A Theoretical Question: EQ vs mic design

  • Thread starter Thread starter mcolling
  • Start date Start date
It's funny, cause what you speak of isn't far from happening.


Digital mic's are slowly becoming a re occuring thing, and while they are expensive now, the technology of the future will make them affordable.


You could have hundrends of presets of almost any type inside the mic. dynamic effects, mic modeling, EQ, time based effects, etc.

You could literally upload any of these on a mic, which means you could have your mic individually gated, compressed with reverb at the mic, not outboard gear.

cool huh?
 
But it's a challenge that anyone with the right mindset could potentially overcome given enough time and dedication to understanding, is it not?
Click to expand...

Thinking about it a little more, I'm not very confident that truly accurate mic modeling can be developed. Maybe mic use can really only be described by chaos theory because of the effect on a mic's sound caused by the infinite variety of placement, room acoustics, mic-to-preamp mating, and user judgment.

Tim
 
The concept is a good one, but it's flawed in one major aspect.

"How would it be possible to replicate the physical components of many different mic and use that in one mic".

I still have yet to see a digital way of replicating a tube mic.

You have to consider that one mic design won't work like every other. Sorta like a one size dosn't fit all. Everything from the circutry on the inside to the design of the casing is essential to the sound of that specific mic.

Different mics will pick up sound in very different ways, so it wouldn't be feasable to try and cram mics at extreme ends of the spectrum (sm57 versus a U87) into one easy to use mic.


However, the digital mics I mentioned are probably the next up and comming thing that may give something close.
 
After all the responses in this thread, I am convinced that a mic modeller that relies on eq and colouration during mixdown could never work 100%. You cannot add sound that is not picked up by the mic in the first place. It seems to me that the digital mics discussed by LRosario would not actually accomplish anything different than the Anteres plug-in. They would just work like a piece of outboard gear attached to the mic, correct?

However, the question is whether a sophisticated piece of software could get close enough. With an affordable but very transparent mic (like the behri b2, say), couldn't you get a good enough signal to run a hypothetically perfect mic modeller, and get a convincing result? Of course the resulting signal isn't going to have the same transient response, but is that really that important if it has the same colouration and eq?

My question is really asking: What percentage of a mic's character owes to the hypothetically recreatable elements of distortion and eq vs. the non-recreatable elements of transient repsonse and response/rise time?
 
As I stated earlier, transient response is a shapeable characteristic, and in fact even Antares Mic Modeller does it.
 
I've got a lot of tool in my workshop - mechanical and woodworking tools. You know those all-in-one tools you see advertised all the time? Well even if they're well made they never work as well as a good quality tool made for a specific job. When I tried the demo of the Antares modeler in the past, it made me think of that.


Dogs only understand visuals that are in three dimensions, so my dog doesn't watch TV because it's two dimensional. She'll sit excitedly looking out the window at birds in the yard, but doesn't see a TV picture of a similar thing as real. Now, she's seen some technologically advanced TV, and that doesn't seem to matter. Maybe mic modeling will reach a convincing level around the same time I can get a TV with good enough 3-D simulation that my dog will watch.

Tim
 
Yes, true. But he said:
mshilarious said:
an EQ cannot recreate a faster-than-recorded transient response. Even modeling algorithms have their limits.
Click to expand...
To me, that makes sense. How can you model something that wasn't picked up in the first place? Maybe you can exaggerate things that are underestimated, but if it isn't there, then it isn't there. I guess you could slow down transient response, but could you speed it up?

If you can model transient response, then wouldn't a fairly effective mic modeller be theoretically possible?
 
The answer is yes within certain constraints. I can make any linear mic have exactly the same frequency response as another so long as there are no nulls in its response and the equality is obtained at one orientation, say on axis for each mic. The trick is to measure the impulse response of the source and the target mics and then "divide" the target IR by the source IR. The "quotient" IR is a transfrer function which if convolved with recordings done by the source will emulate the target. The division can be done in the time or frequency domain.

It is said that the on axis measurements are not good enough but I have found from measurement and from doing this that in reality, so long as the mics are of the same type and diameter, the transformation that applies on axis will also apply to off axis to a good approximation. If you do this with one being a LDC and the other a SDC then it will be less so. The variation of frequency response with angle is pretty much the same with capsules of the same size, geometry and pattern. Even if they aren't, the result can be surprisingly close. Of course, the mounting and housings also affect the angular variation but, still, the results of the procedure are pretty good.

This isn't the same as "equalization" as the word is commonly used implying a knob or slider thingy with a minimum or linear phase response. Using IR's and convolution is still "equalization" but it can realize any arbitrary magnitude and phase transformation.

Even the Antare's Mic Modeler uses transfer functions that are minimum phase so it cannot correct for phase differences but this may not be as important as it may seem. The reason they do that is that minimum phase transfer functions also give minimum latency if used in real time or as an insert on a track for a mix.


