Clarification on Measurement and Microphones

Hi,
I have only been seriously reading up on speaker design for a week but even so it is obvious to me that the USB vs XLR microphone issue for driver measurement and crossover design has been thoroughly worked over. I half understand the difference as it related to timing and the superiority of an XLR microphone in this regard. I also have seen the unambiguous advice from Kimmo Saunisto on using XLR for measurements for use in VitruixCad.
However I also read the ASR guide that everyone seems to refer to: "How to make quasi-anechoic speaker measurements/spinoramas with REW and VituixCAD" - and it advises use of a USB Umik.
As I already have the Umik I am not really keen to buy another microphone unless it is truly necessary. So if anyone can explain the apparent conflict or perhaps clarify/quantify the real world impact of using the Umik I would be most grateful.
Cheers
Bill

 
I think you have read the relevant info. And wish people would read before buying stuff in general. That being said, while you can design your own crossovers without use of a timing reference, it's a bit of a PITA. The REW manual will help you with a workaround though. And there are other references online.

The problem is that, without a timing reference, a system like this cannot calculate the delay in a signal chain properly, because it doesn't know the inherent delay (AD/DA-conversion and on-card processing) nor the external delay (loop time from acoustic center of the speaker driver to the microphone). And you need that for correct phase measurements of every single driver, not for measuring the system as a whole. The workaround is either time consuming or not very exact, like picking a T=0 at 20% of the impulse response, denying that impulse responses are all but similar in rise time.

For once in a lifetime design work the extra trouble isn't that bad. If you're planning to design several speakers, get an old-fashioned XLR mike and a sound card with dual inputs and full duplex DA/AD (most entry models are, btw).
 
Thanks for that - I will look at the REW manual for more on workarounds. I have seen suggestions that this is best done with the mic and timing speaker in fixed locations with the test speaker moved for various measurements. If that is the case it strikes me as fairly laborious but we will see.
Your explanation has certainly helped but it is increasingly apparent to me that the fundamental problem here is that I do not understand how phase is addressed in the design process. I think I need to put some effort into that before returning to this specific issue.
Cheers
 
However I also read the ASR guide that everyone seems to refer to: "How to make quasi-anechoic speaker measurements/spinoramas with REW and VituixCAD" - and it advises use of a USB Umik.
Something to keep in mind about that guide on ASR is that it's tailored to measuring finished commercial speakers, and consequently doesn't get into a detail relevant to speaker design like timing between individual drivers.

It's a really good guide, it's just not all inclusive.
 
I do not understand how phase is addressed in the design process.
When using multiple drivers which will overlap (in the crossover frequency region anyway), their mutual phase difference comes at play. That is because two equal signals (assuming steady state sinusoidal signals) with the same phase add up to +6dB, listening axis. Two equal signals with phase shift of 180 degrees cancel each other out completely. And any situation inbetween gives a sum somewhere in between. So the phase differences of drivers are important in the crossover regions, just to get the levels right. And that is why you need accurate phase and delay figures of the individual drivers.
 
Your explanation has certainly helped but it is increasingly apparent to me that the fundamental problem here is that I do not understand how phase is addressed in the design process.
Hi!

(I learnt this from markbakk, allenb and other members who greatly helped me sometime ago)

Maybe some images and examples can help to explain.

Suppose Freq=4kHz. One period (360deg) has 1/4000 = 0.00025 or 0.25ms
Sound speed in air = 343m/s or 34.3cm/ms.
The length of one period (360deg) of 4kHz sinewave = 0.25*34.3 = 8.575cm
Half of the period (180deg) = 8.57/2 = 4.28cm

If one speaker is positioned 4.28cm backwards, for example, at 4kHz, they will be perfectly "acoustically" out of phase (180deg) even if electrically in phase.
At the crossover region, both speakers will produce the same frequency, but one sinewave reaches the listening point delayed thus causing the phase misaligment.

In any other distance difference, acoustic phase will be anything from 0 to 360 for the same frequency.
And if you change the frequency, the phases will be different for the same distance.

