Noise floor of dynamic and ribbon mics

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

leswatts

Well-known member
Joined
Apr 29, 2009
Messages
269
Location
Tiger, Ga USA
Hi

Has anyone measured or seen  noise data on passive dynamic/ribbon mics? Generally, it's considered too low to spec on commercial products.

I would expect there to be two basic noise sources:

1) the electrical johnson noise of the real part of the mics electrical impedance

2) the transformed pressure noise of acoustic resistances (often called brownian motion noise)

From my studies the self noise of typical powered condenser mics dominates when using a 1nV/sqrt Hz preamplifier. I suspect that is not the case with passive microphones (with a very quiet room).

An interesting complicating bit is the fact that the electrical impedance is a function of the acoustic environment! As an example, some specialized passive microphones I designed last year change their electrical impedance (at certain frequencies) by 4 times or more depending on distance from an acoustic reflector. Therir electroacoustic efficiency is unusually high though.

Les
L M Watts Technology
 
I never bothered to measure individual mics, but designed preamps based on nominal source impedance of the mics (150-200 ohm?).

Low noise bipolars in the .5nV rt Hz region have been available for decades and now even JFETs are getting down there.

I suspect room noise will dominate for most real world applications.

JR

PS One interesting area of exploration is to combine microphones with a horn, similar to a loudspeaker this would give the mic some "free" acoustic gain, in exchange for reduced pattern aperture.
 
leswatts said:
Hi

Has anyone measured or seen  noise data on passive dynamic/ribbon mics? Generally, it's considered too low to spec on commercial products.

I would expect there to be two basic noise sources:

1) the electrical johnson noise of the real part of the mics electrical impedance

2) the transformed pressure noise of acoustic resistances (often called brownian motion noise)
Basically, you've answered your own question. :)
Most dynamic mics have an impedance of a couple hundred ohms, so Johnson noise is -129.7dBu +/-3.
Gaussian noise (as we say in France) is directly related to the mic's sensitivity; regarding this source of noise, it is entirely governed by the noise floor of the room is is situated in. A mic manufacturer wouldn't want to publish a spec that is entirely out of his control, would he?
From my studies the self noise of typical powered condenser mics dominates when using a 1nV/sqrt Hz preamplifier.
The best condenser mics have 6dB self-noise. Considering a sensitivity of 10mV/Pa, that ca. .5 uV noise or -123dBu. The mic will be the dominant noise source for any preamp with a better EIN, which is quite common (most preamps, at max gain have less than -126dBu EIN). 1nV/sqrtHz is equivalent to -134dBu EIN in a 20kHs BW. This is equivalent to 75 ohms resistor. Any mic exhibiting more than 75 ohms impedance will be noisier than the aforementioned preamp.
I suspect that is not the case with passive microphones (with a very quiet room).
As I jus demonstrated, if the impedance is higher than 75 ohms, that surely will be the case.
An interesting complicating bit is the fact that the electrical impedance is a function of the acoustic environment! As an example, some specialized passive microphones I designed last year change their electrical impedance (at certain frequencies) by 4 times or more depending on distance from an acoustic reflector. Therir electroacoustic efficiency is unusually high though.

Les
L M Watts Technology
Then you need to split the bandwidth in finite elements of let's say 2-4 kHz width, evaluate the average impedance, apply the Johnson law, and do an energy sum of all the bands. (En = sqrt (E1² + E2² +...+ En² - I guess you already know that, but someone may be eavesdropping :) )
 
Yeah, John...about the only references I ever saw were just to the johnson noise of the rated mic output impedance. But I think it would be quite different.

I would expect it to be non constant spectral density, with pronounced peaks at various frequencies possibly changing with placement. That might make it much more audible, even with a little room noise.

I have seen wild looking noise spectra for hearing aid transducers.

Last year I was renting an Agilent impedance analyzer for a job, and I could easily see changes in electrical impedance of a speaker just from putting my hand in front of it.

I'd like to measure this, but I'd need an isolation vessel. My guess is that a typical voice coil type dynamic would be noisier than a ribbon of the same sensitivity due to it being much more acoustic resistance loaded.

As far as horns...I have designed multicall horn loaded ultrasonic microphones. I got about 10 db gain with a particular configuration. Required very very close diaphragm spacing to prevent the compression chamber low pass filter frequency from being too low. It was about 40 microns.

Andy Simpson makes dynamics with little low compression maple exponential horns. Looks like they would cutoff at about 5 kHz.

