Trumpet spill suppression

[badEnough]

Hello,

Is there any chance of reducing the spill of a trumpet into the mics of other instruments in post production with some smart algorithm, without destroying any significant part of the sound of the other instruments?

Ideas:

1) In each other channel, automate the frequency setting of a notch filter fed with the result of tracking the fundamental frequency of the trumpet in its channel (forgetting about its harmonics in other channels).

Can anything like that be done in Ableton Live?


2) Is there any "virtual studio cabin" VST plugin, capable of tracking the frequency of the trumpet in its channel and chasing it (by computing FFT or wavelets or whatever) in one or more other channels + optionally its harmonics in any proportion (each mic in its position receives a different timbre), possibly with some time delay tolerance for sound propagation speed, then sum components with inverted phase?

(A treshold value would allow to ignore spill of other instruments into the trumpet mic and not provide negative feedback for it, and eventually it might allow to ignore all but the loudest passages of the trumpet.)


3) ?


Thanks for reading and for any advice



==================================================
Some more details about the scenario:

Because of phase distortion etc., in the final mix the trumpet sounds weird especially when it plays loud, while it is ok listening to its channel alone.

In the bass and drums channels it can be fixed well enough with accurate EQ without substantially "damage" their sound in the final mix (the crash looses some "body"), but the piano would just die with such EQ.

We will be trying closer miking for the piano and different placement of trumpet and sax, it isn't a studio scenario, it's just an auditorium we rent for rehearsing/practicing with our jazz 5tet.

Hardware:

Zoom H2 for the bass on one side and the drums on the other side (4 channels)
Zoom H4n built-in mics for the piano, external mics for the winds (two Shure SM57/58 or more recently two Shure Beta 87A, super cardioid, trying not to have any sources just behind on the axis, anyway the main problem now is the winds and especially the trumpet spilling into the built-in mics of the H4n in the piano).

We are also going to close the coda piano, though that implies diminishing the energy and pleasure of our sessions (besides, the piano player already can't hear his playing very well where the piano is in the auditorium, strange but true, let alone with the piano closed, he won't be happy at all).

The piano itself will always receive something from the trumpet I guess, that is from me, and its strings resonate in some amount, but that should reduce the problem.

I could always play much less loud but I don't have a volume knob on the trumpet, that would change the tone (apart reducing the chance to send energy to the other members of the band and get some energy in return), so playing very low would not be a good solution.

[luddy]

This is a pretty tough one. Seems to me like an automated notch EQ is going to give you all kinds of audible artifacts in the other tracks, but I might be wrong. I'd probably try notching out a fixed band(s) from the other tracks to reduce the bleed. Makes it easier to compensate for the notch filter by nudging up other frequencies.

Anyway, the one suggestion I've got is to try Melodyne editor. In monophonic mode, on the trumpet track, it can track the pitch and produce MIDI from it. In polyphonic mode, on a track with bleed, you can some or all of the trumpet part (maybe) and (maybe) reduce the volume of some of it with the volume tools and other editing tools.

Depending on how dry the other tracks are and how they were mic'ed, another thing you might try I suppose is to painstakingly hand-aligned an inverted copy of the trumpet track, properly adjusted in volume to match the bleed, with one of the other tracks and see if you can cancel some of the trumpet signal out this way. If there is any time-domain smearing of the trumpet signal though (and there almost certainly will be) then this won't work very well, if at all...

EDIT: oh, I just read the bottom half of your post. If you use very directional mics then a lot of the problem would go away. Place the piano mics so that they face the instrument, soundboard, etc., but have their backs to the trumpet. You can play slightly off-axis wrt to the piano player, not exactly at him and not exactly away from him either. Consider throwing up some baffles -- like office dividers -- between say drums and whatever is next to them to cut down some of the bleed.

'Course, another thing to keep in mind is that it is very typical when jazz groups are recorded for there to be lots of bleed between the tracks. It's not the end of the world, it doesn't prevent a good mix from being made unless there are big timing errors that need to be corrected at mixdown or unless the bleed is just out of control to the point where it disturbs the balances.

