Brian Davies DeNoise help needed--Waxcylinder??

[ignatz]

I’ve just begun a 21 day trial of Brian Davies package of applications and am having a problem on a specific point concerning the DeNoise application.

The problem concerns selecting sound samples by mouse–typically a very brief portion of the lead in or lead out grooves of a vinyl or shellac recording.

In Audacity, one would zoom into the waveform a bit to magnify/elongate the noisy portion and then select it with the mouse. Then get the noise profile. Then select the entire song. Then use the sample to process the song.

However, in DeNoise, I see no way to zoom into the wave form, which means the lead in area is displayed on screen as no more than a tiny fraction of an inch in length–making it near impossible to see let alone accurately select just the noise with the mouse.

For that matter I don’t see a way to quickly navigate around in the waveform, largely because of the lack of zooming.

I can’t believe he left this out of the app. So I must be a fool for not figuring out how to zoom or otherwise quickly and accurately select minute portions of a song.

Can anyone help me out?

[waxcylinder]

I don’t have a license for DeNoise (only for ClickRepair). I did experiment a while back with DeNoise (under its 21-day free-trial) when I was digitizing an old blues compilation LP that had been transcribed from 78s to the vinyl.

I don’t recall having to select any samples for noise reduction, just some parameters that could be tweaked. It was not as straightforward to use as Brian’s ClickRepair - but with a little experimentation I got some quite good results - but the parameters needed re-tweaking for each track individually as the 78s that had been used for the ssource material were obviosly different in their noise levels and noise properties.

Does Brian’s manual help at all- see: http://www.clickrepair.net/downloads/DNManual.pdf - otherwise try dropping him a line via email - I have always found him extremely helpful.

Update: I’ve just had a look at Brian’s manual and p14 seems to show what you need.

WC

[ignatz]

Waxcylinder:

Thanks for the response.

I emailed Brian before I even started this thread and he did respond.

He confirms that zooming is not possible.

I have been through the manual you mentioned with a fine toothed comb. Page 14 is the relevant portion as you mention. Notice in the screen shot on page 14:

The wave form is shown in that narrow window at the top of the screen. That window holds 12 seconds of the song in question, left to right. You cannot zoom into it.

The problem is that my lead ins are already cropped to a degree–usually under a half second, sometimes a quarter second.

Since you cannot zoom in, you can’t tell by eye where the groove noise likely stops and the music begins. You have to fumble at it with the mouse and the lack of zooming makes it problematic–on my screen the noise may occupy only 1/16 or 1/32 of an inch wide, left to right.

No, you don’t have to sample the lead in, but he recommends that as best policy. You can’t sample within the song as you would nearly always get some musical content.

The ClickRepair app seems to work pretty well at default settings for removing “frying bacon” crackle.

My offhand opinion at this point, after playing with DeNoise, DeNoise LF, and ClickRepair is that the latter may be the only one preferable to Audacity.

I did considerable fiddlling around with sample files yesterday. Results indicate DeNoise is likely not preferable to Audacity and is considerably more difficult to use, primarily because of the sampling issues indicated in this thread.

DeNoise LF definitely works very well to remove rumble and boxiness, but I’m unclear on whether it is preferable to Audacity’s high pass filtering.

I have just begun to explore ClickRepair and have used only default settings. I’d like to know what settings you use on that app. I’ve tried only “Default 78” at this point. I have not yet critically examined your sticky in this forum.

Thanks for any input.

[pdxrunner]

Hi,

I regularly use all three of Brian Davies’ tools. My usual workflow is export each LP side from Audacity as 32bit WAV (for further processing) and as a FLAC for archiving the raw recording. Once I have the WAVs out of Audacity, I run them through deNoiseLF, ClickRepair, and deNoise (in that order). I’ll them get back into Audacity and import the final WAV from deNoise for any manual cleanup as needed and for track splitting and final export of individual songs into MP3 or AACs (depending on destination.)

I generally haven’t had a problem in deNoise selecting a bit of the recording’s lead-in between the ‘startup’ distortion of setting the needle down or starting the TT and the beginning of the real ‘show’. However, one trick that I have used (very) occasionally is to select and export the bit of lead-in audio from the raw Audacity recording where it’s easy to zoom in. Export it in the same format and bit depth as the real LP side. (Don’t forget to call it something different ) Now, in deNoise, first open the lead-in WAV and use that to sample the noise. Then open the real show and process it without changing the sampling, which will remain ‘sticky’ from the short selection.