Les
L M Watts Technology
 
JohnRoberts said:
PS One interesting area of exploration is to combine microphones with a horn, similar to a loudspeaker this would give the mic some "free" acoustic gain, in exchange for reduced pattern aperture.
This has already been done. French company Melodium had their 75A compression chamber mic, an almost inevitable fixture of church PA in the 60's. A copy of the catalogue can be found here http://poisson.ens.fr/Collection/documents/annexes/a468.pdf . The text mentions that it is almost non-directional (5dB rear-axis rejection), which is not a big surprise, considering the size of the horn, and doesn't mention either the use of a compression chamber, which was mentioned in other litt. They did that in order to increase the sensitivity. As a result, the frequency response was very poor, and largely responsible for the "railway station" effect.
The quoted sensitivity, -55dB for 10 Baries (= 1 Pa) may not look like much, but when you see the rated impedance of 10 ohms, you can see the electromechanical transfer is quite efficient!
 
The best method I know for getting very low noise measurements is lowering temperature in critical parts. Transformers are sometimes used in sensitive applications, and if the thermal noise in windings is too much, you just freeze the bloody trannie with liquid hydrogen or nitrogen :) At 3 K the thermal noise is 20dB down from room temperature.

Back to 300K and reality:

With Lundahl's LL1636 1:20 trannie you would have total input referred noise resistance of only 4.5 ohms. That would produce 38nV noise (270pV / sqrt Hz). Just one 2SK170 J-fet as the first active component would increase the total noise to 45nV. I think this is about as low you will get with normal components, and surely low enough to measure a dynamic mic with 100-200 Ohms impedance.

Sorry for not answering to the question actually, but I just wanted to point out that necessary equipment is quite easy to obtain. Even some commercial pre amps could perhaps do the job, but because in this case you don't have to care about small imperfections in freq response you can use 1:20 trannie with fairly low inductance. But as others have pointed out, perhaps this is already an overkill, but would give you a bit more margin to the mic noise.

-Jonte


 
Abbey, posts crossed in the mail....
What I want to see is curves...complex impedance vs frequency. Of course only resistive phenomina create fluctuation/dissipation noise. Knowes have some papers on this for hearing aid mics, and the noise can be very very colored...it's not white. The impedance is not all resistive either. They did take steps that made the noise more, but whiter and more benign.

The mems people are working on this too...but the mics are still horribly noisy.

I would just like to see curves for studio type dynamics and ribbons. A stong noise peak might be audible even if buried in considerable white noise, because it's correlated.

Les
L M Watts Technology
 
Jonte,

The transformer seems a good idea. I can also mathematically derive noise curves (including pressure and force noise) from a kelvin (4 wire force sense) impedance measurement. Too bad I had to return that agilent unit after the job I used it on was done...but it was $1200/month rental. A $45K box. I am thinking of making one though, just for audio frequencies. Not too hard. The only nasty is having an isolation vessel.

Les
L M Watts Technology

 
> electrical impedance is a function of the acoustic environment

You know why. But for others:

Plot the impedance of a loudspeaker driver in the open; again in a small box. There is a 5X rise of impedance at bass resonance, and changing stiffness via box size changes resonance and impedance curve dramatically.

A series of impedance bumps can be observed if the speaker is coupled to a pipe. Try 4 feet of 4" shipping-tube on a 4" driver, then block the end and try, then saw it down to 3 feet.

> Their electroacoustic efficiency is unusually high though.

Right. That's why we rarely note the effect. A cone loudspeaker's impedance will change when worked in vacuum or when coupled to an infinite tube/horn. However the typical nominal efficiency is ~~1% and so most such impedance shifts are 1%. It can be easily observed near bass resonance on any cone with Qt<1 because the electroacoustic efficiency here is well above 10%. It can be observed on light diaphragm large magnet horn drivers: midband impedance may double on the horn. IIRC, a 2440 is rated 16 ohms, is wound at 12 ohms, but shows more like 24 ohms 200Hz-2KHz if the horn loads well.

Speaker, mike, what's the difference? Oh, some difference, sure, but both must obey fundamental acoustic and electric laws. If efficiency is zero, acoustic load is irrelevant, if efficiency is 100%, ALL impedance is acoustic impedance seen through motor-laws. Real devices have "small" efficiency and "small" acoustic loading effects on the electric side. And note that the electro-acoustic efficiency may be inferred from electric impedance change with acoustic loading (but in real devices, the difference is small and errors abound).

The self-noise of an E-V 635A is near-enough pure resistance noise. It has only slight electric quirks, and is not quite sensitive enough to hear its small Brownian over its electric hiss.

