50 Year Old Analog Audio Noise Removal Issue.

I have been working on cleaning up a 50 year old analog audio and have had great success, until I reached this small section with piano? I don’t seem to be able to remove the hum from the background, regardless of the adjustments I make with the “Noise Removal” tool, unlike previous sections. Is this section damaged beyond repair, or is there an additional adjustment that I am not aware? I have uploaded both samples of Before and After applying the “Noise Removal” filter for your review. Thank you for whatever input and instruction that you can provide in this matter.

Thank you!
Buffalo Bill

One of the rules with Noise Reduction is the show has to be louder than the hum. So that’s a problem.

We have an experimental hum filter here somewhere that automatically pulls out the hum tones normally experienced in, for example, bad cable hum or power supply defect. Conventional problems.



While we’re looking, you’re using Noise Reduction wrong. Make sure you select a Profile from a portion of tape that has no other performance in it. Nothing. No singing, no voice, no background sound, just hum. Then back off on the correction dB number. Start with a small number like 6 or 12 and work up. You’re never going to create studio conditions in software, but you can help.


Scroll down to the Nyquist listing. Third message?


One of Steve’s de-hum plugins is attached to this post … https://forum.audacityteam.org/t/hum-removal-plug-in/13561/1

Or alternatively my dehummer plugin … https://forum.audacityteam.org/t/multiple-notches-with-nyquist-prompt/7583/10 [which is more suitable for extreme cases].

the settings used on my dehummer plugin.gif
Unfortunately the de-hum process does add artifacts and removes some music along with the mains hum harmonics.

I was able to download and install both Trebor’s de-hum and Mains dehummer 2-0. Plugins. I found Trebor’s listed as 50/60 Hz dehumer 2.0 from the Effect drop down. Are both of these files the same, I cannot seem to see a listing for a second filter? And if not, what is the file name assigned by Audacity, so that I may locate it as well. My first track pass I used 1 for Mains Freq. 15.0 Amt. of Hum Removed, and 20 Anti-reverberation. I increased only the Amt. of Hum Removed by gradually increasing the amount by a factor of 5, until I had doubled it to 30.0. This appears to have substantially removed enough of the noise and accomplished and clear audio. Hopefully I understood and used the filter correctly and I think I did?

Thank you Koz

I think Steve’s is named “Hum Remover”, you do have to unpack the zip file to obtain the text file with the name “de-hum.ny” , it’s this “.ny” file that you put in the plugin folder , ( if you put the zipped version, “.zip” , in the plugin folder it won’t be recognised by Audacity and won’t appear in the effect menu ).

If you increase the amount of hum removed from the default setting it has a side-effect of cutting back the higher frequencies.
On this occasion you can get away with that because your recording only contains noise above about 7KHz, but personally I use that dehummer plugin on default settings. [then in this case, cut off everything above about 6 or 7KHz with the equalizer].
Spectrogram View shows no piano music above 6KHz , so frequencies above 6kHz can be removed it is just hiss noise.gif
de-hum.ny (2.17 KB)

Understood. I have updated my personal documentation with your notations regarding using the default settings, and the issues that could arise if the noise or hum was above 7kHz. As you said, “cut off everything above about 6 or 7KHz with the equalizer” I still need to increase my audio wave form to match the rest of the others, but I like what I hear after running 50/60 Hz dehummer 2.0 filter. I’m just not clear on how exactly to accomplish that? I need to decrease the volume, but increase and maintain the integrity of the wave. The previous file called “After 50-60 Hz dehumer 2.0 Recording 1A Count 670.mp3” that I attached as sample from your last review.

I was able to locate Steve’s filter named “Hum Remover”. I might have just overlooked it since I have most all of the plug-in loaded in my Audacity App. Now that I know what it’s called and looking for, if you have any basic rules that I can add to my documentation, that would be cool too!

I’m also now working another section that I have previously removed some noise by using the “Noise Removal” filter on complete track, and I used the “Normalize” feature twice to increase the over all volume of the track. The section of audio is clear but a little too much loudness in the bass. Here is a brief excerpt for you, if you have any suggestions… I’m All Eyes! l.o.l. Oh… and ears too! (:

Thanks again!
I’m learning so… much!

I personally didn’t say anything about noise above 7KHz.

I just suggested Noise Removal might not be the best at removing the particular hum you have and that I knew we had intentional hum removal tools around somewhere.

Then I ducked.

