Reader in Music Technology, University of Edinburgh.
Home page: http://www.michael-edwards.org/uofe
Posts by Michael Edwards
When recording a mix, some of you may have been having problems with clipping when moving from the analogue into the digital domain, i.e. when mixing on the SSL or processing with the analogue outboard before going back into the computer via the Mytek AD convertors. Of course you could just drop the gain slider on the desk, amongst other things, but if you’re looking to do analogue mastering, you may want to use the Cranesong–which has extremely fast attack times–to peak limit and avoid clipping the convertors. In order to do this you’ll need to make sure the Cranesong is set up and calibrated properly. During a mastering session yesterday, Martin, Lauren, and I proceeded as follows
- Using the audiotester app (located on the Mac desktop at present but this may move to the app directory at some point), set up a one second noise burst. This needs to be attacks rather than a constant tone because we’re looking to limit peaks. We actually had to set this software’s output at +5db (not its default -20db), in order to get enough gain from the app to the DAC (i.e. to output just below digital full scale).
- On the patchbay, we routed through the Cranesong as a mix insert on the AWS900. This means that you can then enable and disable the Cransong via the main stereo busses on the bottom right of the mixer. You could also set up an external insert in your DAW if you prefer.
- We set the Cranesong so that we’re not using a preset and can therefore set the attack and threshold etc. manually. To do this, set the preset knob to the V at the bottom (see image below).
- Then set the attack and release to 10 and threshold to 0 so that the compressor isn’t working at all. This way we’re only using the Cranesong as a peak limiter, but of course you could use the compressor circuitry too.
- Make sure stereo link is enabled and adjust the gain. Bear in mind that even in stereo link mode the two gain pots work independently so that you can adjust the stereo image.
- Then it is a matter of setting the peak threshold to the right level so that it is not clipping the Mytek ADC (when you look at the ADC meters on the Mytek). On the Cransesong this is set to 19db. We’ve marked this on the Cranesong for reference.
- Set the meters to GR (Gain Reduction). You should be able to see the amount of gain reduction that you’re getting from the PEAK limiter. A rule of thumb is perhaps not to exceed 2db or so.
Kev and I have just been doing a sanity check, testing the presence of the sub when mixing. When monitoring stereo via the rec or mix buss, or surround via EXTA, we can confrm that turning on bass management splits the signal at 80Hz and feeds the satellites with everything above, and the sub with everything below. When bass management is off, the full frequency range goes to the satellites and nothing goes to the sub, in both stereo and surround modes.
So, the only way to have the sub in your monitoring is to turn on bass management by selecting the MONITOR OPTIONS button followed by BASS then BMN.
Links to PDFs of studio gear manuals:
Radial 8-channel mic splitter
TC Konnekt 32
tc electronic Konnekt32
Manley Massive Passive EQ
This is Kev’s email explaining how to use the SSL total recall. Martin and I were able to successfully save and recall mixes. Fantastic to be able to do this. Once you’ve loaded the saved mix you can hit the little square button to the top left of each fader; you’ll then see the channel controls’ states on the TFT display: move them manually into their saved position (white cursor changes to colour on lock). You won’t be prompted to do this (like on Neve USB equipped devices) but it works nonetheless.
Further details on Total Recall by Christos Michalakos:
By invite only but please contact us if you’d like to attend:
We’re celebrating the opening of the Reid Studio with a short wine reception and live recording session with Paul Harrison’s new band “Breach” on October 4th. Please join us for a glass and some jazz to hear how great things sound in the new facilities. To summarise:
What: Reid Studio Inaugural Celebration
When: October 4th 2010, 5.30-7pm
Where: Reid Concert Hall, Bristo Square (please use side entrance)
Made up of award winning Scottish jazz musicians, Breach are a new trio playing bold original music made up from international rhythms, sinewy melodies, glitchy textures, captivating grooves and flowing improvisation.
