14 Feb
2023
14 Feb
'23
1:42 a.m.
Hey hey,
I have digitised a few last tapes. The sound isn't too bad, but all tapes
appear to be a little too bass heavy. Since I only work in the terminal I
approach the issue with Csound that provides a lot of tools. All kinds of
filters, rms/peak metres, an lufs metre and more.
First: did I miss an obvious commandline tool that would produce a
reasonable solution for that?
There are two main issues that I try to improve halfway automatically. 1.
Sometimes the stereo balance is a little wrong with one channel being slightly
louder than the other. 2. Imrpoving the frequency spectrum, flattening it a
little bit.
For 1. I have started by taking the file's peak of both channels seperately
and simply multiplying them so the peaks match. Over a whole tape side this
should be approximately the same. Or should I go with a different measurement?
lufs integrated loudness, momentary or short term maximums?
For the second issue, I am unsure. I have split the audio into frequency bands
using octaves, more or less. So the first band goes up to 100Hz, the next to
200, then 400, 800 and so on. I have used Csound's tone and atone filters:
atone highpass filter:
https://csound.com/manual/atone.html
and the matching tone lowpass filter:
https://csound.com/manual/tone.html
Then on each band for each channel I run the lufs metre, storing the maximum
momentary and short term values and the integrated loudness.
Though I wonder: are the frequency ranges and type of filter reasonable? If so
how would I approximate a reasonably flat spectrum? Which values to
compare to what? I mean: what kind of measurement to use and should I
compare adjecent bands and extrapolate fitting values from that or
compare one frequency band's measurement to the same measurement on the
unprocessed full-band audio?
Any hints and practical suggestions are very welcome.
Thank you and best wishes,
jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
I'm reaching out to you
To find that you're not there <3
(Britney Spears)
14 Feb
14 Feb
10:48 a.m.
On Tue, Feb 14, 2023 at 01:42:45AM +0100, Jeanette C. wrote:
I have digitised a few last tapes. The sound isn't
too bad, but all tapes
appear to be a little too bass heavy.
The first step should be to ensure the tape player is OK, demagnetised,
the entire tape path clean, and the playback head correctly aligned.
Failing to do that will cause loss of HF, which may even be permanent
(i.e. damaging the tape).
Don't go for maximum digital level. 24 bits has a lot more dynamic
range than tape, so leave a lot of headroom for later adjustments.
Overall level can be raised later, after all other processing.
There are two main issues that I try to improve
halfway automatically. 1.
Sometimes the stereo balance is a little wrong with one channel being
slightly louder than the other. 2. Imrpoving the frequency spectrum,
flattening it a little bit.
Re. 1:
There is no reason why the levels in the two channels should be
exactly equal. If there is a consistent difference due to a technical
problem, then a constant correction is all that is needed.
Peak levels are completely irrelevant for adjusting this. If you do
any automatic correction, use an RMS measurement averaged over the
entire lenght of a song. This should probably be done after any
EQ that may be required. But I'd just listen and adjust the
balance manually.
Re. 2:
There is absolutely no reason why you should have equal power
in octave bands, or any other way to split up the spectrum. No
music I know of has a 'flat' spectrum.
If the FR seems wrong, the first thing to find out why that is
the case. HF loss can be due the old age of the tape or misaligned
or magnetic tape heads. Both LF and HF problems could be the
result of the recording and playback machines using different
EQ standards (the were two in use, roughly US vs European).
In any case, the Csound filters you mentioned are just first
order lowpass and highpass and certainly not what you'd need.
A set of shelf and parametric filters would be the tool to
use. Then *listen* and adjust as required.
For a command line solution, I'd use the filters provided by
the JackParameq class in zita-jacktools. These will provide
up to 8 sections of shelf/parametric EQ. You only need a
few lines of Python to create an application that does the
processing in real time, so you can listen and adjust the
EQ as you go.
You'd need to install zita-jclient, zita-audiotools,
zita-jacktools. Then start python in a terminal.
from zita_jacktools.jackparameq import *
E = JackParameq (2, 'LH333333', 'EQ')
You now have a 2-channel equaliser with 8 sections,
running as a Jack client called 'EQ'.
To set the filters, use
E.set_filter (section, frequency, gain, shape)
section: 0 = LF shelf, 1 = HF shelf, 2..7 = parametric.
frequency: center frequency (Hz) for parametric, or the
corner frequency (Hz) for a shelf.
gain: gain in dB,
shape: relative bandwidth (1/Q) for parameteric, range
0.1 to 10, or filter shape (0.0 to 1.0) for a
shelf. 0 = first order, 0.7 = second order,
1.0 = modified second order.
You can bypass the entire EQ with:
E.set_bypass (True/False)
And set makeup gain with:
E.set_gain (dB)
Csound probably has something similar, but that would
certainly not be the two filters you referred to.
Ciao,
--
FA
2:02 p.m.
Hi Fons,
thanks for your helpful feedback, comments and suggestions. It seems
that I am best served by polishing the recording manually after all.