A note: You can’t actually hear either 50Hz or 60Hz hum. They’re both down near the limit of human hearing not to mention speaker and headphone limits. However, the damage is never from clean power. The interference is always horribly distorted due to lamp dimmers, electric motors, dirty connections, bad extension cords, etc. etc. That’s what gives you the classic harsh buzz that’s so hard to suppress. That can produce interference tones in the same range the human voice uses.

Run away.



That was me (Trebor) who said that : there is only hiss above 6-7kHz on your recording, no music, that is consistent with a recording made with amateur equipment 50 years ago. You could even remove even more : maybe even down to 5kHz in this case.

Try changing the normalize setting to “0dB” : that will make the track as loud as possible without resorting to a limiter.
normalize to 0dB makes the track as loud as possible (without resorting to dynamic-range-compression).png

Another one of Steve’s plugins called a limiter increases the level … http://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#[b]Limiter[/b]
try applying it (maybe repeatedly) on default settings and the amplitude (level) of the waveform will increase …
repeated applications of Steve's ''limiter'' plugin on default settings.gif

My bad. First thank both of you, and sorry Koz for misquoting you and not attributing it to Trebor. If I misattribute one of your comments again in the future, hopefully I won’t? It was only that I wasn’t paying enough attention to who was commenting. I’ll try to keep it sorted out on my end, but again… thank you both so… much for your help on this project.

Trebor, I think I can live with the results from using “Hum Remover" setting at the current level achieved. I understand that because of the type of recording it is, that I could remove more but I agree with Koz that it isn’t necessary in this instance. I first tried using the “Normalize” filter setting to “0dB” as Trebor suggested, but that increased the volume and wave form too much. So I performed an “Undo” and attempted to gradually increase the wave and audio with the Steve Daulton “Limited” as you suggested Trebor. That did the trick! I was able to increase it to match the other sections of the recording. I’m already imagining future final issues with this project, such as making all of the linear tacks consistent with one another. Not to mention how I’m going to compress and make this file small enough to share. I’m guessing this thing will have to be burned to a storage device like a thumb drive, CD, or sent to the cloud? But I’m a long way off from that step. Hopefully you will indulge me in giving you just a little history behind this project. A friend of my Bryan came to me a month or so ago with an Old School RCA Model 1YB29A original four track cassette tape recorder. I was able to pull all four tracks which have the capacity to store a half hour per track. Fortunately there are large gaps on tracks 2-4, but this first track 1 has quite a bit of information on it. So… I don’t really have two hours if recording that I need to clean up, whew! But my end goal game plan is to stich all the tracks together and make one or two files, part one or part two? But it would be way cool just to make it all one recording, so my friend and his family can just listen to the entire recording in a linear fashion. But… like I said, not sure about that step, and besides I’m getting ahead of myself here.

So as I alluded to in my previous post in the third paragraph, I have included another sample, this time of the end of a section this is pretty distorted. (Note: Previous Post Comment.) “I’m also now working another section that I have previously removed some noise by using the “Noise Removal” filter on the complete track, and I used the “Normalize” feature twice to increase the overall volume of the track. The section of audio is clear but a little too much loudness in the bass. Here is a brief excerpt for you, if you have any suggestions… I’m All Eyes! l.o.l. Oh… and ears too! (:” So… is there anything I can do to correct it and bring it back down from peaking at 1000 – 2000 Hz range to an acceptable level? As you can hear from the sample, the distortion is too noticeable.

Again, thank you both for your input and the education you are providing me. I have a professional CD out that I recorded back in 2011 in the studio. If you guys like one of my tunes, just let me know OK? I will send you a free access code to my site for a download. That is… if you like this old guys stuff? I call my style Folk/Rock with R & B influences. So if either of you are interested check it out or let me know, OK?

Thanks again!

PS: This is my favorite piece on my CD. So hopefully you both will enjoy this simple gesture of gratitude from your humble student? I really do appreciate all the hard work engineers do to enhance the musician’s sound. You guys really are the true unsung hero’s working behind the scenes. This front man just wanted to take a moment to thank you both. Now… just in case you have issues with these codes, let me know… OK? Like I said, I’m an old hippy guy, and my music is not well known and pretty obscure. This is actually the first time I’ve even tried send a free access code to someone. I’m old school in that I typically just give someone the MP3 file and let them deal with how they want to store it. So… good luck, let me know what you think too, I don’t get enough critique on my work, for all I know… I really SUCK! But that’s OK, because I love what I do and just hope others will enjoy it. :slight_smile:

Koz – Here is your code.
Gonna Be Alright
Use code: m3id-uavy

Trebor – Here is your code.
Gonna Be Alright
Use code: fl7k-2hyz

Normalize to 0db is maximum loudness , if that’s too loud replace the “0” in normalize with a negative number like “-3” , “-6” is quieter still , etc.

