Topic: Mic mods and tinkering...

[goodcooker]

I bought a pair of Cascade M39 small diaphragm mics for super cheap yesterday and I'm thinking about doing the capacitor swap that people say makes them perform better and with more closely matched output.

Anyone done a mod on a cheap mic? Most of them are based on the based on the Schoeps circuit (for SDCs) or the Neumann U67 circuit (for LDC).

The real difference is in the capsule and diaphragm but I'm hoping to get a decent pair of second stage mics out of the deal for well under $200.

[DSatz]

Hmmm. Mass-produced capsules vary in their frequency response and sensitivity far more than electronics commonly do nowadays, so if closer matching is really a common result (a claim that surprises me somewhat), that could be a function of component choices that you could just as well make with the existing circuitry.

There's no reason that the Schoeps output circuit shouldn't be used with large-diaphragm capsules. The Neumann U 67's circuit is peculiar because the capsule itself is inherently quite bright sounding, so the amplifier purposely rolls off the high end by about 5 dB. There's also a fairly steep low-cut filter below 40 Hz, which was the subject of a patent back in the day. The only reason I can see for using that circuit is if a particular capsule is overly bright--but then again, that applies to a lot of mass-produced capsules. Still, even Neumann didn't use that rolloff in the mikes that they specifically designed for more distant pickup using those same capsules, such as the stereo and quad mikes ((U)SM 69 and QM 69).

I have had several professional condenser mikes modified, with mixed results. Some of the claims for the mods held up in practice but there were "down sides" that weren't mentioned before the surgery occurred (e.g. one pair of AKG mikes got 4 dB better dynamic range, but were sensitive to RFI after the mod; one pair of Neumanns became different- and arguably more interesting-sounding, but had reduced headroom and no longer matched one another well). More often than not, the results didn't live up to what I was assured would occur. Some people who are in this line of development seem to believe in what they're doing, to a degree that isn't always met by the actual results of what they offer. In all cases, no matter how prominent the person was who modified the microphones, I have had difficulty reselling them at all, and have "taken a bath" on the prices I was able to get.

TL;DR: This is a realm of smoke and mirrors, though with some benefits to be gained--but all in all, one is better off starting with good capsules, since they determine 80% or more of the sound quality of a microphone.

--best regards

[Gutbucket]

I've long thought about what might be most effective strategy for optimizing less than stellar microphones that suffer from frequency, phase and sensitivity tolerance variations between units.

What about taking a measurement with each microphone, comparing those against the same made with a reference mic, and inverting the difference to create a corrective filter for each microphone that either directly emulates the response of the test microphone or goes on from there to further tailor the response as desired?  That should effectively match on-axis frequency, phase response and output level within the available dynamic range limit of the microphones.

Most of the work would be in performing the measurements and creating the correction filters.  The recordist would then label each microphone to keep it associated with its own corrective filter, then apply the filters to the files prior to doing any other post production work on the recording.

Many folks here may be familiar with DIY speaker builders doing this as standard practice these days using a measurement microphone and a DSP box or software that applies the crossovers in addition to corrective filtering. So this kind of thing is now within the domain of the DIY enthusiast.

Directly analogous is the TetraMic (and now OctoMic) ambisonic microphone systems from Core Sound which use such a method to linearize and match the responses of its four cardioid capsule elements (eight in the OctoMic) sufficiently such that the ambisonic A-format to B-format conversion applied on the computer after a recording is made works correctly.  Each of these microphones is provided with a set of dedicated corrective filters which are associated with that specific microphone by its serial number.  In fact, I'm overdue to send my TetraMic in for re-calibration, after which I'll have a new correction file to apply to any new recordings made with it.

Such a discussion seems to me to fit the thread title, but I'm more than happy to make a dedicated thread to discuss it if you'd prefer, goodcooker.

