Thanks for your reply.
I played around with 'Tools / Noise Reduction' but cant really find the real 'key' to just drop the constant noise without affecting the actual sound recording.

Maybe if you uploaded a sample of the source somewhere, someone here would give it a listen and respond.

Sound Forge NR is no longer competitive. iZotope RX is substantially better for hiss and noise reduction.

If you can upload a WAV file, I'll be happy to do the noise reduction for you (I own iZotope RX3 Advanced).

johnmeyer - what a nice guy!

The hiss may be to an extent that your reduction-expectations are unrealsisttic. But try Sony NR 2, and take advantage of JM's kind offer, cos that's the best result you will possibly get !

geoff

Given that the hiss is likely to be a defined frequency probably including harmonics, perhaps SpectraLayers may be able to eliminate the hiss - at a cost though, but it is less expensive than RX3/4 Adv (as good as they are) - and a big learning curve as well.

NR2 works decent with din type noise, though iZotope or SL may do better job. depending on the severity. If there's only dialog, rolling off frequencies above 8 or 10kHz can help as well (a common practice regardless of noise problems).
Post a short clip for opinions on what's possible and what's not.

By definition, hiss is broadband without a single frequency and harmonics.

Effective audio noise reduction is an art, the trick is to get just the right noise sample and then adjust the parameters to maximize noise reduction and minimize artifacts. Sony's NR2 is certainly capable but not in the same league as Izotope RX. Nevertheless, find a sample of the audio which contains only noise, generally the longer the sample the better, but at least 250 ms. Capture the noiseprint, use Mode 3 noise reduction and play with the amount of noise reduction and the FFT size (Smaller FFT windows (1024-4096) usually produce fewer artifacts.) It is better to do a little bit of noise reduction and do a couple of passes instead of doing a lot in one pass. Do one pass with maybe 10-12db of reduction, then sample the noise again and do a little more. Like I said, it's an art and you'll get better with practice.

Try overdoing the noise reduction so that you can appreciate what artifacts can be produced. Once you know what to listen for, back off the noise reduction until you get the best compromise between artifacts and residual noise. Residual noise is "natural" and therefore less offensive to me than processing artifacts for the same audibility.

SpectraLayers can reduce noise and there are a lot of other (generally expensive like Izotope Rx Advanced) options for noise reduction, although all of them effect the sound, the more reduction you apply, the more apparent the artifacting. 12-15 dB is a typical amount of noise reduction that sounds clean, It doesn't get rid of the hiss but it reduces it to a desirable level. Learn what can be done with trial offeres of the various software so that you understand there is no perfect fix.

Better to do a good job in the initial recording than to try to fix in post, (although sometimes it just doesn't work out.)

I use Audacity and get great results.

(1) Load file into Audacity

(2) Create a noise profile. Find a small portion of audio that is PURE noise. Be sure it is pure noise and no other sounds. You will have to zoom in to find it. One to two seconds is all you need. Once you find it, highlight it.

(3) Click on 'Effect' and then 'Noise Removal'.

(4) Click on 'Get Noise Profile'. The dialog box will then close.

(5) Do a Ctrl+A to select the entire file and then go back to the 'Noise Removal' and then just click on 'OK'. Do not click on 'Get Noise Profile' the second time. If you do, then the entire file becomes the profile instead of noise only.

It will then remove the noise.

The entire process is much easier than typing this.

The entire flow is finding the noise to use as a noise profile, use the noise removal effect to add the noise profile, select the entire file and then use the noise removal effect to remove the noise.

I tried this in SF and while it worked using a sort of similar method, the results in SF were very poor compared to Audacity. And Audacity is free!

Remember, it is very important to find a small sample of noise ONLY! Any other sound can lead to artifacts and/or poor results. I usually make my small selection and then turn the volume up on my system so I can hear if there are any other sounds there.

Sometimes after doing the noise removal, I can then hear sounds that were originally masked by the noise. So check and see if that is the case. if so, then undo and find another location for the noise profile. Sometimes it takes several tries. The Analyze part of Audacity can sometimes be helpful in finding the noise only. You have to experiment with different DB levels.

All software's that I have tried need a noise profile to remove the noise in the file. It works by taking the noise profile and then applying it out of phase with the noise in the file which cancels the noise. A lot like what happens in a balanced line system with a mixer input. How well it cancels depends on how well you do at getting the noise profile. If the recording was done using auto-gain or compression, then the noise level will vary which will make it more difficult to remove. So avoid auto-gain and compression.

I hope this helps.

If the noise is just that, noise from the camera's mic preamps then you're not going to do any better using SF than Vegas itself. Using SF's Noise Reduction can well give you a worse outcome. Problem with using NR2 on this problem is getting a proper sample of the noise plus it'll almost certainly try to cut frequencies that you want to preserve as noise covers the entire audio spectrum.

Simply use the Track Eq as a High Shelf filter. I'd set the frequency to 6KHz and the roll off as high as it will go with infinite attenuation. As Vegas's Eq doesn't go above 24dB/octave you can get better results by stacking the four filters in that plugin.

To do that select the 1,2,3,4 filters in turn, make each one High Shelf, set its frequency to 6KHz and Roll Off to 24dB/Octave.