PDX

[ignatz]

PDX:

Thanks for that.

I guess your method of sampling in Audacity and exporting that to DeNoise would be a workaround.

However:

How, if at all, do you feel DeNoise is superior to Audacity as a “hiss” and “groove noise” reducer? I"m not referring here to clicks and pops or crackle.

I may buy ClickRepair, but my very limited experience with DeNoise doesn’t show it to be better than Audacity for general hiss reduction and it is certainly clumsier for that purpose.

Any comments?

FYI, my source material is 90% pre-1960 2 channel mono recordings, usually from a high quality source, but usually with a very brief lead in that is hard to grab with a mouse in DeNoise.

The stuff that is ripped from commercial CDs generally does not need to be further processed, but a certain percentage of what I have is direct from 78s and 45s in someone’s collection and therefore can use some processing.

I think you have the right workflow (DeNoise LF, ClickRepair, DeNoise, then Audacity for clean up and export) based on my understanding of Davies’ manuals.

[waxcylinder]

@ignatz

I have been through the manual you mentioned with a fine toothed comb…

We like people that actually read manuals …

@ignatz

My offhand opinion at this point, after playing with DeNoise, DeNoise LF, and ClickRepair is that the latter may be the only one preferable to Audacity.

I suspect that this may be a reasonable analysis – I’m interested to see PDX’s reply though. Note that there is some experimental development work on improving Audacity’s noise removal tools by Marco Diego Aurélio Mesquita – these are not available in the Beta 1.3.12, but are in the 1.3.12 Alpha nightly builds. Marco would be delighted if you would like to test these. See this thread on the forum: http://forum.audacityteam.org/viewtopic.php?f=28&t=35900&start=0&hilit=marco

@ignatz

I have just begun to explore ClickRepair and have used only default settings. I’d like to know what settings you use on that app. I’ve tried only “Default 78” at this point. I have not yet critically examined your sticky in this forum.

Like PDX I run Audacity at 32-bit floating and export a 32-bit file to ClickRepair for processing and re-importing back into Audacity (actually I changed to that after a steer a while back from PDX).

The defaults I use for ClickRepair are: De-Click=30 (I found the default 50 took out a little too much of the real signal), De-crackle=off, Automatic=All (it works so well I can no longer be bothered to intervene), Pitch Protection = on, Reverse = On (this does not seem to impair processing time – and is helpful in avoiding false clicks with sharp attacks, such as percussive noises), Method = Wavelet.

@ignatz: you may be interested to read this workflow that I wrote for the manual a while back (with contributions from PDX among others): http://manual.audacityteam.org/index.php?title=Sample_workflow_for_LP_digitization

And this one on 78s: http://manual.audacityteam.org/index.php?title=Recording_78rpm_records

@PDX: cute work-around for getting the noise-print, I like it

WC

P.S. Off-topic @ignatz: you’re not by any chance a Krazy Kat fan are you?

[billw58]

Just thought I’d jump in here and point out that DeNoiseLF uses a completely different method for noise removal than DeNoise. Like others here, I’ve found that it’s a toss-up between DeNoise and Audacity’s noise removal. But DeNoiseLF is a completely different story. It can remove noise elements below 200 Hz much more effectively, and with minimal damage to the real audio, than DeNoise or Audacity. For that reason alone, I bought a licence for DeNoise, just to get DeNoiseLF.

– Bill

[ignatz]

@waxcylinder

The defaults I use for ClickRepair are: De-Click=30 (I found the default 50 took out a little too much of the real signal), De-crackle=off, Automatic=All (it works so well I can no longer be bothered to intervene), Pitch Protection = on, Reverse = On (this does not seem to impair processing time – and is helpful in avoiding false clicks with sharp attacks, such as percussive noises), Method = Wavelet.



P.S. Off-topic @ignatz: you’re not by any chance a Krazy Kat fan are you?

Yeah, I like that Krazy Kat, but I don’t actually throw bricks.

I just looked at the ClickRepair dialogs and noticed that if you choose “default 78” in the dropdown list and then make the choices WC suggests above, that the “default 78” choice is over-ridden and the dropdown shows only --------. This is obviously by design and shows that the choices are inter-connected.

WC: do you use your stated choices in ClickRepair regardless of source material? The same settings for a 1940 Artie Shaw 78, a 1957 Chuck Berry 45, and a 1987 LP reissue of Artie or Chuck?

