|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
71.181.112.229
Wanted to post my recent discovery of using the Limiter function in Audacity to bring clipped peaks back below the 0dB threshold after normalization. Constructive criticism and alternatives are welcomed and appreciated.I use Audacity to normalize (amplify) the LP recording to levels that are closer to CD rips and digital albums so that changing between LP recordings and digital albums I don't have to make drastic changes in volume levels. I'd been using Normalize in Audacity, setting the peaks from -0.5dB to -1.5dB depending on LP mastering, so the average RMS levels are somewhat similar. Audacity has a Measure RMS tool to check sound levels so that's helpful. But in order to get consistent gain levels it involves reducing (de-amplifying) the highest peaks down to the level of the "typical" peaks, especially for LPs with lots of dynamic range contrast. It's a tedious task and takes way too much time.
Recently I discovered another new function in Audacity, Loudness Normalization, that amplifies the waveform using the perceived loudness levels of LUFS (Loudness Units Full Scale) so it's easy to normalize the files to a relatively consistent average RMS value. I found that using a LUFS value of -15 to -16 gets me closer to digital albums, which are often -12 to -14 LUFS, without pushing the gain too much. That's helpful, but albums with wide contrasts in dynamic range cause the Loudness Normalization process to run those high peaks into the clipping zone. I sometimes have to set the LUFS value at -17 to avoid clipping peaks, but even then there can be some clipping. (See below.) So I still ended up editing peaks prior to boosting gain.
LP SIDE - PRE-NORMALIZE (Avg. RMS = 22.9 dB)
.
LP SIDE - LOUDNESS NORMALIZATION at LUFS 16 with CLIPPING (Avg. RMS = 18.8 dB)
.
LP SIDE - LOUDNESS NORMALIZATION at LUFS 17 with CLIPPING (Avg. RMS = 19.7 dB)
.
However, I did some reading and discovered Audacity's Limiter function can bring those peaks back under 0dB using the Soft Limit setting. Now I just normalize to -16 LUFS and use the Limiter function to reduce the clipped peaks. See below.EXAMPLE TRACK - LOUDNESS NORMALIZATION at LUFS 16 with CLIPPING
.
EXAMPLE TRACK - LOUDNESS NORMALIZATION at LUFS 16 after LIMITER
.
EXAMPLE PEAK - LOUDNESS NORMALIZATION at LUFS 16 with CLIPPING
.
EXAMPLE PEAK - LOUDNESS NORMALIZATION at LUFS 16 after LIMITER
.
EXAMPLE PEAK - PRE-NORMALIZATION
.
I don't typically try to use the limiter with that many clipped peaks but from what I can see (and hear) the end result is basically what I was doing by manually editing peaks. This saves me a lot of time! I sometimes reduce amplitude of a portion of the waveform (i.e., a single song) before running normalization to avoid having so many clipped peaks to limit. I'm down to just a quick edit to remove needle drops and lifts, a quick Loudness Normalization step, and a final Limiter process if needed. My LP recordings are now fairly consistent in volume levels and closer in amplitude to CD rips or album downloads.Comments? Questions or clarification?
Thanks,
TomPS: I will also mention that I got tired of having to deal with clicks and pops as a separate process so I bought a SweetVinyl SugarCube SC-1 plus, which processes the clicks and pops on the fly at 24-192 and sends a S/PDIF signal to the Tascam DA-3000 for recording. I've been very happy with the resulting sound quality of the recordings. A good balance between noise reduction and ultimate transparency.
Edits: 11/10/21 11/11/21 11/11/21Follow Ups:
Hi Tom
I really wouldn't be worrying about losing "resolution" with the LP when setting the level. I would be more concerned about clipped samples and the effect of overload on the input buffer to the ADC.
The maximum dynamic range for human hearing is 20 bits. A typical consumer 24 bit ADC only has about 21 bits dynamic range and this still vastly exceeds the capability of a vinyl LP/phono stage when looking at the signal level relative to the noise floor.
I don't bother adjusting the input sensitivity per record - I set it once according to the cartridge output relative to the worst case maximum output from a test disc so that the overload margin is never hit even on clicks and pops. Typically I would look for a maximum output of -3 or -4dBFS.
I use RX so the functions have slightly different names - "Normalize" in RX is a little different to what you might expect in that it calculates the gain to hit a desired maximum level. However, I essentially do what you do (without the limiting) after editing/processing my file; I set a maximum level of -4dBFS for all my recordings so that I don't risk intersample overs when downconverting my files. I wouldn't go above -3dBFS myself, but you can choose what you want if you aren't using a DAC that does funky curve fitting.
If you really wanted to maximize fidelity, then you would transfer with flat gain and apply EQ digitally as we have talked about in the past.
Regards Anthony
"Beauty is Truth, Truth Beauty.." Keats
Thanks, Anthony,
The ability to edit waveforms to set a desired amplitude is a modern day wonder in my experience, making recording levels less critical than the old days. The only time I pay attention to input levels is if I've changed cartridges and/or phono stage settings. Using a 0.0db input sensitivity on the Tascam I typically end up with peaks at -3 to -4dB, with a range from -2 to -8dB depending on the LP.
For normalizing, I'm using the Loudness Normalization perceived loudness LUFS feature in Audacity to set average amplitude for the entire side. I do take an extra step to reduce or increase amplitude of a song or two if they are dramatically different than the rest of the side prior to normalizing.
I've been using 0.01dB as my peak limiter threshold in post-processing, just to avoid 0db glitches but perhaps I should be using a lower value. I've read that 0.3dB might be a safer margin. I haven't run into any problems with playback using two different DACs but I should check.
I record at 24-192 PCM WAV but then resample to 24-96 FLAC for NAS storage and playback, mainly because the input signal to the Tascam is 24-192 SPDIF from the SugarCube and I didn't want to use the onboard resampler of the Tascam. The sound quality at 24-96 for playback is quite good and takes up much less hard drive space than 24-192 files. I figure it won't hurt to have the finer resolution during recording in case I decide to reprocess certain files using different processes or settings or I find that some recordings warrant the 24-192 resolution for playback.
Thanks again for the comments and suggestions.
Tom
Hi Tom
I sort of do what you are doing for Normalization except I do it manually to ensure that the original dynamic range is preserved as far as possible.
I search for transient pops and clicks that would upset the actual signal peak determination and remove them, then analyse the waveform for loudness in the same way as you - the results give me all the information I need on peak sample amplitude, true signal peak, average rms level etc. With singles or LPs that have sides or tracks recorded at obviously different levels, I compare the average RMS level and then use the Gain function to rescale the tracks/sides to a similar level as the reference. Once combined, I then apply the Normalize function in RX to set my preferred maximum level.
As far as my preferred sample rate goes, I originally went for 24176 as my reference which allows a straightfoward downsample to Redbook and easy conversion to single or double speed DSD. However, I shifted to just going straight to 2496 and not fuss with DSD - I just create a DVD Audio ISO file and playback that back with a PC through my Hifi or burn a DVD if I want to give it to a friend who still uses an Oppo universal player.
I couldn't be bothered chopping up the DSD file into individual tracks!
How old is your Tascam? Sadly my DA-3000 is failing - the headphone amplifier stopped working some years ago. Now, periodically, it will stall and stop writing files to the memory card (which is fine). I bought it around 2014 and got about 5 years out of it before it started playing up. I hope yours lasts better!
I use an RME ADI-2 Pro FS now as well as a Benchmark ADC1 USB (which was bought in 2006 and still going strong!)
Regards Anthony
"Beauty is Truth, Truth Beauty.." Keats
"I search for transient pops and clicks that would upset the actual signal peak determination and remove them..."
I do something similar, i.e., reducing amplitude of the atypical peaks before loudness normalization. But with some LPs I still end up with a number of peaks that get pushed into clipping and use the limiter function to "fix" them. I'm conscientious about overuse of the limiter but I'm still learning. The example I provided in my original post was at the upper end of what I consider reasonable.
My DA-3000 is only a couple years old and I only use it for recording. So far, so good.
Tom
I was going to say that it sounds like you're compressing but Munkie_NL already gave a really good answer.
In my opinion, you don't want to compress the final product (the LP). That's already been done during the mixing/mastering process.
I don't think I'm really compressing the overall music, just the highest peak levels. I've been checking the effect of limiting on particular song tracks to make sure I'm not ending up with distortion, which can definitely be heard when pushing gain and excessive use of the limiter function. It seems to be very much dependent on the quality of the original recording, mixing, and mastering. By using the limiter to reduce the highest peak levels instead of manually reducing them I'm saving myself a lot of editing time. I think I've reached a happy medium but am continually tweaking the process. Thanks for the comment and warning.Tom
Edits: 11/10/21 11/10/21
Wow! I guess my goal is to have my digital recordings of LPs sound like LPs, not like CDs. That kind of defeats the purpose in my opinion.
I use a TASCAM DA-3000 DSD recorder and I copy my LPs in DSD128. I don't use any normalization or limiting and my DSD recordings of vinyl sound spectacular. I've had a number of people tell me that I make the best sounding recordings of vinyl they've ever heard. None of them tell me they sound like CDs. If anything, they tell me my recordings sound just like vinyl and that's my goal.
Good luck,
John Elison
"I don't use any normalization"
If you don't set the recording level so the loudest moment just touches digital zero and then you don't normalize then you are giving away bits. By giving away bits you are giving away resolution and your recording of your LPs don't (can't) sound as much like the LPs as they otherwise could/would.
