panup: Studio Session & LANDR Test

by Panu Pentikäinen (panup at Cakewalk forums)

Alex PortraitAlex ja Armottomat (Alex) visited my recording studio in February. We had five days total to do a fully mastered CD, make promo photos of the band, and record live video footage in the studio for later editing. I’ll describe here how one of the six songs was recorded and mixed.

Drums, bass and the electric guitar were recorded live with one to three takes. Acoustic guitar and demo vocals were recorded, too, but they were re-recorded later over the backing tracks. The drummer was the only one to hear the metronome (standard SONAR audio metronome, time signature set to 1/4); the others had eye contact with the drummer. Although the guitar amp was in another room (the bass was recorded direct), there was no spill other than a faint demo vocal in the drum room mics.

Time is always an enemy when you have to record many songs in a limited amount of time. I decided to make decisions before pressing the R (record) button rather than leaving everything to the mixing phase. I applied EQ to kick drum, drum room and the acoustic guitar before A/D conversion. One of the phrases I hate is: “This sounds like crap now but it hasn’t been mixed yet.” Some people really think that everything can be fixed in the mix! (Although to be fair you often can, because in SONAR we have VocalSync, built-in Melodyne, built-in drum trigger, and AudioSnap).

And although it sounds incredible, now it’s even possible to upload songs from SONAR to the LANDR online mastering service and instantly hear a preview of how the song would sound as mastered. Hearing the demo master may help you to improve the project’s mix.

I sent a rough mix of the song to LANDR and downloaded it for analysis. (LANDR is not a free service, but it lets you download up to two MP3s for free per month. Also, 30 second online previews are always free, which is nice for the test mastering.) LANDR compressed the rough mix, boosted some low end, brightened cymbals and cut some mids around 500 Hz. Compression was very noticeable but the EQing was quite subtle. I felt relaxed after getting a confirmation from another pair of [digital] ears: “there’s nothing drastically wrong in your mix, just go to sleep now.”

Next day we mixed the songs with the band’s leader. Although he said many times that he doesn’t understand anything about sound, I still tried to make his mix sound better after he was already satisfied. After all, when is a mix ready? When all mistakes have been removed, and the balance is perfect? Out-of-tune vocals and lousy backing tracks often can be fixed by editing, but exactly what is the perfect balance in a mix? Here’s the answer (drum roll, please): there is no such thing. One mix may be perfect for me, but sound unusable to you. It’s a matter of taste.

Tastes vary, but there are still some general rules to follow when targeting a mix to the public. Unfortunately everybody does not listen to your music with same speakers as you, in an identical room, and at the same volume. A bass-heavy mix may sound good with cheap earbuds, but awful through a PA system.

According to LANDR’s technical information, their system analyzes a song’s genre and tries to match its sound with similar, good-sounding mixes. I don’t know exactly how it works (because that’s not public information) but I would guess LANDR’s mastering chain includes dynamic EQ, multiband compressor, stereo imaging and finally a brickwall limiter to bring the song’s loudness up.

I found there were three loudness levels to choose from: Low, Medium, and High. The Low option preserves dynamics most while High option makes your mix as loud as possible. I hate excessive brickwall limiting because it sacrifices sound quality. (Here’s an interesting “Loudness War” article in Wikipedia, and one from Sound on Sound magazine.)

I suggest not using the High option by default, and sticking in the Medium or Low Intensity. Medium Intensity is loud enough for SoundCloud and YouTube.

OK, let’s go back to the mixing session. I had about half an hour per song to do the final mix. In that time I had to make all decisions and tweaks by listening to the mix—and checking the client’s face to confirm that he was still happy.

I had promised to include the mastering, too. I bounced songs to stereo wav and sent it immediately to LANDR—this time using the desktop application rather than SONAR’s export option. Meanwhile uploading files to LANDR and waiting for the rendering gave the time to do some mastering myself. I’m used to doing any final adjustments in Adobe Audition and that’s what I used this time, too. (Sorry SONAR to cheat on you but you’re not an audio editor!) I was already satisfied with the mixdown and my mastering chain was subtle:

1)   Fabfilter Pro Q2: Cut @ 38 Hz, boost 1 dB @ 60 Hz, cut 0.3 dB @ 635 Hz

Master EQ

 

2)      Multiband Compressor: compress 0-120 Hz, ratio 1:1.5, 0-2 dB

Master Multi-Band Compressor

 

3)      Brickwall limiter: target volume -12 dB RMS

You can compare mine and LANDR’s Medium Intensity master here:

https://soundcloud.com/panup/sets/alex-ja-armottomat-landr-mastering-test

I matched volumes of the samples manually. LANDR’s version was originally a bit louder than mine.