Paul Harrison (keyboard / electronics)
Graeme Stephen (guitar / effects)
Chris Wallace (drums / percussion)
lapslap are Michael Edwards (saxophones, laptop), Martin Parker (horns, laptop), and Karin Schistek (piano, Nord synthesiser). We recorded free improvisations in the Reid Hall from the 8th to the 12th of September 2010. This was to be our fourth album on Leo Records. On September 11th, the American percussionist and 9/11 survivor Fritz Welch visited us and guested on a 3-hour recording session.
Aware in advance that the decision was insane, Martin and I decided to engineer the recordings ourselves, despite being performers. We wanted to put the new studio through its paces and try out some of its more esoteric possibilities, e.g. ADAT sends from the studio to the hall via the CAT 5 extenders. We couldn’t inflict that and the up to 12-hour days on some poor unsuspecting engineer….
In contrast to our last recording–where we aimed at maximum separation of the instruments to isolate signals–we took the counter-intuitive approach of placing the instruments close together, relying on spot mics for separation where necessary. The theory was that, whatever bleed happened, say, from the sax into the piano mics, if they were close then we wouldn’t have strange-sounding delays i.e. the bleeding signal (if you will) would at least be direct, not reverberant. Having listened to the results I have to say this worked.
We wanted to capture the acoustic instruments with both close and distant miking techniques. To this end we used Neumann U89s in cardioid mode as the main air mics on the sax and horn. These were placed about 15-30 centimetres from the horns’ bells and they were isolated with SE Reflexion Filters. As we’re fans of double-miking horns etc. Martin also used his beloved DPA 4061 omni as a clip-on whereas I opted for an Electrovoice RE-20 as my second on the saxes.
The piano was close-miked with two AKG 414s (in wide cardioid mode) about 15 centimetres from the bass and treble strings. These were mainly aimed at picking up Karin’s inside-piano effects but also blend nicely with other mics to vary the closeness of the sound (if you’re careful to avoid or deal with the proximity effect). We used two Schoeps omnis on the piano as well. These were spaced just over a metre apart and a similar distance away from the piano. Clearly these picked up a considerable amount of horn signals and room as well, but the omni pattern at this distance transduced some amazing bass from the Steinway piano.
The ensemble was captured by a central Schoeps mid-side coincident pair with the cardioid focussed on the piano. The pair was about two metres high and perhaps three metres away from the piano and horns. The horns had clear sight lines to the mic pair and were facing pretty much straight into the sides of the figure-of-eight.
The last pair of Schoeps omnis was in row three of the audience seating, about five seats or so in. The main aim of these was to pick up hall ambience and if we’re lucky these will allow us to get away with adding no artificial reverb to the instrument signals. (It should be added, however, that we did put Altiverb reverb on the laptop signals when we monitored during playing. This really helped the performances and without it the laptop processing would have been too dry to interact with sensitively. We recorded these signals dry though, and will add varying degrees of reverb to them in the mix, perhaps even using Altiverb again with an impulse response of the hall which we took at the close of the final session.)
Placing Fritz was difficult as there was no room in the centre for him to set up the percussion. We opted to put him on the rear raised stage, about 3 meters behind the piano, with clear sight lines to the mid-side and ambient mics. We used a Neumann U89 as an overhead (Martin donated his and took the Neumann TLM 103 I’d brought in for his horn) along with a Sennheiser MD421 Mk2. I was surprised how well matched these were as a pair, actually.
Despite the studio being designed as a 16 channel system we actually recorded 24 channels. This would not be possible on our ProTools HD system (16 channels total with the two Myteks with ProTools cards), but as the 8-channel limitations of the ProTools Core Audio driver had forced us to buy a Lynx AES-16e-SRC card anyway (to support Logic, Nuendo etc.) we opted to record on Nuendo (my preferred DAW) using an Aggregate Device made up of the Lynx and the TC Konnekt 32. We’d mainly thought of the latter as a digital format converter but it’s actually a pretty solid 16-channel firewire sound card too.
I have to admit that I wasn’t absolutely confident that this approach would work so we had a backup plan involving a separate computer to record the 8-channel ADAT stream from the laptops. The idea was to sync the two recording systems through such a primitive device as a hand clap, aligning these when merging the tracks onto one system. Thankfully we didn’t have to do this.