Audacity spectrogram of ''Copy After Noise Removal Recording 1A Count 670''.gif
The loud resonance peaks in that are dynamic : they change frequency with time, so a constant equalization can’t correct them. I believe expen$ive $oftware exists with dynamic equalization which analyses the sound and modifies the equalization around a hundred times a second to correct any moving resonance peaks, but Audacity as shipped won’t do that.

Wow! Thank you again for your private message Trebor, I will check it out for sure. I need to find the legal document and read it to see what it all entails? As always, I appreciate the information and help you and Koz have provided me. I do hope you liked the song in general, since I’m not sure what style of music you personally prefer. I do have a couple other songs that are a little more upbeat. I just glad that you were able to use the access code, before someone on the public side of the form saw it and used it. Then you would have really thought me a dunce. That was a rather ignorant move on my part, but I’m still surprised how many other Audacity users, were interested in reading about my little geeky issue.
I have not heard back from Koz regarding the download I gave him like I did you. Again, someone may get the bright idea to check it out before he sees it? And hopefully he’s just been busy and I didn’t freak him out? I just really appreciate all the help so far, and I pretty sure as I continue to work on that old recording I’m going to get stuck again. l.o.l. I think I going to have to live with that last sample I sent you, I don’t think dialing it back will “Normalize” will help? As you suggested, “Normalize to 0db is maximum loudness, if that’s too loud replace the “0” in normalize with a negative number like “-3” , “-6” is quieter still , etc.” But using a negative number increased the wave sign, but the audio just got louder and more distorted. I think I have to agree with your statement, “The loud resonance peaks in that are dynamic : they change frequency with time, so a constant equalization can’t correct them. I believe expen$ive $oftware exists with dynamic equalization which analyses the sound and modifies the equalization around a hundred times a second to correct any moving resonance peaks, but Audacity as shipped won’t do that.” Audacity is pretty awesome, but I know it has its limitations for now. I do also have Sonar X1 LE that came bundled with my external Roland Octa-Capture soundcard, but the Cakewalk folks who market Sonar sort of pissed me off. I no sooner loaded that software on my other PC, and they upgraded and improved it to Sonar X3 and I would need to pay more to upgrade it. I decided that since I’m still too much of a novice with mixing and mastering, that Audacity would be a better tool to learn on anyway. You and Koz are one of those reasons, I think you guys do a better job of supporting this fee open source software, compared to some of the other closed source apps. But I do like the Roland Octa-Capture soundcard, and since I primarily use Boss and Roland equipment, I had hoped that by using the bundled Sonar software, I might avoid other proprietary issues? So unless you are familiar with Sonar X1 LE and any onboard “Normalizer” or “Limiter” filters with that software that might help, I think I’m just going to have to live with what I was able to clean up.


Wow! Now this is freaking me out? This next section the father was first talking and then his son sings for a little bit. At the start his dad is a muted a little, then his next comment clears up, and then the son sings and there are background overtones that I can hear and see funky Hz range peaks. Why I’m perplexed is that this is all from one original take, so… why it did not all record evenly the same is a mystery to me? You guys are much better at reading the EQ than I am, as a matter of fact. I really need to 101 on understanding Hz an k thresholds and peaks much better. I have a general understanding and I comprehend the subject matter. But… execution is where I’m still very weak. Knowing what to do and how to resolve it in any given situation, is an art that both you Trebor and Koz are way, way more proficient than I feel I’ll ever be? Hopefully I’ll get to be half as good as either of you, but I really do need some guidance with this sample. Take a listen and let me know what you think.

PS: I have already run the Normalize feature twice to increase the overall volume of the track. I then used Steve Daltons “Limiter” three times to increase the volume and wave form, and I Normalized it again. It is a little bit better and louder, but now I can hear a low level distortion of some type above 5k in the 1000 Hz range I think? Again, take a listen to the sample I have included and let me know what one of you guys think.

That sounds like you’ve set the noise reduction too high : it has removed too much from the recording causing distorted speech.

Noise reduction adds processing artifact noises, as a result processed speech can sound computer-generated / robotic , the higher the amount of noise reduction applied the more artifact noises are created. The general rule with noise reduction is use as little as possible.

If you post a few seconds* of the “family recording” before you applied any noise reduction or any other processing, i.e. the raw capture from the tape, we can suggest settings for Audacity noise reduction.