Bob
 
Interesting post Bob, thank you.
A question for you, then: could you hypothesize on why Mic Modeller isn't particularily effective even when applied to source mics that are quite flat? Considering (I think) that it uses a technique similar to that which you were talking about, it should sound close to correct, right? The thing I always find with Mic Modeller is that the modelled differences aren't nearly as large as the actual differences between mics. I'll often find myself thinking, "hmm, is this thing even doing anything?" with Antares, whereas the difference in a new mic is immediately impactful. Though I've yet to run any blind tests, I don't think this is a purely psychological shift. Something seems wrong with the modeller.
 
you'd be better off starting with a sdc, better tranzient responce (i can believe the program can slow the tran responce down, but if it's not thare, you can't assume it back) like an earthworks tc 30.
(ignoring off axis responce problems)

btw, i don't advocate mic moddeling myselfe. maby it'll work some day (think how far midi key boards have come) but not for a llllloooooonnnnng time.
 
bleyrad said:
Interesting post Bob, thank you.
A question for you, then: could you hypothesize on why Mic Modeller isn't particularily effective even when applied to source mics that are quite flat? Considering (I think) that it uses a technique similar to that which you were talking about, it should sound close to correct, right? The thing I always find with Mic Modeller is that the modelled differences aren't nearly as large as the actual differences between mics. I'll often find myself thinking, "hmm, is this thing even doing anything?" with Antares, whereas the difference in a new mic is immediately impactful. Though I've yet to run any blind tests, I don't think this is a purely psychological shift. Something seems wrong with the modeller.
Click to expand...

It all depends on just how good the source inversion and the target model impulse responses are. I've only checked one of their target models against measurements of my own, the AT 4033, and found it pretty close. One thing the Antares models do is give a rather smooth approximation to frequency responses that by my measurement are anything but smooth. They seem to capture the trend but not the detail. It may well be that the sonic impact of a mic is as much a part of the small magnitude details as it of the trend.

It's also true that the Antares models don't attempt to invert or model the phase component of frequency responses. I'm not sure at all how much this affects their efficacy but it's gotta have some effect.


Bob
 
Transient responses are no problem to model _if_ the target doesn't completely drop off a cliff at the higher frequencies. If it's HF response is slight compared to the source then the amount of gain required can raise noise levels to unacceptable levels in the region.

If latency is allowed, there is no problem mapping any transient response shape to another. You can't do this, in general with minimum phase transforms because accurate phase manipulation is essential to making the responses line up temporally (after some latency.)

Bob
 
Guys ... off-axis response is such a huge part of a mic's sound and usefulness.

And as many times as I try to go over it in my head, it just doesn't add up. You just can't take something like an omni-directional reference mic and model things like off-axis response, bleed, etc.

Without an infinite number of reference tones sampled from an infinite number of locations on and off-axis ... sampled at all possible volume levels ... you're just not going to be able to get a good enough approximation.

The best you can hope for would be to have something similar to a POD. A useful tool for certain situations, but not in any way a substitute for the original.
 
giraffe said:
btw, i don't advocate mic moddeling myselfe. maby it'll work some day (think how far midi key boards have come) but not for a llllloooooonnnnng time.
Click to expand...
............................
 
chessrock said:
You just can't take something like an omni-directional reference mic and model things like off-axis response, bleed, etc.
Click to expand...


Of course not, but I think this is only a small part of the equation. Most vocals I record pretty much on-axis. But Mic Modeller can't even do that right. Yeah there's the bleed issue but when you think about it, if you're in a dead room the only bleed you have to worry about is wrap-around sound, and that should figure into the overall frequency response of the mic when modelled with an impulse response.

I think Bob touched on the larger issue here, that Mic Modeller only shows slow trends in response and not high-resolution peaks or dips that perhaps contribute even more to a mic's character. This makes a lot of sense to me.
 
So, what is the interface like on the Antares modeller? Does it have something like a selection for "mic in" and "mic out"?
 
Yup, that and a proximity distance in and out. It also has a "tube saturation" slider which is unrelated to any actual emulation.
 
chessrock said:
Guys ... off-axis response is such a huge part of a mic's sound and usefulness.

And as many times as I try to go over it in my head, it just doesn't add up. You just can't take something like an omni-directional reference mic and model things like off-axis response, bleed, etc.
Click to expand...

You are correct in that. Effective modeling requires that the source and target mic be of the same type, size and pattern. Then, however, the transform that must be applied for good on-axis modeling will also transform the off axis effectively. The way the on axis response of a mic morphs as you go around is not much different between two mics of the same type, size and pattern.

However, applying a mapping between mics that are different in these regards can still produce a result that is superior to the original. That's a matter only your ears can decide.


Bob
 
You must log in or register to reply here.
Back
Top