Do you pay any immediate attention to the condition of the disc?

Or so you just use those settings in all cases for all source material and then re-examine after the fact?

I agree with BillW58 about the usefullness of DeNoiseLF. Someone just sent me an CD containing transfers he had made of some late 1940s 78 RPM discs. The guy knew nothing about how to do it and used a 78 RPM turntable with a 33 RPM stylus to make a cassette of the discs and then transferred the tape to CD. The surface noise was horrendous due to the wrong stylus. I was able to to use DeNoise LF to get rid of a lot of rumble quickly.

I do like the batch processing found in the Davies applications. I have not yet tried batch processing in Audacity.

Has anyone found an app that has much success removing “swish”–that once per revolution high frequency hiss that comes and goes with each rotation on certain 78s–typically a bad pressing? I’m guessing it is all but impossible to treat.

[pdxrunner]

@ignatz

PDX:
How, if at all, do you feel DeNoise is superior to Audacity as a “hiss” and “groove noise” reducer? I"m not referring here to clicks and pops or crackle.

I may buy ClickRepair, but my very limited experience with DeNoise doesn’t show it to be better than Audacity for general hiss reduction and it is certainly clumsier for that purpose.

Any comments?

I’m not sure if deNoise is any better than the current Audacity noise removal, but I really do like the workflow better. What’s great about deNoise is that you can listen to noise only, input , or output while it’s processing. You can also change the settings on the fly and listen to the effect. Once you’ve got the settings you like for that one WAV, then click Abort, and Restart from the Beginning (or whatever that button label is). Turn off all sound for the final processing and it zips through pretty fast. Specifically what I do after doing the noise sample, is turn noise-only sound on, crank the output volume way up, and restart deNoise. I listen for anything musical or rhythmic mixed in with the noise, and adjust the noise-floor slider until I can barely hear any noise. I’ll check what’s getting removed visually by pausing the tool and looking at the display. You really have to read and understand the deNoise manual to make sense of the display. Once I’m satisfied with the settings, I restart the tools without changing any settings. Throughout, I almost always have “Limit Reduction” slider at 9 dB (default was 12 IIRC)

I usually run my needledrops through deNoiseLF and ClickRepair in batch mode (as long as everything in the batch was from similarly produced records and that I recorded similarly with Audacity.) I’ve never run anything through deNoise in batch, because the noise characteristics always seem a bit different, even the 2 sides of a single record (maybe one side got played a lot more.)

One note on my noise sampling ‘trick’. Always put the noise sample WAV through the same processing that you do on the real show. In other words, I would deNoiseLF and ClickRepair the noise WAV before using it to sample in deNoise. You want the sample to be an extension of the real recording.

PDX

[waxcylinder]

@ignatz

I just looked at the ClickRepair dialogs and noticed that if you choose “default 78” in the dropdown list and then make the choices WC suggests above, that the “default 78” choice is over-ridden and the dropdown shows only --------. This is obviously by design and shows that the choices are inter-connected.

Yes indeed Brian’s setting for 78 (or any other of his default settings) work that way - once you make changes it is no longer his default, and thus rightly should not maintain the same profile name. You can create your own variant as custom setting - see the “Custom” selection at the bottom of Brian’s drop-down list.

@ignatz

WC: do you use your stated choices in ClickRepair regardless of source material? The same settings for a 1940 Artie Shaw 78, a 1957 Chuck Berry 45, and a 1987 LP reissue of Artie or Chuck?
Do you pay any immediate attention to the condition of the disc?
Or so you just use those settings in all cases for all source material and then re-examine after the fact?

Basically yes. I did a lot of experimentation and careful listening tests to determine the settings I wanted. And as PDX points out one of great features of CR is that you can use it to listen to the input signal, the output signal or just the “noise” that is being removed - this helps a lot.

I don’t have any 78’s anymore – I used to have a good collection but my parents threw them out when I left home . But I do still have a lot of 45s some of which had a very hard life on my old juke-box and some of my LPs suffered from damp storage in a house I lived in a while back – and for all my vinyl the settings I described work fine.

