I've been listening to this online radio: http://sc-classrock.1.fm:8200/
I really like how all of the songs sound really good. All the details are heard and I can't notice any distrortion or other unwanted effects with the DSP they're using. This is true for low frequency sounds (bass, drums) and for high frequency sounds (cymbals), and everything else too, really. With some of the songs I can hear details that I never heard before and I know those songs very well.
To test, that it's not a placebo or just in my mind, I did an ABX test. I ripped the song "Pearl Jam - Daughter", when they've played it and then compared it to my 320 kbps mp3 rip. They're using 320 kbps mp3 too, BTW.
Results:
foo_abx 1.3.4 report
foobar2000 v1.3.5
2018/08/12 21:06:10
File A: C:\Users\psycho\Downloads\rip.mp4
File B: E:\mp3\p\Pearl Jam\1993 - Vs\Pearl Jam - 03 - Daughter.mp3
21:06:10 : Test started.
21:06:49 : 01/01 50.0%
21:06:59 : 02/02 25.0%
21:07:05 : 03/03 12.5%
21:07:11 : 04/04 6.3%
21:07:20 : 05/05 3.1%
21:07:26 : 06/06 1.6%
21:07:31 : 07/07 0.8%
21:07:36 : 08/08 0.4%
21:07:53 : 09/09 0.2%
21:08:08 : Test finished.
----------
Total: 9/9 (0.2%)
Don't let the filename fool you. It's mp4 only as a container, inside there's an mp3. I used VLC to rip the stream, of course with no transcoding, but by default it used mp4 for the container.
So, I am interested if anyone will take the time and tell me your opinion of what DPS they might be using. Whatever it is, it is good for soft rock songs as well as some of the most heavy ones. A universial solution it would seem, at least to me.
My guess is, they're applying some dynamic range compression, and maybe some EQ with that. But they must've tuned it to perfection for rock/metal music.
Anyway, looking forward to other opinions.
Whatever it is, it is good for soft rock songs as well as some of the most heavy ones. A universial solution it would seem, at least to me.
You are implying that there's something wrong with the original recordings. ;)
DVDdoug, that's not what I tried to say. ;)
It's just that I really like how everything sounds on this stream and up until now I haven't come across a DSP that wouldn't ruin the sound in some way. This one only seems to have positive effect for my ears.
Well, they upped high range a bit, as far as I can hear. Other than that, I think there's some multiband compression applied... and the bass is upped a bit. Not much.
EDIT: I just compared... it's U shape, bass and highs are upped, mids are lowered. Also, I think the stereo separation is narrowed a bit.
I've compared the current song, Silent Lucidity.
hlloyge, very interesting, thank you. I think you got it. Could be that what you describe is just what they do and maybe nothing else. :)
Well, they upped high range a bit, as far as I can hear. Other than that, I think there's some multiband compression applied... and the bass is upped a bit. Not much.
EDIT: I just compared... it's U shape, bass and highs are upped, mids are lowered. Also, I think the stereo separation is narrowed a bit.
I've compared the current song, Silent Lucidity.
Sounds a bit like what a guitarist friend of mine calls an "American scoop" on his guitar amp's equaliser.
Do you think something like what they use could be done with some of these (http://www.foobar2000.org/components/tag/DSP)?
It's called 'smiley face' EQ. Been around forever.
https://en.wikipedia.org/wiki/Smiley_face_curve
Broadcast processors are specific to that application because of their goals. They start with a broadband compressor with moderately slow attack and release to keep levels stable for the following processors. Next is usually a multi-band processor, with the number of bands anywhere from two to ten, with the typical being three or four. However, settings for the multi-band section get pretty complex. You have attack and release for each band, specific crossover frequencies, and several types of action, RMS, Average and Peak, often a combination of them. The compressor and multi-band also include a "release gate" that freezes release if audio levels drop quickly preventing rapid gain increase. Multi-band processors include re-mix controls, basically controlling the output of each band similar to a graphic EQ. The end result can be a form of dynamic equalization, but also is a means of increasing loudness with less processor side-effect.