SM-58 has bumpy impedance, plus a transformer, and may have an interesting noise curve; however most '58 users don't linger near any noise floor.

Condensers damped with acoustic resistance (backplate holes) have Brownian, and this can dominate part of the noise spectrum. But this order of sensitivity requires compromises with other aspects (cost), so the wise designer will stop striving for sensitivity when head-amp hiss falls below acoustic hiss. IIRC, one mike was designed with low acoustic damping, the resulting bump nulled with an electric filter. Of course there is no free lunch, and the lower Brownian component must be judged against electrical complication, and the fact that a "nulled" resonance still rings.

Since most condenser applications "require" a tube IN the mike, it becomes "easy" to boost-up to any convenient output level. Some old 3-tube mikes delivered Line Level. The classic German condensers went 1-tube for size and heat, needed a Mike Preamp at console, but it did not have to be a High gain LOW noise preamp such as ribbon and dynamic mikes use. Most of our Jensen, LL1636, 990, and Macky VLN/YZ trickery is moot when working from a higher-output condenser, a transformerless TL072 is almost good enough.

I think Olsen claimed some of his lab ribbons would hear Brownian, very likely true, though the incremental cost and other compromises was a poor deal for practical users. I believe the ribbon has a small advantage over the dynamic coil, though extreme results mean extreme costs and no musician needs to work near Brownian level.

> Gaussian noise (as we say in France) is directly related to the mic's sensitivity; regarding this source of noise, it is entirely governed by the noise floor of the room

You mis-understand. If you had a perfectly silent room, and a perfect transducer, and temperature above absolute zero, you would hear individual air molecules randomly striking the diaphragm. This is related to the motion of small particles in a non-moving liquid, observed by the Roman Lucretius, the Scot Robert Brown, and probably many others. Einstein touches it in his 1905 paper.
http://en.wikipedia.org/wiki/Brownian_motion

In free space, Brownian is rarely troublesome, but mike damping cavities may concentrate the effect; I never really had to think about it.

> renting an Agilent impedance analyzer
> I'd need an isolation vessel.


To read impedance plots: Noise source, resistor, amplifier, spectrum analyzer ($9 on an iPhone). A large glass jar, or concrete flower-pot, gives considerable isolation; the hard part is sealing the hatch and leads. An old-fashioned pressure-cooker is a classic quiet-pot.
 
PRR,
thanks for the speaker analogy. I was about to type it myself! Your explanation was good.

It's granted that passive dynamics and ribbons have low noise. My supposition is that the noise can be highly colored with some mics, and as such could have audible effects, even with room noise. Consider a single tone source that can be heard buried 20 db below a white noise signal.

I would expect ribbon mics to have a smooth noise curve due to the fact that they operate mostly in the mass controlled region. Possibly a big spike at LF resonance though.

However consider a mic like the SM57 (I know that one pretty well). It has three resonators used to tailor
the response...The air space just behind the diaphragm for High frequency, the resonator cap over the diaphragm for the famous upper mid boost, and the  damped diaphragm  mechanical resonance in the lower mid range. Each of these will make a peak in the real part of the electrical impedance, and therefore a noise peak.

I'm just exploring whether highly colored noise has subjective effects, even though it's low.

To read impedance plots: Noise source, resistor, amplifier, spectrum analyzer ($9 on an iPhone). A large glass jar, or concrete flower-pot, gives considerable isolation; the hard part is sealing the hatch and leads. An old-fashioned pressure-cooker is a classic quiet-pot.

As an acoustics engineer , I have plenty of nice green B&K gear in the lab. I could also make a fine impedance analyzer with the studio digital interface and some minor outboard tronics to allow kelvin measurement. It actually has better performance than the green boxes...it's just not calibrated to be "admissible in court".

As far as an isolation vessel, I note that at my alma mater the mems researchers are using a welded double walled box with vacuum in between. Funny thing is they don't need it (yet)...because the mics have just horrible noise. Hmm...I could borrow it....

Les
L M Watts Technology
 
PRR said:
You mis-understand. If you had a perfectly silent room, and a perfect transducer, and temperature above absolute zero, you would hear individual air molecules randomly striking the diaphragm. This is related to the motion of small particles in a non-moving liquid
I don't think I misunderstand anything. What I'm saying is that "the air molecules randomly striking the diaphragm" are a sound pressure; it produces a noise voltage that is directly a function of the mic's sensitivity. And there is nothing like a perfectly silent room, "at a temperature above absolute zero", just because of the aforementioned thermal agitation.
If you meant a perfectly soundproof room, that goes without saying.
 