If you do set the record level so that the loudest moment just touches digital zero then there is no reason to normalize the file. It already is normalized.
Normalizing does not change the dynamics. It just insures that the loudest moment uses all the bits. That increases the overall amplitude resolution.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Hi Tre',
I thought you were the professional recording engineer. Did I get that wrong? Anyway, it's not possible to improve resolution through normalization. All you can do with normalization is to increase the volume of your recording to match other recordings. Resolution is set in stone when the digital recording is made.
I used to normalize when I copied vinyl at 24/88 with my Alesis Masterlink because I wanted to match volume levels among all my recordings. Now, I use a TASCAM DA-3000, which has its own recording volume control so I try to set it high enough to capture maximum resolution without clipping any of the peaks. Of course, I always fall short by a few decibels in order to prevent clipping. However, since I now record in DSD128, that format provides ample resolution to capture everything vinyl has to offer, even when the peaks are a few decibels short of full-scale.
Anyway, I used to normalize when I used a digital recorder without a recording volume control, but now that I have the capability to adjust volume before recording and even during recording if I choose, I no longer feel the need for normalization. Moreover, it's not possible to normalize in the digital domain when recording in DSD and I believe DSD sounds better than PCM. Therefore, it's no longer even possible for me to normalize my digital recordings. The best I can do in DSD is to crop my recordings and split each LP side into individual tracks. However, that's enough for me because my digital recordings sound better than ever to my ears.
Best regards,
John Elison
"I no longer feel the need for normalization."
Ya, that's the thing, I do want my recordings to be closer to hi-res albums and CD rips. Not exactly, but at least not a 3 to 8 dB difference, which is typical. I don't mind the processing now that it's pretty efficient.
To each his own!
I prefer not to change every sample in the digital recording just to raise the volume level by 3-dB.
If anything, I would rather clip the loudest peak slightly than to recreate every single digital sample through normalization.
Best regards,
John Elison
"... recreate every single digital sample through normalization."Luckily there's an app for that. ;-)
Edits: 11/12/21
Obviously, there's an app! I just prefer not to use that app to alter every single sample by normalizing. Instead, I prefer to adjust the recording volume control on my DA-3000 so the loudest peak reaches full scale 0-dB. I believe I get a more accurate recording that way.
The only reason I used to normalize my digital recordings was because the Alesis Masterlink did not have an adjustable recording volume control. Now that I upgraded to the TASCAM DA-3000 DSD recorder, it has a recording volume control that I can set to alleviate the need for normalizing. Furthermore, I prefer to copy vinyl in DSD128, which does not allow for normalizing anyway. However, I believe DSD is more accurate than PCM and that's why I record in DSD128.
Like I said, "To each his own!"
Good luck,
John Elison
OK, then. I'm not questioning your method of recording and playback, nor am I trying to convince you to normalize the recordings in PCM, just offering another perspective. You have different priorities than I do, that's all. My priority is to record 1500 albums as efficiently as I can; having to sample and adjust recording levels for each LP is not part of that process. If I were recording the best of those LPs as cleanly as possible, trying to capture the highest resolution afforded by the recorder, I'd probably do the same as you and record as DSD128 and adjust recording levels for each LP.
The biggest difference between you and I is that I still play LPs on a regular basis and being able to record an LP while listening to it being played makes it easy. If I were trying to maximize the sound quality of my LP recordings I wouldn't be trying to record while listening in order to avoid acoustic vibrations. You don't listen (through your speakers) while recording, I do. You record to get the best possible sound quality of a subset of your LP collection, I record to eventually have all my LP music available on a NAS drive. Different objectives, different methods, both legitimate.
Best regards,
Tom
> My priority is to record 1500 albums as efficiently as I can
That's a big job and I hope you're successful. I have only 387 LPs recorded that I'm keeping and only 71 of these are in DSD. I'm sure I've copied a lot more over the years, but I guess some of my early copies are not ones I want to keep. Anyway, 1500 LPs is quite a large project. If I were you, I'd copy my favorite albums first just in case I couldn't finish. I've been copying LPs since 1991 and I have only 387 keepers.
Good luck!
John Elison
Yes, recording 1500 LPs is a bit daunting and perhaps I should prioritize my favorites. But what's surprising, is that I've been pulling LPs out to record that I might not otherwise play. It's been fun bypassing the usual favorites to find the obscure records that I haven't listened to in 15 years or more. Sometimes I'm reminded of why I don't listen to them very often; sometimes I'm reminded to play them more often. Many of the former end up in the donate pile; some of the latter go into clear plastic jackets before going back into the collection. :-)
Tom
that you do it while listening at the same time. Recording as it's frequently called, takes time, and you would have to spend a full year/full time to record 1500, give or take. You might exhibit a mental instability after a week.
