My general philosophy regarding audio processing for broadcast is not to
apply anything in the production room but to import recordings exactly as
they are, using normalization to keep peaks consistent from one cut to the
next.
There are exceptions, though; on a Christmas special I produced last month
featuring a local community band for broadcast on WSCS, I was compelled to
apply some compression to "Twas the Night Before Christmas" and "How the
Grinch Stole Christmas" to make the narrators audible.
As for on-air processing, my goal is simply to make everything audible.
The listener should be able to enjoy the station without having to turn
his or her volume control up and down.
Classical music is probably the most challenging content to broadcast.
Beethoven never knew radio and intended his music to be heard in live
performance, so it takes advantage of the full dynamic range of a concert
hall. There's no way to preserve that on the radio without rendering the
low passages inaudible in a typical listening environment, so compromises
have to be made.
At WCRB, where I worked for two decades, we used to air live Boston
Symphony concerts on Saturday nights. We had to switch most of our
processing off line for these; otherwise, you'd hear audience members
coughing and blowing their noses between movements. For live performances,
there's no substitute for a live operator riding gain, with a copy of the
score for reference.
Rob
--
Сквозь грозы сияло нам солнце свободы
И Linus великий нам путь озарил;
Нас вырастил Stallman на верность народу,
На труд и на подвиги нас вдохновил.
On Sun, 10 Jan 2021, Robert Jeffares wrote:
Hi Andru,
the normalization algorithm used by Rivendell is a safe option which works.
It has some inconsistencies which I will identify later.
A peak voltage based setting of the volume control is never going to be
perfect because the apparent loudness of any audio is dependent on a number
of factors.
Composition of the audio, Frequency, reverberation, "processing" of the
source, and the environment of the listener.
Perception; what is sounds like; is not necessarily measurable on a meter,
be it VU PPM or Spectrum Analyser.
Fletcher–Munson did some good work on this in 1933. There has been
subsequent research. Its a subject broadcasters should be familiar with.
I am grateful to my tutors, Dennis Edwards in particular, for quite some
focus on this. Wikipedia is reasonably useful as a starting point.
There are a number of ways of setting audio level by electronic means, none
of them approach the perfection of a good set of ears.
The first consideration for a Rivendell system is the storage and retrieval
of an electrical signal at a level which gives a good signal to noise ratio
and does not introduce distortion.
Digital sampling has made both criteria easier to deliver [although as an
analog purist I don't like sampling at any rate :)]
The playout for automated audio needs to be at a level which has consistent
loudness. There is no way round the need for a curator to audition the audio
and make a professional decision.
Fortunately Albums and Artists tend to be consistent.
There have been any number of attempts to create a device that can deliver
constant and "bigger" loudness. With the exception of the Texar Prism [ I
said I was a purist ;) ] they tend to try and deliver a constant voltage
output on one or more bands of frequencies, at some point making everything
sound the same, with concomitant listener fatigue.
Best example I have ever heard was an Audimax -> Volumax combo through which
the 1812 Overture was converted from a dynamic range approaching 50db to a
dynamic range of less then 3db. There are several stations where I live
which have processing that sounds like a 20m Ham Radio contest station. To
be fair. The hams do it better.
There will always be a device in the programme chain which provides
protection to the transmission system by limiting peaks.
This will not affect loudness unless it is set at some extreme.
In an automated environment a decision has to be made at the ingest point
You have two choices:
1) Make sure the library has all audio stored and configured so it plays at
a consistent loudness. Your VU meter may vary.
This involves a lot of work.
I have a number of scripts which enable me to select audio by artist album
or some other criteria and to play the output on a spare terminal.
I can grade a lot of songs in a short time by listening to the first 20-40
seconds of each. If your recordings are from the same source you can apply
global settings easily.
There is some mysql fu which you are probably well able to apply.
2) Deploy some electronic steam press which takes audio in from the console
[or the input to the console from the sound card] and outputs a constant
level engineered to give as much loudness consistency as the circuitry will
allow. Various playout systems provide this as a feature. Jazler has a
bundled version of Stereo Tools which does deliver a consistent output.
Any resemblance to the original audio is accidental.
In an ideal world the 'right' level will be delivered from the mixing
console to the programme chain.
The transmission will be a joy to listen to.
In the real world there will be someone asking why don't we sound as loud as
'XYZ'?