Well, it was just three tapes with radioplays, so it will only be six
files to fix. And the fixing I intended to do was static, i.e. no
automation as in DAW automation. By automated I meant having a program
that would analyse the recording and suggest a few values for that
static polishing process. It's more a question of personal comfort than
creating files that are ready for public release.
I couldn't do much about the level of the tape I used to digitise the
tapes, since it is a USB tape deck. I can also only assume that it is in
good health, being rarely used, well stored and of rather adequate
quality.
Thanks for all your pointers. I might search for the US vs. European EQ
standards and see if those might already help in achieving a better
sound. Othersie, it's the oldfashioned listen and tweak. :)
Thanks again and best wishes,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
Don't care about money
It doesn't give me half the thrill <3
(Britney Spears)
3:17 p.m.
"Jeanette C." <julien(a)mail.upb.de> writes:
Hi Fons,
thanks for your helpful feedback, comments and suggestions. It seems
that I am best served by polishing the recording manually after all.
Well, it was just three tapes with radioplays, so it will only be six
files to fix. And the fixing I intended to do was static, i.e. no
automation as in DAW automation. By automated I meant having a program
that would analyse the recording and suggest a few values for that
static polishing process. It's more a question of personal comfort than
creating files that are ready for public release.
I couldn't do much about the level of the tape I used to digitise the
tapes, since it is a USB tape deck. I can also only assume that it is in
good health, being rarely used, well stored and of rather adequate
quality.
Cassette tape? I have a pretty good one of those (though with regard to
head azimuth, there usually is no better than the one the recording has
been made on) as well as good sound cards. Dolby B and Dolby C would be
available if the recording had been made with either.
I could offer to try. On the downside, I have a power landline a
quarter mile off which tends to be annoying for unbalanced equipment
(single coil guitar pickups are sort of a non-option, and I have quite a
bit of practice redoing the internal routing of audio hardware such that
magnetic fields don't squeeze between connections where it counts). So
there may be a use case for a very narrow 50Hz notch filter in the
postprocessing.
All the best
--
David Kastrup
9:30 p.m.
Hi David,
thanks for chiming in. I have loaded the files into my DAW and done what I
could. It's not perfect, but perfectly viable. :) I am a creature of luxurious
habits. Some EQ'ing, a narrow bandreject filter for one of the tapes, oddly at
60Hz, some very mild compression and a highpass filter at 30Hz. This is fine
by me now. Maybe in time, someone else will come up with a recording of the
same programmes.
Best wishes,
Jeanette
Feb 14 2023, David Kastrup has written:
"Jeanette C." <julien(a)mail.upb.de>
writes:
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
If only I could trade the fancy cars
For a chance today,
it's incomparable <3
(Britney Spears)
Hi Fons,
thanks for your helpful feedback, comments and suggestions. It seems
that I am best served by polishing the recording manually after all.
Well, it was just three tapes with radioplays, so it will only be six
files to fix. And the fixing I intended to do was static, i.e. no
automation as in DAW automation. By automated I meant having a program
that would analyse the recording and suggest a few values for that
static polishing process. It's more a question of personal comfort than
creating files that are ready for public release.
I couldn't do much about the level of the tape I used to digitise the
tapes, since it is a USB tape deck. I can also only assume that it is in
good health, being rarely used, well stored and of rather adequate
quality.
Cassette tape? I have a pretty good one of those (though with regard to
head azimuth, there usually is no better than the one the recording has
been made on) as well as good sound cards. Dolby B and Dolby C would be
available if the recording had been made with either.
I could offer to try. On the downside, I have a power landline a
quarter mile off which tends to be annoying for unbalanced equipment
(single coil guitar pickups are sort of a non-option, and I have quite a
bit of practice redoing the internal routing of audio hardware such that
magnetic fields don't squeeze between connections where it counts). So
there may be a use case for a very narrow 50Hz notch filter in the
postprocessing.
All the best
--
David Kastrup
10:45 p.m.
A side note to this discussion. Over a period of time recently I
embarked on a project to digitise a large volume of cassette
tapes of live recordings of bands and other acts for which I did FOH
mix. The oldest was from 1984!
I found the quality of the tape was the most important factor. Chromium
dioxide tapes like TDK SA and particularly SA-X survived surprisingly
well, whereas iron oxide tapes much less so with severely degraded high
frequency content. For some of the iron oxide tapes, I found recording
with Dolby off, even if it was on for original recording, then removing
the added noise later could extract a small high frequency improvement.
A few went straight in the bin though. Dolby can be fickle though
particularly when playing back in a deck different from the original
recorder if the heads and bias haven't been optimised.
I recorded the tapes to at least 24 bit wav files prior to processing,
and as a last step used sox to resample to the desired bit depth and
sample rate.
I ran the most restorable recordings through Izotope RX (on Windows) to
remove most of the tape hiss with quite listenable results. I don't know
of a Linux program which can equal RX, and certainly no CLI program.
Maybe noise-repellent is close but I'm not experienced with it.
For consistent levels there is normalize-audio, a CLI program which can
losslessy adjust to a user-defined level or IIRC the default is -1dB.
