String Quartet Panning

[Ed Sharpe]

String Quartet Panning


I am working on a new recording of my string quartet #1 (Yuki):

http://www.sibeliusmusic.com/index.php?sm=home.score&scoreID=45934 (The old recording is there with the score.)

I am playing with various pan positions for the four instruments.

I am not sure if there is a standard seating plan, the best I can find is from left to right is Violin 1, Violin 2, Viola, and Cello. However right I have also seen reference to Violin 1, Viola, Cello, Violin 2 seating.

Right now I have the pan set to Violin 1, Viola, Violin 2, Cello.

Any feelings on anything of these?


Violin 1, Violin 2, Viola, Cello.

Violin 1, Viola, Cello, Violin 2.

Violin 1, Viola, Violin 2, Cello.


How wide should I make the pan?


Ed

[qccowboy]

1 or 2 would be acceptable, although 2 has its weaknesses (the 2nd violin is turned away from the audience by being placed on the "wrong" side of the stage) there are still some groups that use it. 1 is the more common.

3 has no reason whatsoever. it's just... wrong.

[rbowser-]

Hi, Ed

Since you're going for a "as if recorded in concert" recording, I like plan 2:

Violin 1, Viola, Cello, Violin 2.

With the Cello only just right of center. Lower frequencies sound best at center or close to center, but you'll still be in the realm of seating that's actually used in real life.

I wouldn't pan very wide. The wider the pan, the closer the listener/audience member is - and getting very wide would put them on stage with the musicians. Just a slight spread, with the outsides being maybe around 25% from center in the two directions.

Randy

[etLux]

A second on Randy's opinion.

One of the most common errors in rendering a
string quartet synthetically is panning too wide.

Remember, on a real stage, the players are just
a few feet apart.

David
-----
David Sosnowski
www.DavidSosnowski.com

[danpowers]

In my quartet we usually sit this way: Violin I, Violin II, Cello, Viola http://www.customerssuck.com/board/images/smilies/smileywaving.gif

I think that's a pretty common seating pattern.

[Tim Perry]

Hi Ed,

Most of the string quartets that I've recorded in concert were in arrangement 1. While conceptually I like the idea of the violin harmony lines to be right and left, the stereo spread may be too great and some homogeneity may be lost.

I'm assuming that you are placing the sound sources virtually using sample libraries (ie you are not doing a live string quartet recording)? I'll talk a bit about recording a live string quartet, and hopefully you will be able to intuitively translate what I mean as it applies to your scenario. Both cases involve visualizing sound images in a stereo field, and while writing this post I thought of a way to potentially recreate the natural string quartet panning with pre-recorded samples.

I've recorded a number of string quartets in the past, and for stereo string quartet recording I favour an omnidirectional spaced pair configuration with a wide spacing (microphones about 3 feet apart). The amount of reverb and warmth desired is effected by how far the mics are spaced apart, how high they are located, and how far they are from the performers. The stereo spread is obviously effected by the spacing of the performers, the distance of the mics from the performers, and the spacing between the mics in a spaced pair configuration (it depends on other things as well for different mic configurations). A spaced pair microphone configuration encodes stereo cues as timing differences between our ears, called "interaural timing differences", or ITD. This can be replicated in a DAW by placing left and right channels slightly out of phase (more on this coming up).

Panning can be accomplished using amplitude cues, timing cues, frequency cues, or any combination of those things. Panning in a DAW or mixer is usually done by changing the relative amplitude between channels. However, when we hear the stereo spread in real life our hearing system is using mainly timing differences between the ears and frequency cues (part of our ear called the pinna acts as a filter). I probably did not need to mention all of that, but my point is this:

Imagine the stereo field in terms of space and time--and if you want to do highly realistic panning, then I think the following departure from pure amplitude panning would be a great thing to try (this is what I would do if I was panning a string quartet with a sample library--it basically involved virtually mimicking a live recording):



check page 26 of the pdf hosted here: http://www.engr.uvic.ca/~timperry/Documents/Elec499-SurroundSoundIR.pdf (http://www.engr.uvic.ca/%7Etimperry/Documents/Elec499-SurroundSoundIR.pdf). This diagram is from an old project of mine that's not directly related, but there are four squares on the stage diagram labeled "Qfl, Qbl, Qbr, Qfr" (corresponding respectively to 1st violin, 2nd violin, viola and cello). Dimensions are in cm. This is a bit wider than a typical live recording position, but this is where I would llikely place them if I had the mics located at "L1". If I was using a stereo pair (which I probably would use for a string quartet) I would locate the mics a bit closer (maybe 1/4 of the way between the end of the stage and the position labeled "C" and several feet above the heads of the performers, depending on the room acoustics) and move the positions a bit closer together.



virtually transpose this mic configurations stereo imaging to your recorded sample tracks using time delay instead of or in addition to amplitude panning (in which case, the amplitude panning would be less than usual). This could be done using math, or intuitively done through experimentation. Likely the easiest way to do this is to first duplicate each track (if you had a single mono track for each instrument, 4 tracks total, you would copy each one such that you now have 8 tracks). For violin 1, it would now exist in two tracks that you could name violin1L and violin1R. To create the panning, you would first amplitude pan violin1L 100% left and violin1R 100% right. Next, you would offset the track (or apply a latency/delay) so that violin1L starts sounding a fraction of a second (maybe try 10-15ms as a starting point) before violin1R. This will pan violin 1 to the left of center using timing cues (if I recall, a difference of 20ms between the ears will make something sound as if it's completely panned to one side... this information is available online and in a book I have, so if you decide to try this please give me a shout and I can dig up more specific info)



Frequency cues between the two ears can also be toyed with by applying differing EQ to each left/right track. For example, for violin1 in the given set up you would want the track violin1L to sound brighter than violin1R. This is a refinement that may not be necessary for a desirable result (a typical spaced pair recording will not have much in the way of frequency cues, but a highly realistic binaural recording will ~|:hp:~| in order to simulate surround sound the way our head does). For most peoples needs we can skip this step and move onto the next, crutial point...



Create the spatial impression :wow: using EQ, gain, and convolution reverb. Apply EQ on the Qbl and Qbr tracks in such a way that they sound further away than the Qfl and Qfr tracks. I mention a bit about how you can do this sort of thing on pages 3 to 5 of this pdf: http://www.fileden.com/files/2007/11/23/1602607/My%20Documents/Uvic%20Recordings/TimPerry_Elec340-A1.pdf. Reduce the gain on the Qbl and Qbr tracks so that they have slightly less amplitude, and also send slightly more of these tracks to the convolution reverb (ie a wetter sound for instruments that are more distant).



Well, after writing all of that I want to try it myself sooner or later and possibly write something up about it. I hope that this helps you out, or at least brings up some ideas. Does anyone have any insightful thoughts on this? Let me know if anything that I mentioned is confusing (if you are wondering what's with the emoticons... well, this post did not look like an inviting read at first glance, and the crazy eyes do go hand in hand with with what I'm visualizing).

--Tim

[DPDAN]

Above and beyond all other panning methods known to man... ~|
Altiverb by Audio Ease is going to give you the ability to manipulate your
samples into the most realistic environment you could ever imagine.

Simply panning the dry tracks and "adding" reverb to the dry samples
is NOT going to make a convincing mix.

Dan

[Tim Perry]

Simply panning the dry tracks and "adding" reverb to the dry samples
is NOT going to make a convincing mix.


....So so so true, hence my huge shpeal above.

I knew that Altiverb allows stereo placement, but I was under the impression that the positions were fixed (if not, then awesome!). If someone could not justify buying Altiverb (maybe it was food on the table vs Altiverb), then they might attempt to do/find an effect that does what I mentioned (the stuff past the blue text) with the aid of a cheaper convolution reverb. I know that it's still not ideal as it will not be using a different impulse response for each instrument's location, but adding timing cues and depth perception EQ will take care of some of that. The main missing ingredient in my mind would be the first reflection data... so certainly having IRs for each instrument position is the most desirable/practical appoach.

