Topic: !!Stereo Zoom simplified for PAS!!

[Gutbucket]

[Edit- a new thread with both the simplified table below and a revised and extended table and discussion about it can be found here]


Point mics at stacks and the resulting mic angle determines your mic spacing.  Like dinner in reverse, I’ll start with desert, then explain the why and how I got here if you care to keep eating.

To try this improved, but still relatively simple to use PAS mic setup technique you will need-

1) Two cardioid pattern microphones.
2) A mounting system for the microphones which allows you to adjust both angle and spacing between them.
3) A way to determine the angle between the microphones once you point them at the stacks.  You can use the simple calibrated fist method to approximate angles, or any other method you prefer.
4) This table: (I can't figure out how to use the table formating fuction here, so it's spaced with underscores)

Mic angle___Mic spacing
(degrees)__(centimeters)
__110_________9
__100________14
___90________20
___80________27
___70________36
___60________48
___50________64
 

Here’s how to go about it-

1) Determine where you will setup.  You may not have a choice if this is determined for you by the venue.
2) Setup the mics and point them at the stacks (the PA speakers).
3) Measure the angle between the microphones (or the angle between the stacks, same thing).
4) Consult the table to determine the correct spacing between the microphones for that angle.
5) Adjust the spacing to match.
6) Record.

That’s all you need to know to use it. Give it a try and let me know how it works for you.

If you want to know more, I’ll explain the what & why in a few following posts.
Enjoy!

[Edit- attached is a GIF of the same table.  Note that the setup near the center of the list indicated with grey shading is DIN.  So if the angle between the PA speakers as viewed from your recording position is 90° then running improved PAS is identical to running DIN]

[Gutbucket]

Can anyone explain or point me to something that explains the table format here before I post a couple more extensive tables?

Code: [Select]

[table]
[tr]
[td][/td]
[/tr]
[/table]
Or if there is a better way to simply preserve the column spacing for data copied out of Excel and pasted here I'm open to that.

[page]

Mic Angle,

Mic Spacing

---------------------

110	= 9
100	= 14
90	= 20
80	= 27
70	= 36
60	= 48
50	= 64

[page]

so yeah, thats how you do a table, but are you sure about the values? Did you compute that from the stereophonic zoom doc or is that just an example?

[Gutbucket]

Thanks for the insight into the tables, what a PITA.  I think I'll just post an image capture for the next one with mulitple columns.

It's Stereo Zoom data from here- http://www.sengpielaudio.com/Visualization-ORTF-E.htm, since punching in numbers in that on-line calculator is easier than referencing the charts. I'm working on a follow up that explains how I worked this up.  Basically it takes the SRA of DIN and extrapolates that, adjusting spacing to match your PAS angle.  I'm wording the  follow up for users that are unfamiliar with Stereo Zoom, so appologies to you and others here that are already familiar with the concepts.

[DigiGal]

I'm still hoping someone can create an iOS app version of that visualizer.

http://www.sengpielaudio.com/Visualization-ORTF-E.htm

[Gutbucket]

The what & why-

I was explaining the Stereo Zoom concept to someone recently and realized that I could probably simplify usage of it for them to get started by applying it to the popular and straightforward PAS (Point At Stacks) microphone setup technique.  In case you are unfamiliar with the Stereo Zoom concept, I’ll just briefly note that it is a conceptual frame work which explains the underlying relationship between the angle and spacing between a pair of stereo microphones used to record music for playback over a pair of stereo loudspeakers. It allows a recordist to trade level-based stereo cues which are derived from the directionality of the microphone’s sensitivity pattern and the angle between them against time-based stereo cues which are derived from the spacing between the microphones.  The standard microphone configurations such as DIN, ORTF, X/Y, A-B, etc. are specific solutions along a continuum and the relationship between them and how they distribute the sound sources on playback over speakers is explained by the Stereo Zoom.

I recommend reading the Stereo Zoom paper to help understand how that works.  You can download the PDF from here.  I think it is also in the TS reference section.

Playing around with the stereo microphone setup visualizer web app at the Spengpiel audio website http://www.sengpielaudio.com/Hejia.htm, which I believe uses Micheal William’s Stereo Zoom Data for it’s calculations, I plugged in different numbers and wrote down the data to make a few simple tables.  I based all of this on the DIN microphone setup, which is one of the more popular mic setups for concert recording using cardioid microphones.  The DIN setup arranges the two microphones with a 90° angle between them and a spacing between mic capsules of 20cm.  That setup has a ‘pickup angle’ of 102°, meaning that sounds arriving at the microphones from within that angle will be distributed between the two speakers on playback, with sounds at the outer edges of that angle coming directly from the speakers and sounds arriving from within that angle distributed between the two speakers as a phantom image. In Stereo Zoom terminology  the ‘pickup angle’ is referred to as the Stereo Recording Angle (SRA) and expressed as an angle from either side of center- so a ‘pickup angle’ of 102° is expressed as an SRA of +/- 51°. 

The first table I made listed all the different combinations of spacing and angle using cardioid pattern microphones that achieve the same SRA of +/- 51° that we get with the DIN configuration. Here it is:




Table of various two channel cardioid arrangements for ~100º SRA (+/- 50º)
(see attached image below - Table of two channel cardioid arrangements for SRA100.gif)




Theoretically, I can choose any of the combinations of mic angle and spacing from the first two columns of that table and get a similar spread of sound sources between the speakers on playback. They won’t all sound the same though for a number of reasons. 

For one thing choice also affects how sounds arriving from outside the SRA angle are picked up and the mic arrays sensitivity to them as a pair.  Configurations that have the microphones arranged more parallel to each other will favor sound arriving from the front more than the back and sides to a greater degree.  Another aspect is the different imaging of time based stereo verses level based stereo, which is often more of a personal preference thing.

Also on a practical level not all of the choices are equally valid, especially those that have the microphones angled very widely apart.  That’s due in part to the off-axis response of many microphones being less smooth and accurate than the on-axis response, so it’s often better to choose a configuration that points the microphones more towards the sources of sound.  Partly for that reason, and partly because I suspect the SRA relationship will break down at the extremes, I didn’t include solutions for mic angles of more than 110° and spacings wider than 64cm from the simplified PAS table in my first post. I’ll post the whole thing later if that helps the discussion.  But next I want to explain how I got there.

The constant SRA table in this post is interesting and sheds some light on developing a gut understanding the relationship, but I wanted one that changes the SRA to match that of the mic angle in an intuitive way.  More on that next..

[Gutbucket]

My first thought in doing that was to simply take the list of PAS angles and plug that into the calculator as the angle between microphones to find spacing solutions where SRA = mic angle.  But is that really what we want?  Most of the time we want a SRA that is slightly wider than the angle of sound sources (called the orchestra angle in that on-line calculator).  We discussed this a few years ago in this thread on understanding the concepts of the Stereo Zoom.  As noted in that thread, the Stereo Zoom paper mentions that many engineers prefer an SRA slightly wider than the total width of the sources (orchestra angle) and suggests + 10° as a general guide.  I’m using DIN as the reference here, and it’s SRA is +/- 51° or 11° wider than the 90° angle between mics, so I simply added 11° to each mic angle down the list to get each target SRA and solved for that.  So on the PAS table SRA=mic angle + 11°.  Sometimes it works out to 12° or so to end up with distances in whole centimeters.

BTW, I also cross referenced DINa’s  SRA= angle + 11° at a few positions and the solutions are similar, but slightly wider spaced.  The simple solution of SRA=Mic angle+0° produces wider still spacings.  It breaks down at the extreme settings because at one extreme mic angle becomes zero when both mics point directly ahead, but SRA can never reach  zero (the calculator tops out at SRAs of around 9°  by entering outrageously ridiculous inputs) and at the other extreme the mics become coincident as mic angles approach 140°.

The moral to the story is moderation in choosing values in the middle of the range which balance time-differences and level-differences. That’s why I only posted the range of microphone angles of  50° – 110°.  Note that near the middle of the range in the PAS table you’ll find DIN, which has a pretty even balance between the two with a level differences making up 53% of the phantom image shift and time differences making up 47%.

Here’s the full PAS table including all the mic angles, the SRA values from the calculator and the relative percentage of time vs level differences for each configuration-
 

[Gutbucket]

A couple points I'd like to make about the PAS technique.

We sometimes tend to think of PAS as a simple yet often good sounding starting point that's easy for new tapers to use before moving onto more advanced mic configs.  It sure is easy to setup.  Part of the problem with the technique is there has never been a specific spacing specified, the idea has alway been to simply point the mics at the stacks, which is a blatantly obvious solution to any novice before developing a deeper understanding of the complex relationship between mic angle and spacing.

But there are very good technical reasons for pointing the mics directly at the stacks which we sort of overlook with other techniques, beyond the easy setup, as long as the spacing between them plays along! And the spacing playing along is what this thread is about. Here's a few that come to mind:

1) As mentioned previously, less than stellar mics often show their warts in their less than ideal off-axis response.  Great mics are much smoother off-axis and hold their patterns far better, but all mics are most accurate on-axis, so why not take advantage of that by pointing them directly at the two primary sound sources?

2) One of the biggest challenges we face in recording bands in rooms from a location out in the audience is optimizing the direct to reverberant ratio.  That's just more fancy talk that basically means getting more of the music coming out of the speakers and less of the echoy room sound and crowd sound on your recording.  Pointing the mics directly at the PA speakers does that better than any other configuration.

So even advanced tapers with stellar gear and a good understanding of other stereo configurations may find this advantageous.  Many here like applying the Stereo Zoom, and this is nothing more than SZ applied to PAS.  Maybe a more fitting tread title would have been: PAS grows from ugly duckling >  swansong.  If you do try this please let me know how it works out.

[page]

If you do try this please let me know how it works out.

http://www.archive.org/details/isd2011-12-31.ck950.flac16

I ran 65 degrees, 26cm because of the SZ curves. The stack angle was about 45 degrees. Not exactly PAS, but fairly close. I wanted to keep the SRA of DINa but get close to PAS.

Basically it takes the SRA of DIN and extrapolates that, adjusting spacing to match your PAS angle.

gotcha, that was the bit I missed from the original bit of the post, my bad.

So one of the neatest things that I think you've done is the chart effect. If someone does up a set of tables (with ortf/dina/din/nos SRA basis), then it would be pack a set of charts, and walk off to the show and get a good 3/4ths of the way there.

[Gutbucket]

http://www.archive.org/details/isd2011-12-31.ck950.flac16

I ran 65 degrees, 26cm because of the SZ curves. The stack angle was about 45 degrees. Not exactly PAS, but fairly close. I wanted to keep the SRA of DINa but get close to PAS.

Nice, thanks, dig the Dusters.

So one of the neatest things that I think you've done is the chart effect.

Yeah, the idea was to make it as easy as possible to use with a very simple table.  The hassle I see with SZ is deciphering the graphs and applying them to real world recording situations.

[Will_S]

My first thought in doing that was to simply take the list of PAS angles and plug that into the calculator as the angle between microphones to find spacing solutions where SRA = mic angle.  But is that really what we want? 

This is somehing I've done many times with usually decent results.  Thanks for putting it in a table.  One intersting further step would be shadings towarn of large angular distortions or unacceptable direct/reverberent ratios as in the original stereophonic zoom.

Most of the time we want a SRA that is slightly wider than the angle of sound sources (called the orchestra angle in that on-line calculator).  We discussed this a few years ago in this thread on understanding the concepts of the Stereo Zoom.  As noted in that thread, the Stereo Zoom paper mentions that many engineers prefer an SRA slightly wider than the total width of the sources (orchestra angle) and suggests + 10° as a general guide.  I’m using DIN as the reference here, and it’s SRA is +/- 51° or 11° wider than the 90° angle between mics, so I simply added 11° to each mic angle down the list to get each target SRA

This is interesting but when recording a mono PA are the two stacks really the edges of the sound source, or (especially at distances where the angle gets small) should the focus instead be a convincing representation of the audience and room?  My approach from far away is often to figure out how to get at least a 120 degree SRA (so numerically, 120 or less) without "missing" the stacks by too much or changing the direct/reverberent ratio too much (angular distortion is not irrelevant, but is my lowest priority).  If set up far back, this often means a good image of the crowd with the PA sound centered - which strikes me as pretty accurate.

[hi and lo]

Thanks, Gutbucket. Another typically awesome contribution to the forums! I will try this tonight with the grace spacebar and see how it goes.

[Gutbucket]

Thanks! The Spacebar is perfect for this.  In the photos I've seen of it, both spacing and angle are indicated with marks on the bar and mounts so figuring both should be a piece of cake.

Will, you raise good points about persepctive.  I think much of it is preferece and I don't think there is any one correct answer.  I suppose the esthetic is similar in some ways to choosing a photographic perspective with zoom and depth of field foreshortening- a wide angle view of the whole scene verses zooming in for a closer perspective.  That's why it's called the Stereo Zoom I guess.  Where the analogy breaks down is that with photography what is cropped out is eliminated completely from the image, where here we still pickup sounds from outside the SRA, but they are simply pushed to the outer edges of the sonic image between the speakers. So the question then isn't as much about cropping out audience and room ambience as much as about how wide we want the music to sound. We can apply the SZ to choose different angle/spacing relationships which all have similar SRAs to help control the reveberant balance and rear sensitivity.

I generally like a wide and forward perspective, with an enveloping audience wrapping around the sides, but I think the best perspective is one which matches the other qualities of the recording that indicate distance.  If the recording sounds somewhat distant instead of close, say due to frequency balance or direct/reverberant balance, then a big, wide image of the stage filling the whole 'sonic scene' may be less appropriate.

I have similar priorities, pretty much this order of importance for me:
good direct/reverb ratio
reasonable frequency balance
appropriate stereo width
minimal angular distortion of stereo image

[hi and lo]

Thanks! The Spacebar is perfect for this.  In the photos I've seen of it, both spacing and angle are indicated with marks on the bar and mounts so figuring both should be a piece of cake.

Yup, should be perfect. The only annoyance is that you still have to measure capsule separation. The cm spacing on the bar is just a reference and doesn't reflect the actual capsule spacing. Nevertheless, I have an interesting show tonight where this method should work perfectly for testing.

[Gutbucket]

Yeah, I realized the capsule spacing would be larger than what is inicated on the bar as I was typing.  The clincher though is the clear mic angle indication which is more of a challenge to meausure accurately.

[alpine85]

My first thought in doing that was to simply take the list of PAS angles and plug that into the calculator as the angle between microphones to find spacing solutions where SRA = mic angle.  But is that really what we want? 

This is somehing I've done many times with usually decent results.  Thanks for putting it in a table.  One intersting further step would be shadings towarn of large angular distortions or unacceptable direct/reverberent ratios as in the original stereophonic zoom.

Most of the time we want a SRA that is slightly wider than the angle of sound sources (called the orchestra angle in that on-line calculator).  We discussed this a few years ago in this thread on understanding the concepts of the Stereo Zoom.  As noted in that thread, the Stereo Zoom paper mentions that many engineers prefer an SRA slightly wider than the total width of the sources (orchestra angle) and suggests + 10° as a general guide.  I’m using DIN as the reference here, and it’s SRA is +/- 51° or 11° wider than the 90° angle between mics, so I simply added 11° to each mic angle down the list to get each target SRA

This is interesting but when recording a mono PA are the two stacks really the edges of the sound source, or (especially at distances where the angle gets small) should the focus instead be a convincing representation of the audience and room?  My approach from far away is often to figure out how to get at least a 120 degree SRA (so numerically, 120 or less) without "missing" the stacks by too much or changing the direct/reverberent ratio too much (angular distortion is not irrelevant, but is my lowest priority).  If set up far back, this often means a good image of the crowd with the PA sound centered - which strikes me as pretty accurate.

Good points.   Another thing to consider is... Is the best sound that we want our mics to capture actually AT the stacks themselves?  Especially in bigger venues, with longer throw drivers, could the sound be much more balanced and accurate 20 feet out from the stacks?  50 feet?  100 feet?  In general, this is definitely a very helpful guide (and thanks for putting it together, Lee), but it seems like there will be more variables in a lot of situations.   Are you in the sweet spot?  If not, do you want to point the mics towards the sweet spot (which might or might not be directly at the stacks)?  I also tend to agree with Will_S that direct/reverberant ratio is probably a lot more important than angular distortion when you're recording a PA.  (Even if it's a stereo PA).  In fact, if you're only getting angular distortion on the reflections and on the crowd (and not the music itself), is it really a factor at all?

[Brian Skalinder]

Another thing to consider is... Is the best sound that we want our mics to capture actually AT the stacks themselves?  Especially in bigger venues, with longer throw drivers, could the sound be much more balanced and accurate 20 feet out from the stacks?  50 feet?  100 feet?  < snip >  Are you in the sweet spot?  If not, do you want to point the mics towards the sweet spot (which might or might not be directly at the stacks)?

While I agree with your comments for the most part, and recognize there are a variety of factors that come into play, I think there are two notions suggested here (quoted above) that may confuse newbies.  Perhaps it's due to the imprecision of language.  Specifically:

• The "best sound" we want to capture isn't really AT the stacks, or any other single (or multiple) point(s) source, for that matter.  Generally, the "best sound" exists at a particular spot that suits one's recording / gear / configuration / listening preferences:  typically a specific location / height in the room, i.e. the "sweet spot".
• Pointing mics "towards the sweet spot" doesn't make much sense, if one defines the sweet spot in the traditional taper sense:  the (or one of the) location(s) in the venue that sounds the best.

And, unfortunately, much of the time we're unable to select our "sweet spot", since our location is decided for us, in whole or in part, by the band, venue management, physical space, etc.  In other words, we often don't have control over the single most important variable -- location -- and must rely on the other variables we can control:  like mic pattern and configuration.  In that sense, I can see the the chart proving helpful as a quick reference.  (FWIW, I carry cut-outs of the SZ graphs in my gear bag.)

[alpine85]

You're right, Brian.  "Pointing at the sweet spot" is probably not the best way to describe it.  I was just trying to say that there are times where PAS is not ideal.  Here are 2 extreme examples:

1)  Indoor venue, long narrow room with PA stacks in the corners pointing towards the middle of the room.  You are in the back of the room.  You probably would get way too much room sound/ reflections by pointing at stacks - a wider spacing and pointing just inside the stacks may work better. 

2) Outdoor venue, large open field with PA stacks on the sides pointing straight forward.  Not much center fill, but you are stuck way up front (in front of the "sweet spot").  You would probably be better off with a tighter pattern and wider angle on the mics, pointing just outside the stacks rather than straight at them. 

Yeah, I know, I'm over complicating things, when the point of the original post was to simplify ; ) 

[Todd R]

Interesting discussion, thanks Gut!

I've only read as far as Gut's two long threads -- need to get back to this when I can really digest it.  Mapping PAS configurations/spacings to match DIN (in the first thread) makes total sense to me.  I'm missing how that mapping changes if you want to increase the SRA beyond that of the spread of the orchestra (PA stacks).  Seems like if you apply the need to have a wider SRA for DIN, then you would map to that SRA for PAS, and the mapping would be the exact same (recognizing that the only way to insure the SRA for DIN is exactly 10 degrees more than the orchestra width is to move closer or farther, since the angle and capsule distance for DIN isn't changing by definition).

I'll try to wrap myself around that when I can digest it more. In the meantime, I don't think I've ever cross-posted something on ts.com, but the post I just made in the playback section seems to make more sense in this thread:

I tend to believe that most of our 'standard' configs are too wide, even at 90 degrees. I have always loved the 4v sound run backwards on an ORTF bar at 70 degrees and have been experimenting with a narrow angle quite a bit of late.

Yes, I'm with you on that, or at the very least want to try that sometimes vs always doing DIN.

Probably should at least reference Gutbucket's thread on PAS and stereozoom.  I've only had a chance to glance at that, but alot of what really is going on is probably explained there.

Many years back I used to run on occasions a modified NOS pattern, with caps at 30cm spacing and mic angle at ~75 degrees.  I found it worked good in many instances, esp when you only had cards to work with (I had km140s at the time).  Then I got the jklabs setup for the AKG caps, and both had the option of ck63 hypers available and was also working off a fixed DIN kwonbar -- so I got away from the modified NOS*75 pattern.

Lately I've been trying it again more.  A mod NOS*75 pattern with cards, which often is close to pointing just outside the stacks for where I am taping from, I think gives an SRA that is about the same as DIN, but since it is a narrower pattern, gets more direct PA sound and less room reverb.  So the soundstaging is similar to DIN, with less reverberant sound.

I also think something like this (wider spacing, narrower angle) probably works better with brighter mics.  IIRC, larger spacing can lead to more comb filtering on high frequencies, softening them a bit.  So flat mics don't tend to sound as good with larger spacings since the high frequencies are already flat, but softening the highs on mics that have a high freq bump works a bit better.  (Hopefully I've got that all right, going from my memory of having investigated it 9 years ago or so.)

[Gutbucket]

..it seems like there will be more variables in a lot of situations.

Yes, true of course.  This is just another tool to use when appropriate.  No single setup works for every situation, which is part of the challenge in making great sounding recordings and also what makes it far more interesting than simply patching into the SBD if you enjoy figuring out the puzzle.

You raise a good point about what to do to compensate for less than perfect recording positions.  I think this technique can be helpful for less than ideal positions farther back than would be preferable, and does that by it's nature without having to put much though into it.  That's because in those cases it will automatically produce a setup with less angle and more spacing between mics than it will when setup closer, and that trend helps to get a better direct/reverb ratio which I see as the biggest challenge as the recording position gets more distant.

As for off center locations and otherwise odd sound source locations and distributions, one of the most helpful things for me is simply to close my eyes and listen for a while, trying my best to forget what I know about the situation visually, and just concentrate on the sound itself- where it's virtual center seems to be, how wide it is, and how the different sound sources are distributed within that width.  I then adjust my mic setup with just that in mind.  That can produce a significantly different setup than what I would have arrived at by sight.

2 extreme examples..

OK, however in both of those examples the basic idea of this still works in your favor as a good starting point.  It produces wider spacings and narrower angles from farther back, and the opposite up close.  You can improve upon the solution further if you have a good command over the basic SZ principles. 

And that gives me another idea to possibly take this further with an optional second refining step in the process-

Start with the basic improved PAS table which gets things in the ballpark. Then if you want to adjust further, consult a secondary set of tables like the one in my later post which lists all the configurations that achieve the same SRA.  That would be a more straightforward way to optimize other aspects while keeping the SRA the same.  If you wanted to start with a different SRA than the PAS table (which is always the angle between the stacks +11°), just pretend the PA speakers are at the angle you think is appropriate and the same system works for refining the other aspects from there, without loosing the SRA you wanted.

That would just require a set of additional tables, one for each common SRA.

[Chuck]

Great stuff Gutbucket. A couple of years ago, I made myself a laminated sheet that I carry in my bag with the correct angles and spacing to simulate DIN according to the Stereo Zone article. I have found this to be very useful as you mention when pointing mics at PA stacks. It isn't perfect, but it is a useful starting point when encountering unusual taping conditions.

[Gutbucket]

Thanks Chuck, it's encouraging to hear that basically the same idea has worked for you in the real world.

I'm missing how that mapping changes if you want to increase the SRA beyond that of the spread of the orchestra (PA stacks).  Seems like if you apply the need to have a wider SRA for DIN, then you would map to that SRA for PAS, and the mapping would be the exact same..

Todd, I'll have to digest that a bit to better understand what you're getting at.  But maybe this helps- the difference between the orchestra angle and the SRA is always ~11° on the PAS table.

(recognizing that the only way to insure the SRA for DIN is exactly 10 degrees more than the orchestra width is to move closer or farther, since the angle and capsule distance for DIN isn't changing by definition).

Right. So what I've done here is keep the SRA to mic angle relationship of DIN (which is 11°) as the constant. The orchestra angle is dictated by the speaker angle + that 11°, and the SZ tell then tells us what the spacing should be.

[Gutbucket]

Todd, I think what may be helpful for what you are talking about in the post from the other thread quoted above is a set of secondary tables- one for each SRA of interest, each listing the different configurations which can achieve that particular SRA.  I mentioned that a couple posts above as possibly being a good secondary refining step to further adjust things.  The second table I posted in this thread does that to list all the cardioid configs that achieve the SRA of DIN, but ideally I'd make a similar table for each of the SRAs I typically use.

Here is is again, same table-
Table of two channel cardioid arrangements for SRA100.GIF


[edit- heading over to check out that other tread]

[Gutbucket]

Tech note bump.

[snip] But is that really what we want?  Most of the time we want a SRA that is slightly wider than the angle of sound sources (called the orchestra angle in that on-line calculator).  We discussed this a few years ago in this thread on understanding the concepts of the Stereo Zoom.  As noted in that thread, the Stereo Zoom paper mentions that many engineers prefer an SRA slightly wider than the total width of the sources (orchestra angle) and suggests + 10° as a general guide. [snip]

Looking for some possible insight into the "SRA wider than the actual source" preference thing I came across this paper by Wittek & Theile, "The Recording Angle - Based on Localisation Curves", a 112th AES Convention paper from 2002 down-loadable from here:
http://www.hauptmikrofon.de/HW/AES112_Wittek_Theile.pdf

It explains that in the various sets of experimental data upon which the calulations are based, different stereo mic configurations which should have similar Stereo Recording Angles 'by the numbers' actually are quite similar within the central 75% of the total playback image range between speakers. The larger variations between sets of data happen mostly outside the central 75% region, which is where these general calulations begin to break down, so the authors suggest comparson between various stereo mic configurations is best done by looking at at their +/-75% SRA points, not the full SRA.

Not sure exactly how applicable that is to the question of what SRA is appropriate for a given Orchestra Angle, but it's an interesting read and at least explains the somewhat mysterious (to me anyway) 75% SRA indications in the Image Assistant 2.0 visualizer: http://www.hauptmikrofon.de/ima2-folder/ImageAssistant2.html

[MIQ]

Hi Lee,

  Unbelievably I have also been reading the Wittek paper over the past few days and was thinking about the comments about setting the SRA slightly wider than the actual "orchestra".  The main issue seems to be the localization curves different researches have developed, and specifically, the test signals and listening rooms used during the research. 

I ran into this fact a while back when the SRA I computed using the Williams Zoom did not match the SRA given by the Sengpiel web site for a coincident pair.  Since a coincident pair uses only interchannel level differences to steer the image, the difference in SRAs is linked to what interchannel dB difference is needed to image at the speaker locations and not inbetween them (Wittek calls it 100% source shift).  Williams relies on psychoacoustic curves developed by Simonsen to determine this.  Wittek uses curves he developed in his own testing and Senpiel has his own curves it appears.  Below is a little summary I wrote down yesterday when I was wrapping my head around this.  It is the inter channel level difference (ILD) and inter channel time difference (ITD) needed to place the image at the speaker used by the Sengpiel Visualizer, Wittek Image Assistant, and Williams Sonic Zoom.  These values assume that only interchannel level differences exist or only interchannel time differences exist without the influence of the other.

100% Source Shift Sengpiel                    ILD = 18dB      ITD = 1.5ms
100% Source Shift Wittek                       ILD = 15.8dB   ITD = 1.0ms
100% Source Shift Williams (Simonsen)  ILD = 15dB      ITD = 1.12ms

The ILD differences are really different between Williams and Sengpiel and were the source of the coincident mic SRA differences I ran into.  I also looked the SRA numbers each set of curves would give for ORTF and NOS:

Sengpiel ORTF SRA = 96.4 degrees  NOS SRA = 81 degrees
Wittek    ORTF SRA = 102  degrees  NOS SRA = 82 degrees
Williams  ORTF SRA = 100 degrees  NOS SRA = 80 degrees

I like Wittek's approach and goal. He is trying to come up with a "key value needed for directional balancing with stereo microphones" that "avoids confusion or room for different interpretations".  My understanding is that if you used two different mic configurations, both with the same 75% Recording Angle, the distribution of the sound sources should be very similar between your speakers.  It's not until you get to the last 25% of the sound image, played back near each of the speakers, that the imaging would differ considerably.  Of course there are many other variables that make the sound of two different mic configurations with the same 75% Recording Angle sound different, but at least the imaging would be comparable over most of the region between your speakers. 

A couple of other AES papers are making me lean more heavily to using the Wittek psychoacoustic curves over any other.

"Localization Curves In Stereo Microphone Techniques - Comparison of Calculations and Listening Test Results" - Plewa and Pyda
The authors recorded several different instruments as well as white noise at different source positions in an anechoic chamber using several mic techniques.  These recordings were played back over speakers and listeners were asked to judge the location of each sound source.  The results match those calculated using Wittek's localization curves.

"A New Time and Intensity Trade-off Function for Localization of Natural Sound Sources" - Lee
Here the author asked listeners to adjust just the time difference between speaker signals or just the level difference between speakers to place the image (from the center of the speakers) at 1/3rd, 2/3rd, and at the speaker location.  The results lead to localization curves that are very close to Wittek's and if you apply Wittek's localization functions to the data, it fits a little better than the numbers the author came up with IMHO. 

The cumbersome part about using the RA_75% is that it is a 75% playback angle.  So how do you choose what stuff you want to have smeared out to the edges of the speakers.  If you choose too conservatively (put all the instruments inside the RA_75%) you narrow the sound stage between the speakers (all the instruments are accurately distributed but none of them are played back at/near the speakers).  If you choose to put more of the instruments outside the RA_75% you end up with strange imaging for the stuff outside the RA_75% near the speakers.  The reality is that using the Williams SRA is also leading to strange imaging as the playback locations near the speakers.  We want to think of stereo mic configs as taking a clear picture of the sound stage in front of them (perfect lens).  The reality is that when you listen to their recordings over speakers, the outside edge of the lens is distorted and smeared.  You can clearly see this in the Image Assistant in the "Show Localizations" screen.  The center (+/- 75% = no shit, it's based on the RA_75% right  ) of the localization curve is basically a straight line.  Beyond that area the line starts to curve.  We wish it was straight the whole way, but in fact it is not.

Sorry for the crazy long and intense post but I've been tossing this around in my head for the last two days and have not disucssed it with anyone yet.  Looking forward to hearing what other people think.

-MIQ

3/5/12 - edit for ILD numbers

[Gutbucket]

Apologies for the delay in responding, I've been working a tradeshow for the past week and wanted to digest this a bit and respond when I had more time..

Big thanks for the info on the various studies and differences in data sets.  I simply assumed Sengpiel used the Williams data.  I'd like to look more deeply into the Wittek curves and data.

Seems to me we have two separate issues here, and I'm not sure they are related to each other: The first is the question of what SRA is appropriate for a particular 'Orchestra Angle'.  I realize that there may not be a good hard-and-fast answer to that.  It probably depends on a number of things like the type of music, the size of the ensemble, the room, the recording position, the desired image perspective, as well as the personal preferences of the recordist.  The second is the interpretation and comparison of the data from various imaging studies, where things seem to match up more or less within the central 75% of the SRA, so comparison of various techniques is best done by looking at their behaviors within that region.

It's probably a false conclusion to take that to mean that as a general rule, the appropriate SRA is one which is 25% wider than the Orchestra Angle.



A couple comments-
One analogy to this 'distortion of angular source distribution' is image distortion in camera lenses.  Many lenses distort the image at the outer edges of the frame in comparison to the center, higher quality lenses less so.  The difference is that camera lenses 'hard-crop' everything outside the frame, so a lens manufacturer can choose whether to minimize the size and cost of the optics and accept that distortion at the edges, or use larger and more complex optics and use only the less distorted central region.  Microphone arrays may or may not have reduced sensitivity outside the SRA, but they don’t hard-crop or exclude those sound sources, they just radically distort their apparent position on playback. It seems that they begin to do so outside the middle 75% region of the SRA.  But the analogy holds in that both photographer and recordist can chose to deliberately distort (compress or expand) the live image to better fit the desired end result. ‘Stereo Zoom’ as well as stereo fish-eye.

I generally like a very wide stereo distribution verses a more narrow, mono-sounding recording, as long as the image is solid across the center and there isn’t an apparent ‘hole in the middle’ with nothing there, or more commonly an  ‘island in the middle’ with nothing to the immediate left and right of center.  I really don’t mind image distortion of the distribution of sources out near the speakers, other than overly mono recordings which have much too little out there, as long as the central region is solid and blends out seamlessly to the sides.  Many live concert recordings I hear err to the safe side of a narrow, more mono sound, which admittedly is better than wide with a weak middle.

But I really don’t care much about imaging accuracy.  I care more about how I can manipulate the imaging to improve the end result.  I find it interesting to think about and understand the ‘objective’ image distortions of the scene as heard live from the recording position verses the recorded image played back on speakers, and I enjoy going home and listening to hear how close the recording matches what I heard, but I don’t consider duplicating that distribution accurately to be important at all.   Many times I want to purposefully distort the image to something more pleasing if the live distribution was less than optimal.  That might be because there was a less than ideal distribution of live sound sources even if I was able to choose the best microphone location for stereo imaging, or I may have deliberately chosen a different recording position based on things more important than imaging- such as one which better balances the relative levels of the individual sources, or one that has a better direct-to-reverberant sound ratio, or better overall timbre, or any number of things which are far more important to me than playback image accuracy.

[MIQ]

Interesting comments, especially on imaging accuracy vs optimizing the sound at playback.  No stress on the speed of response.  I wrote from Japan and am now back in the States, so I know how difficult it can be to stay on top of forum posts.   

There are plenty of times when the "orchestra angle" is quite a bit larger than the "playback" angle could ever be when listening in the sweet spot of a conventional 60 degree speaker set up.  It's not too easy to get the image outside the speakers  , so you are stuck "compressing" the entire panorama of recorded sounds into the 60 degree playback system.   It's not likely that a two speaker playback system could ever faithfully reproduce the sound image we hear with the 3D ear-brain sensory system we are equipped with.  The thing I'm curious about is how much closer you are able to get with your surround sound recordings.  I know from reading some of your other posts that you've done a fair bit of this kind of thing.  I know it's getting a little (lot) off topic but when you record with a set-up that dedicates mic channels (or tetra mic decodings) to the rear playback channels, is it quite a bit more convincing?  Slightly better?  Just different and cool? 

I had a chance to listen to the DTS NeoX 11.1 surround sound demo room at CES this year.  I didn't hear any "live" music recordings so I'm not sure if having all those speakers would make it more "real" but it was neat. 

Back to the RA_75%, you wrote: "It's probably a false conclusion to take that to mean that as a general rule, the appropriate SRA is one which is 25% wider than the Orchestra Angle."  I bet you are right.  I think though, that Wittek and Theil are proposing that you could get pretty good results by choosing a mic system with an RA_75% that is about 25% less than the orchestra angle.  This corresponds to Fig 9 in their paper.  At least you would have the majority of the sound sources nicely distributed within the speakers and some stuff less accurately distributed out toward the speakers.  Like you wrote, there won't be "one rule fits all situations", especially if there is a lot going on at the edges of the "orchestra". 

-MIQ

[Gutbucket]

Not much talk on-topic in the thread at this point. I’m more than happy to jump track to interesting off topic tangents.

Any playback system outside of a high-order ambisonic or WFS system is going to distort the location of sound sources compared to what is heard live to some extent, it’s just a question of how and if the distortions are acceptable or not. 

I’m of the mindset that it is essentially impossible to reproduce the exact event, and trying to do so is interesting only as an intellectual goal, not an artistic or musically enjoyable one.  I have no qualms about setting up my recording rig or adjusting things later to optimize the recording for how it will be played back, or to ‘fix’ problems with the sound at the live event.  At one level that simply means editing out obnoxious noises, rumble, pops, et cetera; correcting timbre with EQ, and choosing microphones and setups that sound closer to what I consider ‘good’.  Leaving in the hiccups would be more accurate to what was heard there, it’s a judgment call and I aim for better than live in the ways it can be, understanding that some things will always be better live.  Usually the goal is one of transparency and striving for the platonic ideal of what I think it should sound like, often an improvement on the live sound I heard at the recording position in some ways.

What is important in a mono recording?  With mono there is just one channel so in geometric terms the recording is ‘zero dimensional’ like a point. What is important in a mono recording is overall level, timbre, instrument balance, and the direct/reverberant ratio. There is no left/right imaging, although depth can be implied by differences of timbre, level and reverberence. 

Stereo doesn’t make any of those core things less important, it expands things by distributing the sound sources along the line between the speakers.  It’s a line between two points, so it’s geometrically one-dimensional.  Overall level, timbre, level balance and reverberation are still king.  Anyone will notice those things immediately if they are off.  Unless the spatial L/R imaging is grossly messed up, most people wouldn’t hear a distribution problem, especially if they were not at the live event and had an idea of the actual arrangement.  Some more discerning listeners will listen for things like how wide the apparent image is, how even the distribution of those sources are within that width, and how distinct the placement of individual sounds sources are within it.  Yet they still only have an imaginary idea of what that should be, unless they were there at the live event (an even in that case I’d argue that what they want is something pleasing, something close to what they think the distribution should be, rather than something that is as geometrically ‘accurate’ from the actual recording position as possible).

Multi-channel surround is two-dimensional, expanding the one-dimensional line between speakers to a horizontal plane around the listener.  Same hierarchy of importance: Level, timbre, level balance between sources, direct/reverberant ratio.. then farther down on the list envelopment in the ambience and directional source distribution.  But like everything else, the question is not really one of directional or revererent accuracy as much as about presenting something believable and pleasing.  As far as the sound source distribution aspect of multichannel, I love hearing a more solid image across the front, the crowd reaction wrapping around the back, the more tangible sonic fingerprint of the room reverberence. 


But it’s not only better source distubution that makes stereo superior to mono and muti-channel superior still.  Things sound more ‘real’ in stereo than mono, and more so in good multi-channel.  The timbre of many instruments is better and ‘more real sounding’, the direct/reverberant balance clearer and less critical to get just right.  And if it’s not just right I can adjust it afterwards to some extent.  In some ways it’s actually easier to make a decent stereo recording than a mono one, and easier to make a multi-channel one than a stereo one.  There is less compromise required in fitting the recording situation to the limits of the medium.  For example, a recording that would be way over reberberant from a less than ideal mic position becomes more listenable because the listener can direct her attention to what she wants to hear, more like she can at the live event, and ignore the room reverberence behind, even without the limited adjustments I can make to improve it after the recording is made.  It’s somewhat ironic that recording in multichannel is more forgiving in that way.  Of course it’s technically more complex to setup, record, mix and playback, but the challenge of ending up with a good sounding recording is actually made easier.  Same for stereo over mono.

I think it’s quite a bit more convincing, more real sounding, interesting and cool as well.  I wouldn’t make the effort otherwise.  It’s also helpful in making the best stereo recordings I can, assuming I have the time to mess with mixing it down, which I rarely do.

To get the 3-dimentions we’d need to record height information. My recordings with the ambisonic mic capture that to a basic extent, but I have no way of playing back the height information currently, other than pointing the virtual microphones up or down a bit to better optimize the 2D sound.

I haven’t heard any of the 11 channel matrix surround decoders, but I can image that a good implementation, properly set up might work quite well with live recordings by distributing the decorelated ambience information (mostly the L-R difference signal) more evenly throughout the room for a stronger feeling of envelopment and ‘your are there-ness’, even though it makes no improvement in the location of actual sound sources.  If so, that would make it a good example of the value in improving the ‘suspension of disbelief’ even though there is no improvement in actual accuracy.

[MIQ]

Great post Lee!  Thanks for taking the time respond and explain your approach and insights.  A couple of thoughts:

I agree with the hierarchy of importance you've laid out - overall level, octave to octave balance, relative level of the instruments to each other, direct to reverberant levels, spatial distribution of instruments (between 2 or more speakers) and envelopment.  Like others have mentioned though, the performance trumps all of this if it gives you goose bumps. 

Imaging is a funny thing.  A lot of listeners don't pay much attention to it, but once you start to listen for it, it takes on more importance.  To me, I much prefer recordings where the instruments are distinct, and clearly focused in the horizontal line between the speakers (stereo).  I also like the localization I get with the HRTF recordings I've made, but I wouldn't necessarily say they were exactly like being there.

I wish it was more common and easier to distribute/play-back multi-channel recordings.  Like you wrote the results are often very pleasing even if they are not an exact replication of the live event.  I even like the results of some of the surround up-mixes (Dolby Pro Logic II, DTS Neural, etc) on (some) stereo sources.  It may not be "real" or what the artist/engineer originally intended but it can be nice to listen to. 

The problem I see with striving for the exact replication of the live event using WFS or ambisonics is that you end up needing an infinite number of speakers to get it "just right" at maybe a small listening area.  I haven't heard a lot of that kind of stuff but that's my impression. 

-MIQ

[Gutbucket]

I listen for and like hearing distinct positions of individual instruments sources as well- which I think of as a higher-level illusion built atop the foundation formed out of those more basic recording aspects.  Only when those things are more or less correct can I put them out of mind and shift focus to higher level issues like imaging.. and maybe experinece that magic goosebump, spine-tingling 'willing suspension of disbelief' thing.  For me, multi-channel playback takes that one step further in listening for the imaging of the room sound itself, the reactions of people from various directions around the room, and even better location of sources on stage when that's appropriate.  It's facinating and addictive.  You are right on about the quality of the music and performance trumping all of this stuff though.  When considering what this is all really about, a cheap mono cassette recording of a sublime performance beats a technical marvel of a recording made of something musically worthless.  At it's core, it not about technique and gear comparisons but communicating the information and emotion of the music and the performance.

I've never heard any ambisonic or WFS playback systems myself.  I'd love to sometime.  But again, I don't think shooting for technical perfection with zillions of speakers would be an appropriate goal so much as acieving a more convincing illusion.

The primary value I find in all of the matrix surround and upmix techniques for music is the ambience extraction to help create a more convincing diffuse ambient field in the listening room which is one of the big components of a live musical experience.  In that light, all of them are basically fancy versions of the old Halfler difference technique.  In some ways the basic Halfler thing is better in its simplicy and leaving L/R path untouched, as it seems to me that many of the more advanced matrix techniques compromize the direct sound somewhat in trying to get improved differentiation of individual surround source locations which is more imortant for sound effects and movies.  When I do mix down my multi-channel recordings to stereo, I usually switch back and forth between monitoring in stereo and checking the DTS Neo6 and Dolby PLII music mode auto-upmixes of my 2-channel stereo mix.  I can do things to get the decoders to spread the room ambience and crowd reaction out to the sides and around the room with the decoders on, but I want to make sure that I don't do anything that would compromize the straight 2-channel stereo.  Usually I can get things so that it benefits the straight stereo playback as well as the matrix decode, so if someone has things setup correctly and wants turn the decoder on they can, and it's a win all around eitherway.

[mosquito]

Thanks for the thread and ideas, Gutbucket. 

I carry around the SRA charts both on paper and as JPEGs in my cell phone.  The PAS-Zoom GIF files would work well like that too.

[MIQ]

When I do mix down my multi-channel recordings to stereo, I usually switch back and forth between monitoring in stereo and checking the DTS Neo6 and Dolby PLII music mode auto-upmixes of my 2-channel stereo mix.  I can do things to get the decoders to spread the room ambience and crowd reaction out to the sides and around the room with the decoders on, but I want to make sure that I don't do anything that would compromize the straight 2-channel stereo.  Usually I can get things so that it benefits the straight stereo playback as well as the matrix decode, so if someone has things setup correctly and wants turn the decoder on they can, and it's a win all around eitherway.

That is a great approach!  I imagine this involves using uncorrelated and/or out of phase signals you've captured.  Stuff like widely spaced omnis or the Side signal from a MS set up?  It is great that the Hafler technique is like the Side signal of a MS set up (L-R = S). 

-MIQ

[Gutbucket]

That's right. Wide spaced omnis often have enough decorellated information to do so effectively on their own, but increasing the S component of any stereo signal, not just one recorded as M/S, works too. Samplitude makes that simple by allowing switching of the panning feature from shifting level between two channels to adjusting the M/S mix of them to anything between mono and full reverse-polarity, over-wide stereo.

[Gutbucket]

Try figure 8's in PAS outdoors.  They require the least spacing between mics of any pattern.  See here- http://taperssection.com/index.php?topic=177050.0

[Gutbucket]

Duplicate post below, which I just added to the follow up thread to this one- Improved PAS table (printable) - good imaging with high direct/reverberant pickup ratio. Although duplicate posting is discouraged for good reason, I'm also posting it here for those reading this thread who don't make it over to the other one, as I think this is good confirmation of how well improved PAS can work, even when the room sound is not a problem we are trying to work around..

Interesting streaming listening comparison of a few different microphone setups allowing one to listen to the way they capture stereo ambience at Helmut Wittek's Hauptmikrofon website, here- http://www.hauptmikrofon.de/audio/stereoambience.html

Granted none of the samples are of music, but they do provide a useful basic comparison for hearing the differences between X/Y coincident, two near-spaced setups (ORTF and improved-PAS-like), and a spaced omni configuration.

I'm posting the link here because I find I personally prefer the setup labeled "quasi-ORTF" for all samples there except the construction site, and that "quasi-ORTF" setup closely resembles a typical "Improved-PAS" configuration (40cm / 40 degrees) with the microphones angled only 20 degrees away from center - which is a pretty typical PAS angle from the taper section or soundboard area further back in the room.  Only on the construction site sample did I prefer the ORTF sample for it's more distinct left/right imaging width.  For all the other samples I felt the quasi-ORTF samples produced a better balance between sharp imaging (X/Y furthest to that extreme) and natural sounding diffuse ambient openness (spaced omnis furthest to that extreme).


A few comments-

I was listening on headphones.

There is no right or wrong choice here, only personal preference.

I like the improved-PAS-like quasi-ORTF samples here because of their stereo qualities - that is to say, how they reproduce the sound, even though it is not actually being leveraged for the reasons we'd choose PAS!  It just sounds better to me than the other samples.  Where as the primary purpose for choosing PAS is to either simplify setup, or maximize the direct/reverberant ratio as much as possible.  It's very encouraging that it also simply sounds better and more natural to me when in a prefered recording location without the ease of setup constraint.

I like to angle spaced omnis apart from each other rather than pointing parallel to each other, especially if that pair is the only mics I'm using.  That provides some additional level difference information at high frequencies which makes the imaging somewhat less washy and more distinct.  I think that would improve the spaced omni samples here, but the way its been done here more clearly represents the basic differences between setups without that kind of modification.

I wish there was a way to play both the spaced omnis and X/Y samples simultaneously.  I like that setup for live music recording because it sort of gets the best of both worlds.  There was a sample player page at the Schoeps website at one point (may still be up) which allowed similar samples to be played singly or simultaneously.   I don't think it was intended for simultaneous playback of more than one sample at a time but it worked.  It was very interesting hearing the difference between each setup on its own as well as combinations of two setups, as in a four microphone configuration.  It helped confirm my suspected preference for X/Y + spaced omnis over near-spaced + spaced omnis, and over all of the two mic configurations alone.  Best of both worlds from a harmonious combination.