Microphone Modelling

[Maxim Didur]

If I want to simulate one microphone from another using post-processing, what audio parameters should I take into account to get as close a result as possible? (I mean, for example, I need to match the frequency response of the microphone I want to simulate, what else?)

 

[Tim Campbell]

I helped design Antelopes modelling microphone. It really depends on which mics you want to imitate.
Using off the shelf parts I would recommend an edge terminated capsule and a very clean amplifier.
You cannot easily imitate the top end of an edge terminated capsule by using eq on a K67/K47 type capsule. It is easier to do the opposite - eq an edge terminated capsule to emulate a Neumann type.

 

[kingkorg]

If I want to simulate one microphone from another using post-processing, what audio parameters should I take into account to get as close a result as possible? (I mean, for example, I need to match the frequency response of the microphone I want to simulate, what else?)

https://groupdiy.com/threads/townsend-labs-sphere-guts-and-basic-analysis.77464/page-2
https://groupdiy.com/threads/slate-ml-1-vms-microphone-guts-images.79079/

 

[gyraf]

And as the sound character in a mic comes from its off-axis response, which we can not post-process-add, or even estimate, the whole exercise is void. Always was, since the first antares version in 2000.

/Jakob E.

 

[kingkorg]

And as the sound character in a mic comes from its off-axis response, which we can not post-process-add, or even estimate, the whole exercise is void. Always was, since the first antares version in 2000.

/Jakob E.

There are ways to resolve this issue, but at this point it's still proprietary stuff. AA Polar Designer gives some clues, but there's more to it

 

[gyraf]

Interesting...

 

[GeorgeToledo]

And as the sound character in a mic comes from its off-axis response, which we can not post-process-add, or even estimate, the whole exercise is void. Always was, since the first antares version in 2000.

/Jakob E.

I think that if multiple capsules were used, as acoustically transparent of a headbasket as possible, and good programming to interpolate between impulse responses, that these systems could take a big step up.

Still issues with ever doing it perfectly, but that would be a starting point if someone wanted to up the ante.

Manipulating polar pattern after the fact, with a single capsule, would be a smaller optimization. That one’s not really a new idea.

 

[kingkorg]

I think that if multiple capsules were used, as acoustically transparent of a headbasket as possible, and good programming to interpolate between impulse responses, that these systems could take a big step up.

Still issues with ever doing it perfectly, but that would be a starting point if someone wanted to up the ante.

Manipulating polar pattern after the fact, with a single capsule, would be a smaller optimization. That one’s not really a new idea.

George, you are really close to what i suggested in my previous post.

There is one issue impossible to resolve. Variance between the original mics you're trying to model. Random guy on YouTube will pull out his random u47, and compare it to the modeling mic. It would be a miracle if they sounded the same, even though the modeller could be indistinguishable from the mic you used as the target.

 

[Tim Campbell]

Yes. Antelope process both diaphrams seperately with 2 seperate amplifiers within the body.

 

[OneRoomStudio]

It would be interesting to apply similar processing to even more capsules - something like an ambisonic microphone with spacial processing. I can't imagine there would be much you couldn't model with a setup like that.

 

[Tim Campbell]

George, you are really close to what i suggested in my previous post.

There is one issue impossible to resolve. Variance between the original mics you're trying to model. Random guy on YouTube will pull out his random u47, and compare it to the modeling mic. It would be a miracle if they sounded the same, even though the modeller could be indistinguishable from the mic you used as the target.

This is SO true of all the clones out there whether it is capsules, transfomers, complete microphones, etc.

There was just a review where 251's were discussed. A well known engineer contributed to the discussion because he has 2 vintage 251s. He unequivically declared that 1 manufacturer's mic was the only clone that sounded exactly like the real thing. The problem is his mics are some of the brightest 251s I have heard and he has mentioned this in multiple forums. Blackbird studios has I think 25 vintage Ela M 251s and none of them sound the same and I believe only 3 are considered to sound stellar.

 

[Maxim Didur]

George, you are really close to what i suggested in my previous post.

There is one issue impossible to resolve. Variance between the original mics you're trying to model. Random guy on YouTube will pull out his random u47, and compare it to the modeling mic. It would be a miracle if they sounded the same, even though the modeller could be indistinguishable from the mic you used as the target.YE

Yeah, thats why I'm saying as close as possible, not the same. However by "sound parameters" I mean: Frequency response, Transient response, polar pattern, saturation of microphone, are there any other things I need to consider?.

 

[kingkorg]

Yeah, thats why I'm saying as close as possible, not the same. However by "sound parameters" I mean: Frequency response, Transient response, polar pattern, saturation of microphone, are there any other things I need to consider?.

