Removing "background hum"... - Page 2

Z Cheema · #11 (**permalink**) 08-25-2007, 09:47 AM

Grazie is correct, I always ask for fans or Air cons to be turned of for the duration of the recording (if possible) I always explain why that is and then the client understands and usually complies. Failing that but as always now I mic every one up that is involved in any critical talking.

But A hum tends to be low freq like a mains hum, which should be located at about 50 (UK )the noise of the camera being picked up by the Mic was below the 1khz mark.

I am doing some stuff at the moment for some one who has given me a recording with mostly hiss and sometimes the odd voice appearing from it, in this case it is going to be difficult as the S/N ration is to low, come back Dolby B/C

LJR · #12 (**permalink**) 08-25-2007, 11:39 AM

Yes, trickier indeed, although given a hiss is usually a higher frequency than most dialogue you maybe able to isolate on the frequency spectrum instead of looking for a noise print.

Grazie · #13 (**permalink**) 08-25-2007, 11:20 PM

LSR, have you seen the noise print graph in NR? What NR does do, is that you can SAMPLE a section of "pure" hiss, and test against that. All you need is a few micro second samples. You can then spread this further out too.

Grazie

LJR · #14 (**permalink**) 08-26-2007, 02:33 PM

Hi Grazie, While I haven't seen it in Vegas to my way of working there are typically two methods of reducing unwanted noise within a clip though, one is to take a noise print - pretty much in line with what you are describing - where you select a small portion of the timeline which represents the noise and which the software then aggregates and creates a virtual noise print. You can then usually fine tune that through a noise threshold (dB level) and amount the filter is applied. It typically is used to remove white noise (wind in the trees, waterfall, etc) although it can work on single frequency noise too (given a good SNR). The other method is to identify an unwanted sound at a frequency level through using a frequency spectrum viewer and have the tool isolate (and subsequently remove) that. You can usually identify the offending frequency as a line running through the frequency spectrum's histogram. You typically then click the frequency (the line) within the histogram and it selects that frequency to remove.

The first has a strong interdependency on the SNR, the second method does not as it's isolating a specfic frequency irrespective of its amplitude or close proximity to other audio frequencies.

What I was suggesting here is that if the SNR is low and it's difficult to isolate a noise print then perhaps looking at the frequency of the hiss and isolating by that may help.