[goodcooker]

[quote author=Gutbucket link=topic=198572.msg2369375#msg2369375 date=164064455
Such a discussion seems to me to fit the thread title, but I'm more than happy to make a dedicated thread to discuss it if you'd prefer, goodcooker.
[/quote]

I included tinkering in the title to encourage this very line of thought.

[goodcooker]

Hmmm. Mass-produced capsules vary in their frequency response and sensitivity far more than electronics commonly do nowadays, so if closer matching is really a common result (a claim that surprises me somewhat), that could be a function of component choices that you could just as well make with the existing circuitry.

TL;DR: This is a realm of smoke and mirrors, though with some benefits to be gained--but all in all, one is better off starting with good capsules, since they determine 80% or more of the sound quality of a microphone.

I agree 100% with these statements. A lot of the Chinese manufactured mics look exactly the same and are in fact in the exact same machined housings but differ in sound quality from brand to brand and often from mic to mic within the same brand. Whether that is a function of design choice, manufacturing tolerances or something else is anyone's guess.

Significant performance value can be had when swapping components. I used a pair of Oktava MK012 modified by Joly for a while. The capsules were good quality and included serial number specific graphs from the factory but the amplifiers were almost an after thought. It was speculated that since components were difficult to keep in supply in Russia that pretty much any capacitor that was close in value and would fit went in the circuit. Mine were rebuilt with closely matched higher quality capacitors and resistors and it changed the harsh peaky sound of the mics considerably. I did a side by side informal comparison with the hypercard caps and a pair of MK41 and was quite surprised. I found the Oktava to be much more pleasing.

I'm hoping that the Cascade mics that I bought are decent enough to be useful on their own and that changing the cheap ceramic capacitors for better quality closely matched metal film and polypropylene components can improve them. The change is likely to be negligible but you never know until you try. I'm out $20 and an hour of tinkering if it doesn't amount to much. I'll do some informal tests along the way.

These aren't going to be my main mics. I'm going back to MBHO subcardioid capsules next year. Most versatile mic in the tool kit (and best bang for the buck) if you ask me...

[Gutbucket]

One attractive aspect of a correction filter is that there needn't be any modification made to the microphone itself.  But of course it might be used in conjunction with component mods as well.

I do want to be be a realist about what a correction filter will be able to achieve.  Even though one might take this beyond the correction of measurable on-axis flaws and improved response matching between pairs, modification of the overall response as an additional sweetening step is not the primary idea I'm proposing.  That would be more like the automatic application of overall EQ as a first processing step, albeit in a particularly accurate and repeatable way.  What I'd like to focus on instead is matching response across multiple microphones and fixing "correctable" flaws and response irregularities, rather than changing the microphone's general character.  We aren't going to use this to turn a Cascade M39 into a Schoeps MK4, or Neumann KM84, or whatever. 

A microphone's general specifications of self-noise, distortion and SPL cannot be modified using this method, and represent limits within which any correction filter must work.  The "filter insertion losses" of our corrective filter can only negatively impact those absolute specification measures.  Yet within those limits, we can modify the raw response of each microphone using digital filtering, and the steps for doing this will be: 1) measuring , 2) creating a filter based on that measurement, and 3) applying the filter.  Each step done in a repeatable and accurate way.

[Gutbucket]

I can imagine several different ways of doing this, with various levels of commitment and potential reward. Here's the simplest, which seeks simply to more-closely match the response of two microphones with each other without making any other corrections, requiring nothing more than the two microphones to be matched and a "matching EQ" type plugin.

A simple method for achieving level and frequency response matching only-

1) Go out and tape some music.  Use your regular recording setup and make a great tape.  But in addition, bring along that pair of cheap mismatched mics and run those into a second recorder or extra channels on your primary recorder.  Set up the cheap mismatched mics coincident and parallel with each other.  Maybe just gaff tape them together with the capsules aligned and point the pair at the stage.  Set the gain on the preamp/recorder identically for those channels even if they meter differently and roll.  Concern yourself only with getting the two mics closely aligned with each other physically and in terms of recording trim.  Don't worry about where they are pointed, how they are suspended or any of that jazz.