The Aggregate Device (now available to all you Logic and Nuendo users as the “LynxTC” device on the studio Mac) was rock solid. It offers 32 channels of digital input and output over Firewire and AES via the patch bays. The first 16 channels are the Lynx card, the second 16 are the Konnekt 32.
Below are my pre-production notes. Martin and I both have the possibility in Max/MSP to output 4 channels for quadraphonic playback. Thinking of a possible future surround release we decided to capture these rather than record a stereo mixdown.
24 input channels needed: 4 piano, 2 synthesiser, 2 horn, 2 sax, 2 percussion, 2 room, 2 distant ambient, 4 martin laptop, 4 michael laptop (16 mic inputs, 8 digital over the ADAT extenders) so we'll need an aggregate device made from the lynx and the tc konnekt to make 24 channels total I/O. first 8 mics go straight into desk, from there to mytek 1 and from there to lynx at 96k (clocked from mytek1) last 8 mics go into desk, from there to mytek 2 and from there into tc konnekt (running at 96k also, clocked from mytek1) 8 digital go from martin's fface to adat extenders, from there to the adat->aes converter and from there into the lynx, using SRC on the lynx for 48k to 96k conversion. compression: we'll put limiters on one sax and horn mic (SSL dynamics 1&2), and on the piano close mics (Vertigo). threshold c. -3DBFS, ratio c. 10:1, fastest attack poss, so it's only there as a protection should we get an unexpected transient. we use the desk's track busses 1-4 for each of the instruments' mono mixes (for laptop processing and headphones monitoring), routing these to the 3rd mytek running at 48k (internal clock). nuendo is set up to do the headphones mix of the max signals over the desk's track busses 5&6, also routing to the 3rd mytek. 3rd mytek then sends 6 AES to the adat-aes convertor and into the extender back up to the hall. so the adat loop is: michael 4 channels of maxmsp to martin, martin adds his 4 channels and sends all 8 over adat extender. studio routes back over the adat extender 4 mono channels of instruments for processing plus a stereo headphone mix of the laptops. this goes to michael's adat in and he routes the 4 monos to martin along with the 4 laptop; the headphones mix goes out michael's analogue outs to the headphone amp: IN from extender michael 1: piano michael 2: horn michael 3: sax michael 4: guest michael 5: headphones L -> analogue out michael 6: headphones R -> analogue out michael 7 michael 8 OUT to martin michael 1: piano michael 2: horn michael 3: sax michael 4: guest michael 5: max 1 michael 6: max 2 michael 7: max 3 michael 8: max 4 IN from michael martin 1: piano martin 2: horn martin 3: sax martin 4: guest martin 5: michael max 1 martin 6: michael max 2 martin 7: michael max 3 martin 8: michael max 4 OUT to extender martin 1: martin max 1 martin 2: martin max 2 martin 3: martin max 3 martin 4: martin max 4 martin 5: michael max 1 martin 6: michael max 2 martin 7: michael max 3 martin 8: michael max 4
Obviously, running essentially two digital systems at two different sampling rates is not ideal. We had to do this though as we wanted the sonic benefit of recording the mic signals at 96k, even though the laptops were limited to 48k (any higher and the CPU couldn’t cope with what we needed to do). However, the 96k system (Lynx, Konnekt 32, two Mytek convertors) is all clocked from the first Mytek. The 48k ADAT system (Mytek 3, ADAT->AES convertor, two laptops) is clocked from the third Mytek and feeds into the Lynx, which does sampling-rate conversion (SRC) from 48k up to the recorded 96k. This is the reason we couldn’t route the ADAT signal into the Konnekt 32. This would be the ideal choice if everything was at the same sampling rate, because the Konnekt 32 would do the ADAT to AES conversion for us. As it has SRC on its inputs too, we thought we could use it even with the two sampling rates, but it turned out that as soon as we ran it at 96k, it thought ADAT signals needed to be S/MUX’ed so our signals got munged. The Lynx with SRC turned on was the way to go then.
Beware though: the Lynx only does SRC on inputs, not outputs. I was hoping it would be bidirectional so, for instance, in a 96k session we could still use, say, a Fireworx FX processor running at its maximum 48k i.e. SRC’ing both out and in. Not possible I’m afraid.
