Lord Smurf & all - How to remove rumble still remaining after noise removal

[VideoFanatic]

Any idea how I do that in Audition CS5.5?

[jagabo]

Any idea how I do that in Audition CS5.5?

No, I don't use it. I used the Notch filter in Audacity, with a Q value of 10.

[JimmyS]

SpectraLayers Pro from Sony has some very interesting capabilities.

http://www.sonycreativesoftware.com/spectralayerspro

To see how it functions, check out (or download) the webinar.

http://www.sonycreativesoftware.com/spectralayersintrowebinar

There are also a some demos on youtube.

[lordsmurf]

This would honestly take too much time and effort to remove what's a fairly inaudible issue.

Furthermore, it's a home copy of footage. I could understand the effort for the sole surviving copy of something, for the point of library archiving, but not for WWE VHS recordings for personal enjoyment.

The better solution may be to stop listening to it on speakers that poorly overemphasize bass to the point that your videos sound like a car muffler with a hole it.

At some point, you have to end the processing. It's too easy to be OCD.

[VideoFanatic]

WWE never sells their seasons and it's hard to find good quality home recordings. The ones I have are the best quality but even then they have audio problems. You guys keep telling me that it's my speakers that's the problem but I don't believe it is because I listen to different types of content through my Micro system connected to my PC, my £200 headphones and my TV and none have the rumble except these home recordings (other similar home recordings don't have this rumble). Most people don't have high end equipment like yourselves - we can hear the rumble. I can't watch home recordings like this until I fix the audio otherwise I don't enjoy watching them because the rumble really bothers me. I'm at my wits end, if nobody knows how to remove the rumble then I'll have to pay an audio specialist to remove the noise.

Jagabo - I tried what you said in Audacity but it didn't remove the rumble.

JimmyS - I tried SpectraLayers Pro but it's only good for removing obvious sounds like a dog barking or a siren, etc. It's useless for removing general noise. I learned how to use it but the interface is awful (I don't know why these high end programs all have bad UI's!)

Lordsmurf, I noticed this thread here about rumble. The guy mentions "I looked at Lord Smurf's digitalfaq and his audio guide about "remove low rumble and harsh bass". The first file I tried seemed to help a bit". Where is that guide?

[lordsmurf]

I'll have to pay an audio specialist to remove the noise.

Because it's unlicensed WWE content, nobody will touch it. Just an FYI.

Where is that guide?

- How to Restore Audio Quality with Sound Forge
- SoundForge Audio Filter Presets Pack [DOWNLOAD]

... but I'm not sure that's going to help here, for this clip.

That guide needs a major overhaul, too. It was written quite a while ago. While it's not outdated in any way, it needs to be rewritten to reflect methodology and philosophy more than purely based on examples. The presets needs to be updated to V11, which has a few more filters, but V10 is quite good.

[jagabo]

Before hum reduction and after.

[VideoFanatic]

The after sample just seems to have a different noise in place of the rumble!

[pandy]

Normally I would do a noise print of the noise in silence then do 100% noise reduction on the whole file. However the rumble noise does not occur in silence so I can't do that.

This is not correct - one noise sample is OK when noise profile or noise energy (PSD) doesn't change which is not true for VHS - tape have different noise profiles in different places.

Also strategy for some noise (or distortions) can be different - generally shorter FFT have better temporal resolution but worse frequency resolution and opposite for larger FFT - they have better frequency resolution but worse time resolution.

Noise reduction should be multistage, first noise profile for acquisition device and link should be captured (VHS connected but play not active (optionally pause can be turned on)), then profile for silence areas should be acquired for normal play, larger recording (long ones) should be split for few part and independent noise profile should be acquired for each part (best at few places).

Also whole setup should be OK - grounding loops and similar problems should be eliminated before capture not after.

[lordsmurf]

The after sample just seems to have a different noise in place of the rumble!

That's why you have to quit processing at some point. You start to introduce 2-3 new errors for every one error you take away.

I've been working with digital audio since 1994-1996, forget exactly when. About 15-20 years. I was doing audio long before I got into digital video. This aspect has never changed. It's a limitation of processing/filtering audio.

[jagabo]

The after sample just seems to have a different noise in place of the rumble!

No, I simply didn't do as much hiss reduction as you did. Compare my before and after files against each other, not your previous efforts. In fact, the hum sample I posted a while ago was simply the difference (amplified 10 or 15 dB) between the before and after files.