Having said that using a cutoff frequency of 6KHz might be too low even for speech, you need to A/B the results to know for sure but that's easily done in Vegas with the Fx On/Off check box.

One last tip, if you're only noticing this noise wearing headphones or up close to your speakers you're probably hearing a bigger problem than your audience ever will, so go gently trying to fix the problem.

Bob.

FWIW,
The noise reduction tools in the low cost Sound Forge Audio Studio and the Noise Reduction 2.0 suite bungled w/ SF-Pro are quite different. Though they do some of the same processes, NR-2 results are superior.. providing it's used properly. The iZotope RX is then again superior to NR-2.. you-get-what-ya-pay-fer.

Most pro audio folks I've encountered in my 30+ year tenure define "hiss" as high-frequency white noise.

"Most pro audio folks I've encountered in my 30+ year tenure define "hiss" as high-frequency white noise."

Well they shouldn't. White noise has a flat spectrum analogous to white light.

Hiss is random noise with no strong spectral peaks, but which may have a high frequency emphasis. Sustained /s/, /sh/, /f/ and unvoiced /th/ are examples of hiss.

@ PeterDuke,

If you want to see a flat response across all frequencies with a real time audio analyzer, then you would use a pink noise. A white noise will look a bit like the noise that an amp can generate. I have seen noise from certain amps that looks a lot more like a pink noise instead of a white noise. Still, a pink noise is the one that is flat.

The names of the noises defy logic but that's the way it is.

"Still, a pink noise is the one that is flat."

"The names of the noises defy logic but that's the way it is."

Rubbish!

White noise has a flat (uniform) spectral density. As I said before, it is called white by analogy with white light, although in fact light from hot bodies such as the sun and stars have spectral lines, unlike white noise

Pink noise has a spectral density that falls at 3 dB per octave or 10 dB per decade. It is used in measurements because many natural noises such as traffic, shops, etc. have a spectral density that tends to fall with increasing frequency. Pink noise is also a useful signal to use with octave or 1/3 octave filter banks because the output from each filter is uniform. The name "pink" was coined because in comparison to white noise there is more energy in the low frequencies. In the visible spectrum this corresponds to the red end. Add red to white and you get pink.

See Wikipedia for confirmation of what I have said. I worked for many years in a laboratory where we used such noise sources and measured noisy work sites.

@ PeterDuke

I was referring to the way it is displayed on a Real Time Analyzer.

My hearing is very poor so I depend a lot on visual aids to help me with my audio work. I can see what I cannot hear. So I depend a lot on a real time analyzer as well as other audio tools.

One thing for sure, when one does a Google search one will find a ton of confusing results! To make matters worse, it can be a bit confusing trying to explain this. Sort of like the Kelvin descriptions of the colors of light. An illogical mess when you look at the way it actually is VS. the way people call it.

Things I know for sure are when I choose a 'pink noise', my real time analyzer shows a flat response, as in a straight line. When I choose a 'white noise' my real time analyzer shows a non flat response that to a certain extent increases from low to high. That is what I am referring to.

If one wants to equalize a system for a flat response and/or check the frequency response of a recording device then one should use a 'pink noise'.

You should learn the characteristics of the tools you use. Apparently your analyser does not show spectral density but something like 1/3 octave or octave band analysis. What is the make and model of your analyser?

You don't have to use pink noise to equalize a system. You could use any signal provided that you measure both in and out and your system is linear. I agree that pink noise is a good signal to use if you want to ignore the sharp peaks and dips of standing waves you might get with a sine wave.

Hi John,
sorry for not replying right away. I had to attend a family emergency. All is better now.
It is very kind of you to offer to help with the noise reduction however, I do not know how to upload the WAV file to this forum.
I could copy / paste to a 'regular' e-mail !
Thanks and looking forward to hearing from you.
Stef

Stef,

You will need to upload your WAV file to a free file sharing site, and then post the link to that download..

For instance, here is a link to an audio file I sent to my daughter a few years ago because I thought she'd find the song funny:

https://dl.dropboxusercontent.com/u/1561578/Nellie%20Mckay%20-%20The%20Dog%20Song.mp3

You can make this link "clickable" if you follow the directions given in the sticky at the stop of the forum. However, you don't need to bother doing this. Here's what it looks like as a clickable link:

The Dog Song

If you don't already have an account set up at a file sharing site, they are easy to create and use. Here are some options:

Google Drive

If you already have a Google account, you already have Google drive. Just sign in and use it.

Dropbox

Dropbox is easy to use, and a great way to backup and store stuff. You get at least 1 GB for free.

Mediafire

This site is designed for sharing media (photos, video, audio).

If you are an Amazon Prime member, your membership also gives you some free storage.

There are many other options beyond these. I use Dropbox and Mediafire quite a bit, and still haven't run out of the free storage allowance.

http://www.mediafire.com/listen/k28t4i1baky1b59/1_)__4-4-15_SFPro10_-_Sound_Copy.wav

It is always trial and error with me. I hope this one goes thru.
Thanks for your patients and help.
Stef

What is it that you want to KEEP in that example - the random little clucking sounds?
I only ask because there is not much "hissing" evident.... one man's goose is another man's gander...