Tweeter rapid roll-off

[Rolledmyown]

Hi folks, I'm doing some tweaks to my home-made speakers. A number of years ago I bought a pair of Scan Speak D2905/930000 tweeters, which were reputed to be pretty flat in their response well beyond 25K. When I installed them I found that my Behringer 8024 Ultracurve Pro couldn't perform an GEQ on those tweeters at 20K, at which point their output was dropping like a stone. I still have those tweeters and am now using REW to attempt PEQ on my system (no longer with the Behringer installed, I use a laptop PC running JRiver MC31,and a Wadia di122 decoder as DAC and preamp). I found using REW that the output from the tweeter still falls like a stone, as can be seen below (this is a near field, on axis measurement, with only the tweeter connected via its crossover to my power amp. By 22K the roll-off is almost vertical). I'm using a UMIK usb microphone, with its proprietary calibration file (with my Behringer I used similar but different microphone from the same manufacturer). Initially I thought that maybe I had done something awful with my home-made crossovers (e.g., a bad impedance compensation or too much padding) or perhaps I've not attached them to the baffle adequate.. But maybe all is well and what I'm seeing is a measurement error and I don't have to get my soldering iron out!
I've included below the manufacturer's specs for the tweeter. I'm old enough that I can't hear well above 15K anyway, so I'm sure I'm not 'missing out' when it comes to the reproduction of my speakers, but I certainly want to know if I've made a mistake in installation of my tweeters. I've had a search of the specs for my Wadia and I think it says that its output is flat to about 45K. Is my laptop (pre 2010 vintage Vaio) the culprit?
PK

 

[rattus]

What is the sample rate of your audio device?

 

[Rolledmyown]

Good question, My Wadia.adjusts automatically to whatever the source was recorded at, so if I play a FLAC file ripped from a CD it will use 44.1KHz. However, when the REW sweep is used the sample rate appears to be 192KHz..

 

[sam_adams]

If you are using a Umik-1 you will be limited by the sample rate to a maximum frequency of ≈ 24 KHz. If you are using the Umik-1 with the MiniDSP supplied .cal file, there will be no mic correction above ≈ 20 KHz. If the mic response drops precipitously above 20 KHz—and it probably does—you will see a rapid falloff of the response above that point. If you have a Cross-Spectrum Labs calibrated Umik-1, the .cal file will be good out to 24 KHz. Just make sure that you added the sense factor line from the MiniDSP supplied file to the CSL .cal file.

If you are measuring nearfield after Keele, you must have the mic no further than 0.055 inches (about twice the thickness of a credit card) from the diaphragm of the tweeter. Even then you will probably get some severe interference artifacts because the mic is so close to the driver. If you measure from a distance of ≈ six inches, you should have the tweeter flush-mounted on a sufficiently wide baffle. Flush-mounting means that the front surface of the baffle must be at exactly at the surface level of the tweeter faceplate with no gaps or protrusion above or below the baffle surface or around the circumference of the tweeter. Anything else will result in response anomalies—especially at higher frequencies.

 

[Rolledmyown]

Thanks for the thoughts Sam! I did have a small burr around the perimeter of the tweeter, which I removed. No change in the response. There is also a small gap between the tweeter and the baffle, which I'll try and fix. My near field measurements were at about 1" from the tweeter on its baffle, and the measurements I took today (before and after removing the burr) we at the listening position.
My mike is a Umik 1 and I'm using the MiniDSP cal file. I assumed that even though the response was not calibrated over 20KHz the response close to 20KHz is only a few dB off so I was hoping that the performance of the mike at 20-25KHz would be close to that at 20KHz; maybe a bad assumption depending upon what REW does with the file when it gets a signal outside its calibrated range.
In any event, the roll-off of my tweeter starts to drop off well before 20KHz is reached, at which point the calibration should be OK.

 

[dcibel]

Problem is making decisions from near field measurement, and comparing it to IEC / IB measurement from the manufacturer.

Nearfield measurement accuracy is limited by diameter of diaphragm being measured, at high frequency you will get destructive interference from wave from on edge of the cone vs the other.

Back your mic up even 6" and you should see some dramatic difference is measurement response. Measurement for any EQ decisions should be in far field anyway.

 

[ddude003]

I believe the native rate for the UMK1 USB mic is 48kHz... I would set _all_ devices to use 48kHz so there would be no _automatic_ upscaling or downscaling from any hardware or software in the signal chain... The spec for this mic is "Frequency response: 20 Hz - 20kHz +/-1dB with calibration loaded"... You may be able to get a reasonable response to 24kHz, if your entire signal chain is capable of such, with the 48kHz HW/SW settings... For example, I do 0 to 24kHz sweeps with my kit and get reasonable results from 24Hz to 22kHz, which is what my entire signal chain is capable of (spec)...

 

[Anatoliy Gavrilov]

A number of years ago I bought a pair of Scan Speak D2905/930000 tweeters

The tweeters are a few years old - were they warmed up by the pink noise signal before being measured?
At the manufacturer, the sound pressure drop (at 20 kHz) passes at a level of - 2 dB from the average pressure. And in order to get the most similar measurements, you need to repeat the measurement conditions at the Scan-Speak factory.

so if I play a FLAC file ripped from a CD it will use 44.1KHz.

So a CD has a top recording frequency of 20kHz, right?

 

[ddude003]

An audio CD can represent frequencies up to 22.05kHz, the Nyquist frequency of the 44.1kHz sample rate... Ripping an audio CD at the native sample rate would give you a file of 44.1kHz... Depending on the ripping software you may have the options of upscaling, downscaling, possible filters and various output file formats like .wav, .flac, mp3, etc... A CD player may output an analogue signal at 20kHz while a more modern DAC and player software may output an analogue signal of the full 22.05kHz...

 

[Rolledmyown]

Thanks for your involvement folks, I'll keep pursuing this and fill you in if I get anywhere!