Taperssection.com
Gear / Technical Help => Post-Processing, Computer / Streaming / Internet Devices & Related Activity => Topic started by: daze on January 22, 2023, 06:35:58 PM
-
I'm looking at a spectogram from a DAT tape, where the original source file was a cassette tape, and there's a horizontal line running pretty much through the whole tape (although it tapers off on the right channel after the tape flip) a smudge above 16 khz (my research indicates that this could have been caused by someone having their TV monitor too close, or something unshielded, or a sketchy soundcard). I can't figure out the best way to deal with it. In the past I've successfully used the de-hum on Izotop for hum in the 50/60 khz range, but I can't figure out how to attack this line in the 16 khz vicinity. It's not prominently loud but I'd like to figure out how to tamper it down. If anyone has any tips/advice, thanks in advance.
-
This is a job for the Spectral Denoise module.
-
This is a job for the Spectral Denoise module.
Thank you! I played around with that a little but didn't succeed in figuring out how to target the errant noise/line, I'll have to do a deeper dive.
-
This is a job for the Spectral Denoise module.
Thank you! I played around with that a little but didn't succeed in figuring out how to target the errant noise/line, I'll have to do a deeper dive.
You need to find a section of your recording where it's otherwise silent, and select it. Set Spectral Denoise to Highest quality preset, and Learn that passage. It will tell you if the selection is too short.
Then set Amount to 6.0 as a starting point, and Preview with various parts of your recording, toggling Bypass on and off, then toggling Play Noise Only on and off to make sure you're not affecting the music. Experiment with different reduction amounts, making sure you don't get any phasing artifacts.
If this 16 kHz tone is the only offending noise, you could enable the Reduction Curve and draw a very narrow curve that only affects that region.
If you're stuck on it, you could send it to me and I'll have a crack at it. I won't claim to be an expert, but I have a lot of experience with that module.
-
Thank you so much for your clear advice/instruction. Thanks to you I'm seeing some light here, and thankfully there is a few seconds I can use for the "learn" passage; using the most aggressive setting (6.0) I can substantially "fade" that line. I'll have some time tomorrow to play around more and fine tune and try to find the sweet spot where I don't end up with any phasing artifacts. The "reduction curve" isn't very intuitive so I'll have to sink into that more too so I can have it in the arsenal.
-
Not sure what you mean about level 6.0 being "aggressive".
And yes, the reduction curve is clunky, which is why I don't use it often. It was easier to use in early RX versions.
-
If it’s a very tight single frequency line, select it and gain it down. Then you’re leaving everything else alone. I would do that first even if using spectral denoise for other purposes.
-
That was essentially my first thought as well, except using a simple notch filter with a relatively high Q to target just that frequency band. Posting about that option for for those without access to tools like Izotope RX.
-
^ that would be my go to, especially if it’s very clearly a distinct freq/range
-
Not sure what you mean about level 6.0 being "aggressive".
And yes, the reduction curve is clunky, which is why I don't use it often. It was easier to use in early RX versions.
Maybe my verbiage was off, I was just thinking the higher levels are more aggressive and can lend to more artifacting.
-
If it’s a very tight single frequency line, select it and gain it down. Then you’re leaving everything else alone. I would do that first even if using spectral denoise for other purposes.
Thanks. I'll give that a try, I like having lots of options because as we all know, something that works well on one recording could be a disaster on another one.
-
That was essentially my first thought as well, except using a simple notch filter with a relatively high Q to target just that frequency band. Posting about that option for for those without access to tools like Izotope RX.
Thanks. I'm on a very steep learning curve here, and all of this advice is extremely helpful and appreciated.
-
General procedure for tuning a notch filter is to use a single band of parametric EQ. That requires setting three variables: frequency center, filter Q (essentially the bandwith of the filter, equating to how aggressively pointy or smooth and flat the curve is), and gain. Start by dialing in the approximate 16kHz frequency center, set Q to something like 5 or so, boost gain, and determine if that aggravates the problem, making it worse (checking by ear or visually). Fine tune the frequency center to home in precisely on the frequency of the problematic noise line. Boosting helps do this because it further excites the unwanted signal. Fine-tune the frequency setting until the noise is loudest (or brightest). When precisely tuned, the unwanted signal will remain emphasized even a very high Q filter setting. Now change gain to cut instead of boost. The filter is now attenuating the problematic frequency range. Further adjust Q to make the notch just wide enough to cover the bandwidth of the unwanted signal, without effecting the region below and above.
As always, listen carefully to what the filter is doing specifically, but also listen to everything else with a broader ear. It may sound best to only partly attenuate the noise signal instead of trying to remove it entirely. Make sure the cure isn't introducing unwanted problems itself. Often less is more, but it takes careful listening to be sure.
[edited for typos and clarity]
-
General procedure for tuning a notch filter is to use a single band of parametric EQ. That requires setting three variables: frequency center, filter Q (how aggressively pointy or smooth and flat the curve is), and gain. Start by dialing in the approximate 16kHz frequency center, set Q to something like 5 or so, boost gain, and determine if that makes the aggravates the problem, making it worse (checking by ear or visually). Fine tune the frequency center so as to home in exactly on the frequency as the problematic noise line. Boosting helps do this because it further excites the unwanted signal, so adjust until the noise is loudest (or brightest). When well tuned, even a very high Q will still excite the signal. Now change gain to cut instead of boost. The filter is now attenuating the problematic frequency range. Further adjust Q to make the notch just wide enough to cover the bandwidth of the unwanted signal, without effecting the region below and above.
As always, listen carefully to what the filter is doing specifically, but also listen to everything else with a broader ear. It may sound best to only partly attenuate the noise signal instead of trying to remove it entirely. Make sure the cure isn't causing problems like the disease. Often less is more.
Thank you for breaking this down so that a novice can digest and start to grasp these concepts. And I totally agree, "often less is more." Sometimes when one is learning it's tempting to get carried away with some of these tools.