Microphones and Phase

[mortron]

I am still not a mic owner... And just before committing, came across something that I did not realize or maybe still don't understand.

Per another forum post here, they say the MiniDSP uMik-1 is unable to measure phase relationships like how a Dayton EMM6 and a 2 channel interface can. There apparently is a workaround but not ideal. Does that sound right?

Am wrestling with a Cross Spectrum uMik-1 vs an uncalibrated EMM6... As much as flat response is important, I am much more interested in things like phase relationships for crossover design. The accuracy of the uMik-1 is undeniable after calibration, but the almost $100CAD premium (after shipping) for something that gives up an important function seems misplaced. If I just want flat, uMik-1 makes sense but for crossover work the EMM6 uncalibrated would still allow me to extract phase relationships yes?

Does this sound sensible/rational?

[Chris Roemer]

I'd say it pretty much depends on which/how you use your XO design software.
If you're using JB's "PCD" (or a "workalike") then phase (collection) won't matter. You'll generate "minimum phase" files and have to enter "XYZ offsets" for your drivers into the software - to get proper phase "summing".
If you end up w/a mic that can capture "excess phase" (or the phase that includes "time of flight" - based on the distance from your mic to each driver) then you COULD use THAT phase - right out of your files.
While the 2nd method "works", it's no where near as "powerful" as running a model using min-phase data and offsets.

[mortron]

I haven't committed to any software yet. So would you say it's safe to... say... that the calibrated uMik-1 is going to be my best bet then?

[johnnyrichards]

The calibration has nothing to do with the phase, it is a limitation of USB based measurements. You really should get a calibrated mic, USB or otherwise. Two channel measurements are great and preferred, but with some careful measurement techniques you can derive correct phase from single channel measurements. Any margin of error will be down to how accurate you model things and how accurate you measure physical things like driver spacing and distance to the mic. On two channel measurements, those are accounted for during the measurement itself. The real difference is if you accurately derive those things using a single channel measurement, you can virtually move the mic around within modeling software. Using two channel measurements, you can physically move the microphone to see things. Six of one, half dozen of the other.

Actually, if you are new to measuring, the Dayton Omnimic is a good way to start - there is a lot of auxiliary support out there that is specific to Omnimic. I gave it a shot, but I prefer Arta quite a bit, not just because it supports two channel measurements.

[johnnyrichards]

I would also add that modeling software makes assumptions about driver behavior off-axis, this can lead to errata in the polar response as measured versus modeled. For some drivers, no big deal - but for drivers with breakup modes it can dramatically affect things including phase.

Measure measure measure.

[rpb]

I haven't committed to any software yet. So would you say it's safe to... say... that the calibrated uMik-1 is going to be my best bet then?

This is just my opinion. If you think your sim will, or should be able to give you the right x-over the first time, then the choice may be important. I honestly don't know. If you make a mistake along the way, it will not be as predicted anyway. On the other hand, it might match the sim perfectly, but sound wrong to you in your room.

On the other hand, If you have a few extra xo parts, and don't mind additional measurements, and testing, then the mic selection is less important. Start with a sim (if you want to use one ), and connect the woofer portion of the xo. Then you measure it to see if it does what you expect, or want it to. Change a few values to get it close to your target. Then do the same with the tweeter. Remeasure the speaker with both drivers, and look at the sum. Reverse the polarity of one driver, and remeasure. Make changes as needed, and play some music through the speaker(s) to check overall tonal balance, etc.

[rpb]

To elaborate a little. I use trial and error. I don't extract phase, or try to measure the z offset. I measure the raw driver in the speaker box, and then usually add a 2mH coil for a start on the woofer filter, if it's an 8 ohm driver. This takes some time, and a lot of sweeps, but I find it enjoyable. My mic is strapped to a golf club shaft, and I hold it towards the speaker by hand. The xo is on the floor connected with jumper leads.

[dlr]

I haven't committed to any software yet. So would you say it's safe to... say... that the calibrated uMik-1 is going to be my best bet then?

The mic calibration is only one part. The other is whether or not the software can provide a calibrated output. That said, I would always use a calibrated mic for SPL. That way you'll know that you'll have an accurate SPL for meeting your design target.