You might still have expected–and I did wonder–that coupling the separately clocked 48k and 96k systems via the Lynx–even with SRC–might cause dropouts and other nasty little clocking problems. But it seems the Lynx handles this perfectly and whatever it does with the incoming clock and the external clock that’s driving it, it works. Once the system was up and running it didn’t give us a single problem.
When the TC Konnekt 32 firewire cable is in (i.e. when it’s running as a sound card and not just a digital format converter) you can’t change the clock source and sampling rate settings on the front of the hardware. Instead, you have to change them in software, with the TC Near Control Panel (in the Applications folder), on the System Settings page. Set it (and Mytek 2 if you’re using it) to the sampling rate of Mytek 1, and the clock source to external word clock. Similarly, change the Lynx clocking to external and set its sampling rate in the Lynx control panel. (Both this and the TC Near software will start up automatically on the studio Mac.)
If you change the sampling rate of e.g. Mytek 1, you have to change it in the TC Near and Lynx control panels too (and Mytek 2 if appropriate). Always make sure all systems are running at the same sampling rate (unless you’re using SRC). If you don’t, you may not immediately notice problems, but you’ll probably find drop-outs (perhaps as long as half a second), digital burbles or pops, or various other nasty things creeping into what should be a pristine recording.
If you’re having problems getting the TC Konnekt 32 to work as a sound card and locking properly in the LynxTC aggregate 32-channel device, open Nuendo or Logic and first load the TC Near driver as if you were only going to use it as a 16-channel firewire sound card. The sampling rate should then be alignable with the Lynx and Myteks. If that works you can load the LynxTC Aggregate Device and all 32 channels should appear.
The studio is now fully 24-channel compatible and sounds fantastic. Really, I’m not sure I’ve ever heard anything sound better than this. The combination of top-notch mics, SSL pre-amps and analogue processing, Mytek AD/DA conversion, and the PMC speakers, is a real winner.
We’re going to edit the sessions in our home studios using Nuendo and the SSL Duende channel strip plugins (probably no compression though). When we’re ready with the mix, we’ll move to the desk and transfer the Duende settings to the desk’s analogue EQ and use all 24 channels to create an analogue sum (maybe using the Mixbuss compressor too). I’m looking forward to that. I’ll post sound examples and photos asap.
To get to know the various dynamics, EQ, and reverb effects on the TC System 6000 I tried a mastering session with the 5.0 mix files of my piece 24/7: freedom fried for viola d’amore and computer. This was recorded in September 2006 at ZKM Karlsruhe, Germany, by Garth Knox. It was (is?) to be released on the Wergo label with video art by Brian O’Reilly; however it seems to be still mired in legal issues relating to the other pieces on the disc. [ update: someone must be listening: it’s out. ]
We recorded the viola d’amore in surround and even went to the trouble of re-recording the electronics in the same hall. We did this by playing the files through four D&B speakers and recording with a surround mic array in order to capture the ambience and create a more natural sounding mix with the live viola. I was never really happy with the room sound though, so I was curious to see if a little TC surround reverb would help out.
The mastering chain
The mastering was a three-step insert chain process: MDX5.1 –> 5.1 EQ –> VSS 6.1 Generic Reverb
The MDX5.1 dynamics processing was perhaps the most impressive effect used here. It has a radically different approach to dynamics, raising lower levels but not higher, so there’s an overall increase in weight to the sound but the transients don’t get squashed. Very effective; very slick. There’s a soft limiter (which I didn’t use) and a brick-wall limiter too (which I did use, just for safety–the limiting light did flash a couple of times in the piece but no more than that).
5.1 EQ was a joy to use after being forced in the past to EQ surround mixes with an array of three stereo plugins (or even worse: three passes with stereo outboard). Because there was some pretty untamed bass in the original mix (I thought I’d monitored correctly back then…hmm….) I did a pretty heavy high-pass (9DB per octave) starting as high as 68 Hz. I also lifted a touch with a shelf at 8Kz and two fairly wide parametrics at 700Hz and 2.5KHz to give the sound a little more body and presence. I was surprised at how characterless, or rather transparent, the filters were; I was also shocked at how much boost or cut could be added without wrecking the sound (no harshness to my ears).
