Topic: Mic / SB Matrix Time/Phase Shift

[d5]

Apologies if this has been covered before, but I did a search and while it's certainly mentioned in passing, I didn't see a thread directly addressing this.

I've recently started running 4 tracks; stereo audience microphone and stereo SB and I've spent some time calculating what I predicted the time shift would be and then actually making adjustments and it comes fairly close.

On one particular recording, the mic's were set about 2 meters from the stage and approximately 4 meters from the mains. On paper this seems to be about a 11ms shift (11ms = 12/343) to earlier for the aud/mic's. When I make the adjustment by visually comparing the tracks, I come up with about 9ms.

I suppose there are all sorts of phase problems with doing a matrix and this only compensates certain aspects, but was hoping to get other opinions on approaches. Also, i'm using Audacity to make the adjustment. Also, visually comparing tracks is not exactly a science, wondering what others do here.

Thanks,
John

[LoveBuzz81]

Time = Distance/Speed of sound (1,126 ft/s)


sample:

4 ch matrix mix: stereo pair is located 30 ft. f/ the PA and SBD mix.

30 (distance)/1126 (speed of sound) = 26 ms (time)

add 26 ms sample delay to the sbd mix.

[ScoobieKW]

Time = Distance/Speed of sound (1,126 ft/s)

At sea level, with a certain barometric pressure. Due to these variables, a millisecond per foot works to get you close, then lining up by eye is helpful.

if your mics are onstage or close, often a phase flip of the board channels will be enough, take a listen and judge for yourself.

[Gutbucket]

In addition to air-travel time/distance, digital processing and recording will cause some small but significant delay of signal.  That can potentially affect things on both sides of the equation  - on your side for the recorders, which is probably only significant if you are using different recorders for AUD and SBD, and on the SBD side due to a digital board or FOH processing which is probably more common.  It won't be much but it can be several milliseconds.

Calcs help get close, visual alignment helps get close.. yet the best way I know is to get close via those methods then make a final alignment decision by ear while nudging things back and forth a bit listening for what works best.  For me stage banter is the most optimal material to use for detecting alignment issues by ear, we are super attuned to hearing detail and echo on human voice so as to be able to extract speech content, and there generally is no other conflicting content while the band members are talking to the audience so any misalignment becomes easier to hear.

[goodcooker]

I follow the "roughly 1 ms per foot or less" rule then line up by ear. You can hear comb filtering in the bass frequencies when you are off. When the snare  drum sounds tight and the kick drum and bass guitar have good presence with no smear I call it good. I also utilize the wave scope in Wavelab. There may be a plug in for Audacity.

If you are using a coincident stereo mic and the SBD is run in mono you can line them up just by looking at the scope which will show a straight up and down line indicating mono compatibility.

[morst]

Food for thought: In what plane is the source you're aligning? The distance from the upstage kick drum to the downstage lead vocal mic could be several meters.

I often go by drum stick clicks on a quiet song count-in, as the drums are often mic'd with more channels than all the vocal and instrument mics combined.

I also feel that within 10 ms, delay won't be noticed, and probably not within 20 ms, realistically.

And if you err, err on the side of SBD source happening first. It's weird otherwise.

[d5]

Food for thought: In what plane is the source you're aligning? The distance from the upstage kick drum to the downstage lead vocal mic could be several meters.

The stage is wide, but not deep; there isn't really room for a performer to be directly in front of the drums.

My mic's are close to center and usually about 5' to 7' back from middle, upfront stage vocal mic depending where i decide to clamp. I have flexibility here, but found this range sounds pretty good. I the 3rd picture, this would be approximately above the von Trapp beer coster  , but about 2' to the right. My mic's are about 12' to 14' from the main speakers. The decks are a pair of 702's linked, so they start on a common sample and share the word clock from the master deck. I thought I was hearing some smearing and "think" that the time shift helped to clean this up, but I didn't do a good A/B comparison. The time shift i settled on was 5 ms as this lined up visually for vocals.

Note: The pic's are from 2 different shows; first is from Hot Creekin' Acoustics on Feb 11; they have percussions, no formal drum kit, but pic shows the right side of the stage. The other two pic's are from Marks Brothers on Mar 18.

Thanks,
John

[Life In Rewind]

Besides the waveform visual - take note of your levels when making the sync - you'll start to have levels increasing as you get closer to "in sync" - watch for peaks that you didn't have before.

[nolamule]

Calcs help get close, visual alignment helps get close.. yet the best way I know is to get close via those methods then make a final alignment decision by ear while nudging things back and forth a bit listening for what works best.  For me stage banter is the most optimal material to use for detecting alignment issues by ear, we are super attuned to hearing detail and echo on human voice so as to be able to extract speech content, and there generally is no other conflicting content while the band members are talking to the audience so any misalignment becomes easier to hear.

This is a great recommendation! 

[Life In Rewind]

Calcs help get close, visual alignment helps get close.. yet the best way I know is to get close via those methods then make a final alignment decision by ear while nudging things back and forth a bit listening for what works best.  For me stage banter is the most optimal material to use for detecting alignment issues by ear, we are super attuned to hearing detail and echo on human voice so as to be able to extract speech content, and there generally is no other conflicting content while the band members are talking to the audience so any misalignment becomes easier to hear.

This is a great recommendation! 

I agree - I find, visually - once you're zoomed in, it gets harder to decide which little jaggy "peak-lets" to use when picking a sync point - since the peak looks less extreme at high zoom levels - its not as obvious.

[Gutbucket]

And if recording using near-spaced or wide-spaced mic setups, the peaks will quite often not visually align perfectly when zoomed way in due to the slight time of arrival difference from the spacing between the mics.  The waveforms are visually confusing way down at that level.

[d5]

Thanks for all the great suggestions, I plan to work a bit more with this one particular recording. I came up with an 11 ms shift using calculations, but visually 5 ms seems about right. I think the next step will be to take some samples and try a range of shifts, something like 3, 5, 8, 13, 21 so I can do a true compare.

I think if I can get one where I feel is "best" sounding, it'll make future work easier; I do tend record at this club quite a bit.

Thanks again!
John

[Fried Chicken Boy]

I also feel that within 10 ms, delay won't be noticed, and probably not within 20 ms, realistically.

In addition to goodcooker's statement about the "roughly 1 ms per foot" general rule of thumb and then using your ears, I agree with what morst posted.  Many years ago when I was in college, I recall an acoustics professor of mine telling us that the human ear and brain won't notice a delay if it's less than 20 ms.  As long as you have the SBD and mics lined up relatively close, it should sound just fine. 

[Gutbucket]

Slight misalignments of under 20ms are generally too short to be perceived as a delay or echo but may be perceived in other ways - image shifts, image blurring, image width, tonal changes, depth changes, apparent local performance space size, stuff like that.  One reason a unaccompanied talking and drumstick clicks are good signals for detecting slight misalignments is that those kinds of spatial effects are more easily identified as such using those simple "well known" clean sounds heard in relative islolation, whereas for musical signals those attributes might well be part of the original sound of the instrument. 

Unconscience attention is drawn to the earlier of the arrivals, so if a SBD slightly proceeds an AUD the image may sound slightly closer and more present, whereas if the AUD precedes it may sound a bit wider and less center prominent. 

Also, because short delays are not perceived as being a delay, they can be useful for creating stereo interest or for differentiating channels from each other.  One simple but effective pseudo-stereo technique is the application of a short delay in one channel verses the other.  That can be useful when mixing in a single monophonic channel of room ambience or reverb, or making less apparent a short repair consisting of one channel copied to the other to cover an intermittent flaw or dropout.  For my surround playback stuff I sometimes introduce a slight delay into the rear facing mic channel(s) if there isn't enough front/back separation and the recording needs some help keeping the stuff in front from leaking into the surround channels too much without otherwise having to lower the level in those channels.  Likewise, I might introduce an ever so slightly different delay to each surround channel so that a single monophonic channel of room and audience ambience conveys something of a similar openness and richness to using multiple recorded ambient channels.

Hope all this doesn't come across as pedantic, I'm sure may of you guys know this stuff well.

[danlynch]

Yeah, be careful with that "slight delay between channels" stuff.  I pulled a board feed at a local venue (later mixed with mics) and the traveling FOH was a little too heavy in that channel-delay trick.  In post, when I laid out the tracks in Soundforge, I didn't initially notice the delay and after mixing with the mics and using the 1 foot per ms formula, the result was echo-garbage.  After spending a little time trying to figure out WTF, I noticed that the SBD tracks weren't aligned and fixed it (eventually).  It also woke me up to the idea that I was doing this stuff by rote / formula / habit and maybe should be less robotic next time.

[nulldogmas]

Yeah, be careful with that "slight delay between channels" stuff.  I pulled a board feed at a local venue (later mixed with mics) and the traveling FOH was a little too heavy in that channel-delay trick.  In post, when I laid out the tracks in Soundforge, I didn't initially notice the delay and after mixing with the mics and using the 1 foot per ms formula, the result was echo-garbage. 

That's bizarre. Why would the FOH need to impose artificial fake room echo when they have an actual, you know, room to work with?

I'm equally baffled by the occasional soundfolk who lather on the plain old reverb on vocals to absurd levels, especially in fairly wet rooms. But topic drift, I know.

[robgronotte]

Could someone tell me how I can easily delay one channel when doing a SBD/AUD mix in Audacity?

I have found a few tutorial videos which show you how to use the Latency function, but that seems to be for when you are recording a new track with a microphone, not working with two existing tracks.

I guess I could just delete the correct amount of time from the beginning of the AUD track, but it seems like there must be a better way.

[AbbyTaper]

Can't you just add silence to the start?

[robgronotte]

Can't you just add silence to the start?

Yes, I guess I could do that also, but either way now the two recordings won't be the same length which gives me another issue to deal with.

[Gutbucket]

There are various ways of doing it via shifting, adding, cutting, delaying.  Haven't used Audacity in a long while, but can't you shift alignment by clicking and dragging one stereo pair along the timeline while leaving the other in place?  May need to unlock the track first to allow movement, and zoom in enough so that the shift is relatively small (on the order of milliseconds), but you'll need to zoom in to near that level to accurately see the position of the transients anyway.  In that way its easy to play around with alignment while listening to get it just right.

[unidentified]

I am no expert, but I have dabbled successfully more than a few times with synchronizing and mixing sbd and audience sources, and have two suggestions. First, the above comment about using the sound of folks talking as an indicator of accuracy is consistent with my experience,  Any trace of the dreaded "ball-park echo effect" in on-stage banter between tracks in a mix I produce tells me that I have to try again. Second, in open situations in fairly modestly-sized rooms, my best friend is a D Day-style "cricket" -- a brass clicker as used in WWII that you can buy on Amazon for about $10 and which produces a distinct and precise click-clack sound that is very easy to find on the wave forms of the two sources.  Ideally, you click it after you start rolling and before the set starts and then again after the set ends and before you stop rolling.  Works for me but YMMV.

[Gutbucket]

^ I keep a similar dog training "cricket style" clicker in my recording bag.  Little rectangular plastic thing with a spring steel thumb plate in it.   I bought a few of them years ago out of a big bin at a pet supply place for like a dollar or two each.  Great for making audible alignment clicks as you describe at the start and stop.  ..which is brief and generally won't call attention to what you are doing.

The clicker is also useful for hearing how the stereo distribution of your mic-setup translates on playback, if/when you ever get a chance to walk all the way around the recording position at some distance while clicking / calling out each of the 12 clock positions before or after the performance.  Super interesting thing to do.  But unlike a couple alignment clicks, this will call attention to what you are doing and generate looks!  Pick your opportunity wisely.

[unidentified]

^ I keep a similar dog training "cricket style" clicker in my recording bag.  Little rectangular plastic thing with a spring steel thumb plate in it.   I bought a few of them years ago out of a big bin at a pet supply place for like a dollar or two each.  Great for making audible alignment clicks as you describe at the start and stop.  ..which is brief and generally won't call attention to what you are doing.

The clicker is also useful for hearing how the stereo distribution of your mic-setup translates on playback, if/when you ever get a chance to walk all the way around the recording position at some distance while clicking / calling out each of the 12 clock positions before or after the performance.  Super interesting thing to do.  But unlike a couple alignment clicks, this will call attention to what you are doing and generate looks!  Pick your opportunity wisely.

Clearly, great minds think alike, Gutbucket!   

[Gutbucket]

  

[nulldogmas]

Can't you just add silence to the start?

Yes, I guess I could do that also, but either way now the two recordings won't be the same length which gives me another issue to deal with.

As others above have noted, you're not trying to get the two recordings to be the same length; you're trying to get them to align throughout the course of the music, which is slightly different.

My workflow, which I've been following for a mind-numbing number of Merge festival tracks now, is:

1) Import both tracks into Audacity.

2) Trim the start of one track or add silence to the other until the first big spike of amplitude aligns. (I do this visually while zoomed in by an extreme amount, then check to hear that it sounds okay, which it always does.)

3) Scroll to near the end of the recordings, and find a clear loud fast-attack spike (snare drum hit is usually good) in both recordings.

4) Click on the spike in one recording, change the Selection display to "Samples," then copy down the number.

5) Click on the same place in the spike in the other recording, and drag to select everything between there and the start of the recording.

6) Effect > Pitch and Tempo > Change Speed and Pitch, then type the number from step 4 into "New Length."

Done. There will be a teeny amount of drift in any part of the recording after that late spike, but it won't be enough to be noticeable over the few minutes left to run at that point.

[morst]

Could someone tell me how I can easily delay one channel when doing a SBD/AUD mix in Audacity?

I have found a few tutorial videos which show you how to use the Latency function, but that seems to be for when you are recording a new track with a microphone, not working with two existing tracks.

I guess I could just delete the correct amount of time from the beginning of the AUD track, but it seems like there must be a better way.

F5 (function 5) gives you the double headed arrow tool which enables sliding tracks left-right.
If you really mean ONE CHANNEL then you will need to split stereo track first, which is hidden in the downward pointing triangle to the right of the track name

[AbbyTaper]

There will be a teeny amount of drift in any part of the recording after that late spike, but it won't be enough to be noticeable over the few minutes left to run at that point.

Thanks for adding that Neil.  I've seen a number of times where people say there won't be drift "because both sources are digital".  In my experience there is ALWAYS some small amount of drift.

[nulldogmas]

There will be a teeny amount of drift in any part of the recording after that late spike, but it won't be enough to be noticeable over the few minutes left to run at that point.

Thanks for adding that Neil.  I've seen a number of times where people say there won't be drift "because both sources are digital".  In my experience there is ALWAYS some small amount of drift.

There shouldn't be drift in the section where you change the pitch and tempo of one source to match the other, because you just changed them to match. After that they'll still run at slightly different speeds, but it shouldn't be noticeable over a span of just a few minutes.

[robgronotte]

I did mean shifting one source, not one channel.

In this particular case there shouldn't be any drift because both sources were recorded at the same time on the same multi channel recorder, so the recordings are exactly the same length and only issue to deal with is the very slight time delay of the microphone source. I visually found it to be about 20ms, small enough that I can't hear any echo effect, but it was recommended to me that I try to account for that.

After this I do plan to tackle a matrix of two sources recorded on different devices though. It still seems odd to me that there would be a drift in digital recordings, but I will try to check for it.

[AbbyTaper]

I did mean shifting one source, not one channel.

In this particular case there shouldn't be any drift because both sources were recorded at the same time on the same multi channel recorder, so the recordings are exactly the same length and only issue to deal with is the very slight time delay of the microphone source. I visually found it to be about 20ms, small enough that I can't hear any echo effect, but it was recommended to me that I try to account for that.

After this I do plan to tackle a matrix of two sources recorded on different devices though. It still seems odd to me that there would be a drift in digital recordings, but I will try to check for it.

If both your sources went into the same recorder there won't be any drift.

[nulldogmas]

I did mean shifting one source, not one channel.

In this particular case there shouldn't be any drift because both sources were recorded at the same time on the same multi channel recorder, so the recordings are exactly the same length and only issue to deal with is the very slight time delay of the microphone source. I visually found it to be about 20ms, small enough that I can't hear any echo effect, but it was recommended to me that I try to account for that.

After this I do plan to tackle a matrix of two sources recorded on different devices though. It still seems odd to me that there would be a drift in digital recordings, but I will try to check for it.

If both your sources went into the same recorder there won't be any drift.

Yup. And if one entire source is 20ms behind, then trimming 20ms from the beginning of it should make the two align perfectly all the way through, no worries about them being "different lengths."

[Gutbucket]

I visually found it to be about 20ms, small enough that I can't hear any echo effect, but it was recommended to me that I try to account for that.

It still seems odd to me that there would be a drift in digital recordings, but I will try to check for it.

20ms should be clearly audible, especially over headphones.  Maybe not so much as a distinct echo, more of a smearing/flammy/flangy/chorus type effect.  But because that sound that can occur naturally in a reverberant recording from reflections, it will be mostly be apparent via comparison with the same in full alignment (or with either source on its own).  In other words, you may need to play with it while listening to identify it clearly as misalignment.

Different digital clocks run at slightly different rates, but are generally quite consistently so over time.  So once two different sources that were recorded using different clock are initially time-aligned.. and stretched to correct for the rate difference between the two clocks, they generally will remain in sync before and after the actual sync points used to align them.

In this situation both sources were recorded using the same clock, so only initial time-alignment was required for them to remain aligned at all other points.

[morst]

If both your sources went into the same recorder there won't be any drift.

Just to be fully nit-picky; If the speed of sound changes substantially over the course of the recording, be it from temperature or humidity or air pressure changes, there could potentially be measurable drift, though a much lower order of magnitude.
But I don't ever bother compensating for that... I chalk it up with clock jitter as something too small for me to hear in a rock and roll context.

[Gutbucket]

Never really considered that before, but yes that is technically true! ..when the sound pickup locations are spaced apart, because the change in the speed of sound effects the travel between them. 

+T for the super nano nit!

[EmRR]

Sometimes 'aligned' invariably means SOMETHING will display phasing artifacts, and you're picking the least evil version.  This discussion should involve the notion of pre-delay, the timing between a dry sound (source/board) and the reverb return (ambient mics, or in reverb, the sound returning after it's bounced and returned).  A lack of pre-delay implies the room has no size, no matter the amount of reverb decay.   As someone who does this for a living in recording and broadcast, I've never picked the version that's perfectly aligned but then, my mix sources are more and varied and the room versus direct sound is not balanced more toward the room as it is when you and I make an audience/board recording such as is being discussed here.  Even then, any reduction I do in the timing difference is only enough to clean up any sense of echo and tighten it up, but never all the way to matched, usually out in the 15-20mS range, past flanging into echo.  I just make the room 'smaller' as needed to get clarity, and a good sense of space.   There's an additional next-level argument to be made for measuring tempo per song, and adjusting the timing so the delay is tempo related, creating an on-beat short slapback to accentuate the sense of space, with it being tucked in musically, rather than competing.  Overkill, but effective.  It's what you do with reverb pre-delay in a mix of dry instruments and vocals when mixing a record.  Higher the tempo, shorter the timing difference.

Related tangent.  Occasionally I get a mono board tape with no ambience, and often can find a short decay room ambience that gives a little sense of space without obviously being reverb. Usually sounds more like EQ than ambience. Experimenting with pre-delay is critical to the result.  The level of it is often surprisingly low to be both effective and non-obvious, frequently -20 to -36 relative, and you can still clearly hear it come and go with a mute toggle.

[Gutbucket]

+T  Good insights. Another reminder that in audio ears beat eyes.

[hoserama]

Reminds me of an Adele show I did in 2016. They had a main stage and b-stage, and each had it's own PA system. Each on opposite ends of the arena floor. So when she would walk between the two stages, the sound guy would switch PAs. Which changed the timing on the audience recording relative to the IEM recording(s), which stayed constant. It wasn't hard to figure out the two alignment settings when the PA switches were between songs, but there was 1-2 switches while music played that were a lot of fun to try and figure out.