This results in the deployment of expensive multi band processors which are
configured to deliver 'perfect' sound in the bosses car or the bosses
lounge.
A track that cost a large amount of money to produce in a state of the art
studio with the best musicians and the best instruments can be made to sound
just like everything else.
[ and we wonder why people don't listen to radio for music so much any more
:|( ]
The processor will cost more than the console. Good management can avoid
the necessity. Your audience will love you for it.
Hints:
Use rdimport from the command line and set --normalization-level=-13
The default [-13] works on mp3 tracks but wav files are left untouched
unless you have the level specified. [ this may not be the case in v 3.x ]
You can also set segue points and trim levels. $ man rdimport tells all
use --set-string-user-defined= to identify the source of the tracks I add
$(date +%d%m%Y)
mysql> UPDATE CUTS set PLAY_GAIN = '-320' where CART_NUMBER in (select
NUMBER from CART where ARTIST like '%Red Jenkins%') ;
--------------------------------- cute code to get songs by artist specify
artist with '%Artist%' after calling script -----------------
#! /bin/bash
mysql -u root Rivendell -e "SELECT * FROM CART WHERE ARTIST LIKE '$1' INTO
OUTFILE '/tmp/output.txt' "
#remove un needed spaces and text
sed -i 's/\ \ A\ \ //' /tmp/output.txt
sort -rk 4 /tmp/output.txt >sortedList
#take out cart numbers
cut -c-6 sortedList >play.txt
#load into player
cat play.txt
sleep 2
while read i ; do /usr/bin/rmlsend --to-host=192.168.1.111 --to-port=5858
PX\ 1\ "$i"! ; done <play.txt
rm /tmp/output.txt
exit 0
------------------------------snip-----------------------------------------
-------
watch -n 35 '/usr/bin/rmlsend --to-host=192.168.1.111 --to-port=5858 PN\
1!'
There is no short cut to grading the audio for storage and replay. There is
no magic process.
If you are importing from a library which has been ripped from all sorts of
sources you may have various iterations of mp3 and variants of wav upstream
of your audio
The translation from the existing sample rate to whatever your system is
using can cause all sorts of weird outcomes.
By flagging the original source and retrospectively replacing where possible
from an original CD, or a digital transcription of Vynyl, you will improve
the library.
That said a lot of CD reissues leave much to be desired.
I have scripts to correct out of phase channels [ on big name brand CD
issues !!! ] and to re sample audio which has imported at greater then 4db
over -13.
Did I mention avoid mp3 if at all possible? the 320mb/s dubs record
companies supply radio stations still don't stack up against an original CD.
I notice one of my sources is offering wav copies as an option.
Not an answer to your request but perhaps something to consider
regards
Robert
On 9/01/21 11:45 am, Bemis, Andru wrote:
Thanks for your response, Fred. I truly appreciate everything
you and the other developers have done to make such a useful and
well-thought out program.
I'm in the midst of importing a few thousand cuts (studio albums, live
recordings, and voice tracks) for a new Rivendell installation for our
local LP station. I've figured out a pretty good workflow, in which I
set levels while placing voice and segue marks, but RMS normalization
would certainly save a good deal of time in the long run, while
producing more consistent results than my quick eyes&ears check.
Ardour/Mixbus has an RMS normalization function which defaults to 0dB
peak normalization when RMS would otherwise introduce clipping.
Perhaps an approach like this would be helpful in RD. It's much easier
to fix occasional errant recordings after the fact than to have to
manually check levels for all.
I know you have plenty of things on your plate, and don't expect this
is the most pressing, but I did want to bring it up.
Have a great weekend, everyone.
Andru Bemis
557 Chenango Street, Binghamton, NY 13901, USA
(269) 767-8235
he/him
On Thu, Jan 7, 2021 at 11:05 AM
<rivendell-dev-requ...@lists.rivendellaudio.org> wrote:
Send Rivendell-dev mailing list submissions to
rivendell-dev@lists.rivendellaudio.org
To subscribe or unsubscribe via the World Wide Web, visit
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
or, via email, send a message with subject or body 'help'
to
rivendell-dev-requ...@lists.rivendellaudio.org
You can reach the person managing the list at
rivendell-dev-ow...@lists.rivendellaudio.org
When replying, please edit your Subject line so it is more
specific
than "Re: Contents of Rivendell-dev digest..."