At this point I had to print the effects to disc as–running at 96KHz–I had no more processing power on the TC to do the reverb.
As I had a complete 5.0 mix I used the VSS 6.1 Generic Reverb algorithm in 5.0 mode.* I used the Vienna hall preset and first tuned the reverb alone by fading the early reflections and dry sound right down (there’s no wet/dry setting here). I low-passed the reverb drastically, taking out everything above 1.5KHz, as well as everything under 174Hz, if I remember correctly. I then adjusted the decay time to around 1.5secs, where I had the feeling that the individual events weren’t bleeding into each other but there was quite a thick hall ambience. Then I played with the early reflections, brought up the dry level to unity, and adjusted the reverb level to -21db. I was surprised here just how much difference +/- 0.5db reverb made. In any case, without being too present as an obvious effect, the reverb added a real touch of class and depth; it couldn’t take away the original room sound completely of course, but I think it distracted significantly.
Surround channel order in multichannel files tends to cause a lot of confusion, especially in the compressed formats. The ogg vorbis file order for surround is supposed to be as follows:
5.0: front left, center, front right, rear left, rear right
5.1: front left, center, front right, rear left, rear right, LFE
I used Max to encode an Ogg Vorbis 5.0 file at 256kbps. The channel order I then got was L=1,C=2,Ls=3,Rs=4,R=5. Before listening to the piece you should probably route your channels according to the following test file:
For comparison, here’s the opening of the unmastered version:
The whole five-channel mastered piece runs to 14mins32 and yet is under 27MB, which is pretty astonishing really:
Maybe these will be of use to future Reid surround masterers:
ProTools file: TCSys6000MasterAug10
[NB due to wordpress restrictions I wasn’t able to upload a .ptf protools file, so I renamed it to pdf and all was fine (great security!). So rename to .ptf and this should work in PT8]
Screen grabs of Hardware Setup and I/O settings
Notes on the Mytek settings for digital routing to and from the TC
mytek 1: source to digital out: aes (i.e. TC return is on mytek1 aes) source to analog out: dio card1 for protools aes for TC direct (i.e. main analog monitor outs are on mytek1) mytek 2: source to digital out: dio card 1 (i.e. TC send is on mytek2 aes) source to analog out: irrelevant but if dio card 1 will send dry mix to ch 9-15 on desk desk channels 1-6 are outputs select EXTA as MON SRC (RH of desk) set channels 1-6 track busses as 1-6 also (ITU Surround Order: L R C LFE Ls Rs)
* TC distinguish between Source and Generic reverbs in their algorithms. From the manual “Until 15-20 years ago, digital reverb was mostly used as a generic effect applied to many sources of a mix. Nowadays, where more aux send and returns are at disposal, new approaches have emerged. Elements of the mix are being treated individually, adding room character, flavor and depth in more creative and complex ways. At TC, we call this a Source based approach, and we have put more than 30 man-years of development time into design and refinement of Source based room simulation. When Generic digital reverbs were invented, they stretched the DSPpower and memory bandwidth capabilities of their time; and Source specific processing was completely out of the question. Even though we may now consider Generic types to be less than ideal, they still have applications for which they may be chosen instead of their Source based cousins.”
Last Friday, Lauren, Christos, Martin, Ev, and I had a great time trying out an analogue mastering chain on Lauren and Christos’s recent live performance recordings.
We ran stereo out of Logic into tracks 1&2 of the SSL, applying significant EQ right there to reduce bass muddiness and add some high frequency sparkle. We added a little mixbuss compression before going out to the Vertigo for some more compression (AB’ing quite a bit along the way) and finally into the Manley Massive Passive for sweetening EQ.
The character of the Vertigo compression was quite different to the SSL’s mixbuss. Perhaps the Vertigo was more transparent, with a more open top-end. The SSL mixbuss was thicker, seeming to apply more glue. The SideChain filter on the Vertigo had us listening hard: the difference when in or out was very subtle but we agreed that it did affect the bass and ‘thud’ content to a noticeable degree.