As others have noted, measured phase is not a requirement. Even if you have measured phase, for software that can model the off-axis, you'll have to generate minimum-phase anyway, so measured phase is useless for anything other than a single measurement point. What is important is relative acoustic offset. The caveat is that the off-axis will be influenced by diffraction, but for the crossover effect itself, off-axis modeling will still be useful.

This article may be helpful: Finding Relative Acoustic Offsets Empirically

dlr

[mortron]

Thanks. It sounds like the CSL calibrated UMik-1 is what will serve me best.

[dcibel]

If the intent of the mic is crossover design, 2-channel system is the way to go. Take it from someone who has been down both paths, I really wish that no one ever told me that a USB mic was a good idea for design. It sends you down a very limited path of single axis design with a rather high probability for error in phase data from this whole "extract minimum phase" and "determine offset" malarkey. Single channel USB mics should only be recommended for room EQ and measuring complete systems.

Especially for new users, fancy laboratory calibration is not a requirement, simply installing the mic in a mic clip can throw off top end response by 1dB or so. First step is to connect a mic, any mic, and start taking measurement with a 2-channel system like ARTA to start getting an understanding of the process and phase relationships, your questions on this topic will become more clear once you take some real measurements. With 2 channel system maintaining accurate phase relationships is easy - simply lock the FFT window start and maintain constant distance from mic to baffle. No min phase, no offset determination, and now you can include full 360 degrees of phase accurate information in your design, not just a single axis.

[dlr]

If the intent of the mic is crossover design, 2-channel system is the way to go. Take it from someone who has been down both paths, I really wish that no one ever told me that a USB mic was a good idea for design. It sends you down a very limited path of single axis design with a rather high probability for error in phase data from this whole "extract minimum phase" and "determine offset" malarkey. Single channel USB mics should only be recommended for room EQ and measuring complete systems.

Two channel is better, on that we can agree as much because it can eliminate all devices upstream from the driver, i.e. pre-amp, amp and speaker wire. However, there is no "probability for error in phase data" if determining offset is done properly and it's not difficult. I've even built in a section in WinPCD to make it even easier. It most certainly does not "send you down a very limited path". It is no more limiting to single axis design than is two-channel because the measurements must be minimum-phase with accurate acoustic offsets if one is to examine off-axis points with accuracy, not just a single measurement point.

Especially for new users, fancy laboratory calibration is not a requirement, simply installing the mic in a mic clip can throw off top end response by 1dB or so. First step is to connect a mic, any mic, and start taking measurement with a 2-channel system like ARTA to start getting an understanding of the process and phase relationships, your questions on this topic will become more clear once you take some real measurements. With 2 channel system maintaining accurate phase relationships is easy - simply lock the FFT window start and maintain constant distance from mic to baffle. No min phase, no offset determination, and now you can include full 360 degrees of phase accurate information in your design, not just a single axis.

There is nothing "fancy" about a calibrated microphone. It is essential to good measurements because an uncalibrated microphone may (will) have some level of non-linear SPL response. I have seen microphones with much more non-linear top end than the 1db that you mention. Even so, if at the very top end it's not as bad as non-linearities lower than that because it's common for designers to do some tweaking of the top to suit personal tastes. I do that, even though I use not just calibrated microphones, my entire measurement system is calibrated for absolute SPL as well.

If you enter on-axis measurements that do not have correct offset and are not at least close to minimum-phase into a software package, then off-axis simulations will not be accurate in the crossover area. This has been known for decades. The first measurement system I used was LMS, a stepped-sine system widely used in industry back then. There was a basic standard method required to provide phase data, but I never had accurate system response simulation results until the relative acoustic offset idea occurred to me. This was back in 1996 when I was using DOS 3.1 on my PC for design software. That was eventually the basis for my SpeakerBuilder article on acoustic offset, although no doubt I was not the first to come up with this method. But it's bullet proof if not properly.

Measured phase is not a requirement for design. Even though I use a two-channel fully calibrated measurement system, to be able to get accurate simulations in a software package (several of which I have used over the years), I always use generated minimum-phase coupled with accurate relative acoustic offset.

dlr

[dcibel]

Two channel is better, on that we can agree as much because it can eliminate all devices upstream from the driver, i.e. pre-amp, amp and speaker wire. However, there is no "probability for error in phase data" if determining offset is done properly and it's not difficult. I've even built in a section in WinPCD to make it even easier.

Probability for error, followed by "if done properly", and I've seen enough new users do this process incorrectly, myself included to know that there is enough probability to not do it properly.