Listen to the samples yourself, and make your opinion (as long as you like my version more!). This is like the chess match between chess champion Garry Kasparov and the IBM supercomputer (IBM won). Can a robot make better art than a human? Yes, it can…I actually dig LANDR’s master more than my own (see frequency analysis chart of the differences). LANDR cut mids at around 500 Hz more than what I did. Questions began to echo inside my head: Was I too careful? Was I deaf? Why did LANDR cut it more?

Frequency Analysis: Human vs. LANDR

I don’t take it too seriously, though. LANDR has its own style, I have my own. The other mixes I sent to LANDR had kind of similar sound: EQ curve is more hyped than flat meaning that it boosts 60-80 Hz, cuts mid frequencies and boosts treble. LANDR also applies more compression than me. Some mixes sounded much more “professional” as mastered by LANDR but not all—it’s hit or miss, depending on the mix.

There were also problems with the LANDR service; sometimes upload speed was very slow and making preview took forever (hours). It even said it could not master one of my songs because it was already too loud (it wasn’t). One mix just failed to render. I sent a message to their support, and the answer was delightful: the help desk guy mastered my song manually (!) (using LANDR, of course).

LANDR won’t replace good, professional mastering engineers but if you’re about to publish your demo and don’t know how to master it yourself, then LANDR is worth checking out. LANDR’s online previews are free, so there’s nothing to lose.

This post was a story about one studio session in my recording studio. SONAR has been my main DAW since version 6. I have 25.000 hours experience of using SONAR. That’s more than average man sees his wife during his lifetime.

 

Artist: Alex ja Armottomat

Song: Juoppohullun lokikirja

Track list (Name, Microphone, ProChannel / FX Bin):
  1. Kick — AKG D112, QuadEq: cut 4 dB @ 170 Hz, Q 1.2
  2. Snare top — Sennheiser MD 421, QuadEQ: cut 0.9 dB @ 1.3k, boost 2 dB @ 6.7k
  3. Snare bottom — Shure SM7-b: QuadEQ: hi-pass 6 db/oct @ 265 Hz, PC-76 compressor
  4. Overheads — Neumann KM184, wide A-B (muted in the mix)
  5. Drum Room — AKG C414B-ULS, A-B, Channel Tools to narrow stereo field a bit
  6. Hi-hat — AKG C451: QuadEQ hi-pass 6 dB/oct @ 3.8k, PC4K S-Type Compressor: thresh -30, ratio 10:1, fast attack, GR 0-3 dB
  7. Ride — AKG C451: QuadEQ hi-pass 6 dB/oct @ 800 Hz
  8. 12” Tom — Sennheiser MD-441: no fx
  9. 14” Tom — Sennheiser MD-441: QuadEQ: hi-pass 12 db/oct @ 52 Hz, cut 8 dB @ 260 Hz, boost 3 dB @ 3.5k
  10. Bass — TC Electronic RH450 –> solid state preamp
  11. Ac Gtr — Neumann KM184, X-Y: QuadEQ: cut 3.6 dB @ 240 Hz, Q 6, PC4K S-Type Compressor 4:1, slow attack, GR 3-6 dB
  12. El Gtr — Shure SM57 (close), AKG C414 (far)
  13. Vox — Pearlman TM-1
Buses
  • Master bus
    • Quad EQ: HP @ 31 Hz, 12 dB/oct
  • Drum bus
    • Quad EQ: HP @ 43 Hz, -2 dB @ 10k, Q=0.8
    • PC Compressor: Attack 13 ms, Release 0.1 sec, Ratio 4. Gain reduction 0-2 dB
    • Brickwall Limiter (max 2 dB)
  • El Guitar bus
    • Quad EQ: 5 dB mid cut @ 390 Hz, 3 dB boost @ 5.5k
    • Brickwall Limiter (cut peaks up to 4 dB)
    • Softube RC48 Hall, 22%
  • Ac Gtr bus
    • QuadEq: -4.7 dB @ 290 Hz, Q 1.0
    • Softube RC48 Hall, 40%
    • BReverb: slightly modified HiFi Drum Room Preset, wet -11.5 dB
  • Vocal bus
    • Sonitus:Gate: lookahead 10 ms, attack 1 ms, hold 150 ms, release 1500 ms, depth -16 dB
    • Waves DeEsser
    • Aphex Vintage Exciter: Mode Mix1, Ax Mix 4
    • CLA Vocals
    • Brickwall Limiter (max 1 dB cut)
    • Softube RC48 Random Hall, 8%
  • Snare Reverb bus
    • QuadEq: low cut @ 80 Hz, hi cut @ 10k
    • Sonitus:Gate: Lookahead 14 ms, attack 1 ms, hold 80 ms, release 255 ms, depth 6 dB
    • BReverb: Hall 910 ms

Try LANDR