Me, I just play the records because I still have them and even if I play a particular album 2-3 times a year I'm not worried about record wear. I still have records from the 70's that went thru a house fire and they still sound...well, not krinkly, but wonderful. Packing them in a rack properly at age 15 saved about half.
Records are amazingly durable if you take care of them.
I'm not worried about record wear and trying to archive them, I'm just being realistic that I won't want to move all the stereo gear and LPs and CDs when we downsize into a small house or apartment. I know I could simply subscribe to a streaming music service and not bother with recording LPs but I'm sentimental and enjoy listening to my album collection now and will certainly in the future. Plus, many of my most treasured LPs are unavailable online.
I don't worry about my mental stability. Not yet anyway. :-)
I hope you're saying that a 24 bit system, at let's say 6db down, has enough resolution to capture the information on the LP.I hope you're not saying that the system is displaying it's full amplitude resolution when the program is peaking at -6db. It isn't, it's 6db down.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Edits: 11/11/21
I thought I was saying that you cannot increase the resolution of digital recording through normalization. In other words, if the loudest peak is 6-dB below full scale, you will not increase resolution by normalizing the recording so its peak is raised to 0-dB full scale. All you are doing is raising the volume of your digital recording. If the loudest peak is 6-dB below full scale on a 24-bit recording, you have 23-bit resolution. Normalizing the recording cannot increase resolution beyond 23-bits.
Best regards,
John Elison
"If the loudest peak is 6-dB below full scale on a 24-bit recording, you have 23-bit resolution. Normalizing the recording cannot increase resolution beyond 23-bits."
So when you normalize from -6 to 0 db the noise floor comes up 6db?
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
If you believe that normalization will increase resolution, try recording with the peaks 90-dB below full scale on a 24-bit recording and then normalize that recording so its peaks reach 0-dB full-scale. See where you find the noise floor on that one.
Good luck,
John Elison
I'll take that as a yes.
Thanks, I guess I never understood that. Well, all the more reason to make sure to record as hot as you can without allowing digital clipping.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
In the old days, when recording LPs to tape or to a CD recorder, the recording levels were critical in getting the highest resolution from the LP. But now, with digital editing, it's made recording LPs so much more convenient. I no longer have to futz around with setting recording levels for each LP. I may be old, but I do enjoy the digital age.
Toom
I hear you, John, I don't want my LP recordings to sound like compressed digital music. But I don't consider all CD releases to be inferior to vinyl. I have some excellent sounding music ripped from CDs and played through a Node 2i bridge and good DAC. My goal is to have the vinyl recordings be closer in volume to CD rips and digital albums so that I don't have to make such big adjustments in volume levels when switching between albums.
I'm very impressed with the quality of the recordings I'm making using the DA-3000. So much so that more than a few times I've been listening to LP recordings through my Node 2i and thought they were really high quality, hi-res releases I'd purchased. I love listening to vinyl but there's some great music available as hi-res downloads and CD rips.
Tom
Hi Tom,
I own over 40 hi-res DSD256 recordings and they're my favorites. Of course, I own some 16/44 CDs that sound awfully good, too. Anyway, I know what you're talking about. I've been listening to digital exclusively for nearly two years. I might have to connect my turntable again because my daughter wants me to copy some LPs for her. I just gave her a FiiO M11 Pro digital player with a pair of Dynaudio Bluetooth speakers and she loves her new system.
Best regards,
John Elison
Hi, John
Kids these days are able to set up a stereo system so easily. No turntables or tape decks, no amplifier, no interconnects, no media storage. :-)
My sister bought a FiiO M9 for travel and loves it.
I have been writing and recording songs in my home studio for the past 6 years or so. Using Reaper with plugins and Audacity. I use a stereo bus compressor on the stereo mix in Reaper and a loudness plugin which is basically a limiter. In Audacity i use normalization. IMHO you have to use limiters with some restraint, they drastically cut off peaks. Listen to snare drums and cymbals. When the sound gets too compressed it's not nice on the ear, think of the loudness wars since the early 1990s in popular Western music.
"The torture never stops"Greetings Freek.
Have you compared the results between using your loudness plug-in and the limiter function in Audacity? Or are they completely different functions, i.e., the plug-in reduces amplitude before it clips whereas the limiter in Audacity "fixes" clipping?Tom
Edits: 11/12/21
Thanks for the insight. I'm getting pretty savvy of how pushing gain too much can quickly get you into distortion. I'm being fairly judicious with the limiter function but am still getting a feel for how much to normalize and using the limiter to reduce clipped peaks depending on the type of music. I have looked at the waveform of many CD tracks that have been compressed to absurd levels so I have an appreciation of the problem. Even some of the LPs I have show signs of excessive gain and the resulting degradation in sound quality. That and songs that were poorly recorded in the first place.
Tom
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: