While we are having a great discussion about reverb here, I thought it might be useful to start one on a complete workflow for classical music. I realize there is a similar conversation but it didn’t seem to go into details of plugins, dither, exporting etc.
I’ll outline my own workflow and I encourage others to jump and suggest alternatives!
-
Capture of audio via SoundDevices MixPre 6 ii or USB interface such as Focusrite Clarett or Behringer UMC 1820 – Before the arrival of the 32-bit float recording in the latest MixPre series, I was setting levels so that the loudest performance peaked at c. -12dB. In most cases, audience applause was always removed so it didn’t matter if whole-room clapping/whooping clipped the recording. With the 32-bit float, obviously I can now capture everything with no worries. I’m contemplating using the MixPre as a USB interface too as the preamps are excellent though how they compare to Clarett is difficult to say.
-
Microphones – Just to be complete about this from a capture perspective, for microphones I currently use Audio Technica AT4021s (cards) and AT4022s (omnis) as well as sE4400a pairs and most recently a pair of sE8s. I’ve yet to break into the “top tier” of classical microphones (the Schoeps, Gefells, DPA, Sennheiser etc) but I’m really happy with what I have. I do have my eye on either a pair of Gefell M300s or the Telefunken M60 stereo master set but alternatively I also feel a pair of AT4050s might nicely round out my collection. As backups I have two Behringer B2 Pros and four B5s (with the interchangeable caps). While I probably wouldn’t show up to many classical gigs with these as the main pairs (purely out of people thinking I wasn’t serious!), I do know that fine recordings can be had with these microphones. As many have said, if you made recordings with these and said they were KM84s, Schoeps ORTF etc., people would purr over the quality. I see it all the time in blind mic shootouts with guesses all over the shop until the results are revealed and suddenly everyone develops an ability to detect the cheaper microphone. Anyhow, I digress!
-
Import into Ardour/Mixbus – With the MixPre and the Zoom F8 before it, I would run the files through Sound Devices Wave Agent so that the individual channels were turned into proper interleaved stereo files. This might not be necessary given Ardour seems to split them into L and R behind the scenes but it feels easier for me to keep track of the source files and not have to worry about panning in the DAW.
-
Restoration work – This is why I still need a Windows partition. RX or Acon Digital Acoustica are priceless! Spectral edits are key to my location work as there are always coughs, chair squeaks and the like just at the wrong moment. I can also catch many early clappers this way although there always is an instance or two of needing to bring in room tone and artificial tails.
-
Editing – I have grown to love editing in Ardour. The transparent waveforms when dragging are super helpful as well as the default short crossfade which reminds me so much of Samplitude/Sequoia. I can’t stand DAWs where the default is to allow the crossfade to keep extending over the amount of overlap. Unless there are hundreds of edits to do, editing in Ardour is a breeze.
-
Plugins
Loudness: To get loudness in the right ballpark, on Windows I have Bute Analyzer and Multimeter to set a momentary max or integrated. I was most at home with the K-meters (K-20 and K-14) so still working through LUFS for classical. There seem to be various methods. Similar to K-metering, I could set a momentary max equivalent to fortissimo and then decide whether it needs limiting and/or compression based on loudness range (between 10 and 15 is good?). Alternatively I could locate the loudest track and set that to something like -18 or -20 LUFS integrated (based on research by Tidal). I know some folks using -16 LUFS integrated for classical but certainly this feels too much for solo instruments (I find -6dB peaks about right for quieter instruments which equates to -20 LUFS in many instances). For full-scale orchestral/choral, I can set the momentary max as high as -10 LUFS (something resembling K-14 metering?) and it ends up at or below -18 LUFS integrated. This is the area where I can’t help researching common/recommended practices and I’ve not settled on one method as yet.
Reverb: As discussed, a lot of the time I use Bricasti IRs in Fog Convolver in Windows and x42 in Linux. For algo reverbs, I love PhoenixVerb and in Linux GVerb+ gets a lot of use. It’s hard to give specifics but I end up with concert hall or church/cathedral for most of my work set 33% or less depending on how much real space was captured.
EQ: It’s rare these days that I use any EQ especially if it going to be an archival recording. But, when I do, I will use Flux Epure or Tokyo Dawn SlickEQ Mastering in Windows or in Linux , x42 or Harrison’s XT-EQ (now free on their website!).
Compression: Flux Solera or Pure Compressor on Windows (gentle ratio 1:1.1 to 1:1.25, low threshold c. -30dB) and on Linux, x42, LSP and more. I enjoy transparent parallel compression (as per Bob Katz in his mastering book) but these days I try to avoid compression entirely or use regular as it seems to “glue” things together better and gives some more presence. I understand that for many pro classical releases, an external box such as a Jünger is used to create transparent dynamics control (far better than any digital plugin could). I have tried using automation curves but find it to be time-consuming and my ears, at least, seem more bothered by block changes in background ambience than a simple really low ratio compressor. It goes without saying that I’m much more comfortable transparently lopping of short transients with limiters than bothering to draw in volume curves…Life seems too short
Limiting: Purely to catch stray peaks from percussion etc, I set a brickwall limiter to -1dB. On Windows I use Flux Elixir, Bute Limiter or Loudmax and on Linux, it is Loudmax or now x42 while using Ardour export to calculate true peak to -1dB. I’ve only done limited tests on this so far so hard to tell if the desired LUFS integrated would be lowered too much for true peaks not caught on the master buss.
Album ordering: Ardour/Mixbus has the wonderful CD track markers that will always snap to CD frames. It would be amazing to have the Pyramix/Sequoia method of being able to easily rearrange tracks (rare but real time-saver when necessary!). Also useful would be the Pyramix method of being able to change pause length between tracks just by typing in a new value in the CD tab. Other than that, Ardour is as good as any DAW out there.
-
Dither and export formats: Living by Bob Katz’s recommendations, if I export to 24- or 16-bit I will dither. I now use simple triangular dither having experimented with airwindows NJAD. Something didn’t sound quite right for classical (edginess and ear fatigue) and with triangular my files suddenly sound much smoother. Most often in Windows, I have taken to exporting at 32-bit float at whatever sample rate I was mastering in (regularly 96k but increasingly often 44.1k) and run it through Fre:ac or EZ CD Audio Converter which both helpfully allow import of a cue file to create the separate files. I will create 44.1/16 FLAC, 320kbps MP3 (overkill?) and then a 44.1/16 single wav so that I can create a DDP using Andreas Ruge’s
cue2ddp
. I’ve been convinced by various people (including excellent Xiph videos) that final formats don’t need to be higher than CD quality. I reached out to the Fre:ac developer because I don’t know what DSP is used for bit depth change but never heard back. EZ CD Audio Converter uses SoX so super-happy with the quality. On Linux, I use the built-in tools in Ardour/Mixbus. This means Secret Rabbit Code for resampling (excellent!) and then triangular dither. I have yet to really sit down and analyze noise-shaped dither, POW-R etc but I’m sure that there might be some subtle improvements if one listened hard enough.
As a side note, I love the Ardour export analysis graphics that you can view and save as well as the loudness/peak normalizing stuff. Ardour’s ability to export multiple formats at once is a real time-saver and is something that Sequoia 15 only just received (and Magix very awkwardly tacked it on as a separate button from regular export dialog so it feels like an afterthought).
Tagging: I finish up any necessary tagging in Kid3 on both OSes. This is great for dragging in album artwork as well as auto track numbering etc.
Graphics: I do everything in Scribus (1.5 branch) on Linux using GIMP or Pinta as necessary for any image manipulation. I try to use as many open source fonts as I can as well as creative commons images from Pexels.com. I’m tempted by the latest Serif Publisher but I think the stable Scribus 1.6 will be plenty for my needs.
As you can tell, I’m in somewhat of a transition between Windows/Linux and closed/open-source for professional classical audio. When there are serious edits needed for, say, a proper recording session with multiple takes, I will gravitate towards Pyramix 12 for source-destination editing and the crossfade window. Pyramix apparently also has some of the best resampling (Hepta Apodizing) in the business as well as the album publishing which is probably on a par with Ardour’s export features. That said, for most other jobs, I will boot into AntiX Linux and happily do the entire job in Ardour or Mixbus. For a half-way house, I spent a good while using Ardour/Mixbus in Windows enabling me to make use of things like Flux plugins.
I’d love to hear what other people use and how they go from start to finish in Ardour (Linux in particular). I’m convinced that Ardour and Linux are fully ready for most classical music tasks. It’s just that when time is of the essence, I’m just really quick at source-destination and I have a weird love of crossfade windows in both Pyramix and Sequoia that make editing so fun and satisfying. Seemingly impossible edits are quick in these programs but I’m not for one minute suggesting that it couldn’t also be achieved in Ardour with a little more time and effort. And then there’s the spectral editing…