Then the multi-band is followed with a peak limiter, which can be varied in release, sometimes attack as well. In the case of processing for actual broadcast FM, there is also a high frequency limiter that acts based on the 75us pre-emphasis curve used. It's up 17dB at 15kHz, 0dB at 400Hz. Pretty wicked, and unless controlled with peak limiting and HF limiting transmitters easily overmodulate. Broadcast processors next include some very sophisticated clipping and processing. Clipping is necessary to hard-limit modulation because of legal regulations, but clipping in a band-limited system causes overshoot which can be self-defeating, so DSP-based digital processing takes all of that into account too.
Today broadcasters may need three different processing chains, one for air, one for HD Radio (the requirements are quite different) and one for streaming (different again). AM is radically different from FM too. However, all applications follow the general architecture of compressor > multi-band > peak limiting, and diverge from there. If a station is FM and HD, there's also delay included so the FM matches the HD delay so radios can transition between the two seamlessly. If a station also streams, the processor may include sub-chains for each, but more typically the stream is processed separately by the encoding software. Typical encoding software is also manufactured by the same companies that make on-air units so adjustments and results are familiar. Generally, on-air FM is processed the hardest, HD somewhat less because it has no high-frequency pre-emphasis to deal with, an the stream can be somewhat less too. But stations also want to be loud or louder than their competition, so everything is processed at least a bit. Announce mics often are processed separately and individually, and modern broadcast plants are fully digital.
It's called 'smiley face' EQ. Been around forever.
https://en.wikipedia.org/wiki/Smiley_face_curve
Yeah...hardly. It may sound that simple, but that curve was never used alone, it was always surrounded with several other things.
Odd this comes up right now because there is an
article series (https://www.thebdr.net/articles/audio/proc/index.html) being published about the history of broadcast processing. They're up to the second installment, there should be at least 3, but could be more.
dc2bluelight, thank you for this insightful post. ;)
Rock configuration.
I just listened to it, and then compared the songs to the non-processed versions (Spotify.)
I don't like the sound on some of the songs :P They sound boomy and the guitars sound worse. However, on some other songs, it sounds good. I think it sounds good on songs that have a too crushed dynamic range ("loudness wars victims"). On songs that don't suffer from that, the processing seems to make them worse.
Maybe they're doing some sort of dynamic range expansion? I think it sounds somewhat similar to the "crystalizer" filter of ffmpeg, but they also do some v-shape EQ on top of that, I think.
Edit: Nope, I don't think that's it. Some of the songs sounding better is most probably them using an original release of the song, while Spotify has a crushed "remaster" or re-release that sounds worse.
I just listened to it, and then compared the songs to the non-processed versions (Spotify.)
What is "it" and why do you think Spotify is non-processed?
I don't like the sound on some of the songs :P They sound boomy and the guitars sound worse. However, on some other songs, it sounds good. I think it sounds good on songs that have a too crushed dynamic range ("loudness wars victims"). On songs that don't suffer from that, the processing seems to make them worse.
Until we know exactly what you were listening too the above doesn't make much sense.
Maybe they're doing some sort of dynamic range expansion?
IF you're referring to FM radio, "they" are not doing DR expansion, that's going the wrong way for the purpose.
I think it sounds somewhat similar to the "crystalizer" filter of ffmpeg, but they also do some v-shape EQ on top of that, I think.
Edit: Nope, I don't think that's it. Some of the songs sounding better is most probably them using an original release of the song, while Spotify has a crushed "remaster" or re-release that sounds worse.
LOTS of assumptions there. Did you read Reply #8?
To assume you know what "release" or "remaster" was used by listening to a broadcast signal would be likely in error. Again, read what's really going on in that chain in Reply #8.