's sneaky. It appears to let me type 10 and then changes to 1 -- by itself -- if I change one of the other numbers. It's impossible to set "10" with the slider or the text entry box.

Mac/Intel/Tiger

Audacity 1.3.7 which up until now has been well behaved. I was using the DC Removal in Normalize, but I got talked out of it. I'm back doing that.

Koz

[billw58]

This was an issue with Nyquist, and an interaction between what you typed into the box and the minimum resolution of the slider. Steve was involved in tracking this down and fixing it, I believe. It is now fixed in 1.3.10. AFAIK if you type "10" in the box and then click OK, it does what you'd expect. But I don't have 1.3.7 to test that on.

-- Bill

[kozikowski]

It does the same thing in Mac/Intel/Leopard/1.3.7. I Won't be able to try it on other machines for a while.

Koz

[kozikowski]

That's OK. I just wondered if that was a Los Angeles phenomena, or it really did that.

What are the current problems in 1.3.10? Which build are you using?

Koz

[billw58]

What are the current problems in 1.3.10? Which build are you using?

http://wiki.audacityteam.org/index.php? ... lockers.29

The first P2 on that list has apparently been fixed, and I'll be trying it out tonight.
[Note to people reading this after Nov 16 - the P2 I'm referring to might be gone!]

I keep 1.3.9 on my machine just in case the devs break something while working on 1.3.10. But 1.3.10 is, in general, much better than 1.3.9. I download the nightly build and try it out every day.

I think you'll like 2.0 when it arrives.

-- Bill

[kozikowski]

<<<I download the nightly build and try it out every day.>>>

I have a slightly different application. I have to explain to some powerful Hollywood production types why it failed.

Now which one would you use?

Koz

[billw58]

<<<I download the nightly build and try it out every day.>>>
I have a slightly different application. I have to explain to some powerful Hollywood production types why it failed.
Now which one would you use?

I'd use the one I find is the most stable for me in my application, of course. I wasn't suggesting you try the nightly builds, but you asked what I use.

-- Bill

[kozikowski]

<<<I'd use the one I find is the most stable for me in my application>>>

Which would be? If this filter thing is still broken in 1.3.9 and 1.3.10 is still going through Major Changes every half-hour.... You see where I'm going with this? There is no workable production program.

That's why I'm still using and recommend 1.3.7. The filter thing was the first time I found something I couldn't work around easily -- except to use DC Removal in Normalize.

I suppose I could whip something up in the graphic equalizer, but I've had troubles with that at the low end. Correct me, but the light blue line in the equalizer tool is what the tool is actually going to do and it doesn't matter where you drag the little graphic balls.

Koz

[billw58]

The Nyqusit bug (the filter thing) is fixed in 1.3.10.

"The green curve is the one that Audacity actually uses to perform the effect, taking into account the limitations of the equalization algorithm." You can get the green curve to follow the blue curve more closing by setting "Length of filter" to the maximum - this is especially important at the low end. So Audacity is at least showing you exactly what it is capable of.

Whether or not 2.0 will be a workable production program for you ... I don't know.

-- Bill

[kozikowski]

<<<blue curve more closing by setting "Length of filter" to the maximum >>>

What's the metaphor they're trying for? The number of Ls and Cs in the filter; the physical length of the filter laid out on the breakfast table and therefore the Q or pole number? Or did I miss something?

Does it always assume lumpy Chebyshev?

Koz