WC

[ignatz]

@pdxrunner

Specifically what I do after doing the noise sample, is turn noise-only sound on, crank the output volume way up, and restart deNoise. I listen for anything musical or rhythmic mixed in with the noise, and adjust the noise-floor slider until I can barely hear any noise. I’ll check what’s getting removed visually by pausing the tool and looking at the display. You really have to read and understand the deNoise manual to make sense of the display. Once I’m satisfied with the settings, I restart the tools without changing any settings. Throughout, I almost always have “Limit Reduction” slider at 9 dB (default was 12 IIRC)

PDX

PDX:

A couple of questions about your technique:

1: You state “You really have to read and understand the DeNoise Manual to make sense of the display”. I notice that most of the application’s screen space is devoted to Davies’ graphs. This is quite different than Audacity, where I pay attention only to the Waveform left and right channels. Do you adjust your Davies’ settings PURELY based on your ear, or do you in fact let what your eyes see in Davies’ graphs affect your settings?

Offhand, the graphs seem to be a distraction and waste of screen space, to the extent that slider settings and button choices are ultimately driven by ear not eye. I don’t want to get bogged down in trying to make sense of the graphs by eye if in the real world the ear controls all and over-rides the eye.

It appears you operate by ear not eye as well? Do you ever adjust anything based on what you see as opposed to what you hear?

2: You state that you leave the limit reduction set at 9, not the default 12. I am guessing this control is very similar if not identical to the “Noise Reduction (db)” slider in Audacity’s NR menu? In Audacity, I leave this at the default 12 well over 90% of the time. I will occasionally set it higher for a particularly rough 78. And I will occasionally do one pass at 12 and a second pass at 12 or less. You may be enough younger than I that your ears can detect unwanted effects with a slider at 12 that I cannot detect.

As an aside:

In Audacity, do any of you typically set the other 2 NR sliders (frequency smoothing and attack/decay time) at anything other than defaults? I have done very very little experimentation with these sliders.

Any further comments appreciated.

[pdxrunner]

@ignatz

You state “You really have to read and understand the DeNoise Manual to make sense of the display”. I notice that most of the application’s screen space is devoted to Davies’ graphs. This is quite different than Audacity, where I pay attention only to the Waveform left and right channels. Do you adjust your Davies’ settings PURELY based on your ear, or do you in fact let what your eyes see in Davies’ graphs affect your settings?

By that statement I simply meant that the graphs can be a bit confusing if you don’t understand what Brian was trying to show.

@ignatz

It appears you operate by ear not eye as well? Do you ever adjust anything based on what you see as opposed to what you hear?

I operate mostly by ear. I pause and use the graphs to check visually what I believe I’m hearing. Mostly I use them to check to make sure that I’m not removing too much through the noise reduction. When I pause deNoise, and then select a portion from the wave display at the top that contains music, not just silence, I expect the graphs to show a bit at the high end of the freq. spectrum to show some de-noising, a possibly a bit at the lowest freqs as well. I do not want to see a bunch of ‘red’ on the graphs in the middle frequencies; which might mean I’m mucking with the music, and not the noise. It’s kind of a checks and balances process I follow. It’s not and exact science, but more of a gut feel. There are a few records I’ve done that later on I’ve thought I was a bit heavy handed on the noise reduction, and others where I thought my ears must have been asleep because I should have been able to do better.

PDX

@ignatz

You state that you leave the limit reduction set at 9, not the default 12. I am guessing this control is very similar if not identical to the “Noise Reduction (db)” slider in Audacity’s NR menu? In Audacity, I leave this at the default 12 well over 90% of the time. I will occasionally set it higher for a particularly rough 78. And I will occasionally do one pass at 12 and a second pass at 12 or less. You may be enough younger than I that your ears can detect unwanted effects with a slider at 12 that I cannot detect.

I don’t know about that - I’m barely on the shy side of 60 and afflicted with tinnitus. That’s why I use the graphic display as well as my ears to help guide the process. I use 9dB reduction rather that 12 because that felt right to me when I started using these tools about a year and a half ago.

For the record, over the last 2+ years I’ve recorded (and uploaded to iTunes and my iPods & iPhone) 125-150 records. Original material has ranged from 80 year old 78s and to 50 year old stereo recordings that I got from my father-in-law, to my wife’s and my collection of 60’s , 70’s & 80’s rock (and new age ), to few brand new LPs. There’s even a few well used 45’s from the 60’s in there too.

[coonsanders]

hi
i dont uderstand the whhole procedure as to how to use this denoisier…i can get a sample from the lead in on a record but im not sure what should be checked for “write,vinyl,shallck” or when
to use the automatic. options.for example if i was doing 78s what would be the procedure for
this?i tried reading the manuel for thiese things but its so complicated(at least for me)any sugg-
estions would be appreciated.

lenny