2) Come home and listen to your tape from your regular recording setup.  Feel good about making it (or not, as is your prerogative as taper). Okay enough of that, move on to the cheap mic pair.  Pull it up on the computer and listen to one of the two microphones in mono.  Switch to the other (also in mono).  Go back and forth and decide which one of the two you prefer, even if they both sound like shit.  Adjust levels as required in order to make this decision, then return the levels to nominal.  Don't make any other modifications other than level while listening.

3) Mark a repeatable time range in the file.  Pull up the matching EQ and run the preferred microphone channel through it, using its function for capturing the overall time-averaged spectrum response over that time range.  Store that curve with some name that is associated with that particular physical microphone (mic 1, raw) or whatever. 

4) Pull up the less preferred microphone channel and run it through the matching EQ, and capture its overall response spectrum over the same time range as before.  Store this curve with a name associated with that particular physical microphone (mic 2, raw) or whatever.

5) Use the curve matching function of the EQ plugin to generate an EQ curve that modifies the response of mic 2 such that it matches the response of mic 1 across the selected time range.  Save this curve (mic 2, correction).

Now go record something else using these mics, except now using a proper stereo microphone configuration (perhaps you taped the two mics together for the first set or opening band, then ran them in a normal stereo configuration for the second set or main act).  When you get home and offload the files to the computer, apply the "mic 2 correction" EQ file you generated earlier to the mic 2 channel.  Leave the mic 1 channel alone.  Compare the result against making no EQ correction.  Better?  You now have manipulated the second microphone to sound more like the first.  You haven't corrected any flaws they both share, but you've made the output of the two more identical.  Go ahead and do whatever manipulations you do to make your recordings sound good from there.  This correction is simply a first post-production step, and remains optional.

[Gutbucket]

A more advanced technique would use a higher-quality, preferred microphone as the reference, which is better than either of the lesser mics being corrected, resulting in the generation of two correction filters that aim to emulate the response of the preferred microphone.  Gaff tape all three together for the initial step (I'd not tape any more than 3 together in this way though).

Advantages of both these approaches are that the test setup is easy! ..and music, or whatever you normally record, serves as test signal.

The additional advantage of the second approach is the reference response is that of a higher quality microphone. Ideally one preferred for that particular recording situation. 

[edit] This should be advantageous in two ways.  It helps fix some frequency response problems that both target microphones being corrected share in addition to those that only one may suffer. And it tailors the resulting matched frequency response to more closely emulate that of the higher quality preferred reference microphone.

[rocksuitcase]

Your stated method of measuring and comparing, in each case seems a solid experimental approach.
In general, most "Presets" on receivers or DAW based software are based on these types of experiments.

[Gutbucket]

^ The key being the generation of a tailored corrective preset specifically matched to the deficiencies of that specific piece of gear (microphone).  "Close enough" in this case has to be really close, requiring measurement that is specific to the problem in order to end up with something better rather than worse or just different sounding.

This has one great advantage over speaker correction filters and room correction techniques, both of which are now common, by way of inverting a problem those approaches are unable to escape.  With speaker and room correction the test signal is simple and non-dimensional, but one must choose a listening position from which to measure, and the correction will only truly work for that point in space.  From other listening positions it can and often will make the perceived problems worse.  Yes, there are averaging and windowing techniques used to mitigate that, but the problem is rooted in complex geometry differing on the listening end, making it essentially unavoidable.

We would instead be doing something of the inverse. The correction in our case will always provide the same effect on the output past the microphone.  The corollary to the aforementioned problem is the geometry on the opposite, test signal input side.  Using this technique, we can't fix or modify the polar behavior of the microphone, only its overall aggregate response with regard to frequency, phase and level.  So the question becomes whether to try and target the direct on-axis response, overall diffuse response, or what - the complex geometry on the other side of the microphone. This is where the use of the our standard "program material" (live music in a venue as we would otherwise record it) as test signal is not simply a super convenient way to do this, its a test environment that provides the same or very similar expected balance of direct and indirect sound that the microphones will experience when used in other taper recording situations. 