[* Choose an excerpt which includes a second or so where no-one is speaking, which would be silence if not for the presence of tape hiss , we need that “silent” bit to use as a noise-profile ]

I do understand what you’re referring to, and that was one of my thoughts and concerns regarding this section of the recording and over usage of filters. I have taken a sample from the original raw master recording without any noise reduction effects for your review, and it includes a couple of seconds of silence at the beginning as you requested. My strategy here is to edit this section back into the overall edited copy, once I have cleaned it up. Once I have completed the entire recording, then I will do the finale mix down and file rendering. Otherwise, I fear I may have over edited that aspect of this one section in that post edited version? So hopefully by starting from scratch, it will be a better solution rather than attempting to undo any editing mistakes I inadvertently made? Unless you feel it would be better to stay consistently true to the overall recording and use my first and second generation edits? So with that said, here is that sample, and hopefully my logic is sound in attempting to resolve these issues? Like I said before, initially I applied the “Noise Reduction” to the overall recording, and then I used it again while working on just that section and I agree with you that I over used filters. This is the first instance where I have encountered this issue, and hopefully it will be unique and isolated to just this section? And now after listening to the master, I hear hum that might have been better addressed with the 50/60 Hz dehummer 2.0 filter, instead of the “Noise Reduction” that I applied to the entire recording, and then again using the “Noise Reduction” and “Limiter” filters to this section in my secondary edit of this section.

Thank you again!

Hum is the most obvious problem on that and best dealt with using de-hum plugins and/or other notch filters.
Hiss is the kind of problem noise-reduction was designed to attenuate with.
The hiss on that recording is already quite low, I’d be tempted not to use any noise reduction, however a small (3dB) amount of noise reduction seems to be the threshold , beyond that IMO it does more harm than good in this example, see my suggestions below , (click on the image to see it in it’s entirety) …
Filter sequence suggestion for ''Original Master No Edits of Family Recording 1A Count 670(1)''.png
If normalizing to -1dB is too loud you need to turn down Audacity output volume …
if -1dB normalization too loud, then turn down Audacity output volume.gif
-1dB is a typical peak volume level for sound files you’re going to encounter on the internet.

Trebor, Part 1.
First of all let me thank you again for all of your assistance helping me with this project. Your reply to my last post was huge. I was able to use the workflow protocol you provided to clean up five more sections. I attempted to use that same protocol with this next section of organ music and vocals, but I was not able to eliminate enough of the hum. I’m not sure if I’m need to use a higher setting, or if because of the different environment this section was recorded in is giving me new additional issues? It sounds like this section was recorded in a church, where the previous tracks were in a home. But before I started playing around with settings too much, I thought it best I’d send you samples so you can hear for yourself. I don’t see where or how I can send multiple MP3 files with this forum utility, so I’m just going to post two comments back to back, with the second comment noting as such with the second file. The first MP3 file, “Copy Orig. Master NO Edits of Family Recording 1A Count 670.mp3” is the original raw master. The second MP3 file is after I applied your previous workflow protocol that you sent me. This way you have a before and after comparison that you can analyze.
Also, why I’m thinking about it… I was trying to find something in the Audacity help section that describes the “Snap to” function, and how to use it. It would same me a lot of time, if I did not have to scroll back and forth from beginning to end of the track section I’m working on. I saw a YouTube video awhile back and that instructor talked about it, but I don’t fully understand how to execute that short cut. I would really appreciate any instruction you could provide on that edit technique.
So… with that said, I will let you take a listen to my two samples and wait for your reply.

Thanks again!

Trebor, Part 2.
Here is part two of the previous message with the second file “Copy2 After Used Normalize 50_60 Hz dehummer Notch Filter Default & Freq. 5,275.0 & Q Value15.0 Family Recording 1A Count 670.mp3” As the file name states, this is the post production or your workflow protocols.

The tape is running slow on that recording : I can tell than because the 60Hz mains hum fundamental frequency is about 58.6Hz on the recording. So the standard de-hum plugins which only offer 50/60Hz options aren’t going to work properly on recording with a 58.6Hz + its harmonics.

You can adjust the speed of the track so the 58.6Hz becomes 60Hz , then apply the 60Hz dehummer.
try increasing speed by 1,5 percent  .gif
If the tape speed is not constant throughout the recording the standard dehum filters aren’t going to work as well because the hum harmonic frequencies the dehum removes are not where they should be …
Spectrogram shows 60Hz mains hum harmonics are lower than they shound be and a bit wobbly because of tape speed variations.gif