[DPDAN]

Hi Tim, great to see you here :)
the very things that you spoke of are incorporated into Altiverb's stage positioning feature.
IT IS AWESOME!!!!

Check out these video demos I did a while back about using Altiverb.

http://www.northernsounds.com/forum/showthread.php/68209-Altiverb-tutorials?p=627317

I was blessed about four years ago and was finally able to buy one of these Schoeps CMXY4V stereo mics..
After using Neumann's, AKG's and many other great mics, all I can say is... now I know why people use Schoeps mics for classical music. Just amazing! There are two, (matched of course :) side addressed cardioid capsules that pivot to adjust vertical angle and the two capsules rotate with a unique gear mechanism to adjust the width of the pattern. A person can sing/speak and move from extreme left to extreme right and there is no noticeable difference in sound, no phasing issues of any kind. So cool!

OK back on topic :hp:

http://www.dankury.com/photos/SchoepsCMXY4V.jpg

[DPDAN]

Ed, here is a link for "All I Ask Of You (http://www.dankury.com/music/AllIAskofYou.mp3)" from Andrew Lloyd Webber's Phantom of The Opera.

1st and 2nd violin are the Garritan Stradivari,
viola is from GPO and the cello is Garritan's Gofriller Solo Cello

I did dome recording with the Arianna String Quartet and they sit from left to right like this...


...2nd..cello
1st..........viola

This mock up recording I did has the following seating (panning)

...viola..cello
1st..........2nd

http://www.dankury.com/music/AllIAskofYou.mp3
Dan

[Tim Perry]

Great video demos Dan--thanks for making/sharing them. I've used Altiverb, but it looks like I was missing out on that customizable stage positioning magic (er... engineering).

Schoeps are fantasting, and when I did recording for the University of Victoria the secret weapon (albiet used daily) was a pair of CMC 6-U. But even better-- I had this microphone array in my possesion last summer for a research/design project (but no, I don't own it!):


http://www.engr.uvic.ca/~timperry/images/PTY%20Hall%20&%20Array%201%20Cropped-THUMB.jpg

Bigger Version: http://www.engr.uvic.ca/~timperry/images/PTY%20Hall%20&%20Array%201.JPG (http://www.engr.uvic.ca/~timperry/images/PTY%20Hall%20&%20Array%201.JPG)


It's a spatial array that holds five Shoeps CMC 6-U mics (with hypercardioid capsules) and two vertical oriented shotgun mics (not shown in the above photo). Better photos of the rig are on page 18 of this pdf:


http://www.engr.uvic.ca/~timperry/Documents/Elec499-SurroundSoundIR.pdf (http://www.engr.uvic.ca/~timperry/Documents/Elec499-SurroundSoundIR.pdf)

[DPDAN]

WOW that's what I would call TMI :)

I am so glad I only use this stuff and don't have to design it.
WHEW!

you are one smart cookie ! :)
Dan

[raweber]

virtually transpose this mic configurations stereo imaging to your recorded sample tracks using time delay instead of or in addition to amplitude panning (in which case, the amplitude panning would be less than usual). This could be done using math, or intuitively done through experimentation. Likely the easiest way to do this is to first duplicate each track (if you had a single mono track for each instrument, 4 tracks total, you would copy each one such that you now have 8 tracks). For violin 1, it would now exist in two tracks that you could name violin1L and violin1R. To create the panning, you would first amplitude pan violin1L 100% left and violin1R 100% right. Next, you would offset the track (or apply a latency/delay) so that violin1L starts sounding a fraction of a second (maybe try 10-15ms as a starting point) before violin1R. This will pan violin 1 to the left of center using timing cues (if I recall, a difference of 20ms between the ears will make something sound as if it's completely panned to one side... this information is available online and in a book I have, so if you decide to try this please give me a shout and I can dig up more specific info)

Good stuff, Tim. An easy way to implement this time delay panning is with the free Hass Delay (http://www.vescofx.com/vfxFreeHaas) plugin. I've used it to great effect. What it does is takes a mono track, makes it stereo and delays one side (the side opposite where you want it to appear on the stereo stage). According to what I've read, 40 ms is about where the effect transitions from sounding like panning to sounding like a sort of bizarre flanging.