It is no more limiting to single axis design than is two-channel because the measurements must be minimum-phase with accurate acoustic offsets if one is to examine off-axis points with accuracy, not just a single measurement point.

We are not in agreement here at all. Single channel measurement system really limits you to a single axis of design. In order to determine offsets you must keep the measurement axis constant, usually on the tweeter axis. Off-axis behaviour can be visualized somewhat with a piston model, but its missing some elements of diffraction and breakup modes etc that you will get from designing with complete "spinorama" of measurements. As you turn a speaker off-axis, the acoustic centre moves around a bit, measurement data generated by minimum phase extraction will have phase errors vs a 2-channel measurement with a locked window start. With full off-axis data set you can design with accurate reference of power response, early reflections, etc as detailed in that CTA document attached above. Minimum phase is not a requirement and in fact can be avoided completely.

There is nothing "fancy" about a calibrated microphone. It is essential to good measurements because an uncalibrated microphone may (will) have some level of non-linear SPL response. I have seen microphones with much more non-linear top end than the 1db that you mention. Even so, if at the very top end it's not as bad as non-linearities lower than that because it's common for designers to do some tweaking of the top to suit personal tastes. I do that, even though I use not just calibrated microphones, my entire measurement system is calibrated for absolute SPL as well.

Emm-6 is "calibrated" though right, just not by CSL out of the box? It's calibration is just not a very good one. Linearity is important, calibrated by a laboratory or by simple repeated measurement against a "known good" reference mic. I own an EMM-6 which is rather linear out of the box without a calibration file, compared to Omnimic, and I recently moved on to a Line Audio OM1 which is just as linear as either of my calibrated mics without any calibration file at all to compensate.

What it comes down to IMO is that there is a gap between "it can be done" and "the recommended process". I've voiced my opinions, anyone or everyone is welcome to take it or leave it. I moved to a 2-channel system and never looked back, and that is the process I would recommend to all users, new or otherwise. As far as software goes, I started my 2-channel journey with SoundEasy, but have since moved on to ARTA and VituixCAD which is also my recommendation.

[dlr]

I don't disagree that a set of measurements such as a "spinorama" is the best, that requires a significant effort. But a calibrated microphone for SPL is still, IMO, a requirement. Keep in mind that software that systems that can process a "spinorama" haven't been available for all that many years. One could always measure off-axis. It's also not a simple task to make that large series of measurements accurately, especially below about 1KHz if one is constrained to typical in-home rooms. There's also the issue of near-field/far-field integration that is difficult enough for single point measurements, especially for the neophyte. I think you do a disservice to someone just entering into this to totally dismiss what has been used for literally decades in software still in use and in some cases much easier to use as a start into the hobby.

dlr

p.s. If I recall correctly, VituixCAD was not initially free software. Neither is SoundEasy (Bodzio). I still have and use the Ultimate Equalizer from Bodzio Software for my dipole system, but it is a rather complicated piece of software. Mine has been from day one, though not as elaborate and geared more for the new user, kept as simple as I felt I could for an easier entry. And initially to mimic Jeff's PCD. But those two require minimum-phase data if more than one design point is to be modeled accurately.

[mortron]

No wonder speaker design is so tough... Just figuring out which mic to use is hard enough LOL.

It would seem there are adherents to both schools in here and it comes down to experiences and personal preference? If there was just one way to do it it would be known as "the way" I suppose.

I like the idea of a calibrated mic, but things like phase etc are things that jump out to me as being important so it has kind of muddied the waters some for me and my limited knowledge at this point. Always thought calibrated or bust was the best way no questions asked, but some questions change that answer now I guess.

Would hate to be mid project and find out my tools are selling me short. Hence my endless questions about the abilities of the two mics. Will have to do more reading. Thanks

[dlr]

Let me reiterate. I do agree that the best system is a 2-channel setup. However, IMO the primary benefit is that even if not a fully calibrated system (as opposed to just the mic), the 2-channel system totally eliminates any upstream non-linearities. The pre-amp, amp and cabling are all made moot which is a significant benefit. You get measured phase as well, but there are separate issues with phase that are probably best addressed in a separate thread. Add a calibrated mic to that and you'll get much better SPL measurement, whether absolute (i.e. a calibrated system) or relative (requires the same drive level applied to each driver tested).

dlr