Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[klaus]

I am trying to figure out why most manufacturers of condenser mics still include frequency response curves in their literature.
On the one hand, graphs seem to comfort buyers- they seem to add to a product's air of professional, objective, scientifically unassailable cache. And they may even show different patterns between different mics from the same particular company (if we assume a rigorous and consistent application of a precisely defined test standard.)

On the other hand, due to lack of any uniform, comprehensive standardization that would be widely followed and agreed upon in our industry, these graphs and patterns give me no confidence hat they could predict the actual behavior of a microphone in its intended application; namely how well the mic will convert the music it captures into electricity in the real world as we hear it.

In my view, graphs of microphone frequency responses seem to predominantly serve the function of confirming one's subjective sound interpretation, and mostly after an opinion has already been formed and a decision already made.
But there is a minority of users who, due to the now ubiquitous access to ProTools and its visual displays, see something weird on their screen, then send me the mic and ask me to fix the weirdness they saw. Asking them to explain what they heard, they admit that it was something they saw, not heard. These cases point to lag between the easy availability of microphone's response curves on a computer screen and the understanding of what they portray.

There are, of course, gross errors in microphone behavior that can and will show up on a frequency plot. But I have come to the conclusion that gross response aberrations, once spotted by the ear, will be addressed, regardless of what a graph will show.

I'd love to hear from owners who use frequency response curves to make any kind of microphone-related decision, even if it's only making a preliminary selection of the microphone they might end up buying.

P.S.: In the March/April 2012 Tape-Op "End Rant" you can find a polemic by Ethan Winer which claims that we can pretty much measure every audio device's property in an objective fashion by plotting "the standard metrics of frequency response and distortion".

[Malotki]

Klaus

What was the myth ?

I think you'll find the plot is the result of a test the manufacturer uses to test if the microphones meet the standards for sale.  I would imagine most companies are selling for profit and can't listen to all there mics before selling. Anyone buying thinks,  oh it needs the standards required by said manufacturer and promptly forgets abiuts it and gets on with using the Mic....

Where's some people read far too much into the graph,  thinking that the company is selling them a bespoke recording solution to standards set beyond there interest in making as much money for the shareholders as possible.
D_

[Kai]

...Ethan Winer... claims that we can pretty much measure every audio device's property in an objective fashion by plotting "the standard metrics of frequency response and distortion".

Audio, in the electric/digital domain consists of different frequencies moving in a large dynamic range.
Therefore a static, one dimensional measurement can't tell the whole story even in this simpler case.

The more complex a device or a devices function is, the more "dimensions" have influence on the result.
An electroacoustic device like a microphone, or even more, a loudspeaker, has so many aspects that would need to be displayed in technical date and graphs to describe it's performance, that it's impossoble to overlook and judge the quality based on those.

Nevertheless, the frequenecy response of a mic does give me valuable information in certain cases.
The Sennheiser MD421 e.g. has a sudden step up at 2kHz, corresponding with it's very "piercing, present" vocal sound.
What you can't read out of it's graph is that it delivers an extremly dry and well-defined bass.

Regards
Kai

[klaus]

Klaus
What was the myth ?
I think you'll find the plot is the result of a test the manufacturer uses to test if the microphones meet the standards for sale. 

That's another myth. If you ever see an actual graph made on standard B&K equipment as used in most manufacturing, you will see that the graphs provided to buyers are idealized, averaged plots, void of any fine points, and usually idealized to ± 2dB (meaning, a whopping 4 dB variation could be present from one specific frequency to the next.) Sennheiser used to be one of the few companies which included the actual quality control plots with every (upper price range) mic.

[soapfoot]

Many end-users... (thinking particularly of the growing prosumer market) may not know much about interpreting documentation even if there was an acceptable, AES-standardized means of measuring and displaying the results.  Looking at the 'shape' of the graph is rendered even more meaningless when you consider that the scale and even range defined on the X and Y axes can vary-- everything looks "flat" if you zoom out and make the scale coarse enough. 

Marginally more useful are multiple plots of the same microphone under different circumstances using the same scale.  In other words, for example, a clearly labeled graph showing multiple distances can give you a data point for understanding how a particular directional mic responds to proximity.  It cannot, however, tell you how a mic sounds at any distance, and it is not very useful for predicting how one mic sounds relative to another.