Cheers, Roger
19 Feb
19 Feb
11:50 p.m.
On 2/13/23 4:42 PM, Jeanette C. wrote:
Hey hey,
Hey...
Adding to the thread...
Not what you are looking for, but you may want to take a look at this:
https://github.com/trummerschlunk/master_me
This is MasterMe, a realtime (or not) mastering tool. Developed by Klaus
Scheuermann - a mastering engineer & others) for the QSs originally,
used pretty much every week for our Quarantine Sessions streaming
broadcast (I integrate it in my SuperCollider software). Written mainly
in Faust so it could be compiled to a command line client...
One of the cool things it does is leveling to a proper loudness which is
invaluable for our streaming sessions (you can choose the level in LUFS,
and tons of other parameters).
Best,
-- Fernando
I have digitised a few last tapes. The sound isn't
too bad, but all
tapes appear to be a little too bass heavy. Since I only work in the
terminal I approach the issue with Csound that provides a lot of tools.
All kinds of filters, rms/peak metres, an lufs metre and more.
First: did I miss an obvious commandline tool that would produce a
reasonable solution for that?
There are two main issues that I try to improve halfway automatically.
1. Sometimes the stereo balance is a little wrong with one channel being
slightly louder than the other. 2. Imrpoving the frequency spectrum,
flattening it a little bit.
For 1. I have started by taking the file's peak of both channels
seperately and simply multiplying them so the peaks match. Over a whole
tape side this should be approximately the same. Or should I go with a
different measurement? lufs integrated loudness, momentary or short term
maximums?
For the second issue, I am unsure. I have split the audio into frequency
bands using octaves, more or less. So the first band goes up to 100Hz,
the next to 200, then 400, 800 and so on. I have used Csound's tone and
atone filters:
atone highpass filter:
https://csound.com/manual/atone.html
and the matching tone lowpass filter:
https://csound.com/manual/tone.html
Then on each band for each channel I run the lufs metre, storing the
maximum momentary and short term values and the integrated loudness.
Though I wonder: are the frequency ranges and type of filter reasonable?
If so how would I approximate a reasonably flat spectrum? Which values to
compare to what? I mean: what kind of measurement to use and should I
compare adjecent bands and extrapolate fitting values from that or
compare one frequency band's measurement to the same measurement on the
unprocessed full-band audio?
Any hints and practical suggestions are very welcome.
Thank you and best wishes,
jeanette
20 Feb
20 Feb
12:02 a.m.
Hi Fernando!
Feb 19 2023, Fernando Lopez-Lezcano has written:
,,,
...
Thanks for the tip, I will take a look at it. I've never used anything
that is Faust based. What kind of a commandline tool will I get from
this? Will it have a shell interface? Would the options be somewhat
standard or will it require knowing the workings of Faust?
Best wishes,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
All I need is time
A moment that is mine
While I'm in between <3
(Britney Spears)
4:48 a.m.
On 2/19/23 3:02 PM, Jeanette C. wrote:
Hi Fernando!
Feb 19 2023, Fernando Lopez-Lezcano has written:
,,,
...
Thanks for the tip, I will take a look at it. I've never used anything
that is Faust based. What kind of a commandline tool will I get from
this? Will it have a shell interface? Would the options be somewhat
standard or will it require knowing the workings of Faust?
Faust is a digital signal processing language. It can be compiled into
pretty much anything (VST, LV2, SuperCollider, Pd, you name it).
https://faust.grame.fr/
There is an online compiler but I don't think I have used that. See if
your distro has faust available and install those packages.
There will be a bunch of faust2* scripts (you may need to install a
separate faust-tools package). For example faust2jaqt will create a
stand alone jack client with a qt graphical interface. "faust2sndfile"
will create an offline soundfile processing command line program.
So:
faust2sndfile master_me.dsp
will create a "master_me" executable you can run from the command line.
If you run it without parameters it will print what you can use to
change the default values of all internal stuff (there is a lot of
stuff!)...
Best,
-- Fernando
1:37 p.m.
Hi Fernando!
Feb 20 2023, Fernando Lopez-Lezcano has written:
...
There will be a bunch of faust2* scripts (you may need
to install a separate
faust-tools package). For example faust2jaqt will create a stand alone jack
client with a qt graphical interface. "faust2sndfile" will create an offline
soundfile processing command line program.
...
Wow! Thanks for that! I've come across the odd faust code, but I never
knew about this feature or these scripts. This is amazing and great fun.
Best wishes and thanks,
Jeanette
--
* Website: http://juliencoder.de - for summer is a state of sound
* Youtube: https://www.youtube.com/channel/UCMS4rfGrTwz8W7jhC1Jnv7g
* Audiobombs: https://www.audiobombs.com/users/jeanette_c
* GitHub: https://github.com/jeanette-c
There's no time to loose
And next week,
You might not see me here <3
(Britney Spears)
431
days inactive
437
days old
linux-audio-user@lists.linuxaudio.org
9 comments
5 participants
participants (5)
-
David Kastrup -
Fernando Lopez-Lezcano -
Fons Adriaensen -
Jeanette C. -
Roger