Today's Topics:
1. RMS normalization? (Bemis, Andru)
2. Re: RMS normalization? (Fred Gleason)
----------------------------------------------------------------------
Message: 1
Date: Wed, 6 Jan 2021 20:36:43 -0500
From: "Bemis, Andru" <an...@andrubemis.com>
To: rivendell-dev@lists.rivendellaudio.org
Subject: [RDD] RMS normalization?
Message-ID:
<capgxi1avfq11llz_2bk-rjjwu_hrbrblb+ukkliskfccjgk...@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"
I'm new to this list, so if this has been covered in the
past (I presume it
has), please forgive me. Does RD support RMS
normalization, or only peak?
*Andru Bemis*
*Producer & Host*
*THE CHENANGO SESSIONS*
*World-class musicians recorded live in small rooms and
intimate spaces.
Weekly one-hour broadcasts, syndicated worldwide through
Pacifica Radio &
PRX.*
- Website: chenangosessions.andrubemis.com
- Pacifica:
audioport.org/index.php?op=series&series=WBDY+Chenango+Sessions
<http://audioport.org/index.php?op=series&series=WBDY+Chenango+Sessions&;>
- PRX: exchange.prx.org/series/39247-chenango-sessions
- GCR2 Global Community Radio
- KAKU 88.5 FM Voice of Maui (Kahului, HI)
- KBOG 97.9 FM (Bandon, OR)
- KPSQ 97.3 FM Public Square Community Radio
(Fayetteville, AR)
- KPOV 88.9 FM High Desert Community Radio (Bend, OR)
- KWSI 100.3 FM (Grand Junction, CO)
- KXCJ 105.7 FM Illinois Valley Community Radio (Cave
Junction, OR)
- WBDY 99.5 FM The Bundy (Binghamton, NY)
- WHPW 97.3 FM (Harpswell, ME)
- WOOL 91.5 FM Black Sheep Radio (Bellows Falls, VT)
- WPVM 103.7 FM (Asheville, NC)
- WXOJ 103.3 FM Valley Free Radio (Florence, MA)
557 Chenango Street
Binghamton, NY 13901, USA
(269) 767-8235
*he/him*
--
*Message sent from?andrubemis.com
<http://www.andrubemis.com>*
-------------- next part --------------
An HTML attachment was scrubbed...
URL:<http://caspian.paravelsystems.com/pipermail/rivendell-dev/attachments/2021
0106/ff1c2202/attachment-0001.html>
------------------------------
Message: 2
Date: Thu, 7 Jan 2021 08:57:36 -0500
From: Fred Gleason <fr...@paravelsystems.com>
To: User discussion about the Rivendell Radio Automation
System
<rivendell-dev@lists.rivendellaudio.org>
Subject: Re: [RDD] RMS normalization?
Message-ID:
<f3b777fb-735f-4f19-b613-ee255edc0...@paravelsystems.com>
Content-Type: text/plain; charset="utf-8"
On Jan 6, 2021, at 20:36, Bemis, Andru
<an...@andrubemis.com> wrote:
> I'm new to this list, so if this has been covered in the
past (I presume it has), please forgive me. Does RD
support RMS normalization, or only peak?
Peak only.
There have been discussions about implementing RMS
normalization, but it introduces some thorny questions
?e.g. handling exception conditions, such as sample
clipping.
Cheers!
|---------------------------------------------------------------------|
| Frederick F. Gleason, Jr. | Chief Developer
|
| | Paravel Systems
|
|---------------------------------------------------------------------|
| A room without books is like a body without a
soul. |
|
|
|
-- Cicero |
|---------------------------------------------------------------------|
-------------- next part --------------
An HTML attachment was scrubbed...
URL:<http://caspian.paravelsystems.com/pipermail/rivendell-dev/attachments/2021
0107/5cb7be98/attachment-0001.html>
------------------------------
_______________________________________________
Rivendell-dev mailing list
Rivendell-dev@lists.rivendellaudio.org
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
End of Rivendell-dev Digest, Vol 93, Issue 7
********************************************
Message sent from andrubemis.com
_______________________________________________
Rivendell-dev mailing list
Rivendell-dev@lists.rivendellaudio.org
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
--
Communication Consultants 2020 Limited
64 Warner Park Avenue
Laingholm
Auckland 0604
New Zealand
_______________________________________________
Rivendell-dev mailing list
Rivendell-dev@lists.rivendellaudio.org
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev
