May 11, 2020 at 7:42 pm #46257
So today was Week #4 of a KRVM Vinyl Revival show with tracks ripped by me from vinyl to MP3 files. The MP3 files are the best quality that Audacity can make: 320kbps, with all settings at best quality. The uncompressed master WAV files sound indistinguishable from their MP3 counterpart prior to handing off the MP3s to the station.
The MP3 files are assembled by the station into a show. On-air, the assembled show generally sounds somewhere between kind of ok at best, and awful at worst. Mostly awful.
They rip CDs to MP3s for other shows and they sound fine. Why do my ripped vinyl tracks usually sound crappy on-air? I’m normalizing them to -6dB, and using a rumble filter, but that’s it.
What am I doing wrong? Or are they re-encoding my MP3 files at some very low bitrate with lower encoding quality?
Also the MP3 facsimiles sound very inferior to when they played vinyl live in studio.May 11, 2020 at 8:38 pm #46259Andy BrownParticipant
The MP3 files are assembled by the station into a show
That’s where I would look. Something they’re doing in their process is probably the issue. Also, why are normalizing to -6? The audacity default is 0. I normalize to -0.5.May 11, 2020 at 10:13 pm #46260Randy_in_EugeneParticipant
Is there any reason you can’t send them wav files? It’s one less compression point. So what if they take a little longer to upload/download? I recently switched from MP3 to wav on my KOCF show and the sound coming back on the internet stream is noticeably improved.
May 11, 2020 at 10:51 pm #46263Steve NaganumaParticipant
- This reply was modified 3 weeks, 2 days ago by Randy_in_Eugene.
Check out Ocenaudio (also free for Windows, Mac and Linux). Ocenaudio has Spectrogram View which may give you more information on your files.May 12, 2020 at 8:08 am #46266
Thanks for the tip on Ocenaudio. I’ll check it out.
Normalizing to -6dB is what I was instructed to do. Presumably that’s how they normalize everything. Maybe that’s what the calibration of their audio playback chain expects.
When we went into this I was assuming they’d drop the individual MP3 files into Simian as-is and program the playback order. But maybe they’re re-encoding everything into one large file with Audition or SoundForge (they use both). I’m kind of reluctant to press them for technical details because they’re slammed trying to keep all the specialty shows going at a time when volunteers can’t come to the station, and their GM is leaving. I don’t want to be Misstra Know-It-All. But it takes about 6 hours to “produce” 2 hours of material. Record in real time, normalize, clean up the intros and outros, populate some basic metadata, export, and upload. I just think the end result should sound better.
Maybe I’m just too picky.May 12, 2020 at 8:13 am #46267NotalentParticipant
No, You are not too picky, we need more people paying attention to quality audio.
Thank you for caring.May 12, 2020 at 11:07 am #46269nosignalallnoiseParticipant
What Andy_brown said. I assume you’re encoding at constant bitrate, discrete stereo (not joint stereo). They could be reencoding the final show at a lower bitrate, for whatever sick reason.
If it were me then I’d lock everybody in the station and hold them at gun point until they finally pledged to stop doing that! 48/320 or Fight!
FYI, -6dB is 50% peak (switch your output meter in Audacity to linear mode, generate a test tone, normalise it to -6 then hit play and you’ll see) which AFAIK is a fairly standard program level. SRL is -18 dB in digital audio systems which is equivalent to VU 0, or IIRC 250 nWb/m magnetic flux when measured across the head gap, but that’s tape. Presumably they want a little more headroom in case somebody’s program audio exceeds the station’s preferred peak level. That’s just my wild guess based on nothihg, anyways.May 12, 2020 at 11:48 am #46270
Constant bitrate. 44.1 kHz 16 bit.
I’ve been using joint stereo. I did some research on that a while back and the general consensus (which may be wrong) is joint stereo is preferred although it may not make any real difference. I’ll switch to discrete for the next batch.
Another wildcard (and I’m sure this will elicit come opinions) is I’m using phono preamps with USB output. One of the preamps is a relatively cheap ART USB Phono Plus. The other is a high-end Music Hall PA 2.2. The Music Hall supports up to 96 KHz 24 bit, but I’ve been running it at 44.1 16. The ART is strictly 16 bit but can do either 44.1 or 48.
I can hear no difference between the preamps and spectrum plots look virtually identical on the same test recordings after normalizing.
To be continued…May 12, 2020 at 5:37 pm #46275DarkStarParticipant
Which MP3 encoder are you using? They are not all equal in quality.May 12, 2020 at 5:44 pm #46276nosignalallnoiseParticipant
I really doubt it’d be your preamp causing the problem. (FWIW, I use a stereo receiver with the tape-out connected to the line-in of a sound card or an….. sorry.) I think we can safely narrow it down to two factors. (1) Their decoder doesn’t properly handle joint stereo (don’t shrug it off; I once had an APEX DVD machine with an earlier (crappier) firmware that made it put sum on the left channel and difference on the right) or (2) they’re downbitting when preparing the show for broadcast and playing the reduced bitrate file over the air.
I’d probably bet on condition #2 these days. Probably some volunteer who doesn’t know what they’re doing, or doesn’t care, and is using the default bitrate (usually 128) and transform accuracy settings (usually medium) for the export. Not much you can do except beat them upside the head and ass with a rubber chicken, and rinse and repeat until they finally get a clue. I don’t even use 128 Kb except for mono material!
Condition #1 was more common 15-20 years ago than now, when the MP3 format was just coming of age, but it’s surprising the amount of decrepit equipment decades past its prime there actually still is in the wild, in regular operation.
He’s using Audacity for editing, which probably means he’s exporting using LAME since the one interfaces with the other directly. Current version of LAME is 3.100.May 15, 2020 at 7:57 pm #46309Greg_CharlesParticipant
Who is the chief engineer of KRVM now? Is Bob Rathman still alive? Sparks? I see Ichabod on occasion.
Anyway, it might be worth pursuing the -6db normalize thing. What kind of processing at KRVM is happening? Did anyone ever donate say….an old Orban OptiMod to KRVM?
I will say that when mastering projects, 90% of clients want their CD’s as loud as when Brian Big Bass Gardner did it. Unfortunately (and excluding Brian) when a loud CD hits the standard broadcasting processing it can actually make it more quiet and or dull…depending on the Orban and its settings, and how the CD was mastered. Most CD content I’m thinking of is in the -6 to -5dB RMS, but ultimately those are just numbers and often meaningless due to the actual content.
How about take a “good sounding” CD-ripped MP3 track from KRVM. Then copy the identical track from vinyl using that “rumble” filter on and also just for kicks without it? I use Wavelab Pro 9.5 & 10 for error checks and analysis, but I’m guessing other tools should give some clues. Obviously, the CD ripped “good sounding” track might be mastered from a different ME than the ME who mastered and cut the vinyl, but looking at the KRVM processing, plus the CD vs.vinyl should manifest some clues. Personally, 1st generation 80’s CD’s usually sound the best to me…before the loudness wars. I can’t comment on newly cut vinyl because I haven’t bought any.
If an MP3 was created at 320 44.1 stereo Lame or Fraunhofer, I can’t possibly understand how a .wav could translate via broadcast better than this formatted MP3. Of course I don’t know what you have for post processing either. When I do general CD vs. MP3 320 comparisons, the content is usually a jazz or classical CD into a McIntosh C32 to the MC 2205 to B&W 803s in my untreated but large open living room. The differences such as reverb tails are quite subtle. I would never hear these through any broadcast stations I can think of.May 15, 2020 at 8:17 pm #46310
KRVM hasn’t had a full-time engineer since about 2014. when Randy Larson was both their GM and engineer. When he left, they started using a contract engineer. Dylan Hicks.May 16, 2020 at 9:28 am #46318NotalentParticipant
Greg, Even a 44.1/320 MP3 will sound inferior on the air to a .WAV… The main reason being “cascading codecs.”
The modern air chain has several codecs in line… Therefor it is very important to start with the highest possible quality file.
Consider even though you can’t hear the difference between the high quality MP3 and .WAV it is still true that the MP3 file contains 20 to 30 percent of the original data.
I don’t know how old KRVM’s studio and transmitter infrastructure is, but think about it… If the OP’s MP3’s are sent to the station and then assembled into an new file that is one pass through another codec, if that file is also an MP3 then you now have approximately 20% of the data left from the original MP3. The new MP3 is either burned to a disc for playback on a CD/DVD player (another conversion from D to A) and then played through a digital console… or put into an automation system as a file through a sound card which may also be converting to a standard file type… The digital console may or may not have a digital output being used… The signal then goes to an audio processor and into the HD radio encoder… The STL, the Exciter, and finally to the transmitter. All of which will be a combination of analog and digital. Each step in the chain will further reduce every file in quality.
By the time all those conversions take place you will notice the difference.May 16, 2020 at 4:18 pm #46327
About 5 or 6 months ago KRVM installed a profanity delay which digitizes everything, mostly unnecessarily. It’s yet another conversion that does nothing good for audio quality. I understand why they did that but I doubt it’s been used even once.
I’m going to give them WAV files for the show on 5/25 (5/18 is already submitted as MP3s). No idea what their reaction will be. Heh. If it’s a problem I’ll have MP3s ready anyway.
About a year ago they indulged me by airing a WAV file from vinyl as a test and everyone agreed it sounded great. Angry Eyes by Loggins and Messina. But there wasn’t any followup and Angry Eyes is still in rotation as an MP3 in Simian.
I guess “Ain’t nobody got time to mess with WAV files”.May 16, 2020 at 8:07 pm #46328
BTW I am not criticizing anyone at the station for how they do things. I don’t know squat about radio production or their workflow or especially their workload at this time. Forty or so volunteers can’t do their specialty shows live at the studio. Some of the volunteers are not computer literate at all. Yikes.
So one thing I’ve learned about doing vinyl for radio is that when the entire vinyl track is more or less at the same level, normalizing to -6dB is fine.
But something like Foo Fighters Come Alive is problematic. The first 2 minutes are very quiet, like a low vocal accompanied by guitar. Then all hell breaks loose. Normalizing entirely to -6dB based on overall peak levels lowers that quiet section too much. Applying a little compression to that section helps. It’s just radio and not classical music either. Make it sound “good”.
- You must be logged in to reply to this topic.