I am sorry; I mis-understood "the noise floor of the room" to mean blower rumble etc.
 
Well, great discussion. Thoughtful posts.

I don't think I misunderstand anything. What I'm saying is that "the air molecules randomly striking the diaphragm" are a sound pressure; it produces a noise voltage that is directly a function of the mic's sensitivity.

Ok, Abbey. I understand. If you had a 100% isolated warm room filled with air would it have a sound?
Is thermal energy sound? I'm guessing, but I would say kinda. You must have a dissipative system to detect it, but a limit might be the characteristic impedance of plane wave propagation, 430 Rayls.
However a practical wide bandwidth microphone would have an acoustic impedance vastly higher than that and therefore predominate in noise generation. Seems that our microphones will make rooms cold though...

www.en.wikipedia.org/wiki/Brownian_motor

PRR, the baxendall link was enlightening. As I expected the cardioid mic has the biggest motional impedance induced bumps in it. It's self noise is most colored. The resistance controlled omni dynamic and damped backplate condenser have smooth motional impedance, but it can be significant. Here's what happens when one takes steps to lower it:

www.bksv.com/doc/Bp0389.pdf

My feeling so far is that the typical 1nV/sqrt Hz head amp drives studio microphone design.... Once you reach about a 1db noise factor no further electroacoustic design compromises are made. Does this mean
attempts at designing lower EIN ( than about 1 nV/sqrt Hz) head amps are useless exercises? I would say no...if you want to use less sensitive smaller mics like mems.

I find the whole flutuation/dissipation noise theory fascinating. I'm an engineer, not a statistical physicist... I know the formulas for  electrical resistance, acoustic resistance, and mechanical damping noise and can measure and calculate voltage, pressure, and force noises. But I really never delved into
the fluctuation /dissipation theorem and it proofs. But I am now. It has little bearing to this discussion, but I can't stand not knowing and understanding something fully. So I am studying Langevin and Nyquist.
Hey, it proves the corpuscular theory of matter, too!

I design microphones for a living, but not studio microphones. So I am learning here.


Les
L M Watts Technology
 
leswatts said:
I would expect there to be two basic noise sources ... in the ribbon mic:
1) the electrical johnson noise of the real part of the mics electrical impedance
2) the transformed pressure noise of acoustic resistances (often called brownian motion noise)
Look to old Olson paper. He had measured acoustical noise (noise of real part of
radiation impedance) in the ribbon microphone. Note, that He had specially develloped
magnetic circuit and in his system electric noise (noise of the ribbon resistance) still dominates (it was
some 1 dB SPL or so).

When the transformer is properly engineered (= noise of winding + barkhausen noise is
comparable, even lower with the electric noise of ribbon resistance, it must be some hunderths of miliohms,
not ohms !!!) main noise source is ribbon resistance.
I have observed effect, that when contact between ribbon (alluminium) and tranny primary winding
(copper) is not metalurgical, noise of the contact dominates over other sources.
It may affect performance impact of ribbon mics, that they are noisy.
I do mean, that it holds only for improperly ribboned (or re-ribboned) mics, where
contact noise dominates (contact noise was some +10 dB greater than noise of all system in my experiments).

Let you solder together ribbon mic input tranny primary wires together.
Then measure ribbon mic in the vaacuum and compare noise values.
When measurement in the vaacuum will be worser than 3 dB, you have serious problem with contact noise.
(Note, you do not need vaacuum if ribbon frame with contact can be removed from the flux.



 
Thanks xmvlk.

I have often wondered about ribbon contact quality since it tends to have rapid surface oxide growth.
My dad just bought a 60 year old altec (western electric) 639  ribbon/dynamic microphone and it works perfectly, so it must have microwelded metal/metal contact still. The reason he bought it was because before college he had a job installing and tuning the ribbons!

I would not be surprised that ribbons have electrical noise dominate, since they have low acoustic resistance other than radiation resistance.

I have read that barkhausen effect and certain glassy solids do not follow classical fluctuation theory in their noise production, but can't comment until I learn more.

Les
L M Watts Technology
 
leswatts said:
Well, great discussion. Thoughtful posts.

I don't think I misunderstand anything. What I'm saying is that "the air molecules randomly striking the diaphragm" are a sound pressure; it produces a noise voltage that is directly a function of the mic's sensitivity.

Ok, Abbey. I understand. If you had a 100% isolated warm room filled with air would it have a sound?
Is thermal energy sound? I'm guessing, but I would say kinda.

