<<I'm experimenting with noise removal and some PCM M10 3 recordings in Aud=
acity (beta 1.3)? now myself, as a stopgap solution till I can get some ste=
reo mics.
My initial attempt, taking the noise sample from the recording, left the re=
cording full of strange little hiccups and funny noises (but no hiss). I as=
sume this was because I couldn't find anywhere where there wasn't at least =
one little bird calling faintly in the background.>>
I generally get better results with manually setting parameters on my noise=
reduction plugins than is possible with the 'learn' function. Use a much s=
maller sample if you're going this route. 'Chirps' are the result of overly=
aggressive broadband noise reduction. I don't know the program you're usin=
g but if there are settings for amount of noise reduction try rolling it ba=
ck considerably. Two passes through the plugin at 2.5 or 3 dB of noise redu=
ction will sound considerably better than one pass attempting to take out 5=
or 6 dB of noise. If that amount of noise reduction still results in too m=
uch audible noise you probably can't make it much better without serious de=
gradation. For this type of noise reduction to work well you need a fairly =
good contrast of intended or foreground sound relative to the unwanted syst=
em noise.
<<I then made a recording of just hiss at the same recording level with the=
recorder under some cushions, and I used that for the noise sample for the=
noise removal. After experimenting with the noise removal settings, I star=
ted getting much better results, but still not right.
At first I thought it was ok, but after listening a few times, I noticed th=
at the reverb had gone, leaving it sounding very dead. I increased the atta=
ck/decay time and got some reverb back, but now there was a little burst of=
hiss each side of each loud bird call. More fiddling with the other settin=
gs, and I've improved it a lot, but the best way to improve it seems to be =
to reduce the amount of hiss removed!>>
See above. From this description it sounds to me like the background noise =
is simply too great to be a good candidate for this sort of dynamic noise r=
eduction plugin. I would precede the noise reduction process with some care=
ful filtering. I tend to hear most broadband electronic noise as most troub=
lesome in the region very roughly centered around 4kHz to 6kHz, so I use a =
subtle bit of low Q bandpass filtering in that area, rather than lowpass fi=
ltering higher up. A few dB of hiss reduction with EQ in this range can eas=
e the task of the noise reduction plugin later on.
<<I haven't given up yet, but given the amount of time I've spent on it now=
, and the imperfect results, I can see why people just shell out for better=
microphones.>>
And better noise reduction algorithms. But yes, better mics plugged into be=
tter mic pres to start off the process with quieter recordings eases the ta=
sk of any subsequent noise reduction processing a lot.
<<I assume that if I come up with some settings I think sound ok, people h=
ere will be able to point out other noise removal artefacts I hadn't notice=
d.>>
The main thing is the flangy, phasey chirps you've noticed, & the dynamic p=
umping around isolated sound events. If that program has the ability to mon=
itor in real time just the noise that it is removing you can quickly get a =
sense if it is damaging the signal in any way.
<<That said, at this stage I still think it's worth a try if all you have i=
s the recorder's internal mics. I hadn't noticed the hiss at all in other r=
ecordings I made in noisier places, so it's only in very quiet places that =
it's a problem.>>
This sort of issue is entirely dependent on the nature of the specific reco=
rding & the actual signal to noise ration obtained. Sometimes you're lucky =
& it cleans up readily, & sometimes not & there really is nothing that can =
be done to make it appreciably better. The only that is certain is that the=
best initial recording you can get will require the least amount of post p=
roduction work.
Scott Fraser
|