-Luddy

[crumhorn]

You could always try the old back to back microphone technique. Use 2 directional mics close together but pointing in different directions and phase reverse one of them.

Any sound originating close to one of the mics will be picked up much more strongly by that mic and so will not phase cancel, but sounds coming from across the room will tend to reach both diaphragms at the same time and cancel each other out.

For your existing recordings you could try slightly delaying the trumpet track to bring it into phase with the version picked up on the vocal mic.

[badEnough]

Thanks a lot for your replies!

ShelLuser: I went for info about M4L, quite impressive! I had been thinking about trying something not real-time in mathlab, but M4L would probably eliminate the need for anything like that.
The hamburger is quite fair, OK, I hope the other guys who replied are ok with it too, I mean one for each one, not one shared among the three of you

This is a pretty tough one. Seems to me like an automated notch EQ is going to give you all kinds of audible artifacts in the other tracks, but I might be wrong."

Right, I tried by hand on a few notes, poor result (the worst part of the problem might be coming from harmonics rather than the core sound) not worth risking artifacts on the other instruments.

I'd probably try notching out a fixed band(s) from the other tracks to reduce the bleed. Makes it easier to compensate for the notch filter by nudging up other frequencies.

I'm doing that on the bass and the drums tracks, but the involved frequency range is too important for the piano.

Anyway, the one suggestion I've got is to try Melodyne editor. In monophonic mode, on the trumpet track, it can track the pitch and produce MIDI from it. In polyphonic mode, on a track with bleed, you can some or all of the trumpet part (maybe) and (maybe) reduce the volume of some of it with the volume tools and other editing tools.

That must be quite an amazing tool, I'll have to investigate.

Depending on how dry the other tracks are and how they were mic'ed, another thing you might try I suppose is to painstakingly hand-aligned an inverted copy of the trumpet track, properly adjusted in volume to match the bleed, with one of the other tracks and see if you can cancel some of the trumpet signal out this way. If there is any time-domain smearing of the trumpet signal though (and there almost certainly will be) then this won't work very well, if at all...

I had been trying with no success, probably because the trumpet arrives with a rather different timbre to the other tracks.

BUT your answer encouraged me to look for further information, I found something simple yet to me enlightening nearly at the bottom of this document, where it says "time alignment":

https://ccrma.stanford.edu/courses/192a ... phones.pdf

"... delay the close-miked sources until they align... about 1 msec per foot..."

Actually, I found out that delaying the trumpet track by 9 msec I don't have that nasty effect any more.
It's not spill suppression but that pretty much solves the main concern (should I change the subject of this post to "SOLVED - etc." or would this give origin to another topic?)

EDIT: oh, I just read the bottom half of your post. If you use very directional mics then a lot of the problem would go away. Place the piano mics so that they face the instrument, soundboard, etc., but have their backs to the trumpet. You can play slightly off-axis wrt to the piano player, not exactly at him and not exactly away from him either. Consider throwing up some baffles -- like office dividers -- between say drums and whatever is next to them to cut down some of the bleed.

Yes I guess in some moments I was alas pointing to the piano and those must be the worst moments.

But why wouldn't I want to turn exactly away from the piano either?

We were thinking of trying (this Friday) rehearsing in a live concert like disposition, winds players turning their back to the rhythmic section.

'Course, another thing to keep in mind is that it is very typical when jazz groups are recorded for there to be lots of bleed between the tracks.

Exactly what I was thinking about today, some pictures of historic sessions...

It's not the end of the world, it doesn't prevent a good mix from being made unless there are big timing errors that need to be corrected at mixdown or unless the bleed is just out of control to the point where it disturbs the balances.

-Luddy

I guess sound engineers eventually do some tests about how to mike a session... admittedly we are doing it on a per week per session basis

Which brings me to the fact that this Friday I'd like to try what crumhorn suggested.
Very interesting. Looking for some schematic drawing, I only found this text so far, that's exactly how you described it.

http://www.dsprelated.com/showmessage/32271/1.php

The noise-canceling microphone is relatively simple. While not
completely effective, it does a pretty good job. It consists of two
microphones back to back, in essentially the same sound field as far as
distant sounds sources are concerned and so connected that their outputs
largely cancel.

Thanks again to you all!

[luddy]

BUT your answer encouraged me to look for further information, I found something simple yet to me enlightening nearly at the bottom of this document, where it says "time alignment":

https://ccrma.stanford.edu/courses/192a ... phones.pdf

"... delay the close-miked sources until they align... about 1 msec per foot..."

Actually, I found out that delaying the trumpet track by 9 msec I don't have that nasty effect any more.
It's not spill suppression but that pretty much solves the main concern (should I change the subject of this post to "SOLVED - etc." or would this give origin to another topic?)

Right, that's the thing about bleed, it's mostly a matter of damage control. The delay probably actually brings the tracks very close to being in perfect alignment. If you're really lucky, it might actually work out that the 9ms-delayed trumpet track is phase-reversed w.r.t. the other tracks so you're both aligning and canceling out some bleed.

But why wouldn't I want to turn exactly away from the piano either?

Because the pianist won't be able to hear you haha.

I guess sound engineers eventually do some tests about how to mike a session... admittedly we are doing it on a per week per session basis

Yeah, they do lots of tests to be sure that bleed is minimized. They also listen for phase problems between mics that are close, especially multiple mics on the same instrument (like two piano mics). You know about the 3:1 rule? It says that if mic A is 1 foot from the source it's micing, then mic B should be at least 3 feet from mic A. (Substitute any distance for 1 foot.) It basically means that the farther a mic is from a source, the farther other mics need to be from that mic. The converse is that the closer you get mics to their sources, the more mics you can pack into an area. For jazz recordings you usually don't want to get the mics too close to instruments -- it destroys the natural feeling of space that you find in most jazz recordings -- but that in turn means that mics have to be pretty far apart.

-Luddy

[badEnough]

Right, that's the thing about bleed, it's mostly a matter of damage control. The delay probably actually brings the tracks very close to being in perfect alignment. If you're really lucky, it might actually work out that the 9ms-delayed trumpet track is phase-reversed w.r.t. the other tracks so you're both aligning and canceling out some bleed.

I'm not so lucky.

Today I'll measure distances and check that the trumpet mic (Shure Beta 87A plugged into one of the two H4n ext mic channels, the other is for the alto sax) is placed at the same distance from the piano mics (H4n built-in mics) and the drums/bass mics (H2 built-in mics).

But why wouldn't I want to turn exactly away from the piano either?

Because the pianist won't be able to hear you haha.

right

You know about the 3:1 rule? It says that if mic A is 1 foot from the source it's micing, then mic B should be at least 3 feet from mic A. (Substitute any distance for 1 foot.) It basically means that the farther a mic is from a source, the farther other mics need to be from that mic. The converse is that the closer you get mics to their sources, the more mics you can pack into an area.

(While looking for information I had found the more conservative version of it, 4:1.)
I think we were very largely respecting the rule, the problem is the trumpet frequencies and projection.

For jazz recordings you usually don't want to get the mics too close to instruments -- it destroys the natural feeling of space that you find in most jazz recordings -- but that in turn means that mics have to be pretty far apart.

-Luddy

OK, I'll keep that in mind (this confirms that I can relax about also hearing the drums a lot into the piano track... that's how I'm syncing the H2 to the H4n - a "clap" by the drummer at the beginning and at the ending of each take - and it does not represent any problem, the piano EQ is not conflicting with a good drums EQ).

The auditorium we rent is pretty large (same price as horrible tiny rooms we used to rehearse in before, with cheap synths with broken keyboards, vs. concert piano... it's just a slightly longer journey).

Again, thanks a lot!!!

-badEnough