I started with your original sample, not your hiss reduced sample. Because the hum was more noticeable there, and frankly, I couldn't stand listening to your hiss reduced sample. I ran a low pass filter to soften the hiss but had no intention of entirely removing it. That was the "before" file with the obvious hum. Then I took the before file and ran the notch filter sequence to reduce the hum. If you compare the before and after clips you'll see the only major difference is the hum reduction. My intention was to show you that the hum could be reduced, not to perform a full noise reduction of the original.

[budwzr]

Is this any good?

[VideoFanatic]

The samples with rumble in were obviously bad quality as you can also hear a high pitched noise. I rarely get a problem like that where noise reduction causes a high pitched noise. I just posted that as the rumble was easy to hear in that sample. Due to the problems with those samples it's difficult to hear what improvements you guys have made. So I've updated my 1st post with new samples. This time the new sample just contains rumble and the audio quality is better. Guys, could you please try to remove the rumble again on my new sample and could you please post your fixed samples for me to hear. I appreciate it, thanks.

LordSmurf - I appreciate what you're saying but I believe this rumble can be removed without lowering the quality of the audio.

Normally I would do a noise print of the noise in silence then do 100% noise reduction on the whole file. However the rumble noise does not occur in silence so I can't do that.

This is not correct - one noise sample is OK when noise profile or noise energy (PSD) doesn't change which is not true for VHS - tape have different noise profiles in different places.

Also strategy for some noise (or distortions) can be different - generally shorter FFT have better temporal resolution but worse frequency resolution and opposite for larger FFT - they have better frequency resolution but worse time resolution.

Noise reduction should be multistage, first noise profile for acquisition device and link should be captured (VHS connected but play not active (optionally pause can be turned on)), then profile for silence areas should be acquired for normal play, larger recording (long ones) should be split for few part and independent noise profile should be acquired for each part (best at few places).

Also whole setup should be OK - grounding loops and similar problems should be eliminated before capture not after.

Pandy - I do 100% noise reduction all the time without problems where all the noise is removed and I get no artifacts. The earlier sample I posted is a rare exception. I don't have the original VHS tapes, I only have these VHS to DVD conversions.

I don't really understand what FFT is even after reading your explanation! Audition has a default FFT Size of 4096 and a default Noise Print Snapshots of 4000. Would using different values prevent quality loss or distortions such as that high pitched noise?

[pandy]

Pandy - I do 100% noise reduction all the time without problems where all the noise is removed and I get no artifacts. The earlier sample I posted is a rare exception. I don't have the original VHS tapes, I only have these VHS to DVD conversions.

Then You don't have real noise but noise shaped with desired signal (codec which encoded noise based on your audio) - single profile no longer works, 100% noise remove is to radical - instead subtracting noise energy from signal you nee mild, perhaps adaptive equalizer that remove some low energy (bellow some threshold) from spectrum - in analog domain similar filter was named DNL.

Audition have something named Hiss Reduction - this one can be OK, also Noise reduction but small FFT size 128 - 512pt max and very mild noise level removing settings, bellow 50% for sure. It is better to frequently hear results - currently only you ears can judge final quality...

I don't really understand what FFT is even after reading your explanation! Audition has a default FFT Size of 4096 and a default Noise Print Snapshots of 4000. Would using different values prevent quality loss or distortions such as that high pitched noise?

FFT size is always trade-off between time and frequency resolution - large FFT have high frequency resolution but worse time resolution. For signals compressed with lossy codecs i think short FFT are better (they give less artifacts), also noise reduction level should be smaller that 100% - try to encode noise (various levels) with lossy codec - this shows how noise is seen by encoder - compare signal before and after encoding - signal no longer will be noise.
IMHO 4096pt FFT is to big but anyway You and Your ears/brain are the judge...

[jagabo]

I don't hear any rumble or hum in your new filtered sample.

[VideoFanatic]

Can you not hear any difference between these two files? I can hear a low sounding rumble-like noise like in the sample you listened to before where you said you could hear the rumble.

• Audio After Noise removal (rumble still present)
• Audio After Rumble & Hiss Removed (ECW June 13 2000)

[pandy]

Rumble & Hiss Removed (ECW June 13 2000) this one is destroyed by FFT - there is lot of echo and chirping, sound is without noise but also signal is destroyed...