[fma]

… for example, a clearly labeled graph showing multiple distances can give you a data point for understanding how a particular directional mic responds to proximity.  It cannot, however, tell you how a mic sounds at any distance, and it is not very useful for predicting how one mic sounds relative to another.

In a similar sense, some conclusions about the directivity of microphones might be drawn from the frequency responses at different angles - if they've been measured and provided (and not idealized).

Usually, I don't recommend polar diagrams as a substitute for them. They are much harder to read because polar diagrams are standardized to 0dB at 0° angle for all given frequencies. Therefore the real frequency response of the microphone does not enter into the polar graph.

fma

[Dinogi]

Was it Walter Sear who said he didn't care if a mic had a sawtoothed frequency response, if it sounded good on what he was using it for?
I always got the impression that many of the great mics were great because of their particular anomalies.
I also believed that the art of voicing a microphone went far beyond textbook engineering, and became more about preference and purpose.
If any of the above is true, then wouldn't an over generalized frequency chart only serve to influence the uneducated?
Dean Giamette

[klaus]

As virtually ALL published frequency response plots for condenser mics reflect imagined and assumed listening preferences (totally flat, with a slight rise in the "air" department, and a slight drop in the "mud" flats), it's no wonder that they all look pretty much the same- from beginner's Neumann to Holy Grail Tele (see below)

The fact that neither of these response curves looks anything as smooth in real life, once the "averager" imposed by the sales department is removed, is a minor inconvenience on the way to self-confirmation.

[saint]

Graphs of various TONES mean diddly squat in the real world. The ONLY thing that means anything with ANY mic, new, vintage, used is what it SOUNDS like with program... PERIOD! Brands have reputations, but you MUST take every single device in your chain (mics, outbaord eq's, compressors, etc., etc., etc. as if it were a person... same clothes (model); different faces (sound)!

[mike zietsman]

I agree Klaus,

surely it would be more fair if it were standard to not do any smoothing on the graphs and also if manufacturers would release things like frequency/resonance graphs and frequency/phase response graphs?  I feel like those are much more determining factors of a microphones sound. the famous ribbon mics have pathetic looking frequency response graphs (even with smoothing) but sound amazing due to the almost complete lack of audible resonance (which also makes eqing them a dream)

ultimately it doesn't matter what the graphs say though... it depends how the mic sounds and there is no substitute for listening to it and deciding for yourself...

[Kai]

.it would be more fair if it were standard to not do any smoothing on the graphs

It's not possible to measure a classical frequency response without smoothing.
Tha AC signal needs to be rectified to DC for measurement.
This process necessarily incorporates some amount of smoothing.

The question is, how much makes sense.
Very fine irregularities can be the result of imperfect measuring environment (room, source, structure and principles used for measuring).
Measuring a mic always means comparison with an "ideal" reference mic in front of a never ideal audio source (loudspeaker).
That's just where the problem starts.

There a two basic methods for this:
- #1 Side by side: in front of a hopefully perfect symetrically radiating loudspeaker.
- #2 replacement: measuring the reference first, then the specimen under test on the same place and using the difference.

This done in an anechoic room or using TimeDelaySpectrometrie to remove room influences.

I do measurements using both methods, #2 is by far superior, but does need some smoothing to deliver senseful results.

A usual LDC- or SDC-mic does not have a "sawtooth" frequency response, btw.
The response is quite smooth by itself.
All of the factory supplied graphs from Schoeps, Neumann, AKG and Beyer I controlled were realistic and by no means "oversmoothed" to hide anormalies.

These graphs should not be mixed with the catalog data, those are "idealized" by some amount to show the avaraged specs, not of a single sample.

So "factory supplied graphes are oversmoothed" might be another myth.

Regards
Kai

[klaus]

Can you name companies, aside of Sennheiser, which used to supply individual graphs for each mic, that still do this? And if that is still done by some manufacturers, are the measuring parameters published?

For example: "Measured at a distance from the sound source of xxx centimeters, at a sound pressure level of xxx dB SPL @1khz, measured at the capsule..." etc.

Without measurement norms and references to which all manufacturers agree, the inclusion of a frequency plot, even one of an individual mic,  is untrustworthy as predictor of the sound of a mic.

On the old PSW forum we had a thread where David Josephson, who heads and AES panel on microphone norms,  said as much: only measurement parameters which are scientifically defined and agreed upon by all manufacturers are a meaningful contribution to the objective knowledge base of a mic.

[Kai]

Can you name companies, aside of Sennheiser, which used to supply individual graphs for each mic, that still do this? And if that is still done by some manufacturers, are the measuring parameters published?

It's been some time that I bought a factory new mic - my studio is stuffed with mic's.
So without further investigation I can't tell.

I've noticed that some companies, like Schoeps and B&K / DPA build their mic's with so small individual tolerances, that even the catalog graphs tell the whole story that such a graph can.

B&K (now DPA) used to give all information, even with curves measured at different distances (30cm + 1m)
Level is standardized at 94dBSL with all graphs I've ever seen, although it wouldn't affect the response if another one was choosen (within certain limits).

I found the frq. resp. information alway useful for preselecting a mic in the mid and high range. Still there are great differences in the texture (smooth, hard) even if the graphs are comparable.

At LF the graphs are almost useless.
Specially directional mic's do all sound different here, even if the response looks similar.
Other factors like directivity and pulse response are more important for LF sound.

One has to admit that it's close to impossible to do valid measurements of directional mic's in the LF range below about 100Hz.
You need a very big anechoic room, and still you compare apple with pie - reference mic's are always omni and thus do respond to sound pressure (static air pressure).
Opposed to that directional mic's respond to sound waves (= moving air), so it's hard to build a valid setup.

For myself I most times "solved" the problem by using a certain Neumann U87 as reference, and just do relative comparisons (I wouldn't even call that measurement).
Still the results don't tell much about quality of bass, just quantity.

Regards
Kai

[soapfoot]

When I first began recording, I purchased secondhand an Oktava ML-52 ribbon mic that came with an individual frequency response graph for that exact mic, including the company logo (looked to be official documentation, not work done by the previous owner).  I am not sure whether this came from the Russian Oktava factory, or one of the many domestic "premium" re-sellers that attempt to add value to these microphones with modification and selection/testing.

[leswatts]

Well, I had to re register to post on the new forum.

With my new products I do supply an individual frequency response test for every single unit.

It's useful information but of course a single on axis response does not tell all.

A few random thoughts...

Certainly many published curves are "drawn by the marketing manager". They are virtually useless.

Much of the character of a microphone is determined by its off axis response. In particular the scooping effect well off axis of a single D directional unit. At a typical 1 meter distance Lf tends to go figure 8 (even with cardioid) and HF above delay element cutoff is pressure response. Thus the scoop.

LF measurements are problematic due to channel effect and other issues. The only way to get accurate data is to know both pressure AND particle velocity at the measurement point. A single calibrated reference pressure mic is not enough.

Microphones are not ears. I'll give an simple example.
I think most of us are familiar with the "bass suckout" in a typical room when the listener is at the room center
and a speaker is near a wall. Resonances result in a hole in the bass region that is all too easily heard. This is a pressure minimum. Our ears hear this because they are pressure sensors. But say we place a directional mic at the listening location. It's all or partly a pressure gradient sensor unlike the ear. So that pressure minimum is a pressure gradient (particle velocity) MAXIMUM, so the mic hears the hole as a peak.

So, an on axis frequency response measurement does not tell all. But it gives very useful information. It tells us
that we have a properly tuned diaphragm/backplate and delay chamber. If we didn't test individually the products could be all over the place.

We choose to share that information with the customer. There is a risk that the actual data not looking much like
"drawn by marketing manager" curves might have negative connotations but we believe it's the proper thing to do.

PS...I can't read the captchas...is it me? ;^)

Les Watts
L M Watts Technology

[klaus]

Thanks for your thoughtful reply, Les.

What still remains a mystery to me is what would be desirable to see in an ideal frequency graph.

Let's assume for a moment that slanted, self-aggrandizing distortions in the measuring process are eliminated, and the mic is measured by an independent party which pays close attention to any of the many treacherous variables that can falsify the outcome of the process. Let's also assume minimum smoothing of the curves is done as the signal (sine wave, most likely) is swept across the spectrum.

