Do these measurements look right?

[Giallograle]

I've recently started using REW to measure the frequency response of a couple of drivers. It's a very powerful tool, so I've followed a couple of guides to get settings right, and played around with them when they didn't seem to be working well. However I'd appreciate if any other users could take a look and let me know if these settings look ok for my purposes.

I'm measuring the frequency, phase responses and impedance of a pair of Seas H254 tweeters and Kef B200 SP1039 mid/bass units in my Roksan Darius speakers in order to set the parameters for an active crossover.

I'm running REW V5.20.13 in Ubuntu on my Macbook. For measuring, I have a Line Audio Omni1 microphone, a Motu M2 USB interface and phantom power supply (one tip - I had to set PulseAudio to ignore the Motu to keep a working connection), and I built a loopback connection.

I set and calibrated the microphone input by a process of trial and error to minimise noise, and with my multimeter for exact levels, as the Motu has separate level controls for left and right inputs and headphone out which I used to power the speakers. I run the laptop on its battery when measuring to minimise mains noise. I'm measuring near field response from as close as I can get, around 5mm to the centres of the drivers; REW estimates about 30mm compared to the loopback signal.

Repeated runs with my latest settings give consistent results.

After adjusting the windows, I measured the B200s with a left width of 5.0ms, and a right width of 100ms. The Seas H254s with a left width of 2.0ms and a right width of 50ms. I also used the 'Estimate IR delay' function in Controls to correct the measurements.

Both drivers show quite large dips then rises (see final image); the B200 from 1750Hz to a peak around 3.5kHz, which corresponds to a jump in their impedance probably a result of a standing wave across the cabinet. Kef specify a 25-3,500Hz response. The Seas is supposed to have a 3-25kHz response, but has a 7db dip at 9.5kHz, and after a +3db peak at 13.3k rolls off rapidly; by 17k it's also down 7db and by 20k it's down 12db.

These are not current units although they've been replaced in the last ten years, however I'm wondering whether the results I'm measuring are correct, or whether the dips could be an artefact of my settings? Be grateful if anyone has any thoughts on this.

TIA

John

Impulse responses - I set the windows to include all of the 'noise' from the initial rise:

Frequency responses of all four drivers look closely aligned:

 

[AudiocRaver]

Not sure if this will help of not:

• The impulse responses need to be spread out more with a finer time scale to be useful. Even then, they might not show anything particularly vital at this stage of your work, just a fyi.
• 5mm might be too close for meaningful measurements. Extreme near field driver measurements can get the mic tip into a zone where the air movement is chaotic and gives somewhat problematic measurements, whereas farther away the air movement and frequency response settle down and make more sense. That is a different use of "near field" and "far field" than we usually run into, although valid in the world of driver design, as far as I understand.
• I am not an authority here, this is what I remember from my reading a few years back.
• Other than these comments, your measurements appear reasonable

 

[Giallograle]

@AudiocRaver Thanks for the advice.

The impulse responses need to be spread out more with a finer time scale to be useful.

Do you mean a shorter window to focus on the transient response? That smooths the curves, particularly at lower frequencies, but keeps the basic shape the same - see below. Or do you mean something else?

In case the headphone amplifier was running out of power, I ran a series of lower-level sweeps from the same position. All have exactly the same shape, SPL lower each time by the reduction in measurement level, so it doesn't look like an issue with the amplifier.

I moved the microphone to around 50mm and ran three sweeps, settings unchanged which look identical. The original dip at 7k is slightly smaller and sharper but there are a few more dips which look like artefacts of the measurement, perhaps reflections as these would be around 0.7ms.

Here is the original Seas frequency response from 5mm, vs three at 50mm, overlaid. The spl is more variable and the dip is shorter but still there. I'll try a bit further away again to see if I can get rid of the dip.

Here is the impulse response, using the same settings which capture the noise before and after.

Reducing the window to the transients makes little difference to the frequency response - the curved lower SPL ending at 600k:

The Seas datasheet for the H254 is on the Seas website (can't post the link) - note they've switched the descriptions.

Here are their response curves. Sound pressure and distortion on-axis:

Sound pressure on- and off-axis and impedance - a 4db peak on-axis around 10+k, but no dip:

The Line Audio Omni1 is measured in free-field so calibrated and optimised for sounds arriving on-axis, +-1db, 20-20kHz, >100Hz at a range of 70cm.

I've been using it on-axis. However my measurements differ significantly from Seas'.

Is there anything in my settings or procedure which I could change to achieve a better result?

 

[Giallograle]

I've measured at a series of distances up to 1m and played with the settings some more, which has improved the higher frequency results. The responses still look like they contain artefacts from the measurement process rather than the driver.

Shortening the window to 1ms either side of the impulse (a 1k wave, or 345mm) has helped remove possible reflections, but has reduced resolution to 1kHz, and the curves still have a series of waves which seem to become relatively larger as the distance increases.

It seems difficult to get accuracy at higher frequencies.

 

[ddude003]

Just a shot in the dark here... Any idea what room conditions were when those data sheets were taken? Were those from an anechoic chamber, a well treated room or just a random living room? And what about your current room conditions? Even if your only a meter or two away, the room can make for some interesting responses... Holding the mic or using a professional mic stand? Speakers on the floor on stands or table?

 

[skid00]

Room reflections will play a huge part in the smoothness of the curve. You'd have to suspend the speaker 20 feet in the air, or use an anechoic chamber, to match the mfr's curve.

 

[Giallograle]

Seas' frequency response curves are from an anechoic chamber.

Mine are from a small room; the speakers are on stands and the mic is on a decent stand. There are soft furnishings; the suspended wood floor is carpeted; the walls are reflective. However the minimum 1ms window I tried corresponds to a distance of 345mm so with no reflective surfaces within 345mm I thought measurements above 1kHz should be ok?

Multiple sweeps overlay almost exactly, and all display a series of 'waves'- the lower frequency ones move up in frequency as distance increases and the higher-frequency ones seem to be static. What could be causing this?

At the moment I'm not clear that reflections are causing this problem but if necessary, I could take the speakers outside to measure them.

 

[ddude003]

So my next guess would be that your headphone amp is not cutting it driving the speaker(s)... The frequency response look strangely similar to that of a headphone frequency response... Is that top line on the second chart showing that the Seas tweeter is running in the 5 to 7 ohm range across the frequency range? Any reason you don't use a power amplifier for your testing?

A quote from something I had read on the subject... "While some loudspeakers require very little power to produce audible sound a headphone amplifier is designed to play into an easier, higher impedance load. Speakers will draw too much current and you will get distortion at best and possibly damage your headphone amp."

It would be interesting to compare frequency range/ohm maps on the speaker vs a headphone and how the headphone amplifier responds with its power output given the various loads... See http://www.atma-sphere.com/en/resources-paradigms-in-amplifier-design.html the voltage paradigm vs power paradigm...