Haven't we already discussed all the parameters in the two previous threads you started? The biggest challenge would be to measure all the aspects in order to match them. Not even Slate, or Townsend Labs have done this 100% right. I don't mean to be condescending, but you can't do this on your own. For starters try to figure out how to match frequency response, or read the two threads i suggested in order to get the best out of what's already there. And leave alone the "transient response" aspect of it, it's an oxymoron, and should be least of your worries. It would also help if you explained exactly what you're trying to do.

 

[Maxim Didur]

Haven't we already discussed all the parameters in the two previous threads you started? The biggest challenge would be to measure all the aspects in order to match them. Not even Slate, or Townsend Labs have done this 100% right. I don't mean to be condescending, but you can't do this on your own. For starters try to figure out how to match frequency response, or read the two threads i suggested in order to get the best out of what's already there. And leave alone the "transient response" aspect of it, it's an oxymoron, and should be least of your worries. It would also help if you explained exactly what you're trying to do.

Well, basicly I have a one phrase recorded on diffrent micropones, through the same audio interface, in the same room and position. Lets say I have this phrase recorded on AKG p120 and Neumann U87 AI. My goal is: How can I make phrase recorded on AKG p120 sound like Neumann U87 AI, as close as possible. What microphone characteristics I need to consider to make it sound as close as possible. What should I try to match (Except of frequency response cause I already know about it, what else is very important)?

 

[kingkorg]

All the aspects discussed in the previous thread. You can match them using matching eq and pink noise as sound source in a dead room. But since p120 has different capsule/headbasket and very different off axis response they will sound similar just at the modeled position and distance. You need a mic with k67 capsule and similar headbasket in order to model u87.

 

[soliloqueen]

there is no such thing as "the" frequency response of a microphone. a frequency response is a 2d slice of a 5d response (distance, angle, frequency, amplitude, time). You can match two frequency responses, and that will ensure that if you take two slices in the same circumstances out of each microphone's responses, that will match, but that's it. most modeling mics use models like this.

a frequency response isn't a characteristic of a microphone, it's a measurement of a microphone. that's like saying that a photo of a lake is a characteristic of the lake. as you add more types of measurements that account for more dimensions, you can start to approximate the source closer and closer (also recording sound of the lake increases the feeling of immersion that you are at the lake) but that would be very difficult. in order to add dimensions to the mic sim, you'd need to do some pretty insane things. I can think of a few ways to do this, but I'm not gonna do r&d for free for a stranger.

 

[Maxim Didur]

there is no such thing as "the" frequency response of a microphone. a frequency response is a 2d slice of a 5d response (distance, angle, frequency, amplitude, time). You can match two frequency responses, and that will ensure that if you take two slices in the same circumstances out of each microphone's responses, that will match, but that's it. most modeling mics use models like this.

a frequency response isn't a characteristic of a microphone, it's a measurement of a microphone. that's like saying that a photo of a lake is a characteristic of the lake. as you add more types of measurements that account for more dimensions, you can start to approximate the source closer and closer (also recording sound of the lake increases the feeling of immersion that you are at the lake) but that would be very difficult. in order to add dimensions to the mic sim, you'd need to do some pretty insane things. I can think of a few ways to do this, but I'm not gonna do r&d for free for a stranger.

I emailed you

 

[Cqwet Dbdfte]

I'm new to microphones, but have some comments.
The Neumann U47 design is well designed for its time. Looks like a further development of an earlier Western Electric mike.
U47 tubes were selected for low noise and had a high (DC) heater voltage (which likely had low ripple). Half the heater current was burned up in a resistor. Goodness! The heater power also provided fixed bias for the tube. Avoiding a standard 6.3V or 12.6V heater avoided AC heating, clean DC heat back then was not an option. Not an issue today. (I have seen HV DC used to power up a series of heaters in old HP oscilloscopes, great when one tube goes out, the chain breaks.)
There are several tubes that could be used with DC heat, utilizing one of the "spare" wires in the 7 bundle. The VF14 operates with low plate voltage, and low plate current, not exactly the most optimal point for this or any tube. But it worked. Had they had JFET's back in '47, I think the tube would have been avoided.
The output transformer design what I've seen looks pretty simple, but is evidently good enough.
Some have reported somewhat weak bass response for the mike, getting worse with aged tubes.
Is this because of too low grid bias current, or some fine balancing act between capsule's weak current supply and some grid action?
What is the BV.08 output voltage at some high, or max, usable level?
The diaphragm tension and thickness would likely have a considerable effect of the output, as would may other factors like hole geometries, and material used.

 

[Philbo King]

About the only realistic way to simulate a mic is to use an Ambisonic mic with better bandwidth than the one you're trying to emulate.

 

[gyraf]

..and on top of that, all its individual capsules will have to have perfectly flat or near-perfect pattern-vs-frequency response, i.e. completely linear off-axis response, as this is otherwise a significant deep-embedded source of coloration that we can't access separately later or correct for

Proximity-related behavior is a whole chapter by itself, and does not lend itself to easy simulation

/Jakob E.