It's a mess - that's the reason why you need timed references in your measurements when you build your crossover, cause you will sum all speaker frequency and phase response. With timed measurements, all the phase effects will be considered (polarity, crossover, geometry and speaker acoustic phase changes) since REW detects the exact distances (in milimeters) that each speaker is from the microphone and register them in the .frd files.

In XSim or other crossover simulator, it will read those distances and take them in consideration.


1744656575377.png


REW measurement distance
1744657096259.png
 
Thanks for the responses guys!
To be clear I broadly understand phase and constuctive and destructive interference both electrical and acoustic. What I have no idea of is (a) how these are accounted for by the crossover and (b) more specifically how do I visualise the phase alignment and go about optimising it.
RonB made these two comments in this regard: you will sum all speaker frequency and phase response and In XSim or other crossover simulator, it will read those distances and take them in consideration
I was expecting a processs step where i was required to somehow look at phase curves and align them but are you saying that the design software includes both electrical and acoustic phase information in determing the summed frequency plot so that essentially all I have to do is optimise this combined SPL curve?
 
I was expecting a processs step where i was required to somehow look at phase curves and align them but are you saying that the design software includes both electrical and acoustic phase information in determing the summed frequency plot so that essentially all I have to do is optimise this combined SPL curve?
Correct. All that phase information should be captured in the measurement for the software to use, and that is why the absolute timing between the raw drivers needs to be captured when you do your measurements. The software will still produce a result if that timing information is missing, but without it, it cannot be known if the simulation is reflecting reality.
 
  • Like
Reactions: ron68
I was expecting a processs step where i was required to somehow look at phase curves and align them but are you saying that the design software includes both electrical and acoustic phase information in determing the summed frequency plot so that essentially all I have to do is optimise this combined SPL curve?
Yes, the crossover simulators, such as XSim, do everything for you, as long as you provide the .frd (frequency response + phase) file with timed measurements and the impedance file (.zma) of the speakers in the enclosure where they will be installed.
All your job is to work on the crossover elements, the filters, by defining the inductors, resistors and capacitors to feed each speaker.

The design tool will add everything together and present the final sum for you.
 
Thanks again guys!
That is great - still more to read but the basics are falling into place now.
I have been primarily looking at VituixCad. Fantasic software and quite comprehensive guide but it assumes you already know what you're doing. When this sint the case it is very easy to get lost - can't see the forest for the trees so to speak.
I did buy Dickerson's "Cookbook" many years ago but I seem to have thrown it out in a big clean up just recently - kicking myself as it is just what I need to understand the concepts better.
 
I am leaning toward getting an XLR mike - I have enough to get on top of without wrangling workarounds with the Umik and adding unneccessary variables.
I gather that mike is the same as the Sonarworks XRef20? Is the XRef20 preferred over the Dayton EMM6 or are they much the same?
 
They are quite the same, the Xref is a bit cheaper. Word is that the (individual) Dayton calibration files often are off, but I don’t know if they are better or worse than the ones Sonarworks provide. I had mine calibrated afterwards, here we have a few companies that offer this service at a reasonable price. Technically they are all based on what Behringer sells as ECM8000, those come without individual calibration files though and can be way off at high frequencies.

If you’re serious you’d check out Earthworks, NTI or other (semi)professional brands, but those set you back a few nice homebrew speakers pricewise. No need for that really imho.
 
I agree my skill level (and potentially only one set of speakers) - dont't warrant investing in a Earthworks or similar. I was initially leaning toward the Dayton but some searches left me with the impression the Sonarworks calibrations just might be a little better. Anyway found one at a good price so I have ordered the Sonarworks.
Started gluing up the boxes today - should be ready to start testing as soon as the mike arrives.
Thanks for the assistance.
Cheers
 
And that is why you need accurate phase and delay figures of the individual drivers.
From a practical standpoint, the main reason we need those references is because we have no real idea what the rest of the system(audio interface, ADC/DAC, power amplifier, etc etc etc) is actually doing.

This is exactly why USB microphones tend to fall short.

In fact, latency or delay can literally vary from one measurement to the next, depending on your operating system, drivers, or other variables.

So even a simple time loopback reference is objectively better than having none at all, especially because they are very cheap to implement!
Obviously this is not needed when just looking at individual drivers, but really essential for designing crossovers.