What shape of the curve would best represent what we like to hear in a good mic? A ruler flat response? Boosts or cuts at certain sections of the frequency band? A prominent peak or trough somewhere? And, if we want to include a polar pattern of the mic, do we really know that off-axis uniformity is a desirable feature of a good mic- not only visually ideal but acoustical as well? Has "uniform off-axis response"  been correlated and shown to be indicative of other desirable measurement features in a mic?

The mind would lead us to conclude that "flat" and "uniform off-axis response" is best. But that assumption may not take into account  that a microphone does not behave like ears and brain. (You showed with your pressure/pressure gradient experiment, that this may not be the case. I could cite numerous other examples of how the ear processes sound unlike a microphone, and how a good microphone's processing is at best a euphemistic approximation of our hearing.)

What I keep coming back to is this:
So far, visual representation in form of plots or specifications has failed to be a useful tool in predicting a microphone's sound and quality.

If a recording mic were to be only used as a faithful reproducer of sine waves fed into a loudspeaker, we would have less of a problem with the truthful visual representation of that event. But our goal is much more ambitious: we would like to visually validate what we hear out of the actual mic, so that that visual representation of sound can be a useful tool for the design, selection and marketing of microphones, which it still massively fails to do, all utterances of pocket protector types to the contrary.

I have my doubts that single event (sine wave) plotting and tracing will ever give us enough information about a microphone's ability to process complex waveforms in the time domain. Our ears are still best suited to predict a mic's usefulness as a tool to capture music.

The low priority for listening education in our business, and the resulting fear of many engineers and musicians to trust their ears to make esthetic decisions continues to fascinate me- topic for another subject.)

[klaus]

The impotence of comparing frequency plots of mics to shape sounds becomes apparent, latest when you considering pair matching.

I just finished modifying/upgrading a pair of new U87Ai mics for a prominent classical pianist. The mics sounded a bit different in timbre from one another- not much, but enough so, that I was not able to eliminate the deviation and optimise the sound by tweaking low or high end response in the amps. It was clearly the capsule character imprint that made the difference: one was a bit forward in the mids, but slightly attenuated in the highs. The other was silky throughout, with a little bit of a 'blah' factor in the intelligibility range. Not really missing anything there, but just not lively enough for my taste. 

Now, explain anyone to me, how what was desirable and what was not in these two mics could be detected by looking at a frequency chart, and what one could do about it to correct the shortcomings of both mics/capsules with the help of the chart? While the phenomena were clearly audible, a frequency chart would not be able to correlate this to what I heard, and offer a path to a solution. 

I ended up splitting the better sounding of the two capsules, so that I would have two front sides with similar timbres, then fine tuned capsules and mic amps a bit, to the point that both mics now sound indistinguishable from each other to my ears.

Facit: The outcome, rather than the starting point, would probably be the more telling and interesting exercise to plot a frequency graph of: how close would these plots now be? Or would they still tell different stories, while our ears judge the mics identical in sound?

[halocline]

The low priority for listening education in our business, and the resulting fear of many engineers and musicians to trust their ears to make esthetic decisions continues to fascinate me- topic for another subject.)

Ear training, what you call listening education, is typically the most painstaking and difficult skill in music education in general, not just in audio engineering. Developing the skill to distinguish between very fine gradations of sound is not something that can be simply learned through reading or conceptual understanding, it's based on (usually) years of effort, drive to learn, and opportunity to have continual access to source audio material in order to develop the increased familiarity necessary to make increasingly better decisions based on actual sonic perception.

There might be a useful analogy between written language and frequency plots; the information is there, but the expressiveness is typically found in elements not captured by written or visual representation. Maybe it's a dumb analogy...I don't know that much about frequency plots to really judge.

[Kai]

...what ...could be detected by looking at a frequency chart... a frequency chart would not be able to correlate this to what I heard,...

What's your measurement setup and method?
Maybe it's too different to your listening tests, which you once described as:
you do use your own voice as source and a Sennheiser HD 414 without foam as monitor.

When I hear sound differences between same model mic's I usually can track them down to the frequency response taken from 2 different distances with directional mics.

Regards
Kai