Good luck, Ed!

Rob

[Tim Perry]

Thanks for posting that link Rob. Hass Delay looks like a nice little free plugin that could come in handy to simplify panning via timing cues. Some mono recordings could start to come to life with just some basic tricks using this (plus a convolution reverb).

I think that bringing the soundfield to life is one of the most overlooked things in home recording (likely because it's not easily to do on a budget).

[Ed Sharpe]

Thank you all for the time and the replies. It has help me a lot, even though a lot of it went right over my head.

Thank you again.

Sincerely,

Ed

[DPDAN]

.... even though a lot of it went right over my head.

don't feel bad Ed, Tim's post had some stuff I have never even seen,
let alone understand, or have the capacity to understand :) he he he

Dan

[Tim Perry]

Hi Ed,

I hope that I did not make the string quartet panning seem more complex than it actually needs to be (but I probably did). I have a hard time being concise with topics that really interest me (like this one)--start typing a bit and then a half hour later there is a conglomorate of too much material on the page.

The main thing that DPDan and I were putting across is: find some way to go beyond pure amplitude panning. I mentioned a way that you could do it manually, but it would require some work. DPDan mentioned that Altiverb can take care of all of that for you and makes it very intuitive. raweber posted the free Haas Delay (http://www.vescofx.com/vfxHaas) plugin that allows you to do time delay panning (taking care of the main point I was talking about).

[Ed Sharpe]

Dear Tim,

No worries about too much information. I appreciate the time you took and it did help.

...

The main thing that DPDan and I were putting across is: find some way to go beyond pure amplitude panning. ...


I get lost at even simple statements like that. But not worries, I am slowly learning.

Note: I am an old school composer, everything still starts on paper. All the "recording" stuff is new to me. The only times I have been in a recording studio is when I am on the "other side" playing. As for DAWS I only know enough to be dangerous :wow:

[Tim Perry]

Amplitude panning is what your DAW uses when you pan something left or right. Stereo tracks in a DAW are made up of one track panned 100% left and the other panned 100% right. The idea of going beyond amplitude panning could be as simple as using the Haas Delay (http://www.vescofx.com/vfxHaas) plugin instead (or in addition to) the panning controls on the daw. But then, simple does not necessarly mean easy.

Unfortunately, I don't know an easy recipe (other than maybe Altiverb) for realistic string quartet panning. The suggestion that I gave in an earlier post should work, but it may be a lot to go through. The most important step of that process is whats mentioned in bullet point 2 of the long post. Doing that or using Haas Delay (one or the other, not both) will give you time delay panning.

But, maybe it makes more sense to take it one step at a time and not wory about this stuff for the time being. It all depends on how ambitous your goals are for this project.

[Peter Jeffrey Gale]

An easy way to implement this time delay panning is with the free Hass Delay (http://www.vescofx.com/vfxFreeHaas) plugin. I've used it to great effect.


The idea of going beyond amplitude panning could be as simple as using the Haas Delay (http://www.vescofx.com/vfxHaas) plugin instead (or in addition to) the panning controls on the daw.

Let me see if I've got this right ... so because of the Haas effect I can use the free Haas plugin to create L-R panning with no loss of volume?

... yep that seems to work a treat with almost no effort at all! ... it even takes a stereo signal so there's no need to bother splitting the track into separate L & R tracks.

... and presumably I can use the pre-delay in my convolution reverb to create a front to back "panning" effect?

That seems to work quite well too!

Thanks for lead, guys! :)

That'll do nicely until I can afford Altiverb. ;)