We didn’t really compare the SSL and Manley EQ, rather we used them for quite different tasks: tonal balancing on the SSL as the first step in the chain, sweetening on the Manley as the last step. Spoiled? Yes.
To go out of the desk to the Vertigo and Manley we first used the handy SSL Mix Insert point on the patchbay before plumping for the actual Mix out (mainly so we could take advantage of the mixbuss compressor–we weren’t sure where this was in the chain actually*). We ran the Manley output back into the computer by patching into the DAW inputs directly (i.e. the top Mytek). This allowed us to record the mastering chain directly back into Logic.
By setting the outputs of the recording in Logic to channels 3&4 on the desk, and routing these to the SSL’s Record buss instead of the Mix buss, we were able to flick the monitoring source on the desk to compare pre- and post-outboard signals**. If we’d have added a clean send from Logic to some more desk channels we could have compared pre-SSL EQ too for some A/B/C comparison, but time got the better of us.
* Update: Seems like we did the right thing as, according to the signal flow diagram, the mixbuss compressor comes after the mix insert.
** In retrospect it might have made sense to have reversed this i.e. sent the mastering chain (channels 1&2) to the record buss and the rest to the mix buss, but according to the signal flow diagram it doesn’t make any difference.
The two blue units have ProTools cards; the grey unit does not and is intended primarly for analogue conversion connections between the desk and the digital only TC System 6000.
In all setups the top Mytek is the master, the other two are slaves (as are the TC System 6000 and the Konnect 32). If you are using ProTools, start Hardware Setup, select the 192 I/O #1, and set the clock source to internal. Unit #2 should also be visible (if not, see below) and automatically have the same source. If you then change the sampling rate in Hardware Setup, both blue ProTools Myteks will show the new sampling rate on their front panels (but you’ll have to manually set the grey unit if you’re using it).
If you’re not using ProTools with the Myteks, even when they are word clocked, you still have to manually set the sampling rate on all units.
If the word clock light is flashing on the slaved ProTools Mytek (i.e. it’s not clocking properly), you’ll probably find it’s also not showing up in the ProTools Hardware setup–until it does, it won’t lock to Word Clock. To get it to show up you can try starting the CoreAudio driver and/or Protools; if that doesn’t work, shut down the Mac and Myteks, ensure the ProTools cables in the back of the Myteks are properly seated, fire up the Mac and Myteks again and all should be well.
Digital Connections from ProTools via the Myteks to the TC System 6000
We do this by using a dedicated ProTools 5.1 Buss, mapping audio channels through this, picking up the returns in another buss, and sending them out of the analogue outputs on a third buss to monitor via the desk.
Although 16 Protools I/Os show up for each unit, only the first 8 are valid for each Mytek unit.
The following will no doubt change once the AES patchbay is in, but at the moment the middle Mytek will send via AES to the TC and the TC will return via AES into the top mytek.
In I/O in Protools:
#1: input: AES1-8, output: analogue 1-8
#2: input: AES 1-8, output: AES 1-8
On the top Mytek: set the “ADC Source to Digital Out” to AES in order to pick up the send return from the TC. With the “DAC Source to analog out” set to DIO Card, you’ll be hearing the ProTools output via channels 1-8 on the desk (which is what you want); set it to AES and you’ll be listening to the FX return from the TC before it’s altered in ProTools (handy for monitoring).
Middle Mytek: Set “ADC Source to Digital Out” to DIOCard 1. The “DAC Source to analog out” is irrelevant here but if you set this to DIO Card it will send the dry (i.e. pre TC) signal to channels 9-15 on the desk.
TC System 6000
Make sure the clock source is Word Clock: go to System (?), press the clock source and move the right fader down until you see Word Clock. At sampling rates above 48K you’ll have to engage “Double Speed” before the unit will lock.
To map AES inputs through the algorithms and back out again, go to Routing (?), select Labels, then use the two left faders to connect inputs, and the two right faders to connect outputs.