Hiss Removed but Rumble Still Present (ECW June 13 2000) - this one is better but FFT noise removing is audible and this sound badly...

[VideoFanatic]

Pandy - In this file "Hiss Removed but Rumble Still Present (ECW June 13 2000)" can you hear the rumble? It sounds like you have Adobe Audition CS5.5. Do you know of a way to remove the rumble without damaging the audio?

Should I use a different FFT setting to prevent quality loss? If so, what setting should I use?

[jagabo]

Can you not hear any difference between these two files? I can hear a low sounding rumble-like noise like in the sample you listened to before where you said you could hear the rumble.

• Audio After Noise removal (rumble still present)
• Audio After Rumble & Hiss Removed (ECW June 13 2000)

Yes, I hear a difference between those two files. The second one is terribly over filtered. There is chirping, phase distortion, a horribly reduced spectrum distribution, and it's hollow sounding. It sounds like it was recorded under water. But there is no significant buzz, hum, or rumble in the first one. Even with a 20 dB bass and lower midrange boost.

[VideoFanatic]

Jagabo for the original file where you could hear the rumble, what settings did you use in Audacity? Also I know you said you didn't want to completely remove the rumble as it caused problems but if I did, what settings should I use?

Is there also any way to remove the high pitched noise?

budwzr - I listened to your file but the rumble is still there.

[jagabo]

I used the notch filter in Audacity to remove the buzz. I use a Q factor of 10 and applied it at 60, 120, 180, 240, 300, 360, and 420 Hz. Maybe at 480, 540, and 600 too, I don't remember for sure. Do you still hear your "rumble" in this file?

[VideoFanatic]

Yes but I only hear it occasionally now, it's no longer constant. I know you can't hear it but I can! Is there some stronger settings I could use?

[pandy]

I will start from:

remove DC (high pass filter 10 - 20Hz) or center file to remove DC bias,
remove grounding loop remains (fundamental + harmonics - notch filtering),

This is first steps before any FFT processing, also block selection should be done on ZERO crossing points,

Rumble effect is in my opinion related to lossy compression - watch those spectrograms http://www.baudline.com/solutions/codec/index.html - real picture of the how noise is seen when only one frequency is superimposed with noise - imagine now complex interactions between signal and noise transferred trough lossy codec...

I can't give ready to use prescription but from my experience for such signals, shorter FFT seems to be better, overall rather lower than higher level for noise settings is recommended (something between 10 - 40% not 100% for sure), if this is transferred from VHS then there is no sens to push frequencies over 8000Hz (do some lowpass filtering, to "enrich" audio some spectrum band replication filter can be used like "crystality" in ffdshow).
Once again - rumble and chirping sounds are result for complex relation between noise modulated by sound and encoded by lossy encoder - using to strong FFT settings to remove "noise" in unavoidable way create audible problems.

[jagabo]

Yes but I only hear it occasionally now, it's no longer constant. I know you can't hear it but I can! Is there some stronger settings I could use?

Repeat the sequence.

[VideoFanatic]

I used the notch filter in Audacity to remove the buzz. I use a Q factor of 10 and applied it at 60, 120, 180, 240, 300, 360, and 420 Hz. Maybe at 480, 540, and 600 too, I don't remember for sure. Do you still hear your "rumble" in this file?

In Audacity 2.0.0 I go to Effect, Notch Filter. I see this window:



I don't see any place to enter multiple numbers?

[jagabo]

I don't see any place to enter multiple numbers?

You run the filter multiple times, once (or more) at each of the indicated frequencies.

[pandy]

For Audition you can use FFT filter and preset for 60Hz ground loop.

[VideoFanatic]

I tried Audacity using Jagabo's method. I also tried 2 passes. It still had rumble and just changed the sound of it.

I tried Audition using the FFT filter and preset for 60Hz ground loop but I got the same thing.

[jagabo]

Try a high pass filter with the cutoff set at 600 Hz, and rolloff 6 dB/octave. If that doesn't work try a sharper rolloff and/or a higher cutoff.

But since nobody knows what noise you're talking about nobody will be able to help you.

[VideoFanatic]

I also tried the previous method you mentioned on the file where you said you could hear the rumble but it didn't remove the rumble.

Try a high pass filter with the cutoff set at 600 Hz, and rolloff 6 dB/octave. If that doesn't work try a sharper rolloff and/or a higher cutoff.

How do I do that in Audacity?