Les
L M Watts Technology
Not kinda; it's real sound, it's mechanical energy originating from thermal energy. I know it is difficult to grab instinctively, but it's exactly the same as a room, perfectly opaque, with no light source in it, has some infra-red radiation, unless it is at absolute zero, or the very basic fact that a black body emits light depending on its temperature, and again, it's not kinda light, it's real light.
 
Not kinda; it's real sound, it's mechanical energy originating from thermal energy.

Ok, I think I can buy that. I'm not really trying to get an intuitive picture of it...rather a rigorous one by going through the theorems. I haven't seen so many Lorentzians and Hamiltonians since  I was in school. ??? :eek:

Here's what I meant by kinda:

The theorems say that resistance is fundamentally made possible by fluctuation. If you have a large resistance, you must also have large fluctuation to make it possible. Now air has a characteristic impedance, and it's a real number ( a resistance) for plane wave propagation. Microphones often have very high acoustic resistance with a corresponding high fluctuation across it. Since they are usually a huge acoustic impedance mismatch with air (much much higher) the fluctuation from their resistance dominates.
However if a microphone could be made with very low mass low resistance components such that it was an impedance match to the air and gobbled up all radiation intercepting it, you would still be left with a transformed radiation resistance from the impedance of the air as it's electrical impedance. The noise of that I think would be your warm room noise.

I'll consider the above a guess until I fully understand the fluctuation dissipation theorem and it proof.
Then I'll know for sure. I hate to not know, or guess.

Les
L M Watts Technology.

 
> Seems that our microphones will make rooms cold though...

Yes, this follows. We have added a path from a closed room to an outside destination. If that outside destination receives energy (which we want), the total energy in the room must decrease.

Conversely, mike-amp hiss in a reversible transducer must warm the room.

For profit, we fuel musicians with twinkies and coke, inject them into the room, add energy far in excess of what our recorder needs, and must suck it out with fans, and more noise... the thermal-noise heat balance is lost in the mess.

> less sensitive smaller mics like mems.

You have to catch some air.

You can recover real power from a large loudspeaker near a loud source. Passive (no battery) telephone systems are practical, if the efficiency is resonant-peaked in critical speech band and you talk real loud and close.

Conversely, a small device captures very small power. You reach a point that the output is not recoverable (or not at desired quality) even with electronic amplification.

So you hope to find a compromise size: small enough for convenience and cost and directivity, yet large enough for ample power output.

This has usually put us between 0.3" and 2", with the smaller mikes optimized for extended HF (or compact toys) and higher noise level.

It may be interesting that the human ear is a 2" cup with a 0.3" hole. The earhole in cats and Corgis is only slightly smaller, even though the creature is 1/3 my linear dimension. Microphones do not stray far from the Creator's blueprints. It's all air. The hardware (fleshware) must be sized to suit the air.

MEMS is micro-silicon, right? So the largest structure is about as small as I can see. My naive thought is that this is too small to approximate human hearing without some significant re-writing of laws of nature. Perhaps a horn, but a wide-band horn is a very large thing. If the price is low, an array makes much sense.... but how can an array of structures be cheaper than a punching of saran-wrap?

Ah: signal voltage goes up as square-root of diaphragm area. Noise voltage goes down as square-root of FET area. Diaphragm costs $X/acre, FET costs $Y/acre, customer will pay $Za $Zb $Zc for noise at level a b or c.

With MEMS on Silicon, you -might- find that you want over half your die as FET rather than diaphragm. But I doubt it, unless diaphragm construction is frightfully costly (FET is cheap).
 
leswatts said:
The theorems say that resistance is fundamentally made possible by fluctuation. If you have a large resistance, you must also have large fluctuation to make it possible. Now air has a characteristic impedance, and it's a real number ( a resistance) for plane wave propagation. Microphones often have very high acoustic resistance with a corresponding high fluctuation across it. Since they are usually a huge acoustic impedance mismatch with air (much much higher) the fluctuation from their resistance dominates.
However if a microphone could be made with very low mass low resistance components such that it was an impedance match to the air and gobbled up all radiation intercepting it, you would still be left with a transformed radiation resistance from the impedance of the air as it's electrical impedance. The noise of that I think would be your warm room noise.

Les
L M Watts Technology.
I entirely agree with the above. I think we're talking about the very same thing in different terms. Basically, the process you're describing is the physical analysis of the electromechanical (mechano-electrical?) transfer function. All I'm saying is that this transfer function is entirely described by the sensitivity/frequency response and impedance curve of the mic (assuming it is minimum-phase)
 
Back
Top