So we don't need to really do much out of the ordinary to setup a good, applicable, and generally repeatable test situation, except to go tape something like we typically would, using the music in that acoustically complex geometric environment as test signal, which is representative of how we will use the microphones anyway.

[rocksuitcase]

the use of the our standard "program material" (live music in a venue as we would otherwise record it) as test signal is not simply a super convenient way to do this, its a test environment that provides the same or very similar expected balance of direct and indirect sound that the microphones will experience when used in other taper recording situations.

Your overall premise seems sound.
The quoted portion is, to me, the key concept as to why your method would yield specific measurements correctly.
By having the test signal comprised of "typical AUD taper signals", both diffuse and direct, the amount of filtering post capsule should be relatively constant .

[Gutbucket]

Thanks for thinking this through with me.

In thinking about it a bit more, I realize that when the initial "test signal measurement step" is performed, the direct/reverberant ratio at the recording position will determine which reference mic response measure will influence the corrective filter most strongly once applied to the target mic.  For example, if the three mics are tapped together and pointed toward the stage at an outdoor show where direct sound from the PA tends to be dominant in comparison to the reverberant sound as heard at the recording position, the on-axis response of the reference mic will be what is most closely emulated. Whereas if that measurement step is performed in a more reverberant indoor space at a recording position that tends to be dominated by random reverberant sound arrival from all directions, the average reverberant response of the reference microphone will have a greater influence on the corrective filter.

What's the take-away of that? Well, one might make several sets of corrective filters initially, measured in different acoustic taping environments, then choose whichever one works the best.  Or, one might want to be able to select which filter set to use based on the acoustics of the current recording situation.

[rocksuitcase]

Thanks for thinking this through with me.

giggity! it is fun to think.

What's the take-away of that? Well, one might make several sets of corrective filters initially, measured in different acoustic taping environments, then choose whichever one works the best.  Or, one might want to be able to select which filter set to use based on the acoustics of the current recording situation.

Then you start heading toward digital microphone post processing, tetramic territory. ? having several pre-sets you can use and/or analyzing the sound in real time and creating the filter out of that. (which is my sense of what you ACTUALLY can achieve by post processing the tetramic multi channels)

[Gutbucket]

I don't want to confuse things with a tangent, but a closely related application I've thought about which may potentially benefit greatly from the technique mentioned above is the improvement of stealth rigs.  In that case, mics tend to be mounted in weird ways in weird places, sometimes muffled, often with their response influenced by closely adjacent boundaries and other obstacles.

I wonder about using a preferred open-recording stereo rig, properly arranged and setup, as reference.  Perform the measurement recording while standing beneath that, running the stealth rig.  In this case the reference mics and target mics would no longer be coincident, so there would be phase randomization above some frequency, but not sure that matters if the filtering only targets frequency response and level.  Perhaps a diffuse test signal should be used, such as applause.

I run four identical miniature omnis and the way I have them arranged and mounted creates a somewhat different response in one pair verses the other, and to a lesser extent, in one of the two of the second pair.  Ideally I'd like to correct for this so that when I do post processing I begin with an equal timbrel and level balance across all four channels from the start. As it is, I EQ them in pairs as well as individually by ear as necessary to achieve the same end result, but it would be preferable, save time, and quite possibly make for an audible improvement as well, to start with them all leveled and flat with respect to each other.  The caveat is that the stealth setup needs to remain the same each time.

Similarly, this could be used to transparently compensate for the muffling effects of windscreens.  One filter for big Shures, another for dead rats.. based on the measured response of the same mic without those installed.  That's probably unnecessary though, as its easy enough to compensate for by ear with a relatively simple EQ curve.  Would be interesting to see the actual effect of them in a visual curve however.