In addition to Spot's excellent advice I might add that you should use 'Mode 3" in the Sony Noise Reduction plugin. It doesn't reduce quite as much noise, but does a much better job of avoiding the flanging artifacts.

John

I sent the sample (sundancemediagroup.com).

Thank you very much!

>>...'Mode 3"...

Got it.

I was experimenting with the different modes and see what you're saying.

1.) I guess if you have those artifacts, is it pretty safe to say that there's not much else you can do?

2.) Does the noise reduction plug in pretty much stand by iteslf, not requiring much pre or post effects?

Thank you very much!

The noise reduction plugin stands by itself. Most advice (in help file and in various tutorials) is to not reduce more than 20dB per pass. Also, pay attention to where you get your noiseprint sample.

Spot, in his in-person tutorials, talked about using really short samples (way less than a second). I am not sure that is really the way to go (mathematically, it doesn't make sense to me), but the results in his tutorial were impressive, so I've tried experimenting with short samples when I take the noise print.

If you have any AC hum, my advice is to first (before NR) take 2-3 passes with a notch filter at 60, 120, 180, 240, etc. until you kill that. After that, use the NR plugin.

I would never actually use 100dB in one pass -- ever. However, I often set up a loop, crank Reduce Noise up to 100% (temporarily, while looping), and then play around with the "Noise Bias." Reducing that will decrease the flanging (and other artifacts, such as loss of high frequencies), but you will beging to hear the noise returning. I also click "Keep Residual Output" to see how much "good stuff" I am losing. When you click on this (which of course is always just a temporary setting), you will hear the stuff you are going to eliminate. If you hear lots of your original source, try lowering the noise bias. (You can also click on the Noiseprint tab and try raising and lowering the noise curve with the little up/down arrows next to the reset button, but I think this pretty much does the same thing as the Noise Bias slider).

Finally, before I change the Reduce Noise slider back down below 20dB, I run through Window FFT size. The 2,048 setting almost always produces the best results, but once in awhile I get better sound with 4,096. You can try the other FFT settings while the preview is looping just to get a feel for what they do.

So, once you've reduce the noise by 15 - 20 dB, get a new noise sample, and repeat the process. On subsequent passes, you don't need to repeat all the fiddling with the other settings. Just leave them the same, pick a Reduce Noise amount that seems about right, and then click OK. If the result seems a little lifeless, you can add some High Shelf in the NR itself, or do it after the fact using one the EQ functions.

Spot is the audio expert in this forum, so listen to him. I am just a self-taught guy. My only big-name client was Jerry Lewis and I restored a whole bunch of 1955 transcontinental telephone recordings that were made on 1950's acetate dictation disks. The NR plugin was a huge help.

Well, using the noise reduction plugin is sort of an art. It is highly dependent on selecting the "right" noise sample. Also, some times it is more effective to run the noise reduction several times using smaller increments. If you want to achieve 30 db of reduction, run the plugin three times at -10db of reduction each time instead of -30db all at once. Of course, you need to run the "capture noiseprint" each time.

John

EDIT: John Meyer beat me to the punch with a much more detailed explanation. Follow his advice.

Thank you very much. I will definitely be keeping this on file.

Questions: When you say do, say 20 Db per pass, do I understand you to mean, reduce noise by 20 DB, then do it again (another pass) before saving the file to file? Do this multiple times and see how it goes? Removing X db a little at a time? I believe that is what you are saying.

2.) Is the "notch filter" a seperate utility? If not, exactly what is it?
Edit: Never mind. I found it on usenet. Thanks anyway.

3.) I did notice the Keep Residual Output and wondered exactly what that was for. That is really slick. Thanks for the good explanation.

Thank you again for your help.

John, if you wanna know part of the secret...
I look for noise that first of all, doesn't allow the meters to move, hence the short sample. I know, mathwise it doesn't work out. John Feith, the guy that wrote this plug in originally, used to tell me I was crazy, but he couldn't argue with my result.
But the REAL secret, is finding the short sample that's in the same frequency range as whatever you are cutting the noise out of. Getting a sample from just "anywhere" doesn't cut it. you gotta hunt around a little. A noisy fridge, or a humming light, will share some of the same frequencies as the voice surrounding it. Find those points, and sample them so the meters aren't moving when you check the sample.
therein, lies a big part of how I do this.
Kent, after listening to that file, I think you could also up the signal to noise, which would help TREMENDOUSLY. :-)
I've returned the sample in two different forms.

>> I think you could also up the signal to noise...

Could you explain what that is? Basically, how would I do this? (Sorry.) Does this mean to simply have more voice than background noise, thus the increase in S to N ratio?

If so, the thing is, there really doesn't sound like there's a lot of background noise at all. Man, I hope I'm not losing my hearing that bad....

Thank you very much for your time and help.

>> I think you could also up the signal to noise...

Bump. Any update on this? Thanks!!

could be payback time kentwolf.

I, unlike DSE am the forum's posterchild audio ediot-- and even I now know what the signal to noise reference means...

since I picked up Now Hear This: Superior Sound for Digital Video and Beyond, (still on discount at www.vasst.com.)

The signal to noise reference is in regard to how your recording was acquired in the first place-- Garbage IN, Garbage OUT... The S2N reference can only help you get better sound on the next project.

What a great thread tho- kentwolf! Take it to the bank alone and your money ahead- pick up NHT and your even better off.

And no, I don't work for Spot. yet.

bestx3,

bt

Kent,
I didn't answer it here, because I thought I'd answered it in our mail exchange.
Basically, getting the mic closer, beefing up the bulk of the mic, and figuring out a barrier between the mic and the very close by computer, would make a tremendous difference to lowering the noise floor and increasing the apparent signal.
For those that haven't heard the sample, it's a thin-body, maybe plastic body mic with lots of body resonance, and it's near a very noisy computer. The noise floor is around-30dB so it's quite loud with about 20dB of dynamic range between the noise floor and the peaks.

Wow, Spot. Your NR technique is amazing. I just tried it and was floored by how clean the reduction was. Too bad I just sent my latest project to the duplicators. This technique would've really cleaned up some video I shot in an orange juice factory.

I'm going to buy your book as a thank you!

Thanks to the Johns too. If you guys had a book, I'd buy yours too!