I can not comment on this particular design, or crossover, but i always measure fr response with tweeter both polarities, in phase, out of phase, just in case. Physical allignement plus generic crossover, you never know what you get. Better to be sure. To measure is to know. Plus its not that hard or expensive, its usefull for future, and its fun. REW or Holm are both free, umic is cheap.
I agree, the phase is unpredictable, without measuring and simulating you can't make a good sound. Clio Pocket is one of the options, its cheap but reguire post procesing with microphone calibration file, result than is very good.
I believe so, and it costs nothing to evaluate on your own. If I could measure it I could give some better answers.
HiVi drivers usually have good response behaviors. This makes them behave fairly predictable in a crossover that is intended for their impedance. Power response is the unknown factor in the equation, as we don't know the off axis slopes of the drivers on the baffle, and the midwoofer may benefit from a Zobel impedance correction circuit to steer it more accurately with the intended slope, but stuff like that comes with time, measuring and modelling.
I use Vituix CAD and XSim3D, both free and effective. To measure, I use a MiniDSP Umik-1 with REW and ARTA software. These do not expire nor require any SaaS subscription garbage.
When I designed a time aligned, linear phase 2-way, there was considerable planning to ensure it behaved as desired. I was repurposing a cabinet with a flat baffle, so I used trigonometry to calculate the angle and timing, designing it so a listener sitting on an average couch/chesterfield would be on the correct axis. At 2.5m distance the tweeter was time and mostly phase correct with the woofer. Although offsetting the tweeter backwards would have been better for the alignement, it turned out very good. You can achieve the same thing with your speaker by changing the listening distance, height, and crossover.
Room treatment is also important, and with those room dimensions of 12x11ft if it's untreated it can have some big adverse affects. It's all a process, as I say now.
A hack...
If you can trace the crossover schematic - you should be able to see either an L Pad (resistor pair just before the tweeter - one in series with the tweeter, the other in parallel) or maybe a series resistor up front in the tweeter crossover.
You could reduce the value of the series resistor to increase tweeter output if that seems dull to you.
I call this a hack - since an off the shelf crossover is never designed to work with specific driver combinations. Look at the raw impedance curve of any "8 ohm driver" and you'll see it isn't 8 ohms flat throughout its frequency response. It may be for example 5 - 6 ohms at the crossover point, and depending on voice coil inductance, could be rising up in frequency making the actual crossover point and attenuation unpredictable.
Your best bet is to get the overall frequency response measured to see how good (or bad) it is, then make a crossover redesign decision.
PS: the 92dB sensitivity sounds rubbish. If the D6.4 is ~ 88dB sensitive, then with 3 - 4 dB of bafflestep compensation, then your overall sensitivity will be more like 84-85dB, not 92dB
If you can trace the crossover schematic - you should be able to see either an L Pad (resistor pair just before the tweeter - one in series with the tweeter, the other in parallel) or maybe a series resistor up front in the tweeter crossover.
You could reduce the value of the series resistor to increase tweeter output if that seems dull to you.
I call this a hack - since an off the shelf crossover is never designed to work with specific driver combinations. Look at the raw impedance curve of any "8 ohm driver" and you'll see it isn't 8 ohms flat throughout its frequency response. It may be for example 5 - 6 ohms at the crossover point, and depending on voice coil inductance, could be rising up in frequency making the actual crossover point and attenuation unpredictable.
Your best bet is to get the overall frequency response measured to see how good (or bad) it is, then make a crossover redesign decision.
PS: the 92dB sensitivity sounds rubbish. If the D6.4 is ~ 88dB sensitive, then with 3 - 4 dB of bafflestep compensation, then your overall sensitivity will be more like 84-85dB, not 92dB
I'll do the measurement and come back to you guys. Is there a cheap and reliable calibrated measurement mic?
Yes - the MiniDSP UMIK-1 comes with a per mic calibration file. I think this would be accurate enough for your needs (it's what I personally use).
Calibration is not relatively important. See that your mic can readily keep time if you want to do crossover work (preferrably not USB mics).
I'd rate calibration more important than keeping time. most omnidirectional capsule based mics are not flat below about 200Hz nor above 1KHz and can have as much as 5dB non linear variation. this will make you "over read" a tweeter (climbing response) and over pad the treble,. This cannot be compensated.
Timing however can be. If you take 2 separate measurements of drivers with a USB mic with a different beginning impulse, you can still align these measurements.
For each measurement, discard the measured phase and derive minimum phase (drivers being minimum phase devices)
Take a combined measurement of both drivers in parallel. It will be wonky as = no problem
Put your separate derived measurements into your crossover software per driver and draw a network with both drivers in parallel
Adjust Z offset of the woofer, until the summed response of both individual derived driver responses = the combined measurement "wonky" response.
this gives you an accurate acoustic offset for accurate crossover design without having accurate measured phase.
there are other techniques with REW using a reference signal I can't recall.
Timing however can be. If you take 2 separate measurements of drivers with a USB mic with a different beginning impulse, you can still align these measurements.
For each measurement, discard the measured phase and derive minimum phase (drivers being minimum phase devices)
Take a combined measurement of both drivers in parallel. It will be wonky as = no problem
Put your separate derived measurements into your crossover software per driver and draw a network with both drivers in parallel
Adjust Z offset of the woofer, until the summed response of both individual derived driver responses = the combined measurement "wonky" response.
this gives you an accurate acoustic offset for accurate crossover design without having accurate measured phase.
there are other techniques with REW using a reference signal I can't recall.
The crossover primarily needs relative data. Any mic can make an accurate cross. The global voicing needs human input regardless of the mic used.
Not all "off the shelf" crossovers are bad - for a select amount of drivers. Attenuations may be required.
Re. topic > I would never have tweeters below ear height.
But global voicing can be problematic if your mic is providing a non-flat response. That is - if you are aiming for a traditional tilted in room response (-1dB for every octave above 1KHz).
I'd rather a mic show me what the speaker is doing rather than what the mic is doing. This isn't about bad data leading to the inability to design a crossover, , it's about having reliable data so you can design a target crossover matching reality.
I'd rather a mic show me what the speaker is doing rather than what the mic is doing. This isn't about bad data leading to the inability to design a crossover, , it's about having reliable data so you can design a target crossover matching reality.
But a stock UMIK-1 is usually not very optimal for this, because it comes with a batch calibration file only. If you want an accurate calibration for the exact mic, you need to buy e.g. from Cross-Spectrum Labs. The stock (batch) clalibration can have a deviation of even 2-3 dB from reality, okay this is still better than without any calibration but could be better.
Hi,
did listening experiments with speaker height, few quick observations: obviously some frequency response issues ensue, but for stereo image speaker below ear is fine, stereo image appears at natural height. If speaker was higher than ear level however, the stereo image felt being too high. So, if speaker is not at optimal height, it is better to have it lower than higher, in my opinion 😉
Last edited:
Clio Pocket 2.2 and separately ask for microphone calibration file, its paid too. But with booth you are ready for speakers work.
Speakers designed for floor placement will experience a reduced low frequency balance if lifted above of the floor by stands. Alternatively, tilting the speakers will lift the treble balance, while maintaining the bass balance. Whether that’s subjectively preferable or not would have to be evaluated by audition.